RENEWABLE ENERGY; WHAT ARE THE LIMITS?

Ted Trainer.
Faculty of Arts, Univ. of N.S.W., Australia.
This attempt to assess the potential of renewables is being updated
as new information is received. Critical feedback is welcome.
4/2004

It is commonly assumed that rich countries will be able to meet their energy demand from renewable sources. However the following evidence on existing and probable future efficiencies and costs indicates that it will not be possible to derive sufficient electricity or liquid fuels to sustain the present high per capita rates of consumption from renewable sources, let alone those growth will require. There must be a transition to reliance on renewables, but a sustainable future cannot be achieved without significant reduction in current material "living standards" and in gross economic activity. However, advocates of "The Simpler Way" argue that a radically alternative society based on frugal lifestyles, zero economic growth and local economic self-sufficiency could defuse global problems and provide a high quality of life.

Contents.

The context.
PV solar electricity.
Storing energy.
Solar thermal electricity.
Wind energy.
Liquid fuels.
The "hydrogen economy".
Conclusions: How much energy might we get from renewables?
The significance of the commitment to growth.
What about "dematerialization" and transition to a service and information economy?
What about technical advance and "factor four" reductions?
Conclusions.

THE CONTEXT.

In the last three decades considerable concern has emerged regarding limits to the future availability of energy in the quantities required by industrial-affluent societies. More recently Campbell (1997) and others have argued that the energy source on which industrial societies are most dependent, petroleum, is more scarce than had previously been thought, and that supply will probably peak between 2005 and 2015. (Fleay, 1995, Ivanhoe, 1995, Gever, et al, 1991, Hall, Cleveland and Kaufman, 1986, Laherrere, 1995, Duncan, 1997, Bentley, 2002, Youngquist, 1997.) These people argue that non-conventional sources such as tar sands and shale oil will not make a significant difference to the situation. The world discovery rate is currently about 40% of the world use rate. The USGS (2000) has recently arrived at a much higher estimate for ultimately recoverable petroleum, but this would only delay the peak by some 10 years.
_______________________

*This paper is an extension of earlier discussions: Trainer, F. E. (T)., (2002), " Can solar sources meet Australia's electricity and liquid fuel demand?", The International Journal of Global Energy Issues, Trainer; F. E. (T.), (1995c), "Can renewable energy save industrial society?", Energy Policy, 23, 12, 1009-1026, Trainer, T. (F. E.), (1995a), The Conserver Society; Alternatives for Sustainability, London, Zed Books.

If the discussion is expanded to take into account the energy likely to be required by the Third World the situation becomes much more problematic. If the present world population were to consume energy at the rich world per capita rate world supply would have to be 5 times its present volume. World population is likely to reach 9 billion by 2070. If 9 billion were to consume fossil fuels at present rich world per capita consumption rates all probably recoverable conventional, oil, gas, shale oil, uranium (through burner reactors), and coal (2000 billion tons assumed as potentially recoverable), would only last about 20 years. (Trainer, 1985.) As will be discussed below, when the universal commitment to economic growth is added, the magnitude of the problems associated with the future availability of conventional energy sources become much greater.

The alarming nature of the energy predicament is made most graphic if considered in relation to the greenhouse problem. In a technical report for the IPCC (Enting et al, 1994, 2001 electronic version) estimate that to stabilize the atmospheric concentration of carbon dioxide at 650 PPM, twice the 1970 level, annual emissions must not exceed 8-12 GT/y by the end of the 21st century. Such a target is much too high as we are now 30% above pre-industrial level of 270 PPM and serious effects are becoming evident. However if this target is taken and world population rises to 9 billion then the per capita emission allowable will be approximately 1 ton. Yet the present Australian emission per capita from fossil fuel burning is 3.6 tons. In addition there is another 3 tons per capita released from land clearing, making a per capita total of 6.6 tons. (Enting et al show that for more acceptable targets emissions must be cut to zero and held there for decades.)
 

Thus the per capita use of fossil fuels should be cut to a small fraction of the present Australian amount. Clearly consumer-capitalist society cannot be sustainable unless vast quantities of energy can be derived from renewable sources to almost entirely substitute for fossil fuels, and to cope with continued economic growth. If not then a sustainable society must involve dramatic reduction in energy use.

Given this context in which there are grounds for expecting increasing and extreme energy scarcity in coming decades, there has been a strong tendency to assume without question that renewable sources can substitute for fossil sources. Because Australia receives more solar energy than most other developed regions of the world it is also commonly thought that Australia will be more able than most to meet its energy demand from solar sources. The following analysis concludes that with respect to the two crucial energy forms, electricity and liquid fuels, this assumption is mistaken, both in relation to existing costs and difficulties and to what is likely to be achieved by technical advance in the foreseeable future.
 

Unfortunately those most familiar with the problems in various renewable energy fields and their limitations tend not to be the best sources for realistic assessments of problems and potentials, given their interest in leaving a favorable impression of their field. Claims are often unduly optimistic. Predictions of costs have to be taken with caution. "Cost over-runs" that emerge when projects are attempted can be the result of glowing estimates designed to persuade investing authorities to sign on for uncertain ventures. Attempts from within the field to critically assess the potential are quite rare.
 

The basic question is whether renewable energy sources can provide virtually all the energy we need. When we hear that a particular country already derives X% of its electricity from the sun or the wind it seems a simple matter of continuing the trend until most or all of the energy demand is derived in the same way. However it is misleading to focus on the contribution a renewable source is playing when it is merely augmenting supply largely derived from coal and or nuclear sources. In that situation the significant problems set by the variability of renewables can be avoided. When the sun is not shining or the wind is not blowing more coal can be burned. However our problem is to develop systems in which almost all energy used comes from renewables, and that means we have to provide for large fluctuations in energy production and for the need to store large quantities of energy, and these problems make a significant difference to the viability of renewables.

It should be stressed that the following analysis is not an argument against the development of renewables. The final section argues that in a sustainable world we must live on renewables and that we can live well on them, but only after radical transition from capitalist-consumer society to "The Simpler Way."

PV SOLAR ELECTRICITY

Flat plate collection systems will be considered first.
 

The potential for solar electricity supply must be examined primarily in relation to the task of meeting winter demand. The following derivation assumes an ideal Australian site, at the tropic of Capricorn where the average daily solar incidence on a horizontal plane in winter is approximately 4.25 kWh/ squ.m. (University of Lowell Photovoltaic Program, 1991.) (For convenience "square meter" will be indicated by "m" hereafter.)
This means that the sun would be approximately 35-40 degrees from vertically overhead throughout most of winter. Thus the incidence of solar energy on panels set at optimum inclination would be 5.18 kWh/d in winter, and collectors set at this angle will be assumed for the following discussion . (Note that this maximises the achievement for winter performance but to maximise annual performance the tilt would only be at half this angle.)

It will be assumed that for 8 hours a day electricity from solar PV plants will be supplied directly, and for the other 16 hours it will have to be stored before being supplied to consumers. Night time electricity demand is about one-third lower than daytime demand (Mills and Keepin, 1993) so in the following discussion supply from a power plant will be assumed to be at the rate of 1000MW for 8 daylight hours and 670MW for the other 16 hours. Although efficiencies above 25% are being achieved in the laboratory the efficiency of PV cells in use is reported by Kelly (1993) to be approximately 13%. (Evidence that actual performance is lower than this is given below.) At 13% efficiency each square meter of PV collection area would produce .67 kWh per day in winter in central Australia.

A 15% loss of this output in transmission from the inland generating site to the coastal consuming areas will be assumed (derived from Ogden and Nitsch, 1993), along with a 7% loss for inversion from DC to AC current. Czick and Ernst, (2003),say that the loss would be 16% with today's technology but that with HVDC systems it could be 10%. Hansen (2004) figures for the present loss rate correspond to about a 190% loss for a 1000MW power station transferring power 1000 km. Losses could be reduced if generating plant was located close to users, but for Sydney winter solar incidence would be about half that in Central Australia.
From these figures the overall efficiency of delivering electricity directly to consumers in the daytime would be 10.27%. In other words to deliver 1000MW, solar energy equivalent to 9737MW would have to fall on the collecting surface. Therefore to deliver 8 hours x 1000MW directly, 77,896MWh of solar energy would have to fall on the collector each day.

The most significant problems for solar electricity supply are set by the need to store energy for supply at night. Storage in the form of hydrogen gas will be assumed here. (Compressed air storage will be discussed below.) Other options will be considered below. The significant problems deriving from the occurrence of a series of continuously cloudy days will be ignored in the following analysis; obviously much greater storage capacity would be required.

The energy efficiency of producing hydrogen gas from electricity will be assumed to be approximately 70%. (Commercial supply in the US is currently via methane reforming at 65% efficiency.) Again a 15% loss in transmission and a 7% loss in inversion will be assumed. Generation of electricity by burning the hydrogen gas will be assumed to be 40% energy efficient. A higher figure for future fuel cell technology is discussed below. The combined effect of these efficiencies would mean that for each kWh of solar energy falling on the surface only .029 kWh would be delivered in the form of electricity after storage; i.e., the process would only be about 2.9% energy efficient. Thus the need to store a unit of energy increases the collection area required by a factor of abut 3.5.

To meet the 670MW demand for the 16 hours of the day when the sun is not shining via a 2.9% efficient process, 373,519MWh of solar energy would have to fall on the collection surface each day. Adding the direct and the night time figures indicates a need for a total of 451,416MWh to fall on the collecting surface each day. At 5.18kWh per square meter the collection area would have to be 87 million square meters. Each square meter of collection area would deliver on average .2 kWh of electricity per day.
 

PV module cost.

The current wholesale cost of PV panels is approximately $5-6(A) per watt , half the retail cost. (BP Solar Australia, 2003, Largent, 2003.) For the large Victoria Market project completed in 2001 the cost was $6/W. (Origin Energy, 2003.) Cost figures can be confused when US costs are taken, in view of the exchange rate at the time of discussion. The value of the Australian dollar used below is the c May 2003 $A value of a little over half the US dollar.

Cost claims vary this makes analysis difficult, especially regarding what elements have been included (e.g., manufacturer’s profit?).

The "balance of system" cost.

The "balance of system" cost, i.e., the cost of mounting panels, connecting wires, control devices etc., is probably the most important, but in general a rather uncertain factor in estimating the viability of PV systems. It has generally been assumed to yield a total system cost that is approximately double the cost of the modules. (Kelly, 1993, p. 300, Commissioners of the European Community, 1994, p. 24.) Solar Energy Systems (2003) estimate that BOS costs are around 43% of total system cost (personal communication.) However they also state that the installed system cost for grid connected systems is $12.50/W, indicating that balance of system costs make up 60% of the total. Largent (2003) says balance of system costs are 60-70% of final system cost. BP Solar, Australia, 2003 advise that balance of system costs make up 40-70% of total system costs. For the Austrian Energy Park 66.8kWp system the balance of system cost was 63% of the total cost. On the other hand Hansen (2004) says that for thin film systems BOS can become c 20% of total cost (although it is not clear if all elements have been included in this figure.)
 

These figures are for non-tracking systems. Systems in which the panels change their angle throughout the day to track the sun collect some 30% more energy (at low latitudes but at high latitudes there might be no difference at all; see Reichmuth and Robison, undated, Fig. 2, p. 3.), but have much higher balance of system costs. For example each of the 15 meter diameter tracking modules in the 10kWe Washington State system (Reichmuth and Robison, undated) uses 6.7 tons of steel, and costs $20,000-$25,000. Each of these supports 80m of PV panels, indicating a cost of $250-312/m for steel alone.

Reichmuth and Robison (op.cit, p. 4) state that conventional wisdom re the flat plate (as distinct from concentrator systems; see below) is that tracking is not justified due to the additional mechanical complexity involved.

Again, the BOS cost is the most uncertain factor in estimating total PV costs, often because it is not clear4 whether all elements have been included. It is the factor most capable of invalidating the following cost conclusions. It has seemed best to be guided by cost figures for total/final installations actually completed, as noted above. These cases seem to indicate that despite some lower claims, the BOS costs for a completed system tend to be equal to the cost of the modules.

If we assume 75 Watt panels, i.e., 150 peak watts per square meter, at $5(A) per watt, the cost per square meter would be $750 for the panels, and if BOS costs are equal to panel costs, then the cost for the whole system would be $1500 per square meter. Therefore the cost of a generating plant 87 million square meters in area would be $130.6 billion. ( Profit and operation and management costs for both PV and coal have not been included.)

How does this figure compare with the cost of a coal fired plant?

NSW Power authorities do not seem willing to give a clear figure for the current cost of construction for a coal fired plant of 1000MW capacity. However the cost of the recently completed Mt. Piper power station in N.S.W., Australia, was $800 million. (Pacific Power, 1993, p. 104.) In 1997 the 2000MW Loy Yang plant in Victoria sold for $4.9 billion, indicating a sale price of $2.45b per 1000MW. ( Sydney Morning Herald, 2003.) Note this would be much more than a current construction cost. Garlic (2000) gives a cost per KW of $1000-1440. Garlic, (2000) states costs per kW of $(A)1000-1440.

Coal for 20 years will be assumed to cost $2 billion. (Garlic, 2000.) Therefore the total cost of the fossil fuel option will be assumed to be approximately $2.8 billion. Thus the PV solar option would cost approximately 47 times the cost of the coal option. (Taking into account externalities, especially the environmental costs of coal use, would reduce this figure.) If a 30 year plant life is assumed the multiple would be 33.

Other cost factors.

The discussion to this point has dealt only with the cost of constructing the collection area, and there are many other factors that would multiply the final lifetime cost for the total system many times. The cost of construction plus fuel accounts for only about 28% of the present price of electricity generated by coal-fired plants. Following are several additional factors which would significantly increase the cost of the solar plant.

a) Operations and management costs, especially the cost of regular cleaning of the large collection area. For wind systems O and M costs over plant lifetime add approximately .7 of construction cost.

b) No provision has been made in the above estimate for the extra capacity needed to cope with extended cloudy periods. On clear days the home lighting system referred to at c) below generates around twice as much energy as is required, yet difficulties experienced in cloudy periods would not be eliminated if generating and battery capacity were doubled. In large scale systems the problem might be avoided if there was sufficient alternative generating capacity available in cloudy weather, such as hydro power. However this solution generally involves the problem of duplication of plant which will remain idle some of the time.
To provide storage capacity for a cloudy day for the output of a 1000MW power station must be (8 hr x 1000MW + 16hr x 670MW)100/2.9 = 645,517MWh. This is 1.43 times the amount derived above where storage for only the 16 night time hours is required. This means that collection area and cost for a system that can supply through 3 cloudy days in a row must be able to collect and store energy capable of generating 5.3 times as much electricity (deliverable as electricity) as it must deliver in a 24 hour period when storage is for only one night.

c) The actual performance of PV systems in the field can be well below expectations deriving from theoretical considerations, when all extraneous factors capable of affecting output have had an opportunity to operate. Theoretically electricity generated from wood fired steam plants should be produced at c 33% efficiency, but Hohenstein and Wright (1994, p. 162) provide figures showing that for the entire US electricity via wood system the actual performance was only 22%.
 

PV panel performance can be lowered by imperfect alignment, dust and water vapor in the atmosphere, dust on panels, ageing of the cells, losses in wiring and inverters, loss due to protective covering glass (Kelly, 1993, p. 300) and the heating effect of sunlight on the cells. The nominal ratings usually quoted derive from tests in ideal laboratory conditions which do not include the above factors. Especially important for systems not connected to the grid is the fact that when output exceeds demand or storage capacity much of the energy being generated can’t be used and has to be dumped. Similarly, a large scale system capable of meeting all demand in mid-winter would have approximately twice the required capacity in mid-summer, given that solar energy incidence is about twice as great in summer. Knapp and Jester (2001, p. 45) say that "system loses" due to wiring resistance, inverters etc., typically reduce output by 20%.

A home lighting system monitored in Sydney, at 34 degrees South, with a nominal rating of 11% efficiency on a cloudless summer day provides as useful energy only 5.7% of the solar energy falling on its surface. This includes the loss due to battery storage. Winter performance is even lower, because the sun is on a lower angle, shines for a shorter period, and its energy has to travel through more atmosphere. This is a tracking system. Systems involving stationary panels would be around 30% less efficient. These figures do not include losses due to the dumping of more than half the energy collected in summer when batteries are full. (yet battery capacity is too small for convenient supply in winter.) Because the average daily power delivered per panel is c .2 kWh, it would take about 70 years to pay back the c $500(A) panel cost, which is only c 25% of the system lifetime cost including batteries, if the energy was sold at the same price as coal-fired electricity is sold from the power station.

Data published in 1999 by BP Solarex (Corkish, undated, Ferguson 2000a) on a 390 square meter system in the UK, a 805 square meter system in Switzerland, and a 7960 square meter system in Toledo, Spain, show that over approximately three years the output of these systems was around 6-7% of the solar energy received by the respective collection areas.

The large Victoria Markets system installed in Melbourne in 2001 performs at c 11% efficiency. A smaller, 1.26kW system installed in Melbourne, with panels normal to the sun in mid winter, delivered as electricity only 8% of the solar energy falling on the panels, averaged over the 2.5 mid winter months. (Renew, 2001.)

An inspection of data on actual generating performance from the US Solar Electric Power Association (2002) also indicates that delivered electrical energy from recent large scale systems is often c 8% of the incident solar energy.

d) The energy cost of constructing the plant must be subtracted from its lifetime output before we can discuss the amount of energy it would actually deliver.

PV cell manufacturers usually claim payback periods of c 3 years. (Corkish, undated.) Knapp and Jester, (2000) report 1.8 years for thin film CIS and 3 years for silicon modules. However these figures are usually derived from performance under ideal laboratory conditions. As is noted in c) above many factors reduce panel performance below these levels and this means that real payback time in the field will in general be much longer than might be expected from the manufacturers' statements. Ferguson’s (2000a) estimates that for the Toledo system referred to above the energy needed to produce the panels would be .25 of the energy the system will produce (over an assumed 30 year lifetime in this analysis.) For the UK site the fraction was .38.
 

The figures usually stated for payback refer only to the energy cost of cell production. (Knapp and Jester say their figures relate to module production.) The dollar cost of PV cells is only about 40% of the cost of the panel or module when glass, aluminum or steel framing and wiring etc. are included (Kelly, 1993, p. 304)l, although this is probably not a good guide to energy costs. As has been explained, module cost is typically only half or less of the whole system dollar cost so the energy costs for the balance of system must be added before a realistic system energy cost figure is arrived at.

A full emergy accounting would also include the energy cost of constructing the factories, deliveries to it, mining of materials, retailing of the cells, the energy cost of plant lifetime operations and maintenance, etc., for the PV modules and for all components of the balance of the system. In other words the total emergy cost of the PV system includes the energy cost of all the work and production that would not have taken place had the plant not been built and operated for many years. Such estimates are not available but total energy costs are likely to be considerably greater than for the cell production costs that are usually focused on in discussions of PV payback.

The Knapp and Jester study seems thorough. If its figures are taken, and if the energy cost of the balance of system is equal to half that of the modules (an uncertain number), then it would take about 4.5 years to pay back the energy cost of producing a silicon cell system, i.e., 22.5% of the energy output of a plant with a 20 year lifetime, or 15% of the output of a plant with a 30 year lifetime.

e) The basic cost calculation above does not take into account the plant's down time for repairs, breakdowns and general maintenance. If it is assumed that it would be out of operation 30% of the time, a typical figure for coal fired stations, then the necessary area and cost for a plant to deliver 1000MW constantly would have to be multiplied by 1.43. However PV plants are likely to be in operation for a much higher proportion of the time than coal-fired plant. If down time is 10%, the above cost, area etc figures must be multiplied by 1.1. (Repairs to solar systems might be carried out mostly at night.)

f) The cost of building and operating the hydrogen production, pumping and storage systems would be considerable. To store the hydrogen to meet night time demand would involve a huge storage volume given the low energy density of hydrogen. To retrieve the 10,560MWh from hydrogen via a process that is 70%x40% efficient would require storage of 37,700MWh of hydrogen. At 3kWh per cubic meter, the volume of hydrogen would be approximately 12 million cubic meters, or a mine shaft some 1,300 km long. Of course the gas would be compressed reducing the volume but increasing energy and plant costs. Even liquid hydrogen has only 25% of the energy density of petrol. (The difficulties in "the hydrogen economy" are discussed below.)

g) The cost of the plant to convert the stored hydrogen to electricity would have to be added. This would be comparable to the cost of a coal-fired power station (assuming the hydrogen is used as fuel to generate steam. The fuel cells of the future will probably be more efficient but at present are very expensive.)

h.) The performance of PV cells degrades over time.

i.) Most of the silicon for production of cells currently comes from scrap left over from computer industry, and would cost more if it had to produced specially for the solar industry.

j.) The cost of the capital that would have to be borrowed to build the plant, i.e., the interest to be paid, might double the total construction cost figure from all the above factors combined. A coal-fired plant produces around 122.6 million MWh in its lifetime (assuming it is out of operation .3 of the time), so for a $2.8 billion construction plus fuel cost the cost of the electricity produced per kW is 2.28 cents (or 1.52 c for a 30 year life.) In Australia it is sold by the station operators at around 3ckWk. However, the 1998 Australian retail price of domestic electricity was 10.1 cents per kWh, which suggests that profit, operation and management and interest costs (and distribution costs, which PV can avoid, but only by incurring other costs; below) can be expected to multiply the cost of electricity due to plant construction cost by a factor of 4 to 6.

k.) A decision to build large scale solar generating plant with the sort of costs under discussion here will obviously not be made until the cost of energy from other sources ceases to be cheaper than the energy generated by these solar plants. We must assume therefore that the cost of the energy required to build all components of the solar plant including cells, balance of system and all contributing factories, deliveries, trucks, tools etc., will be approximately the same as the price of the energy it will generate, which it has been indicated would be very high. Given that energy-intensive materials make up much of the construction cost, the cost of the plant would be far higher than that assumed in the above derivations, which assume present energy costs for construction and materials.

Combining these factors would indicate that the initial $130.6 billion cost estimate might have to be multiplied several times.

Dollar payback periods.

Although not central to the present discussion it is of interest to note the long times required for costly PV systems to meet their dollar construction costs. A 450W system offered by Pacific Power for $8500 (including the $2500 subsidy from the Federal government) would probably produce about 2kWh a day in Sydney (annual average). Coal fired electricity can be sold from the generator at 3-4 c per kWh in Australia. Thus if electricity generated by the three modules sold at the usual electricity price annual earnings would be $365x2x.03, i.e., $25.55, and it would take 400 years to earn the purchase price.
 

The Victoria Market system yields comparable figures. The $1.75 million system is expected to produce 290Mwh per year, which would sell for $9,600 at the price of coal-fired electricity. At this rate the system would take 182 years to pay its capital cost.
 

These have been comparisons with the price of electricity generated from abundant and thus cheap coal, and do not take into account the environmental costs of coal use. However these long payback periods indicate the magnitude of the increases in electricity cost that would have to be accepted in an economy based solely on renewables.

In their commendable efforts to stimulate the development of renewables governments have given very generous subsidies (said to be 48Euro cents/kW for German PV electricity, some 28 times the Australian cost of coal-fired electricity.) It is not surprising that the Australian government is now considering abandoning its subsidy scheme.

What difference might technical advance make?

The assumptions made within the above analysis are apparent and enable derivation of the conclusions that would follow if different assumptions about efficiencies and costs were made. If it is assumed a) that cells with 20% actual operating efficiency in the field (as distinct from nominal peak watt rating), compared with the 13% taken above, b) a cost of $2 per watt for PV cells, i.e., a 60% reduction, c) fuel cells producing electricity from stored hydrogen at 60% efficiency, then the cost of the plant to deliver 1000MW would only fall by about 60%, i.e, to the region of 20 times that of a coal fired plant plus fuel or a of nuclear plant. Note that this refers only to the plant needed to send the energy from the collection field, partly in the form of hydrogen and therefore does not include the cost of plant to convert the hydrogen into electricity.

At present fuel cells could be four to six times as costly per kW of capacity as conventional energy generating plant. US DOE gives a multiple of 10 for car engines. Manger (2003) says the cost is 40 times that for advanced diesel car engines. In addition fuel cell life has been reported to be only 200 hours.

The cost of PV cells has fallen significantly over the past 3 decades, but the trend seems to have flattened out now. (Kelly, 1993, Durning, 1997, p. 27.)The cost for the Victoria Market system was $6/w (higher than that assumed in the above analysis.) If the cost per square meter of PV technology fell to zero the cost of the large collection area required in the above discussion would still be very high. If the PV material was sprayed at no cost onto 6 mm toughened glass at the mid 1990s wholesale price of approximately $60 per square meter, the cost of the glass alone for the above 87 million square meter collection area would be $5,220 million. (Littlewood 2003 estimates the cost of PV glass in 2003 at $50/m, and at $70-80/m for curved glass for concentrating systems.)
 

In other words the "balance of system" cost sets a difficult limit when the collection area must be large, and one that is not likely to be greatly affected by technical advance as structures are simple and major breakthroughs in their design are not likely. As has been noted, in the early 1990s the BOS cost per meter seems to have been about the same as the cost of the panels, i.e., at present c $750/m. (Again it should be noted that BOS cost is the most uncertain element in the analysis of PV systems.)

Almost all of the materials cost of cells is due to aluminum, glass and silicon; for silicon cells it is 85% and for thin film technology it is 97%. (Knapp and Jester, 2000.) Thus there would seem to be little scope for cost reduction from advances in the solar technology involved, although increased scale of production might make a significant difference to overall costs.
PV roof cladding systems.

PV roof cladding systems.

Integration of PV cells into roofing etc. material would reduce balance of system costs, e.g., for support structures (and roofing replaced.) It would also avoid transmission loses and costs which make up one-third of the retail cost of US electricity, but only if systems are elaborate enough to be completely independent of the grid. Such systems would involve the excess generating and storage capacity needed to cope with long cloudy periods. They would increase some costs, especially for storage in many small units each with its own batteries and power conditioning equipment such as inverters and regulators and petrol driven backup generators.

Replacing roofing with PV panels sets the problem of whether the solar incidence where the house is located is adequate. For instance in Sydney, 34 degrees South, in winter the solar incidence is 2.78 kWh per day, only 2/3 of the 4.25kWh per square meter per day in central Australia where large scale centralised PV systems would be ideally located.

Rooftop collection surfaces are fixed in orientation and on average rooftops differ considerably from ideal orientation, and are subject to shading by other structures. It is likely that only about 40% of the surface of an average house roof would have an orientation enabling effective use as a solar collector in winter. In mid winter in Sydney the mid day sun is 56 degrees from vertically overhead, so a roof surface facing North with a 12 degree slope will be 44 degrees from ideal inclination. However because it is angled somewhat towards the sun it would intercept about 1.2 times the 2.78kWh/m/d incident on a horizontal plane in Sydney in mid-winter, i.e., 3.3kWh/m/d. This is only .78 of the 4.25kWh/m/d falling on a horizontal surface in Central Australia at the Tropic of Capricorn at that time of the year.

At 13% efficiency, 40 square meters of PV panels receiving 3.3kWh/m/d would generate 17kWh/d, about equal to household demand, if storage issues are ignored.

At the $(A2004) price for BP panels, $1292/m, the panels would cost $51,680. Over a 25 year lifetime and in a region with an annual average insolation of 5kWh/d, they would generate 235,000kWh, so panel cost per kWh generated would be 22c. This is relatively low but BOS costs are not included, and access to the grid for storage is assumed.

To supply the same amount of power as was assumed above for a single centralized 1000MW PV plant (i.e., 1000MW for 8 hours without storage, and 660MW for 16 hours via storage), solar energy intercepted must be (1000MWx8hr x1/.13 for generating efficiency) + (660MWx16hrx1/.13x1/.28 for .4 efficiency of fuel cell generation after storage in hydrogen at .7 efficiency)…i.e., 351,637MWh(th)/d. If one square meter intercepts 3.3kWh/d, collection area must be 111million square meters. This is probably about 100m per house, some 2.5 times the 40m available area assumed here for a roof, for one power station.
To meet Australia's total electrical demand, 700PJ, would require the equivalent of about 30 power stations each of 1000 MW capacity, and therefore 3330 million square meters of collection panels. This is approximately 12 times the area available on all residential roofs (making the above 40% assumption and again ignoring factors a-i. above.) To also fuel a car via rooftop PV panels would be to more or less double the magnitude of the task.
 

The main problem is that the option is available only for that proportion of houses that can be integrated into a grid running mostly on some other energy source. Much less than half a house’s electricity demand could be met while the sun is shining, even ignoring cloudy days, and a rooftop PV house would have to draw from the grid for 16 hours a day. It would not be sensible to have a wholly-rooftop generating system that is capable of meeting the demand of all houses at night (via batteries0but then remain idle during sunny days; that would be to build two sets of plant when one would do. It would seem therefore that rooftop PV might make good sense for an individual householder now, but that only a small proportion of houses could have rooftop PV before causing the above whole system problem. In other words, rooftop PV does not seem to be capable of making a big contribution to the task of moving off coal.

Concentrator PV technology.

Large reductions in PV costs are promised by the development of cells that receive sunlight focused from reflectors, enabling the area of PV material to be much smaller than the area over which solar energy is collected. Cells capable of concentration factors of 1000 to 2000, and over 25% efficiency, are being developed. The ANU cells are 22% efficient (Smeltink, 2003.)
Swanson (2000) discusses the fact that although this approach has been under development since the early 1980s, it has not been taken up enthusiastically. One reason is that it is not as suitable for the many small and stand-alone tasks that the more simple flat plate technology is being used for. Concentrating systems are more complex, involving tracking, and thus best suited for bulk supply, and here their high cost has been the main impediment.

A experimental 20kWe peak system operating at Rockingham in Western Australia is in the early stages of operation, although performance has been reported (2003) as disappointing so far. The best daily output recorded, 75 kWh/d, represents an efficiency of about 5.5% (i.e., electrical energy produced as a percentage of solar energy falling on the collector). I have been unable to get any cost figures (especially balance of system costs) from the developers, although these would not be a clear guide to costs for eventual large scale production.

An experimental system at Australian National University (Corkish, undated) involves a concentration factor of around 40, i.e., the area of PV cells required is only 1/40 of that over which sunlight is collected. However the cost of the cells has been reported at $(A)65/W, (personal communication from ANU), which is 13 times the cost of normal cells.

Sala et al. (2000) report on an experimental 480-kW system . Efficiency is reported at 8%, over a year. Total plant cost was Euro 2.13 million, or Euro 4,437 per Wp. Remarkably the PV receiving module cost is given as US 81cents/W. ( (I am still attempting to clarify the contradiction between this and the ANU cost given above.)

The stated costs per watt for concentrating cells can be misleading. They are far lower than for flat plate PV cells, e.g., 80c/W vs $5/W, tempting one to ask why aren’t they used in flat plate systems. Apart from the fact that they do not work as well at one sun concentration, their cost would be much higher for use in such a situation. The situation seems to be that one square meter of concentrating reflector focusing 1000W of solar energy on concentrator cells operating at 38 suns and 35% efficiency will deliver 350W from 260 square centimeters of cells. Thus at 80c/W the cells would cost $280, or $1.08 per square centimeter. The 10,000 square centimeters of cells in a 1 square meter flat plate system would cost $750, or 7.5c per square centimeter. Thus it would be much more expensive to use concentrator cells in a flat plate system. (Smeltink, 2003, confirms this general account but reports that some cells cost 68c/W.)

The overall cost of concentrator systems will be determined primarily by the balance of system cost. As has been noted, for systems which do not track the sun this is usually assumed to be about as much as the cost for normal PV cells per meter. However concentrator systems must track the sun, so structures will have to be fairly substantial, involving supports for collecting surfaces, machinery and control systems, moveable in at least one dimension and capable of withstanding strong winds. Costs for these items are not likely to fall greatly due to technical break throughs as they already involve relatively simple structures.

Unfortunately it has not been possible to find clear and confident general figures for the balance of system costs for tracking systems, either for PV or solar trough. The support structures for the two would be similar if the heat exchange components of the latter are excluded, because in both cases a frame supports a parabolic or fresnel reflector and the whole assembly must be capable of movement about at least one axis (for seasonal change.) Note that because it has a U shaped cross section the area of the trough or concentrator reflector has to be greater than the area of the solar radiation intercepted. Strebkov (undated) states that the ratio is between 2 to 1 and 2.4 to 1. (Web pictures often seem to show lower ratios.) This effect does not occur with flat collectors and tends to increase the costs of trough systems. For the Rockingham, Western Australia project the curved glass for the reflectors cost $70-80 per square meter. (Littlewood, 2003.)

The cost of the SEGS VI system's collector was $(US)487/m (or about $(A)812/m), although this included heat collection apparatus. Strebkov (undated) says the cost of the collection field for central receiver solar thermal systems is $(US)200-600/m, although this would not be a good guide for trough systems.

In their discussion of another proposed trough system Brackman and Kearney (2002) state that the collection field would make up 45% of the total cost. Again unfortunately this figure includes heat absorption equipment, but it again indicates that the balance of system cost in PV concentrating systems is likely to be far more than the cost of the PV components. These figures are sobering since they indicate that for trough thermal and concentrating PV systems the equipment needed in addition to the heat absorption system or PV cells costs at least twice as much as they cost.

From Brackman and Kerney's figures the absorber would probably have been around only 4% of the total system cost so when subtracted the "solar field" cost would come to about 37% of the total plant cost. If we leave the cost of the power block out of the total figures and think of the remaining balance of system as comparable to that required for a PV concentrating system then this suggests that for PV concentrating systems the BOS would come to around 65% of the total cost.

Haberle et al estimate that for a 50Mw peak fresnel trough system in Egypt the reflector plus absorber add to only 7% of total cost, a remarkably low figure. This suggests that the rest of the plant that would be needed in a PV concentrations system could cost about 4.5 times the cost of the reflector plus absorber. (In Strebkov's example this reflector plus absorber cost does not include the power block which was 28% of total cost, nor "service and other costs" which were 36%.)

The overall costs given in the account by Haberle et al. seem to be surprisingly low; i.e., Euro77 million total cost for a 50MWpeak, 450,000 m system, (i.e., only Euro171/m but $(A)290/m). However the figures for the collection equipment are helpful re the problem of estimating PV concentrator BOS costs, i.e., again indicating that BOS cost is high compared with that of the PV component.

The cost breakdown given by Sala et al. states that the cost of the "structure and tracking" and mirrors came to Euro327/m, or $(A)556. (However the rest of the BOS came to another Euro 180/m, making a total BOS cost of Euro 507/m. or $(A )862.) In other words the BOS was 61% of the total cost.

Tyner (2003) says collector costs for troughs in use are $(US)125/m ,so c $(A)250/m, assuming one-axis tracking, but $(US)200/m.

From this diverse and rather unsatisfactory evidence on trough systems (above and see further below) it would seem that the collecting structures for concentrating systems would cost in the region of $(A)300 per square meter. Thus collector costs seem to constitute only a remarkably small proportion of total cost for solar trough systems, indicating that even if PV concentrator technology becomes very cheap the balance of system cost for very large collection areas will remain very high. For instance at $300 per meter the BOS cost of the 87 million square meter 1000MW flat plate collection system referred to at the beginning of this paper would be $26 b. (Note that system assumed 13% efficiency, whereas the efficiency of trough systems reported here has been closer to 9%, suggesting that the $26b figure should be multiplied by 1.4.)
Ignoring the storage problem.

If we take the above figures for a PV system providing 8 hrs supply during daylight and multiply by three, 233,688MWh would have to be collected per day to supply 1000MW for 24 hours. The collection area and thus cost of the system would be 52% of that for the system providing for storage, i.e., around 24.4 times the cost of a coal-fired power station. Note that this is for winter supply, not annual average.

STORING ENERGY.

Compressed air storage.

Storage of energy in the form of compressed air is claimed to be much more energy efficient than storage as hydrogen. Sorensen (2000, P. 568,552) gives two figures; 40-50%, and 65%. Hansen (2004) says 75% of the energy used to compress the air is retrievable, without the use of added heat. (He says that if gas is used the energy return on the total air plus gas input energy can be 85%.) Thus to retrieve 1kWh 1.33 kWh must be invested, and to store the 670MW x 16 hrs needed when the sun is not shining, at the energy efficiency assumed above would require investment of 101,264MWh. Total daily electrical energy delivered would therefore have to be 179,161 MWh, and collection area and power station cost would be c 179,161/451,416, i.e., 40% of the plant cost involving hydrogen storage. In other words plant cost would be much lower but still some 19 times that of a coal-fired station. To this would have to be added the dollar cost of the equipment to generate electricity from compressed air, the storage facility, and the transmission of the power to the storage sites and then to users. All these items would also impose energy losses, raising the above estimates of dollar costs.

The major drawback with CAES is to do with the very large storage volumes that would be required to store significant quantities of energy.. Sorensen (2000) says 15MJ can be stored per cubic meter, i.e., 4.16Kwh, meaning that to store night time demand from a 1000MW plant would require 2.4 million cubic meter storage area, i.e., a mine shaft around 240km long.

Marden (2004) shows that to store 1GWh in air at 10 atmospheres would require a 953,000 cubic meter space (that does not leak), and the task would take a large compressor ( 1 t air per minute) 9 days. Therefore to store the night time demand from a 1000MW power station, 10,720MWh, at 50 atmospheres, would require a 2 million cubic meter cavern, and would take 1944 hours, i.e., 39 days.

If the task has to be carried out by a PV plant generating for 8 hours a day, which is 1/243 of 1944 days, it would require compression capacity of 243 t/min, which is quite unrealistic.
There would probably be far too few caverns or old mines large enough for this form of storage to enable bulk electricity supply via intermittent sources.

For smaller applications, such as vehicles, Doty (2004b) estimates that a 120 gallon tank could store .576 kG at 15 atmospheres, for $(US)730, which is 85 times the cost of a diesel tank containing the same amount of energy, 40 times as heavy and 200 times as voluminous.

Other storage options

Energy storage via thermochemical processes would seem to be about as efficient as hydrogen gas storage (possibly somewhat less; Kaneff, 1992, p. 43.), although for large scale generation there would be a significant problem of storing very large volumes of gas temporarily. Storage of energy via methane reforming or ammonia recombination is more energy efficient than storage via hydrogen, yet these processes would require one cubic meter of gas storage per 1.54kWh, at normal pressure. Thus to store the energy from a power station for the 16 hours when the system was not generating would require a mine shaft approximately 1500 km long, assuming 60% energy storage efficiency. Obviously gases would be compressed to reduce space requirements but this incurs energy costs, discussed below with respect to the "hydrogen economy".

The vanadium battery promises a higher storage efficiency, initially 87% but this will deteriorate with recharge cycles. However current estimates of world potentially recoverable vanadium resources indicate that far too little exists for a world supply and storage system, especially when automobile demand is added to electric power demand. (Erickson, 1973, Trainer, 1995.)Pumped storage vs. hydrogen storage.

Overall "in and out" efficiencies for operating of pumped storage systems have been reported from around 60%, although some claim that 80% might be a reasonable average. The Queensland Office of Energy estimates 70%. If it is taken as 80%, 1.25 units of electrical energy would be required to provide 1 unit after storage. The efficiency of hydrogen storage and retrieval might be taken as .7 ( for hydrogen generation) x .5 ( for probable future efficiency of generating electricity from hydrogen via fuel cells), i.e., .35%. Thus 1/.35 = 2.86 units of electrical energy would be required to provide 1 unit after storage.

However the fact that pumped storage is much more energy efficient than hydrogen storage does not make such a big difference when the task is to store the 16 hours x 670MW output of a solar plant required over night. The hydrogen system must produce 8h x 1000MW for day time direct supply, plus 16 h x 670MW x2.86, i.e.,
38,659MWh.

A pumped storage system would have to collect 8h x 1000MW plus 16 h x 670 x 1.25MWh i.e., 21,400. Thus the system with the hydrogen storage requires collection of only 1.4 times as much energy as a system with pumped storage. Whereas the hydrogen system analyzed above would be 47 times as expensive as a coal fired system, if the geography and infrastructure permitted pumped storage the system would still be 33 times as expensive.

If dams are not available close to where the solar energy is collected energy must travel to the dam and then from there to where the electricity is to be used. There are few if any dam of any significant elevation anywhere near the best solar collection sites in the flat centre of Australia. Electricity generated there would have to be transported long distances to dams, then long distances to the main consumption regions, adding energy losses to the whole system. Note that for pumped storage two large reservoirs are needed, fairly close together, one high and one low.

SOLAR THERMAL ELECTRICITY.

The most promising solar electricity option seems to be solar trough thermal. DeLaquil et al (1993) report that costs for central receiver and dish-Stirling thermal systems are 1.14 and 1.43 times as expensive as for trough systems. Manci (2003) says the corresponding ratios for the costs of electricity produced are 1.6 and 2.5

From the Sandia website ( www.energylan.sandia.gov/sunlab/program.htm ,) report of 1997 figures for the SEG VI 30MW system (Table 4), 57 GWh/y were generated from a plant costing $(US)119.2 million some years ago, after subtracting 1/3 of the power delivered which was generated from gas backup. A coal-fired plant operating at .7 capacity would generate 6132GWh/y, i.e., 108 times as much electricity. This indicates that the cost of a solar trough system capable of the same output would be $(US)12.8 billion (ignoring storage), i.e., $(A)21.1b.

However the annual average solar incidence at the SEG VI site is very high, 7.9kWh/m (probably exceeded on less than 5% of US land area, located at the South West corner). This is almost double that for the winter incidence in Central Australia, indicating that for mid winter supply from the latter site the comparable cost of a 1000MW plant would be in the region of $(A)42 billion. A PV plant large enough to generate 6132GWh/y, without storage, at 13% efficiency and at a 4.25kWh/m/d site would be $(A)48 billion. The comparison is made difficult by the fact that the figure for the trough plant includes all costs and that for the PV plant excludes factors a-i above.)

These rough estimates suggest that trough systems might cost half as much as fixed plate PV systems.

The figures given by Brackman and Kearney (2002) for the 1991 performance of SEGS IX, 483,960 m in a region where incidence averages 8kWh/m/d, indicate an efficiency of only c 7%.

Solar thermal systems involve the problem of "start up" threshold or delay. DeLalquil et al (1993) report that solar energy incidence must rise to over 300W/m before electricity is generated, even then at a low efficiency. At Sydney in winter solar incidence is over 400W/m for only 2 hours a day. (Morrison and Litwak, 1988.) In Central Australia it is above 400W/m, 500W/m and 600Wm for 6, 4 and 2 hours respectively. There would also be start up delays after the passage of cloud (unless there is salt storage provision; below.)

However Grasse and Geyer (2000) provide a valuable plot (Fig.22.) from SEG VI for the solar incidence, collector efficiency and generating rate, for a cloudless mid-summer day in 1997 in which incidence reached 1000W/m. The sun rose at 6.45 but there was no electrical output until 7.30 when solar incidence had risen to c 700W/m. At about 8 a.m. electricity output had reached around 75% of maximum but solar incidence was 800W/m. Peak generating output was only reached at 9 am when incidence was 1000W/m. Solar incidence fell to zero at 8 p.m but generation fell from its peak at 6.30. (There is less delay at the end of the day than at the start, presumably because at the start the system has to warm up.)

Also of interest is the fact that the system involved salt storage and because it is therefore important to collect all energy generated through the day the system is large enough to collect 48 MW for a short time around mid day although it averages only 30 MW for the day. This again is the general problem that variable renewables set; i.e., the need to build much more collecting capacity than the plant averages.

The start up problem probably confines trough systems to regions where long hot days are most common. PV systems seem viable though very costly in central and even Northern Europe but trough systems would seem not to be.
 

As with fixed flat plate collectors, solar thermal trough systems and PV trough concentrators suffer a "cos effect". Receiving surfaces are normal to the sun only at mid day and early and late in the area of sunlight they intercept is a fraction of the mid day area. This factor contributes to the start up delay. Whereas a tracking PV system can generate at almost full capacity as soon as the sun rises above the horizon, at this time of the day very little solar energy will be falling on troughs set on an East-West axis, because the sun is incident on them at a very low angle. (Dishes and troughs can be tilted to very low angles without shading each other but only if spaced very widely, setting other problems and costs.)

Within the above discussion of possible BOS costs for PV concentrating systems it was seen that reported BOS costs for trough systems seem to range from $US300 to $800. It is not likely that the balance of system costs for solar trough and PV concentrator systems will fall markedly, given that the technology involved is simple, involving supports and adjustment equipment for the reflectors. "There is little scope for future performance improvements or cost reductions for solar trough systems" (Commissioner of the European Communities, 1994, p. 25.)

If trough systems can only reach maximum efficiency for electricity generation in regions where solar incidence exceeds 800W/m for many hours a day they will be confined to restricted areas. This is not to say that they cannot make a valuable contribution in wider areas, such as pre-heating water for coal powered stations.

Heat storage in molten salt.

More recently solar trough designs have included provision for storage of heat in molten salt enabling solar systems to generate for several hours after the sun sets. The "in and out again" loss of energy has been reported at 15%, whereas for storage as hydrogen and conversion to electricity via fuel cells it might be only 65%.

Haberle et al (2003) say molten salt storage at 307 degrees is being used but there is no cost effective system in place for 390 degree heat. The lower temperature is associated with c 28% efficiency generation (Dey, 2003.) (I have seen an unrecorded recent reference to salt storage at 500+ degrees.) Mills (undated) reports amonia and rock bed heat storage systems at $(US) 673/Wp, which seems to be a considerable cost, although the meaning of this figure is not clear.

Systems for storing heat in salt have only been developed to provide for a few hours. To provide for longer periods would involve very large additional collection and storage plant. Thus these systems cannot help with the problem of generating on cloudy days.

WIND ENERGY

It seems clear that some regions of the world will be able to derive a considerable fraction of their electricity from the winds. However because of the lack of publicly accessible information on wind mapping in Australia it seems that little can be said with confidence regarding potential electricity generation.

The Sustainable Energy Development Authority's website estimates that in NSW 1 GW could be derived from wind. However in November 2002 demand was 11.5GW.

Mills’ study (2002) concluded that Australia has a large potential wind resource, but most of it is not useable due to "exclusion factors", notably long distance from grids. The cost of building lines to wind farms must be included in the cost of providing wind electricity.

Evidently the CSIRO now has good wind mapping information for NSW, but has not made it public. (Some information is given below.) However they have said that sites must have at least 8m/s average wind speeds, and the Federal Renewable Energy subsidy of 4c/kWh before generation becomes economically viable. (Personal communication.) This is surprising given that wind is usually thought to be economically viable in areas with over 7m/s winds.

The American Wind Association (2001)has said that three times present US electricity use could be derived from wind. Unfortunately many statements like this have been made but they leave important issues obscure, such as whether class 4 and 5 wind regions are included as potential. Class 4 winds are said to have 90% of potential wind energy, but it is far from clear whether their use will ever be viable.

A study reported in Planet Ark for June 2003 claims that US potential is far greater than previously thought when 262 ft towers are assumed, compared with the 164 ft towers in use today. Generating costs equal to those of coal fired electricity are claimed. I have not been able to clarify the nature, costs, problems associated with towers of such height, including possible storm failure rate.

A study by the Commission of the European Communities, (1994, p. 34.)concluded that "…realizable on shore technical potential is …about 350TWh, 23% of the Communities total electricity demand in 1990.

Capacity

It is commonly assumed that windmills will perform at 25% capacity on average; i.e., that a 750MW will have an average output of 188MW. Caution is required here. Firstly a mill's capacity is primarily a function of its location. Very good sites enable a mill to deliver over a long period 35% or more of the peak output it is capable of. However average capacity in the Netherlands, Denmark, Sweden and Germany has been reported as 22% (OPT Journal, 2003.) The average capacity achieved by Californian mills in 1990 was 18.6%. (Elliott, Wendell and Gower, 1991, p. 56.) In 21997 and 1998 UK ,capacity averaged between 24% and 26.7%. http://www.cprw.org.uk/wind/winstat.htm#rolling%20capacity

As with PV, performance in the field seems to yield efficiencies well below those one might expect from theoretical analyses, or from lab tests in ideal conditions. Although windmill efficiency can be expected to improve, the sites first used will tend to be the best ones, indicating that we could expect average capacity to decline over time as less ideal sites have to be used.

Penetration

As with other renewables it is a relatively simple matter to introduce wind power within a system primarily based on non-renewable sources, enabling adjustment of the coal or nuclear generating rate to accommodate fluctuations in the renewable source. However when the wind contribution rises beyond a certain proportion of total demand problems arise, especially the need to leave some of the renewable component sources idle part of the time. It is commonly assumed that in good wind regions wind might be able to supply 20-25% of electrical energy produced by the system before a penetration problem arises.
Denmark is reported to have such a problem even though wind has only a 13% penetration, resulting in much energy being dumped at certain times, and much having to be sold at low prices. (Country Guardian, 2002.) This problem is said to have arisen regarding 34-45% of wind generated electricity in 2000. Germany is said to have problems with 15% of supply coming from wind. (Duguid, 2003.)

Denmark's extensive development of wind energy has been facilitate by the fact that its neighbours have made much less investment and have therefore been able to buy Denmark's surplus when it was available. In a renewable energy world there would be less scope for this. Denmark's problem suggests that the 25% penetration in good wind regions commonly assumed might be optimistic.

Subsidies.

The considerable penetration achieved by renewable energies in some countries has been due in part to large subsidies. While these are desirable in order to enable development of these industries, they can give a misleading impression regarding the viability of the technologies. Coal fired power can be produced for 2-4 c/kWh in many countries, yet in Australia Pacific Power pays home owners 10c/kWh for power fed into the grid from home rooftop systems. In Denmark the subsidies are "very large", 10 billion DKK per year, around DKK .45/kWh and the price of wind electricity is 4 to 5 times that of other electricity from other sources. (Country Guardian, 2002.) In Germany the subsidy for PV power is reported to be 48Euro. cents per kWh. Worldwatch (2001-2, p. 46) reports PV power in Germany receiving a 10 year interest free loan plus 50ckWh. I have a report that in the US the subsidy is 3.3c/kWh.

Although we should be willing to pay much more for renewable energy the question is at what point costs would become too high. We might be able to cope with a 5 fold increase in price, but a 10 fold increase would seem to be quite prohibitive.
 

Figures from a proposal by Babcock and Brown for a 200MW South Australian wind farm throw a little light on what seems to be a precarious financial situation. (Sydney Morning Herald, 17th July, 2003.) The project will cost $450 million, and will sell electricity at 8cWh. Thus over 25 years and at 25% capacity income will be $1051 million. At the probable loan repayment rates ( from personal communication) interest on capital borrowed will probably be $250 million. Operations and management (at 2% of capital cost p.a.) will be c $225 million. Cost will therefore be in the region of $960 million, i.e., not much below total lifetime income. Annual earnings would therefore seem to be $3.6 million, or .8% of invested capital. Assuming a 30 year lifetime and a 30% capacity factor would improve the outlook, but these figures make it difficult to see how projects could be viable without a subsidy that enables 2-3 times coal-fired generating cost to be charged. (The above estimates are not made with great confidence.)

The problem of variability

As with solar energy, wind energy varies considerably over time. This is not such a problem if non-wind generators can be turned up when winds are low. However the question this paper is primarily concerned with is whether renewables can meet almost the whole of demand, which sets problems to do with storage and over-capacity. Ferguson (2003, p. 3.) notes how energy dispatchers in the UK need firm commitments from wind farm operators regarding the amount of power they can deliver 4.5 hours ahead. Because the wind farms can't be very certain about this and because there are penalties for falling short, they tend to aim low and in one recent year ended up delivering only 86% of the energy wind farms generated. Similarly Denmark sometimes has to give surplus power away. "A couple of years ago we even had to pay Sweden to take it." "Most of Europe can lie under high-pressure with not a breath of wind for days. In winter these conditions bring frost and fog, so demand for heat and light soars." (Duguid, et al, 2004.)

More important is the large variation in wind energy and therefore capacity achieved from summer to winter. In Denmark, Germany, the Netherlands and Sweden the winter capacity of windmills in 2000 averaged 33% but the summer capacity averaged only 15%. In August 2000 German and Netherlands capacities were actually only 8% and 7%, after averaging 38% and 35% in February. Thus in these two countries capacity varied by a factor of 4 or 5, meaning that a system capable of fully meeting summer demand might be 80% idle in winter. This would seem to set formidable if not fatal problems for an energy supply system that cannot feed into a grid powered by nuclear or coal sources.

For Europe as a whole Czick and Ernst (2003) report that windmill capacity varies from 55% in February to 12.5% in May, and averages under 18% for the four warmest months of the year.

Europe as a whole has a 2.5 to 1 variation in wind energy from winter to summer, much the same as in the US. For Australia the variation is between 1 to 1.4 and 1 to 1.8 ( http://www.iset.uni-kassel.de/abt/w3-w/folien/magdebO30901/folie_41.html ) In addition there can be significant variation in wind averages from year to year, up to 25% according to the World Energy Council (1994, p. 152.) The Australian CSIRO reports that annual wind averages at a location can vary by a factor of two. ( www.cssiro.au/weru .)

The areas required

The area over which windmills must be placed to equate to a 1000MW power station is quite large. If 750 kW mills of 80m diameter are placed at 10 x 5 diameters, lose 13% of energy due to array interference and function at 25% capacity, then 6135 mills spaced over 1963 square km would be needed to deliver 1000MW (and three times as much in August 2000 in the Netherlands and Germany, given the low capacity factor discussed above.) This estimate does not take into account losses in connecting wiring and power conditioning equipment, nor in transmission from wind farms to users.

Sorensen (2000, p. 435) says at 10 diameter spacing 10% of capacity is lost due to array disturbance, and that in large farms a higher overall loss occurs as wind disturbance effects tend to compound if many mills are sequenced.

Europe probably has 120,000 square km of Class 5 land and above (7.5m/s or better average wind speed), which would enable the number of windmills corresponding to 61 power stations. The actual number possible for densely settled Europe would be much lower, due to the savage effect of the "exclusion factors" discussed below.

Again in the US and Europe where considerable development of wind energy has taken place, performance figures currently reported for windmills will be associated with the best sites. As time goes by further development of wind farms will tend to be in less ideal sites, hence the overall capacity factor for the wind system might be expected to be lower than at present. (However, improvements in technology etc will tend to improve it.)

The area of Class 5 or better winds in the US would enable the equivalent of about 240 power stations, again ignoring the exclusion factor. US electrical energy, approximately 12.97 Quads in 1999, equates to about 433 power stations (operating at .7 capacity.) Again the effects of exclusion factors and losses in long distance transmission from the best US wind regions to the Eastern and Western cities would have to be added.

CSIRO modeling for NSW, Australia, indicates that within the best 90,000 square km of he state there are 550 square km with winds over 8m/s, and 7000 over 7m/s. (This is via a 2003 personal communication from the NSW Sustainable Energy Development Association but CSIRO has confirmed that the figures come from their recent mapping.) At 1963 square km per power station this would correspond to .28 and 3.56 power stations, again ignoring exclusion factors. NSW peak power demand corresponds to about 16.5 power stations operating at .7 capacity. Note that there would probably be additional suitable areas outside the 90,000 squ. km surveyed, but probably not very much as this area would have been taken as the most promising area for wind generation.

The CSIRO 2003 Wind Resource Planner’s Guide states that to lower the threshold from 8ms to 7ms would multiply available area by 20. Taking into account the fact that 7ms winds have 2/3 the energy of 8ms winds, if all of the increased area could be used this might equate to another 3.3 power stations. (www.csisro.au/weru, p. 28.)

Australia's total electricity demand in the late 1990s was 700PJ, or 22GW. This is equivalent to the output of 32 power stations functioning at .7 capacity. It was estimated above that a wind farm of 1963square km is required to replace one coal-fired 1000MW power station, or of 1374 square km to replace one coal fired power station operating at .7 capacity. This indicates that the area of windmills to provide Australian electricity demand would be 43,968 square km. If one-sixth of Tasmania, the best wind region, could be covered with wind farms the state might provide the equivalent of 32 power stations, although losses in transmission to northern users would have to be taken into account. If one third of a 30 km deep strip along the approximately 700km Victorian coast was devoted to wind farms the output might be equivalent to another 14 power stations. The evidence on exclusion factors below suggests that much more than 2/3 of suitable land would be excluded. There are good wind resources at the SW tip of Western Australia but I am assuming that 4-5000km transmission to the Eastern states is not feasible.

Exclusion factors

Surprisingly large proportion of the areas with good wind generating potential have to be excluded from use for a variety of reasons, primarily pre-existing use, and distance from electricity grids. It seems that for these reasons on-shore sites in Denmark, where wind supplies only 13% of electricity, is close to the limit due to these exclusion (and other) problems. (Country Guardian, 2002.)

The 1997 US EIA/DOE study (2002) came to the remarkable conclusion that "…many non-technical wind cost adjustment factors … result in economically viable wind power sites on only 1% of the area which is otherwise technically available…"
Elliott (1994, p. 8.) estimated that siting constraints would limit wind to providing 10% of UK electricity demand. Elliot, Wendell, and Gomes (1991) state that 75% of the class 7 wind area of the US would have to be excluded from use. . Sorensen (2000, p. c 311) reports surprisingly that even for offshore wind areas (near Denmark?) only 10% can be used in view of other uses, such as shipping lanes.

Offshore wind potential.

The American Wind Energy Association (2000) estimates US off-shore potential as 1/7 that of on-shore potential. The former is more expensive to construct and maintain.

Correlation of winds at different sites.

If all the mills in a supply system are located at the one site then all will be idle when the wind ceases there. However if the mills in the system are at sites distant from each other it is likely that some will have good winds when others are idle; i.e., it is likely that the correlation between wind strengths at different sites will tend to be low, and if the sites are very distant it might approach zero. This means that for a wind system spaced over a very large area some of the mills are always likely to be generating.

If the correlation were zero, the system’s output would be the same as the capacity factor for any one mill given the region’s average wind pattern. So it might be 25-30%, meaning that the system will deliver .25-.33 of the energy indicated by the total peak capacity of all the mills within it.

Geibel (2000) indicates that for large European areas the correlation might be .2. However the time period over which wind speed is measured makes a difference to this calculation, and this could be important. If the wind speed used for the calculation is an average taken over 12 hours, correlations tend to be higher than if it is a 5 minute or 1 hour period. (That is, if we ask what was the weather like here and there across the whole region this morning, usually it will tend to have been similar.) Generators and dispatchers have to worry most about variations over long time spans; i.e., it is not good for them if the winds this morning tend to be down everywhere in their region, even if there is little correlation between the speeds at the various mills within their region. Of course sometimes the whole of a continent can experience mild weather for days on end. I have not been able to get data on actual, recorded correlations typical of large operating systems. "Germany, with up to 15% of its power now wind-generated is approaching the same threshold. …it’s buying balance power on the market – at up to 20 times the wholesale cost – and selling surplus power very cheaply." (Duguid, et al., 2004. P. 15.)

Note that even if these are low, seasonal and annual wind patterns vary considerably, the total wind output in summer can be significantly less than in winter, and the total this year can be significantly different from last year, and the absence of correlation between winds at sites within and between regions does not help with this problem. Nevertheless the hope among wind enthusiasts is that the correlation effect might reduce the need for storage to the point where compressed air, or pumped storage etc might enable wind to provide all/most electricity.

Inter-Continental systems.

Czick and Ernst (2003), discuss the possibility of linking the whole of Europe to regions such as Siberia several thousand kilometers away in order to overcome problems set by wind variability within smaller regions. The correlation between wind speeds falls as the area considered increases. At one point in time low winds might affect all mills in a small area but at that time good winds will probably be blowing in some other regions far away. The closer the correlation between winds within a given region approaches zero the closer the system will come to having constant electricity output (at a level corresponding to the average capacity factor for mills in the system.) (I have not been able to find evidence on the actual correlations that occur within specific regions; CSIRO Australia is reported to be working on this.)

Czick and Ernst argue that such a system would reduce the variability of supply to about 10% and enable the associated need for storage to be met by pumped storage using existing dams.

Systems of this kind would involve losses due to sending large quantities of electricity several thousand kilometers. Czick and Ernst state that at present these losses would be 16% but could fall to 10% given construction of HVDC lines. Transmission lines would probably be limited to 5GW each. The cost of these would have to be added to the cost of the wind energy system. Czick and Ernst estimate that HVDC transmission adds 30% to windmill costs.

A report from Electronix Corporation, Western Area Power Administration (no documentation available) says that 500KV lines capable of carrying 660MW cost $(US)600,000 per km, substations for 250 KV lines cost $160/kW, and undersea cable for 250MW lines cost $400,000 per km. These would seem to be substantial additions to the cost of long distance wind energy supply. Arnold (2003) reports that 5Gw HVDC lines from coal powered stations would add 40% to generating cost, at $2b (US0 for 5000km. Where lines are buried provision for heat dissipation would add to costs; some 100W/m. (I am trying to get costs for the Bass Straight line being constructed.)

However the cost of the feeder lines from windmills to the HVDC line would be substantial, given that a 5GW line would have to be connected to some 35,000 mills in a network over 10,000 square km. The connections between the mills would probably require some 17,000km of wiring.According to one estimate if 5GW HVDC lines cost $1000/kW this would add 40% to the cost of coal-fired power. (gerhaus@aston.ac.uk)

The problem of conductor size and weight also has to be taken into account. One report stated that for copper the diameter of the cables would have to b e 27cm and for aluminum 36 cm. Transmission lines now have steel cores, but weight and cost implications for large scale power transmission have not been explored here.

These large scale systems would also encounter the problem of seasonal variability mentioned above. In winter there is about twice as much wind energy as in summer.

Czick and Ernst, indicate that for the intercontinental system they consider (from Europe to Kazakstan, and from Siberia to Mauritania) output would still be 50% higher in winter than for the 4 summer months, and November output would be lower than the winter average. Czick and Ernst say this system could supply 30% of base load demand, if it had a non-wind backup capacity equal to 26% of the rated power of the windmills. This is a surprisingly large backup requirement for a system that is only capable of reducing supply from coal or nuclear by 30%.

Note also the political and moral difficulties that such a system involves. It would harvest for Europe the wind resource from an area some 5-6 times as large as Europe, in order to meet only 30% of (present) European demand. Surely the many people living between Mauritania and Kazakstan would also like access to energy harvested from their lands. In a just and sustainable world some energy exporting might be acceptable but the figures Czick and Ernst give indicate that there is no where sufficient wind in this large area to provide European per capita electricity consumption for all people living within it.

The resource and energy costs of wind systems.

The resources used in mills alone are considerable, and their Energy Return On Investment should be taken into account. A recent report on an Australian mill of 30 m diameter stated that it contained 237 tons of material above the ground, with another 15 meters depth of concrete under the ground. The energy content of these materials would b e in the region of 10GJ or 21,900MWh. This is 10% of the energy a 500kW mill would produce running full time for 25 years, so when down time is taken into account 1/7 of the mill’s output would have to go into paying back its energy costs of production.


Evaluation of wind systems should also include the resource and energy costs of connecting windmills together into farms and connecting these to the national grids. If we take the earlier estimate that windmills spread over 1374 square km would be required to equal the output of a 1000MW power station operating at .7 capacity, then the length of the connecting wire between these mills would be 3,435km.

A complete accounting would include the resource and energy costs of operations and management over 25 years. (Should the cost of workers’ travel to work be included?)

Conclusions regarding Wind energy?

Again it is difficult at this stage to state confident conclusions about the potential of wind energy. In many regions , especially Europe, Canada, New Zealand, Central US, and Crete, it will clearly make a considerable contribution to electricity supply, but even in Europe problems of variability, integration and availability of space seem likely to limit the contribution to a small fraction of present demand. There are several very optimistic claims re US potential, including a Worldwatch claim that it could supply all US energy, not just electricity. (Such claims often refer only to the energy in the wind, not the quantity that can be harvested and delivered when and where it is needed.)

However Tyner (2002, p. 13) concludes "…under the most optimistic assumptions, the analysis suggests that wind power is capable of furnishing only a small fraction of the net energy needed to power the US economy…"
Long distance and inter-continental transport of energy via hydrogen seems to be ruled out by the high losses involved (see below), and HVDC seems more viable for long distance transport but involves high costs and significant losses, and sets problems to do with equity (not within the US.)

Australia's prospects seem to be much less promising than Europe's. The resource might be quite large but most is presently far from grids. Again if these were constructed their cost would have to be added to total wind system costs, and losses in transmission would be significant. The above estimates re areas required seem to indicate that wind cannot meet current more than a fraction, say one-third, of demand. The rapid growth in demand for electricity will be commented on below.

LIQUID FUELS.

The second of the two crucial energy sources for industrial societies is liquid fuel and the potential solar source of this is biomass. The limit here seems to be much clearer and more severe than for electricity despite the fact that evidence and estimates on some of the basic variables again differ considerably.

Biomass yields and quantities.

The limits to liquid fuel production are not primarily to do with the energy return ratio (considered below). They are to do with quantity, i.e., the areas of land available and the associated yields.

Non-plantation sources

Non-plantation sources are far from sufficient to solve the problem. Lynd estimates that idle US cropland could provide only 14% - 28% of current US transport fuel (1991), even making the extremely optimistic assumption of 21 t/ha biomass production. (US corn plant growth is 15 t/ha with intensive application of fertilizer, water and pesticides on good soils. US average forest growth is only around 3 t/ha/y.) Di Pardo (undated) says that only 10% of US cropland is the maximum amount that could be used to produce cellulose biomass inputs.

Lynd estimates that 186 million tons of waste biomass (dry) could be collected in the US (at under $56/t, 1994 dollar, the higher of two costs examined). Lynd (1996, p. 412.) says this would yield 20 billion gallons of ethanol, which is only equivalent to 6% of US petroleum consumption.

The Oak Ridge National Laboratory says US forest wastes could provide 8Quads (8.4 EJ), whereas all US energy is around 85-90 Quads. (ORNL, undated.) This seems to be a very high estimate, equivalent to almost 2t/ha/y from the entire 250 Mha of US forest, not taking into account energy costs of harvest and delivery.

Plantations.

The plantation question should be seen in terms of what areas are likely to achieve what yields per year, via procedures that are sustainable over very large areas in the very long term. World average forest growth is around 2 t/ha/y (FAO, Undated) and the Australian average forest growth rate is probably well below the world average rate. However Mason (1992) says pine grows in Australian plantations at around 4t/ha/y on average and Bartle (2000) reports mallee harvest at 7.5 dry t/ha/y. Some Australian plantations achieve 10-12 t/ha/y growth, but these are in select small regions where conditions are unusually favorable. Giampietro eta al (1997) say woody biomass can be harvested at 8.5 dry tons/ha/y, but this would assume relatively favorable growing conditions.

Australia's forests total approximately 41 million ha but the potentially harvestable area might be only 20 million ha when water catchments, national parks and the wishes of private owners are taken intro account. Nilson et al (1999) conclude that in general possibly 40% of existing forest areas might not be accessible to biomass harvesting, being on steep slopes, near creeks or on private land or protected catchment. (These restrictions would not apply to plantations established specially for biomass use.) In addition note should be taken of the fact that if Australia were to be self-sufficient in forest products local production might have to be increased considerably. (Imports are $2.7 b p.a., while exports are only $1.2b p.a..)

Also, approximately 6 million t/y of wood are presently being harvested p.a. for domestic heating in Australia. Current Australian and world timber and fuelwood demand are probably well beyond maximum sustainable quantities.

These figures indicate that Australia’s existing forests are not likely to be capable of providing the large quantity of biomass required. (Estimates are given below.) As will be explained below, plantations for energy production are not likely to solve the Australian problem. Currently there are only about 1 million ha under plantations in Australia and its relatively poor soils would probably place severe limits on the extent to which this area could be increased and continuously cropped. Mercer says Australia might increase plantations to 10 million ha. (1991, p. 81.) The required area (estimated below) should be considered in comparison with the 23 million ha of pasture and the 21 million ha of cropland presently in use in Australia.

Yields.

Optimistic conclusions on the potential for biomass typically make very high assumptions regarding achievable biomass yields. For example Lynd (1996) and Foran and Mardon, (1999) assume dry weight yields can be 20-21t/ha/y, and these can be maintained year after year. Such discussions usually make reference to instances where yields of this order and greater have been achieved in specific locations or experimental conditions. For instance the Oak Ridge National Laboratory reports on switchgrass, willows and poplars in the US growing in experimental plots at 11-15 t/ha/y. (McLaughlan, 1999.) However for very large scale biomass production large areas of land would be required and it is not plausible that large areas with such yields can be found in the US, let alone in Australia with its poorer soils. (American agricultural yields per are around twice Australian yields.) Personal communications from ORNL state that these high yields are likely from only about 20 million ha of US cropland. ORNL (undated) estimates that only 8 Q would be available for fuel production in the US (presumably not including potential plantations.) Hohenstein and Wright (1994, p. 187) found that only 91 million ha of US farmland could yield an average of 5 t of biomass per ha per year. Graham (1994, p, 187) concluded that 88 million ha of US farmland will be available by 2030, but 75% of this will not be suitable for bio-energy production, meaning that only 16.2 million ha will be available.
Consider the following yields for Australian agriculture; wheat, 1.9t/ha/y (i.e., grain; total plant biomass might be 3 t/.ha/y), fodder, 3.5t/ha/y, overall agricultural production excluding sugar cane, 2 t/ha/y. (US wheat straw is 3.3t.ha/y; Pimentel email.) In other words biomass yield from Australian cropland, which is obviously the best growing land available, is under 4t/ha/y (...after the application of 3.5 million tons of fertilizer and considerable pesticide and irrigation inputs.) (A.B.S., 1997-8.) It should also be noted that c15% of biomass harvested is lost in six month storage (Wright, 1994) and that biomass energy production is likely to take fertilizer applications comparable to those in agriculture. US corn production takes c 135 kg of nitrogen per ha per year, and wheat 60 kg. Panney and Mason (1994) estimate that biomass energy plantations will require 50-60 kg of Nitrogen per ha per year. These energy cost equivalents have not been included in the following derivations.

The Australian CSIRO Beyond 2025 Report (Foran and Mardon, 1999) argues that biomass energy for Australia could come from the areas that need to be replanted to remedy Australia's dryland salinity problem. However dryland could be expected to have biomass yields that are a small fraction of those for average Australian cropland. Nevertheless Bartle (2000) expects coppicing of Eucalyptus mallees to yield 5-7.5 dry tons of feedstock per ha per year, although there is at this stage little evidence on the areas that could sustain such yields or how yields will stand up over time. Continued harvesting from nutrient poor soils could be expected to lead to deterioration in growth rates, or to require fertilizer application thereby adding to the energy costs of production. Morrow (19 ) reports that half Australia's farmland should be fallowed.

Bugg et al (2002) claim to have carried out the first thorough study of biomass land potential in the Australian state of NSW. They give figures for "capable" and "suitable" land, although the definitions are not clear. "Capable" appears to include areas in use for other purposes, and Bugg (personal communication) indicates that some of the "suitable" category would not be accessible for biomass production (because of prior uses, unwillingness of owners to put the land into biomass production, etc., see Nilson above on the severity of exclusion factors.)

Their Table 2 lists as hardwood production under "suitable", .6 mha @ 20+t/ha/y, 1/7mha @ 18t/ha/y, 2.3mha @14t/ha/y, 4.9mha @ 10t.ha.y 8.7mha @ 5.5t/ha/y,31.9 mha @ under 3t/ha/. The total yield, c 190mt/y, includes 41 mt in the last category that would not be economic to harvest for biomass in view of the low yield per ha, and another 41 m tons that would probably not be economic to harvest . When the exclusion factor consideration is added it might be reasonable to assume that the maximum harvest that could be taken for biomass energy production in the state is under 100 million tons. (This makes the invalid assumption that no more of the available land will be put into plantation timber production or woodchip exports.)

Little confidence can be put in an Australian figure derived from this study, but it would probably not be more than four times as great.

In view of the fact that replacing liquid fuel by biomass would require much more wood than this (below), only a negligible difference would be made by adding forest and agricultural wastes., for instance even if plantations grow to 10 m ha.

Berndes, Hoogwijk and van den Broek (2003) review 17 studies of global total biomass yield potential. Unfortunately these differ greatly, in assumptions and conclusions, and some seem to involve quite implausible growth rate assumptions (e.g., 46-99t/ha. which Berndes et al say are not supported.) However inspection of the core plot of estimates is instructive. This includes a yield by area graph for world grain production, sloping down from c 7t/ha to meet the base line at 700m ha. If a line is drawn parallel to this curve but at twice its height, to represent a total potential biomass production, the total yield under this line approximates that for the average of the 17 estimates plotted on this Fig 6. This would represent a plantation yield of c 12 t/ha on a small amount of the best land, tapering to zero yield on the last of the c 1500 million ha this line in effect assumes for biomass plantation. This is equivalent to an average yield of 10 t/ha from 600 million ha, i.e., a total yield of 6000 million tons. This is 120EJ of primary energy, compared to the current world primary energy consumption they state, i.e., 410 EJ. If converted to methanol it would yield a net approximately 6,000 million tons x 34 gallons of petrol equivalent per ton (this assumption is discussed below), i.e., 204 billion gallons, or 4.7 billion barrels, which is 18% of present annual world crude oil consumption. FAO (undated) indicates that the present forest harvest is about 6.6 billion tons, which is close to the 6 billion ton yield figure derived above from Berndes Fig. 6.

Algae as a biomass source.

In ideal conditions some species of algae grow at very high rates, up to 30 times the rate for land plants. Sheehan (1998) claims 50g/m/d, (which equates to 180t/ha/y although he does not say this growth rate can be kept up for a year.) Reference is made to a proposed scheme intended to harvest 67t/ha/y, more or less equivalent to sugar cane. The oil content can be 40%. Of special interest for energy production is the possibility of using sea water in large shallow desert ponds. 200,000 ha are claimed to be capable of producing 1 Quad, or 8.4 EJ of biodiesel. (Presumably this is a gross output. The claim is puzzling; if a 50 t/ha yield is assumed and algae have the same energy content as wood, then the gross production would only be 160PJ, only 2% of the claimed amount.). In any case that output corresponds to a photosynthesis rate of 7% p.a. When it is growing fast corn achieves c5%, but averages only .3 over a year. (Pimentel, 2004). Sorensen (2000, p. 311) says algae on reefs average 2%, but this could be raised to 3.7%.

Sheehan points out that yields are more like 10g/m/d in field conditions, as distinct from the lab. A major problem is that constant high temperatures facilitate high yields, but large scale energy production would involve large open ponds in deserts, where temperatures fall at night.

Cost estimates reported vary considerably, but the equivalent of oil at $(US)65-100 is quoted. Sheehan does not give energy costs of production.

One difficulty is that the conditions which increase growth rates reduce oil content. Starving the algae of nutrients raises their oil content. Another is that the sunlight conversion rate and therefore efficiency of the process is highest in low light levels, e.g., 10% of full sun.

Another major consideration is where would inputs come from for the large scale production of this biomass? Advocates refer to use of nutrient rich waste water from agriculture, but far greater quantities would be needed. Around 40% of the input material must be carbon dioxide and therefore the process could be coupled to coal-fired power stations, but it is not clear how far CO2 would have to be transported to hot regions for large scale production.

Mardon (2004), who has worked on various biomass input sources, including algae, for the Australian CSIRO, says that the energy cost of the process is so high that the energy return is negative. "…the energy required to grow (and more particularly to harvest and process) the algae is considerably greater than what you can get out of it." Winter growth rates were found to be slow. "…filters are not an effective way of harvesting them, so a lot of energy is required for centrifugation. Even then, the cell mass is very wet, and some form of dewatering may be required…" "…our field work showed that it was not a practical as a way of harnessing solar energy." Mardon also notes that ponds are prone to contamination, and require aeration.

However Briggs (2004) gives remarkably contradictory estimates re EROI, claiming that it could be 10 and even 20. (I am in the process of trying to resolve this puzzle. Briggs is unable to release detail due to patent considerations.)

Finally using power station CO2 would not affect the impact of that carbon on the atmosphere. It would end up in the atmosphere after the biodiesel was burnt. This factor alone would seem to disqualify large scale use of algae for the production of liquid fuels, except where grown without artificial carbon input. (I have not yet obtained an estimate of what the growth rate might then be.)

The photosynthetic limit to yields

The general limit on biomass growth and therefore on energy production from biomass is set by photosynthesis. In natural ecosystems only about .07% of the solar energy input becomes stored as energy within plant material, although in special agricultural situations such as sugarcane growing the figure can rise to .5%. (Pimentel, 1997, p. 14.) For a region averaging 5kWh/m/d of solar energy, natural vegetation would be storing energy at the rate of approximately only 1.4 kW/ha ( i.e., average continuous flow over 24 hr). Pimentel notes that not all of this will be harvestable as the plant will need to use perhaps 40% for its growth processes. (1997, p. 14.) This factor will be ignored below. This 1.4kW rate of solar energy capture might be compared with the average per capita US consumption rate for all forms of energy combined of approximately 10 kW.

Energy Return On Investment, (EROI).

Crucial in assessing the potential of biomass energy forms is the difference between the amount of energy produced in the required form and the amount of energy that has to be used to produce it. Two issues need to be distinguished here, firstly the proportion of the energy in the input biomass that ends up in the liquid fuel, which could be defined as the gross output, and secondly the amount of energy it takes to achieve this, which enables us to derive the net output, and thus the ER.

The situation is somewhat confused by the fact that in some processes some of the energy required can come from biomass that is put into electricity generation, etc., and thus reduces fossil fuels required, and does not use up some of the liquid fuel produced. This is the case with the discussion of methanol by Foran and Mardon below, and it can lead to misleading conclusions re ER . These situations might best be discussed in terms of the ER on fossil fuel invested, bearing in mind that this has been reduced by increasing the total biomass input (feedstock plus fuel) into the process. Thus it is important to attend to both the ER, and the amount of biomass being used per unit of liquid fuel output. The total energy input might be relatively high yet the process be viable, if there is "abundant" biomass to enable a considerable amount to be used to provide energy for the processing. That is a high total energy cost and low ER calculated with this total cost in the denominator) might be acceptable again if there is much readily available biomass to use. This was the case with wood-fired steam engines. Even though their energy in/out efficiency was only c 5% they were quite viable in situations where much wood was available, but they did not do that much work per ton burnt.

What proportion of energy in the biomass ends up in the liquid fuel?

The 1.4kW/ha flow of "potentially retrievable energy" into the feedstock, derived above from the typical photosynthesis rate, is equivalent to an annual quantity of 44,150 MJ/ha . This is the energy content of 2.2 tons of wood, equivalent to 800 liters of petrol. In some regions photosynthesis is much higher than average, but this 800 l gross figure would seem to indicate the upper limit for the gross energy output from a liquid fuel production process based on very large areas of land and therefore average photosynthesis. (For the net output the energy cost of production must be subtracted; see below.)

Ethanol production at present results in about 1/3 of the energy content of the input biomass ending up in the ethanol (Lynd, 1966, Australian Bio-fuels Association, 2003.) This is the equivalent of 267 liters of petrol gross per ton of biomass. Lynd (personal communication) predicts that it will become possible to convert up to 56% of the energy in the biomass to ethanol, corresponding to a gross yield of 448 liters of petrol per ton of feedstock.

How much energy is needed to produce the liquid fuel?

The production of liquids from biomass usually has a low (sometimes zero or negative) net energy return on energy invested; i.e., it might require more energy to be put into the harvesting and distillation etc. than is available in the resulting fuel. Conclusions from different analysts vary significantly.

First it is important to consider how the accounting should be a carried out. For example should useful waste energy from the process be subtracted from the input energy before a net energy cost is arrived at. This could be appropriate if that waste energy can be used in the process. Evidently there are no possibilities for this in the production of ethanol from corn, but where cellulose materials produce ethanol or methanol the lignin waste can be used to produce some of the electricity needed. It is not clear in Lynd's account what difference including electricity produced by lignin waste would make to the net energy required for ethanol production.

Secondly should we be concerned only about the input energy that must be in the form of liquid fuel, and subtract only this from output in order to arrive at a net energy return figure for liquid fuel production; i.e., should we ignore non-liquid fuel inputs? This might be acceptable if the non-liquid energy inputs needed are easily derived from other cheap and abundant sources. However in a sustainable energy world stretched for energy the large volumes of non-liquid inputs would also probably have to come mostly from biomass, so it seems appropriate to subtract all input energy costs from gross output energy when deriving an EROI figure.

The electricity could in principle come from non-biomass sources independent of the ethanol plant (i.e., other than generated from the lignin by-product). However from the earlier discussion electricity supply will be a major problem so it will not be assumed here that surplus electricity will be available from external sources for liquid fuel production.

Pimentel and Pimentel, (1998) conclude that for ethanol produced from corn "...about 71% more energy is used to produce a gallon of ethanol than the energy contained in a gallon of ethanol." (See also Pimentel 1984,1991.) Ferguson says the net energy capture of biofuels is "..so low that these methods are barely viable." (Ferguson, 2000b.) Ulgiati (2001) concludes that the energy return from ethanol produced from maize in Italy is .59, rising only to 1.36 when energy credits from waste are maximized. Slesser and Lewis say the return is .3 from acid hydrolysis and .125 from enzymatic hydrolysis. Giampietro, Ulgiati and Pimentel (1997) conclude from their review that the net energy return ratio for ethanol ranges between .5 and 1.7, again apparently without taking into account energy needed to deal with the waste water. (However Ulgiati, 2001, estimates this at only 1.7% of the ethanol energy.)

Lorenz and Morris (1995) argue that recent technical improvements now enable a positive net energy ratio for ethanol from corn, but only if energy credits for non-ethanol outputs are given.

More recently Shapouri et al (2002) have stated that the energy return for the production of ethanol from corn is 1.34. Pimentel criticizes this analysis for not taking into account all energy costs of production. However this figure is derived by subtracting from input energy the energy that would have been required to produce useful output co-products, such as corn meal. If the energy content of the non-liquid fuel co-product is disregarded Shapouri says the ER falls to 1.08. This is the relevant figure for our purposes, i.e., assessing the viability of ethanol as the major or sole source of the most crucial energy source, liquid fuel. However taking 1.34 means from corn at 7 t/ha ethanol production would be equivalent to 127l of petrol per ton of biomass input p.a.

Pimentel's recent study (2003) concluded that to produce a unit of energy in the form of ethanol, from corn, takes 29% more energy than the ethanol contains. If energy credit is given for the dry distillers grain output from the process, the deficit is still 20%. This study took into account emergy inputs, and details criticism of the Shapouri et al study.

However most if not all of these estimates derive from studies of the production of ethanol from corn. Lynd (1996) argues that cellulose inputs such as wood and grasses can have a energy return of 4.4 (1996, p. 439) , and over 7 in the long term future. Without disputing these figures, they are misleading and require careful interpretation. As noted above attention must be given to how energy return is defined, and which definition is most appropriate for our purposes of understanding the liquid fuel problem. Lynd's figure includes the energy output not in the form of ethanol. About 40% of the energy in the cellulose input biomass ends up in un-fermentable lignin, which can be burnt to produce electricity. Lynd says the electrical energy produced is equivalent to 20% of the energy in the ethanol, so the thermal energy in the lignin is about 60% of that in the ethanol. Again our concern is only with the ER situation regarding the production of liquid fuels, meaning that we are not consoled by the fact that other forms of energy might be derived from ethanol production. (However the elecltricity needed in the process could be generated from co-products.) Thus the ER might be 4.4 overall but for ethanol production alone it is only 2.75 from Lynd's account.

Lynd's figures indicate that one ton of biomass input (20GJ, a high value – some inputs are c 15GJ/t) will yield 6.6GJ of ethanol. Given that the ER is 2.75 then the energy needed to produce this ethanol is 2.4 GJ. Thus the net ethanol output would be 4.2GJ, equivalent to 128 liters of petrol per ton of input biomass. This is indeed the figure Lynd states in two sources for current technology. (1996 and 2003.)

It is not clear how the energy required as a liquid fuel to deal with the large volume of waste water has been taken into account in Lynd's figures. Giampietro, Ulgiati and Pimental, (1997, p. 591.) state that there would be 13 liters of high BOD sewage for each gross liter of ethanol produced, (1997, p. 210), requiring energy for treatment equivalent to 50% of the energy in the ethanol. Ulgiati (2001) says the figure rises to 33.58 liters per liter of net ethanol, i.e., after the energy cost of producing the ethanol have been deducted from the output. (However again he estimated the energy cost at only 1.7% of the energy in the ethanol.)
The large differences between Lynd, Shapouri and Pimentel regarding ER seem to remain unreconciled at this point (personal communications). They are probably due in part to the fact that Pimentel’s reference is mainly to corn as the feedstock and to existing production systems whereas Lynd is discussing cellulose inputs and theoretical possibilities as no plants of this kind are in commercial operation. (Lynd, 1996, p. 431.) Net ethanol production.

Again clear and confident conclusions are elusive as estimates vary significantly. Shapouri stes out conclusions on energy return from ten studies, ranging from –33,500BTU/gal to +30,600BTU/gal.

Giampietro, Ugliati and Pimentel (1997) say that the ER for ethanol is negative, which would mean that no net fuel energy can be produced.

From the figures reported by Foran and Mardon (1999) it can be estimated that ethanol can be produced at a net energy yield equivalent to 98 liters of petrol per ton of dry wood feedstock.

From the above discussion of Lynd's figures the net yield seems to be equivalent to 128 liters of petrol per ton of input, when credits for co-products are ignored.

Shapouri concludes that the EROI is 1.34 when the energy content of co-products is counted, but when it is not EROI falls to 1.08.

Methanol.

Methanol is claimed to be a more promising option than ethanol, (however note toxicity, below.) No commercial plants producing methanol from woody input material are in operation so clear conclusions are elusive.

Ellington (1993) provides an analysis based on current energy costs, taking into account emergy factors such as steel and concrete used in construction of plant. He concludes that for each ton of woody biomass input with an energy content of 18.89GJ, 9.95 GJ of methanol can be produced (i.e., 53% of the energy in the input biomass ends up in the methanol), but it takes 5.4 GJ to produce it. The EROI is therefore 1.84. Each ton of input biomass yields a net methanol output of 4.55 GJ, equivalent to 129 liters of petrol.

Foran and Mardon (1999) conclude that the methanol option will in future yield more gross energy in liquid form per ha of wood than ethanol. From their Table 4.3 figures, for an input of 2.2t of feedstock, plus .4 t to meet some of the energy cost of production, 1 ton of methanol gross (22.4GJ) can be produced. One ton of input biomass (feedstock plus fuel) yields methanol equivalent to 151 l of petrol, net.

Foran and Mardon say 48% of the energy in the feedstock ends up in the methanol, which is a relatively high figure in relation to other discussions. The Australian Biofuels Association say 33% of the energy in wheat inputs end up in ethanol.

As noted above, the calculation of an ER for the process is somewhat complicated by the fact that the material used as feedstock is also used to provide one of the energy inputs. The fossil fuel energy cost of producing the ton of methanol is 9.4 GJ, so the net output can be regarded as 13 GJ ,and the ER on the investment of fossil fuel is 22.4/9.4 = 2.38. However this is achieved by the additional investment of .4 t or 7.8GJ of wood, so the total ER might be stated as 22.4/17.2 = 1.3. Such a low ER might be acceptable given that the process uses "only" 9.4 GJ of fossil fuel, … if a lot of biomass is available. For our purposes, attempting to assess the capacity to replace petroleum and gas via methanol, the main issue then becomes not he ER but the amount of methanol produced per ton of biomass input, and thus whether there is enough land.

Berndes et al. (2000) conclude that future technology could derive 272 liters of methanol, equivalent to 136 liters of petrol, from one ton of cellulose biomass. It is difficult to evaluate their account. They assume energy required at 1/3 that assumed by Ellington, and Giampetro et al. The difference regarding electricity required at the plant is large, 3.89 GJ vs .5 GJ per ton of ethanol produced. It seems from their Table 1 that the .5GJ refers to electrical energy and should therefore have been accounted as 1.5GJ(th) (…although they assume 50% efficiency in production of electricity from biomass; again their discussion mostly assumes future technologies and efficiencies that might be achieved.) The footnotes b and c under Table 1 are not clear, dealing with how inputs are accounted, but they state that one way of accounting that could have been used would have cut their net yield by one third.

The differences between these conclusions are not so great. For the purposes of the following discussion it will be assumed that one ton of biomass can produce the equivalent of 140 liters of petrol.

Unfortunately there seem to be significant problems regarding the toxicity of methanol, especially with respect to motor repair. This factor has been reported to have led BMW to abandon R and D on methanol technology.

The demand for liquid fuel.

US petroleum use (in the mid 1990s) was approximately 6.6 billion barrels or 277 billion gallons per year. (Youngquist, 1997, p 187.) Transport was taking approximately 212 billion gallons or 801 billion liters. (US Department of Energy, 2000.)
In 1998-9 Australia used 1681 PJ of petroleum and 881 PJ of gas, a total of 2562PJ or 128GJ/person. (Australian Bureau of Statistics, 2000.) Combined petroleum and gas consumption is the equivalent of 20.5 billion gallons or 77.5 billion liters of petroleum. (Note that the energy consumption rate is growing; see below.)

Can the demand be met?

If we assume methanol equivalent to 140 l of petrol can be produced net from each ton of biomass, giving no energy credit for energy co-products that can be used in the process, then to meet the Australian oil plus gas demand of 2562PJ would require an input of 551 million tons of biomass pa. If we assume an average yield of 7 t/ha, 78.5 m ha would be needed, which is almost 4 times all cropland and twice all forest area. Such an average yield is highly unlikely from such a large area of Australian soil.
Bugg et al. (2002) indicate that if all "suitable" land in Australia was harvested the total annual growth yield would be 180 mt. It was estimated above that a realistic harvest might at best be well under half of this amount. From this would have to be deducted future expanded timber plantation yield, which might account for the whole amount (for instance plantations have been predicted to expand by 9 million ha and if yield averages 7 t/h this item alone would be 63 m t.)

To meet the US petrol transport consumption, 212 b gallons, 2.3 billion tons of biomass would have to be harvested pa, and at an average yield of 7 t/ha this would require 330 m ha. This is about two times all cropland and 1.5 times all forest. To include US gas consumption would increase the biomass needed to 8.5 b tons. Note that in 1980 174 million tons of wood were already being used for US domestic heating. (Pimental, 1988, p. 189.) These figures align with Pimentel's conclusion that US energy use of 85 Q is 30% greater than the 54 Q of total solar energy captured by all US vegetation. (Pimentel, 1998, p. 197, 1994.) By 2003 US use had risen to 96Q.

Khashgi , et al. (2000) point out that present US ethanol production is equivalent to .8% of gasoline use, and is grown on 1% of US cropland, meaning that some 120% of all cropland would be needed for a gross production of US gasoline. From this the energy cost of ethanol production would have to be subtracted. At another point they say only 14 m ha might be available for energy production in the US by 2030, and this might produce 4.8 EJ, gross. US 1998 total energy use was 90EJ, indicating that 262 million ha would be needed for gross energy output, some 1.65 times all US cropland.

Ferguson (2003) takes Shapouri et al's figures and shows that one-third of US cropland would provide only 1.2% of the average US energy use per capita, 9kW (i.e., net energy yield.) This is a remarkable figure (the derivation set out seems quite sound) given the very high net yield of energy assumed, 18.3GJ/ha/y from corn. (It is high because a considerable fraction of this total is an energy credit given for a co-product of the process, which I argued above should not be given when the concern is the liquid fuel account.)

Most regions of the world seem to have much less capacity than Australia to meet liquid fuel demand from biomass. The Australian total cropland, pasture and forest area per capita, 4.9ha, is much higher than for most regions of the world. The figure for Europe is 1.6 ha, Africa 3.3, USA 2.8, Asia .55, and for the world 1.43. Brown (2003, p. 329) gives even bigger multiples; Australian cropland per person is 3.5 times the US figure and 9 times the European figure. Australian cropland plus pasture per person is given as 15 times the US figure and 57 times the European figure, although these numbers might be rather meaningless given that Australia has a very large area of very low yield "pasture".

Khashgi ,et al. (2000) refer to Johansson's estimate (1993) that 350 m ha might be available globally for biomass energy crops, and that this could yield 80 EJ. (Presumably this is a gross figure.) Global fossil fuel use is given as 320EJ, four times as high.

Johansson's conclusion roughly aligns with the conclusion arrived at by Berndes above; the average production from the estimated world plantation potential, 6000 million tons pa, would yield the equivalent of 20% of current world petroleum consumption.

FAO figures (1991) point to similar conclusions. If cropland, pasture and forest areas are added, then the per capita averages for the US, Europe, UK and World are 2.4, 1.24, .4 and 1.5 ha. (These figures are probably too high, because the definitions seem far too generous, e.g., "forest" is 20% minimum tree cover. Thus the Australian total includes very large areas of very poor pasture and woodland, yielding a per capita figure of 30 ha…which at least again indicates Australia’s relatively fortunate situation.)

If all of this land was used to produce methanol at the equivalent of 150liters of petrol per ton, from wood grown at 7t/ha, then the per capita liquid fuel that could be produced in the US and the world would be 98GJ and 61GJ, respectively. This might be compared with Australia’s current per capita oil plus gas use per capita of 117GJ.

Of course all that land is already fully committed, indeed much of it is overworked, so to provide 9 billion with their liquid and gas fuels from biomass could require finding an additional 22 billion ha capable of yielding 7t/ha/y…on a planet with only 13 b ha of land. Note also that the world figure rules out the possibility of dense rich world countries with little land such as the UK importing their liquid fuels from the Third World.

Can biomass production for liquid fuel production be economic?

Australia's hay/fodder production averages about 4 t/ha, and 30 bales/ton, i.e., 120 bales /ha, and this would sell (pre-Australian 2002-3 drought) for about $550 gross income/ha. Australian Bureau of Agricultural Economics figures indicate that the cost of production is around $270-300/ha, meaning net income is c $270/ha.

If we assume land produced woody biomass at 7t ha and therefore from the above analysis yielded methanol at the equivalent of 7x132= 926 liters of petrol per ha, this would retail at the present pre-tax price of petrol energy for $370. When the dollar cost of producing the biomass, converting it into methanol, and distributing and retailing the fuel were deducted, it is likely that the income to the grower would fall well short of covering costs.

The Australian Biofuels Association acknowledges that farmers could not make a living producing inputs to ethanol production. At present ethanol is produced in Australia largely from "free" inputs supplied by the wheat and sugar industries.

Most indicative here is the fact that power stations pay c $20/t for coal, i.e., 80 cents/GJ. If wood biomass sold for 80 cents/GJ, a ton would cost $16 - 20. At 7 t/ha yield, gross income would be $102 - 140/ha/y. Again fodder producers in Australia gross around $400-500 per ha, and make c $270 /ha net income. Pimental (2004) says firewood sells for $600/t in the US, again suggesting that the price for use in liquid fuel production would have to be much higher than coal.

The point is that for biofuels to be economically viable against today's petrol prices, yields and/or dollar costs would have to be much higher than is likely for very large scale biomass production, which would have to use much more than the high yield lands available. At 5t/ha/y, biomass inputs to processing would have to be 4 - 6+ times as expensive as coal per GJ for biomass production to be as profitable as hay production.

This has the important implication that land with potential yields under 7 t/ha would not seem to be economically viable, cutting savagely into the area that could be used according to Bugg’s findings. (On the other hand, much higher prices for energy will come to be accepted, driving the threshold yield down and increasing the area economically viable.)

Conclusions on liquid fuels.

These considerations indicate that although a large volume of liquid and gas fuel could be produced from biomass, it is not plausible that this source could provide more than a large fraction of current Australian demand.

It should also be noted that if petroleum becomes scarce there will be feedback effects making the biomass situation more difficult. For instance if there is less fuel available and at higher cost then irrigation, transport, fertilizers and pesticides will become more scarce and costly and agriculture will tend to become more labour and land intensive, and agricultural produce will become more costly, reducing the availability and increasing the costs of inputs to biomass production. There will tend to be a shift from energy-intensive building materials such as kiln-fired brick, aluminum, steel and plastics to timber, again increasing pressure on biomass sources. Looming water shortages and the impact of the greenhouse problem will probably significantly reduce biomass production. Also global economic development is accelerating the rate at which people are moving to cities, where per capita energy and resource consumption is higher. (However the proportion of meat in Western diets could be reduced considerably, freeing much land for the production of biomass.)

The implausibility of biomass meeting the present liquid fuel demand indicated by the foregoing figures is reinforced by comments from others.

Giampietro, Ulgiati and Pimentel, (1997) find that to produce only 10% of US energy via ethanol would require 37 times the commercial livestock feed production. They say that providing US food plus energy via biomass would require 15 times the existing cropland, 30 times the agricultural water consumption, and 20 times present pesticide use. For Japan the cropland multiple would be 148. (p. 591.) "...none of the biofuel technologies considered in our analysis appears even close to being feasible on a large scale due to shortages of both arable land and water..." (p. 593.) Their discussion does not take into account pollution control measures required to deal with ecological impacts, notably the large quantities of nutrient-rich waste water. For these and other reasons Giampietro, Ulgiati and Pimentel conclude "...biofuels are unlikely to alleviate to any significant extent the current dependence on fossil energy..." (1997, p. 588.) Ulgiati (2001) also concludes from a detailed analysis of ethanol from maize that this is "…not a viable alternative."

Finally the ecological implications of large scale, intensive, continuous biomass production are unknown. Some would argue that nutrient removal equates to soil deterioration in the long run.

Thinking in "Footprint" terms.

The magnitude of the problem is made clear when expressed in "footprint" terms. (Wackernagel and Rees, 1996.) At the above output of 140 liters of petrol equivalent per ha /t/y, Australia’s per capita 128GJ/y petrol plus gas consumption would require 7.i ha.) In addition Pimental and Pimentel (1997) estimate that 2.2 ha of forest would be needed to yield the 10,000kWh of electricity used by one person in a rich country per year. Thus per capita liquid fuel plus gas plus electric energy production from biomass would require 5.9 ha. To this must be added the productive land area needed for food, water, settlements and pollution absorption etc. However the total global amount of productive land per capita available is approximately only 1.2 ha. If population rises to 9 billion (and the present rate of productive land loss ceases), by 2070 the per capita area will be approximately .8 ha.
What about the "hydrogen economy"?

It is widely assumed that the ultimate solution to the energy problem will be via "the hydrogen economy". There are persuasive reasons for co