A critical reassessment of the prospects for 100% renewable energy.

Ted Trainer

6.2.2017

In a number of articles over the past decade I have attempted to assess the possibilities and problems involved in 100% renewable energy supply, generally arriving at negative conclusions. In recent years the field has changed a great deal but remains complex, unsettled and controversial. The following notes outline the uncertain understanding of the field I currently have given my attempts to make sense of the information that has become available over the last few years. It is being circulated in the hope that critical feedback will enable an improved overview. Although long and somewhat detailed the discussion is easily followed and checked; all assumptions and derivations are clearly visible. I want to stress that the conclusions are not being asserted confidently; they can only be the implications I see within the evidence I have found.

The more recent information indicates that some of my earlier impressions were probably too pessimistic, e.g., with respect to Australian power supply, but this reassessment seems to me to strongly confirm the general view that 100% renewable electricity supply would at least be very expensive, and that 100% total energy supply could not be achieved at anything like an affordable cost. 

The most important recent development is the emergence of simulations based on actual climate data. Many analyses have concluded confidently that 100% renewable supply is possible without making any reference to the weather patterns in the regions under discussion. These are of little or no value.  The basic issue for any serious analysis is not whether 100% supply is possible.  Just as it is possible to meet a household’s total power demand using torch batteries, the issue is at what cost.  For a renewable supply system the answer to this question depends heavily on the amount and cost of the back up capacity required to enable supply through periods of low renewable energy availability, and this cannot be estimated without information on longer term weather patterns.

It is disappointing that the field is confused by a great deal of prejudice and readiness to make dogmatic assertions.  This is especially evident in pronouncements from green sources.  The dominant belief is that industrial-affluent-consumer society can be run on renewables, that the required alternative technologies are available but powerful fossil fuel interests and weak politicians are blocking their implementation. Many leaders of public opinion who should know better are willing to make dogmatic assertions without examining the field carefully.

It is of the utmost importance that the potential of renewable energy be clarified. If it can’t meet all demand there would seem to be only two options; either make an enormous commitment to breeder reactors, or dramatically reduce energy demand via transition to some kind of Simpler Way, (i.e., a society with very low levels of resource use and GDP, no economic growth, and mostly small, highly self-sufficient communities.) TSW perspective does not regard technical advance as capable of solving the big global problems. Because renewable energy has been seen by most as the major technical fix enabling continued pursuit of rich world ”living standards” it has been especially important for those holding TSW perspective to try to determine whether or not it is likely to enable rich world ways for 9 – 10 billion people.

ELECTRICITY SUPPLY

Two groups have recently carried out the first analyses of the (more or less) total Australian power supply task based on detailed weather information.  These find that the production cost of power (not total energy) would be c. 10 – 15 c/kWh in one case and c. 20 c/kwh in the other.  However both these pioneering studies of this complex issue inevitably involve a number of assumptions and simplifying omissions and the following discussion attempts to take into account some of these considerations in order to determine the implications firstly for the production cost of electricity and secondly for the retail price, under all-inclusive and real-world conditions.

Elliston, Diesendorf and MacGill (2012, 2013) deserve much credit for apparently being the first to attempt a national analysis based on actual weather data, recently made available by the Australian Energy Market Opertator. Their general finding is that 100% power supply can be achieved at a cost of around 10 – 15 c/kWh. For coal-fired power the cost per kWh produced for plant plus operations and management (O and M) and fuel is around 3c.

However the following discussion is based mainly on the more recent analysis by Lenzen et al. (I am listed as an author but made a very minor contribution.) The analysis concludes that the production cost of electricity might be in the region of 20 c/kWh, and possibly 30.3 c under fairly common conditions. They say the second figure’s “… scenario comes close to what would be implemented in the real world”. (The effect of the Elliston, Diesendorf and MacGill figure on the cost findings derived below will be considered.) The conclusions of both studies state the pattern of generating units which would minimise the cost of supply capable of meeting demand with high reliability, given the weather patterns for the year 2010. The Lenzen et al. findings are set out in their Fig 3 as a matrix of options that might be selected among, depending on the amount of biomass capacity within the system (…ranging from the present c 1.7 GW to 15 times that amount) and the price put on carbon (… which they find must rise to $500/t before carbon is driven out of the generating mix.).  The main plot enables a production cost in cents per kWh to be read off for any combination of these two variables.

The reasons for working here with the Lenzen et al. analysis rather than that by Elliston, Diesendorf and MacGill include

Š      The fact that it does not select locations in advance but uses an approach which in effect assumes that power stations can be set up everywhere to compete to sell their output according to their cost of production determined by the pattern of solar and wind energy at each of the different sites,

Š      The quite high dependence on wind in the EDM approach, up to 58% of supply (and 68% in Riesz and Elliston, 2016.) Trembath for instance regards 30% as the limit. Wood et al (2013, 1-13) say problems and costs rise to unknown levels over 30%. Lenzen et al. refer to a number of other studies expressing reservations re higher levels.

Š      The somewhat simplified transmission system (understandably) assumed in the initial study: Lenzen et al. make more detailed assumptions.

Š      Inclusion of geothermal in Riesz and Elliston, (2016.) (See below on prospects for geothermal.)

Š      The degree of dependence on CSP. (See discussion below of concerns about overestimation of CSP contribution.)

Š      Use of biomass gas for backup seems quite desirable. There is uncertainty re the accounting of the energy and dollar costs and efficiencies assumed for this path (which the authors note are uncertain.) Given the importance of back up for renewable systems, considerable space will be given here to documenting concerns re this option.

Some question whether the approach is will be useful at all. One source says “...no viable technology has been available to produce refinery grade syngas from biomass.” (Syngas Technology, 2013.) The AETA review of renewable technologies reinforces doubts about the viability of the path and does not attempt to estimate capital costs. It gives only seven lines to it, including the words, “No significant progress has been made on full scale development of such plants and none is anticipated in Australia in the near future.” (p. 53.)  Weisbach et al. say, biogas-fired plant are “…clearly below the economic limit with no potential of improvements in reach.”   (2013, p. 24.) Lenzen (2009) and other technical discussions say that cleaning several kinds of impurities out of the gas sets a “major technical difficulty”.  The gas is at low pressure and has to be compressed, which would affect net energy output. Raman and Ram, (2013), report that the overall efficiency of this path from biomass to power is only 16%. However Larsen (2014) says it might in future be raised to 45%. Timensen et al. (2002) estimate the gas production step could be 50% efficient. I have used the AEMO figure, 26%, (Crawford, et al., 2013) for generating electricity via biomass combustion, not gas turbines. Wood (2012, 8-2) says gas production requires”…substantial further improvement before they become alternative commercial options for energy production.” They point out that the differing chemistries of differing input materials requires somewhat different treatment processes. They also note the need for many small generating sites, to keep the transport distances down, but this increases the total cost of connections between generating sites and the grid. Thodey (2013) adds that small biomass gas generating plant is less efficient, around 20 – 25%, when the efficiency of the gas production system (potentially 67% at best, according to Van der Meiden, Veringa and Rabou, 2010, and Mardon, 2012), and the energy costs/losses in producing the biomass, trucking it, drying, and returning ash to the fields, and in producing, cleaning and compressing the gas are all taken into account the overall energy efficiency of biomass-gas-electricity component is likely to be well under 30%.    In their CSIRO study for AEMO James and Hayward (2012) say the energy efficiency of gas production from biomass is under 50%, If this is so the biomass-to-gas efficiency would be around 25%.  However Thodey (2013) says the yield is only 0.7 - 1 kWh of gas per kg of woody biomass, i.e., an energy efficiency of biomass-to-gas conversion of around 13 - 20%.

Similar uncertainties surround cost assumptions. It seems clear that the EDM costing takes into account only the (very low) capital cost of the gas turbines. They state c. $730/kW, which is the approximate figure AETA gives on p. 34 in their section on gas generation. That section is not on the complete biomass-gas-electricity generation process or its costs.  The capital cost of the full biomass-gas-electricity system would add to the turbine cost the costs for the gas producing plant, including removing impurities and compressing and pumping, and all the machinery and other inputs going into biomass production, harvesting, drying, and transportation. McKendry (2002) says biomass-gas producing plant capital costs are around 2/3 those of gas-electricity generating plant but Worley and Yale from NREL (2012) state higher figures for gas production than for gas turbines. Their biomass input and plant cost figures mean that plant capable of producing gas for a 1000 MW gas turbine would cost $1,005 million, 50% more than the cost EDM state for the gas turbine, assuming 60% efficiency for both gas production and electricity generation.  In addition the availability of biomass year after year in a land prone to severe and protracted droughts, and floods, is variable and uncertain.

Given the above doubts about the viability of producing gas by gasification or pyrolysis the system might have to be based on biomass burning to generate power via steam, which is the only path AETA discusses.  AETA (p. 52) states the capital cost of biomass CHP as $5,000/kW.  In other words if gas generation is not feasible the capital cost of the generation link alone might be almost 7+ times that which EDM seem to be assuming for the whole sector. In addition this approach would have slower ramp rates than a gas system, more like those associated with coal-fired generation, and would therefore be less adept at plugging sudden gaps.  That would have further implications for additional capacity to deal with intermittency.

The full real-world production cost?

There are several considerations complicating the drawing of implications from the Lenzen et al. production cost findings for the probable price firms and households would have to pay for electricity. Lovegrove et al. (2012 p. 192) points to the difficulty of predicting costs and on p. 109 note that for the (much less difficult) issue of CSP capital costs, ”Large differences between original cost estimates and actual installed costs have been common.” They point to for instance the possibility of significant future price rises for resource inputs to construction.

1. Lenzen et al. only include capital, O and M, transmission and “fuel” costs in their total for production cost. The Prieto and Hall (2013) study of the Spanish PV system attempted to take into account all costs, including usually overlooked factors such as plant security vehicle fuel use. For coal fired power the above four factors add to about 3 c/kWh, but when all other factors are taken into account, including company tax and profit, the wholesale cost rises to about 8 c/kWh (…and then many factors multiply this by approximately 3 to yield the retail cost.) We would need similar figures for a complex, multi-form renewable production system. Efforts to tally comprehensive cost inventories do not seem to have been made in EROI studies for renewables, except by Prieto and Hall for PV.

2. The cost assumptions are commonly quoted estimated 2030 values. These are generally around one-third lower than present costs (and a challengeable 50% lower for CSP, see below.) Although commonly used the set involves assumptions re expected reductions and “learning curves” that are open to doubt. For instance Wood et al. (2012), and Hinkley et al. (2011), reports no fall in CSP plant capital cost as built capacity went from around 90 MW cumulative to around 1200 MW, i.e., despite around a 10-fold increase in plant capacity built. Renewable Energy Focus, (2010) states the same general view and expects no fall to 2025. Fig 9. From Bollinger and Seel (2014) shows an increase in solar thermal cost over time. A report by the US Electric Power Research Institute (2010) expects no fall in capital cost for PV, wind or CSP until at least 2025.

Predicted falls are usually based on “learning curves” which project cost falls for each doubling of production, but due to the recent success of PV commercial interest in CSP has not taken off, and therefore the doubling period for it is likely to be quite long, meaning slow cost falls to 2030-50, and these are not likely to halve capital costs by then. In addition CSP mostly involves relatively simple technology and well established construction engineering so it could be that major cost-reducing breakthroughs are less likely.

For wind, the California Energy Commission (2014) does not necessarily foresee cost falls, for this relatively mature technology. Continued falls for PV are commonly predicted and likely to occur but Feldman et al., 2014, report several estimates indicating tapering, indicating around 20% falls to 2040, as distinct from 33% by 2030 as assumed in the AEMO and ETA tables used in this study.There are reports that subsidies for Chinese module production are being phased out.

Another cost uncertainty concerns the increasing difficulties and costs associated with providing relevant materials.  (Dierderen, 2009, Sveredrup, and Ragnarsdottir, 2014.)

3. As Lenzen et al. note, the cost figures used assume the exchange rate which held at the time, i.e., the Australian dollar cost of the imported plant would be  $1A = $1US. It has since fallen by some 25 - 30%, meaning that the Australian dollar cost of plant, which would be mostly imported, would be over 1.4 times as high as has been assumed.

4. Lovegrove et al. (2012, p. 22) estimate that remote area construction of renewable plant (in his case solar thermal) would cost 10% to 20% more than the commonly quoted figures which assume construction at US and European locations. Most of the locations in Australia would probably be remote.  He also says that the initial constructions would cost involve an additional perhaps 15% cost for technologies that have not previously been built in Australia. In addition regardless of locational issues Australian construction costs seem to be considerably higher than overseas costs, due in part to poor Australian productivity growth. (Freebairn, 2017.)

5. The study did not attempt to take into account the embodied energy costs of generating or transmission systems. This was wise given the uncertainty and disputation in this field, especially for PV. However an ER of 18 - 20 is commonly assumed for the most cost effective sector, wind. This suggests that when all inputs are included for all technologies, along with transmission systems, an overall system cost might be up to 10%.  The figure assumed below is 7%.

6. The year 2010 is unlikely to have been the worst ever for renewable generation.  My (somewhat superficial) examination of BOM and AEMO data for 3 solar sites and 5 wind sites spread across central to eastern Australia found that the average value over the three sites for the lowest solar radiation on record for each of the twelve months was 17+% below the 2010 figure. For wind the figure was 32% below the 2010 value.

7. The capital cost figures at the beginning of an enthusiastic renewable building program would be present costs, generally one-third higher than those anticipated for 2030. Thus even if costs fall as assumed the average cost of all plant built before 2030 would be perhaps16% higher than those assumed.

8. In view of the 2016 South Australian blackout caused by storm damage to the grid, and the subsequent loss of the Victorian interconnector, one wonders how robust the long distance transmission provision assumed by Lenzen et al. is, especially in view of the expectation that in future the greenhouse problem will increase extreme weather events. Would a sufficiently strong grid add significantly to overall system cost?

9. Palmer (2017) has pointed out that an ideal system as indicated by an analysis of the kind Lenzen et al. carried out cannot now be built, because much renewable capacity has already been built in locations other than those the study finds to be ideal. For instance wind farms have been put where grid access is favourable at present, not where a big picture analysis would indicate is best.

10. There are important issues to do with assumptions underlying the role of the solar thermal component which would significantly affect performance and system cost.  These will be considered in more detail below.

If only the factors identified in points 3, 4, 5 and 6 above are taken into account the production cost derived from the Lenzen et al. study is likely to be multiplied by 2.3, i.e., to between about 46 and 69 c/kWh. (This assumes Lovegrove’s 1.2 for remote construction but not his 15% first project factor.) If point 7 above is also taken into account the multiple increases by 1.15, to 2.65. It is plausible that if all these factors could be taken into account the multiplier would be at least 50% higher than 2.3.

It is difficult to speculate on what the real-world, practical production cost might be for mix with a much less than maximum use of biomass. (Lenzen et al. do not see the maximum use plotted as desirable/achievable; see below.) The case Lenzen et al. discuss as “… what would be implemented in the real world”, which has a production cost of 30.3c/kWh and assumes c. 30% of electricity would come from wind, 18% from CSP, and 18% from biomass. (This seems to be close to that represented in Fig. 3 by the fourth column from the left.) However that cost figure was generated by assuming a high biomass cost and presumably this cold be avoided simply by capping the system’s biomass use. Nevertheless these considerations indicate that the “real world” production cost this study indicates would be significantly higher than 20 c/kWh (before taking into account any of the above 10 factors.).

            The resulting retail price?

The very important but uncertain issue set by these figures is what retail price for power might a production cost in this 20 - 30 range lead to? The first question here is what wholesale price might these production costs result in.  At present the production cost of coal-fired power is c. 3c/kWh when all relevant factors are taken into account and this becomes a wholesale price of around 8 cents when all other factors operating within the firm exert their influence.  The most important of these would seem to be company profit and tax. (Some accounts seem to add insurance but others include this in O and M costs. Interest on borrowed capital seems to be routinely included in the capital cost of plant.)

For renewable supply the wholesale-retail difference would not simply be this  5 cent difference. If the cost of producing power becomes much higher than in the past then the cost of the power needed for all the operations and processes involved in building and running power generating plant will be much higher.  This includes the cost of the materials needed to build generating and supply equipment, because their production requires energy. In addition costs will pyramid. For instance the higher cost for steel etc. to build turbines will contribute to generating company costs, and the wholesale price of the power that company sells to retailers will be set higher to cover these and make a profit. The retailers will then put their profit margin on top of all their costs, and these costs will include the higher profit amount the wholesaler adds on, etc.

What all this will add up to would seem to be impossible to estimate at this stage, but the important point is that if the power production cost at the beginning of the chain of multipliers is at least twice the present amount as the Lenzen et al. analysis indicates then the retail price is likely to be significantly greater than the new production cost plus the present 22c/kWh for all steps between production and consumer.

If the Lenzen et al. analysis is accepted along with the multiplying factors in the above four points, and if the present difference between production cost and retail price remains at around 25c – 3c =  22 c, then the retail price would have to be at least (20c x 2.3) + 22c = 58 c/kWh. For the selected scenario Lenzen et al. discuss the sum would be (30.3c x 2.3) + 22c = 83c/kWh. These figures are respectively about 2 and 3.5 times the present Australian retail electricity price, which is already high compared with other OECD countries.

To these figures would have to be added the significant multiplier effects of the other 6 of the 10 factors noted above, especially to do with CSP (below) and these could produce a much higher ratio of retail to production cost.  Thus a ratio in excess of 3/1 would seem to be plausible, and this is likely to be significantly economically disruptive (…given the many other escalating  economic difficulties including skyrocketing debt levels, rising inequality, peak oil, the coming of the robots, and especially increasing climate related agricultural and environmental costs). However much depends on issues to do with the solar thermal contribution, and the following discussion indicates that revised assumptions in this domain would add significantly to system cost and retail price conclusions.

            The solar thermal contribution.

The most important of the complicating factors concerns CSP (concentrated solar power). Given the complexity of the modeling and computing tasks involved in the Lenzen et al. analysis it made sense not to attempt to take into account the uncertainty involved in the following aspects of CSP performance.  As a result there is a strong case that the solar thermal assumptions made in the study are too generous/favourable, by a considerable margin.

Various optimistic renewable energy supply scenarios, notably from Beyond Zero Emissions (Wright and Hearps, 2010), rely heavily on solar thermal generation. It could be argued that the prospects for this to be a major contributor have been overestimated and have receded in recent years. Unfortunately the companies developing the technology do not release detailed data on performance enabling clarification of core issues, especially the DNI/efficiency issue discussed below. De Castro (2017) provides evidence that performance of the US Ivanpah and Crescent Dunes units has been less than expected. For the latter a capacity factor of 0.5 - 0.7 was anticipated but de Castro says the average has been 0.12. Both the Elliston, Diesendorf and MacGill and the Lenzen et al. studies found that it would not be desirable for solar thermal to play more than a secondary role.

The performance of the large scale Ivanpah CR system (392 MW) has been disappointing, leading to doubts about its financial viability. (Symonds 2016, Danko 2015, Deitrich, 2016.) However its owners claim that it will take four years to come up to full performance. The published figures on its winter output have been so low that it does not seem wise to attempt conclusions; perhaps the system has not yet been operated at full capacity in winter. Similar concerns arise re winter output from the Spanish Gemasolar project. (See Trainer, 2014b.) Note that Gemasolar has 15 hour storage (also assumed by Lenzen et al.), but the general system power storage task could at times be to meet much of total demand over several days. (See Fig. 5 in Lenzen et al.) In addition Gemasolar is able to generate 15% of output using gas and the available output data is likely to include such use, raising further concern about its low winter performance (below.). Thus the extent to which CSP could be relied on when most needed for back up purposes in a strictly zero-carbon regime, is problematic.

The intention within the industry is evidently for solar thermal power plants to make its significant and profitable contribution in summer when the resource is most available. Thus the technology is likely to be an attractive proposition for generating companies, but it is not likely that it will be of that much value in helping to solve the main problem for supply systems, which is meeting winter demand. The reasons are discussed below.

There are three important elements in the Lenzen et al. study which are questionable but could be refined when the approach is further elaborated. It would seem that plausible adjustments here would raise system final costs significantly. 

1.  The capital cost assumption.  Doubts about this question have been noted above. The figure given for 6 hour storage (from AETA 2012 and ‘Scenario 1 2030’ in AEMO 2013) represents a large (56%) fall from the present cost given by AEMO. The 20 MW Gemasolar Plant in Spain was the first to be equipped with15 hour storage, assumed by Lenzen et al., and its capital cost has been reported as $419 million US, or a remarkable $21,000/kW. (Wilson, 2011.)  One would expect this first-of-a-kind cost to fall in future, but as noted above, over the past ten years or so there seems to have been no tendency for CSP capital costs to fall. Thus an appropriate future capital cost estimate for CSP is more uncertain than for wind or PV and the much lower than present figure used in the study (almost 50% lower) needs to be taken with caution.

2. The efficiency of generation.  Lenzen et al. assume 30% efficiency, based on a table given by Lovegrove et al. (2012, p. 49.) However the figures stated there are 12-15% for troughs and 20-30% for central receivers, and the latter are labeled “concepts”, presumably meaning values thought to be achievable by technical advance.  The maximum CSP efficiency ever achieved was 31.25%, by a US small scale experimental dish system in ideal conditions. (Concentrating Solar Power, undated, Wikipedia.) The record was made on a bright sunny but cold day, providing maximum difference between working and ambient temperature, which suits heat engines. This is not the situation that applies most of the time, especially when air conditioning demand peaks. Large scale use of CSP envisages use of troughs and towers, not dishes and these have lower efficiencies than dishes.

Lenzen et al. note that the analysis is based on present technologies, not anticipated future improvements. Thus it is appropriate here to consider some evidence on current CSP efficiency, which for operating trough systems is only around 14% and for central receivers around 15+%. For instance Lovegrove, Zawadsky and Coventy, 2007 state the solar to electrical efficiency of the operating trough, tower and dish systems they discuss as 10.59%, 13.81% and 13.94%. They anticipate a future efficiency of 19% for big dish thermal.  Hinkley et al. (2011, Fig. 7) state 14.58% for a CR. The SAM (theoretical, not actual) example 100 MW CR unit is predicted to generate 365 million MWh p.a. from a c. 1 million m2 collection field, at Daggett California where solar DNI averages 7.5 kWh/m2, indicating an efficiency of 13.4%. Kearney 2010 reports 15.5% for central receivers, and predicts16.2% for 2050 (p. 50.) Viebahn, Kronschage and Trieb (2004) predict 14.7% for troughs and 15.5% for CRs. The Abengoa PS10 unit under construction is expected to have a conversion efficiency of 17%. (Abengoa, 2016.) In 2014 the Gemasolar CR in Spain, three years after commencing, had an annual efficiency of 11.9%. (Marin, 2015.) It is permitted to use gas to generate up to 15% of output so if the quantity used in that year was deducted an even lower solar to electrical output efficiency figure would have been arrived at. De Castro (2017) reports that the expected and actual efficiencies for Solar 1, Crescent Dunes and Andasol 1 were 14.4% vs 12.6%, 15.4% vs 4.1% (sic), and 15% vs 11.5%. The Solarstor (2017) devices have a gross solar-electricity efficiency of c 21%, so the net figure after subtracting parasitic energy would be somewhat lower.

Thus it would seem that the assumed value is twice as high as most figures currently given.   Most important here is the fact that this figure applies to ideal conditions, and not to the periods of low DNI when CSP is called on to plug gaps.

3.  Net output and efficiency in winter. It is important to clarify how CSP generation efficiency and output vary with seasonal change, because the Lenzen et al. findings depend considerably on the contribution that CSP can make in periods when the renewable resource is at its lowest.  Fig. 6 representing five bad days shows that due to the lack of wind and PV input CSP is called on to provide a lot of electricity, almost always over 15 GW or 6e5% of average demand, and at times apparently up to about 24 GW, 64% of (peak) demand. (This is what Fig.6 seems to show but 19.4 GW is stated in the text.)

In this initial investigation it made sense to minimize the already very complex computational task by assuming that CSP generation efficiency would be the same at all seasons and levels of DNI and times of the year. However this is not the case and future applications of the approach might attempt to use more complicated assumptions, although this would probably be quite difficult to do.  As will be explained, the actual efficiency of a plant in winter is in general significantly lower than in summer but varies according to the priorities underlying its design.

First it is necessary to take out the effect of plant parasitic energy consumption, i.e., the energy needed to run the plant.  Lenzen et al. make the common and valid assumption that DNI must reach 200 W/m2 before any net electricity can be sent out. (For some dishes it is closer to 400 Wm2.) Taking out parasitic energy produces a ratio of electricity sent out to DNI that declines with DNI (to zero at 200 W/m2.) The important question is, is this the rate of decline observed in actually operating plant net output to the grid? If the actual fall off in the ratio is greater that that produced by the siphoning off of parasitic energy then something else is also reducing the efficiency of energy sent out to DNI, and the working/simplifying assumption made by Lenzen et al. i.e., that all DNI over that level is converted at the same efficiency, will overstate delivery.

The FOLLOWING evidence on actual performance indicates that the decline is indeed significantly faster than can be explained by the loss of parasitic energy. It can be seen from a plot of the power curve for a dish-stirling that the parasitic effect is around a 4% drop in net output for each 10% drop in DNI. However observed drops are greater than this. One source (unnamed, undated, p. 5 – 36) reports that for each 1% reduction in DNI, output from troughs falls more than 1%. Table 3 in that document shows that at Jodhpur DNI is 81% of the Barstow level but output (per unit of input energy) is only 75% as high. This general loss of efficiency is evident in ten of the 11 cases set out in the table. Modeling of SEGS by Jones et al. (2001) shows that on one selected day DNI was about 36% below its value on another day but electrical output was 56% lower. The energy taken by parasitics would explain only about an 11% drop. Similarly Geyer (2008) reports trough output dropping 48% when DNI dropped 41%. The Grattan report’s comparison between probable Brisbane and Melbourne CSP performance reveals that the 17% lower average DNI in Melbourne would be accompanied by a 40% fall in power output. (Wood et al, 2013, 4-14.) Plots in Siangsukone and Lovegrove (2003) show that a 20% decline in DNI was accompanied by a 49% decline in output.

There is also some indication that as DNI falls the effect reaches a critical/terminal point even though the DNI is still at a reasonable level, at least for troughs. Fig. 14 in Odeh, Bahmia and Morrison, (2003) shows that solar-to-electricity efficiency falls significantly with kWh/m2 of solar radiation, and at the low end of the plot seems to begin to plunge. They say, “…a SEGS plant could not operate efficiently if the daily radiation is below 22 MJ/m2”. The (somewhat superficial) examination of selected sites in Trainer 2013b found that in winter in Central Australia DNI does not often rise much above this level.  The BZE proposal assumes extensive use of CSP in regions with much lower winter levels than this.

It might be expected that since dishes can be pointed directly at the sun at all times the above effects would not be observed.  However surprisingly Fig. 82 from Kaneff (1991) shows a quite steep decline for a dish-thermal system. When DNI falls 40% from peak, to 600 W/m2, gross electricity generated falls 64%, so net output would be lower still.

A marked effect is also evident in the evidence via a personal communication from the generating company that operates Gemasolar in Spain (Marin, 2015. This aligns with the plot given in NREL undated.) This shows that the ratio of DNI to electrical output in a mid winter month yielded an efficiency of only 9%, around 75% of the annual value of 11.8%, (and 67% of the mid summer value, 13%.) For a drop of 24% in monthly DNI output fell 71%, i.e., the reduction in efficiency was far greater than could be attributed to  lower DNI. This is a surprising reduction given that the geometry to do with solar angle and reflectors is much more favourable for CRs than for troughs.

Two considerations indicate that the actual Gemasolar solar performance in low DNI would be considerably worse than the above 9%.  As was noted above, the system is allowed to generate up to 15% of output from gas. Most of this use would be in winter, so it is likely that the output generated from solar energy alone in winter was well below that yielding the above 9% efficiency figure. (NREL, undated, notes the need to clarify this issue.)

Secondly, the figures are monthly totals and the NREL plot shows that in winter at this location there are periods of high DNI and output within winter months, contributing most of the monthly output. However these were not capable of replenishing storage, aligning with the findings by Elliston et al. for Australia. The crucial task for solar thermal in 100% renewable scenarios is to be able to provide a lot of electricity (two-thirds of average demand for at least five consecutive days in the Lenzen et al. case) in those periods when DNI is low.  It is possible and likely that the Gemasolar winter monthly totals mask many long periods when purely solar input was contributing little to output.

The clearest data available seems to come from Solarstor, operators of the Australian Lake Cargellico CR. (Solarstor, 2017.) The device’s average efficiency in summer is twice its winter value. ”Field efficiency”, i.e., the ratio of heat collected to heat hitting heliostats, falls sharply as DNI falls, that is, faster than the fall in DNI. For instance when DNI drops 30% from 1000 W/m2 to 700 W/m2 heat collected falls by more than 50%, and at 400 W/m2 “field efficiency” is 61% below its level when DNI is 1000 W/m2 and heat is being collected at only 15% of the 1000 W/m2 rate. (Adding turbine efficiency would yield an solar-electricity ratio around 5%, and subtracting parasitic energy demand would lower this value.)

Thus there is considerable evidence that the generating efficiency of central receivers in poor conditions is quite low.  The importance of the issue is evident. In these scenarios CSP is called on (along with biomass) to deal with the back up problem and if its performance in winter and/or poor DNI is as low as these figures indicate then large increases in capacity, system cost and the production cost per kWh would result.

A major cause of the summer/winter effect seems to be the geometry of the field and receiver, determining that less of the DNI entering the field is collected and delivered to the block. This is due to the intention behind the design of the plant. Personal communications with NREL and CSIRO make clear that any CSP plant will have been designed to optimize performance given specific conditions and tasks, and these always include maximising output in summer when the resource is most available.  This means reduced efficiency in other times and conditions. For troughs summer output is maximised when they are laid out on a N-S axis, because the sun can be kept as close as the latitude allows to normal to the trough all day. However in winter that angle would be 44 degrees worse, and therefore much less (of the reduced winter) energy per m2 would be being collected.

For central receivers the relevant factor would seem to be heliostats shading each other. In winter at Daggett, the best Californian site, at midday the sun is at an angle of about 56 degrees from directly overhead. Thus heliostats on the half of the field located on side of the tower closest to the sun would be a long way from normal to the sun and therefore contributing relatively little. Those on the other side would be at a good angle but would be casting long shadows. A system designed to work best in summer would have heliostats spaced as close to the tower as possible, minimizing angles of reflection then, but increasing the shading effect in winter. In addition 15 hour storage requires a high “solar multiple” so that much more energy can be collected for storage than the turbine can use during the day. This means a very large collection field. The maximum solar multiple practically workable for a 100 MW plant seems to be around 2.5, meaning collector field size would have to be quite large and some heliostats would be kilometres way from the tower.  This in turn weighs against spacing reflectors at any greater distance in order to minimise shading.

This account in terms of solar angles aligns with the finding by Chhatbar and Meyer (2012) that each degree further from the equator a plant is located, efficiency declines about 1% (again a greater rate than the parasitic loss would explain) although they did not discuss causes or the angles and shading issue. 

If more or less correct this account means that Lenzen et al. have been overly optimistic in assuming that all DNI over 200 W/m2 is converted to electricity at a constant efficiency. If the analysis had assumed the Solarstor device’s field efficiency in poor conditions then the CSP contribution represented in Fig.6 would have required more than three times the 61 GW capacity found to be needed (or would have require resort to be made to PHS etc.) At the assumed capital cost of $6,256/kw this would have added $13 b/y (or $26/b/y) to system cost, raising production cost to 26 (or 36.3) c./kWh.

It should also be noted that Elliston, Diesendorf and MacGill generally find CSP to be of relatively low value in winter. Even though they assume 15 hour storage they say that in winter recharge of storage generally cannot provide for more than 5 hours supply.  Similarly De Castro (2017) reports on a four day period when all Spanish CSP could generate at only an average 1.5% of the peak rate. Unfortunately these findings do not align with the pattern evident in the Lenzen et al. five difficult day plot, which shows storage functioning strongly throughout the 15 night time hours for the consecutive days.

Resort to PHS would significantly increase system cost. In Table 1 in Lenzen et al. the 2030 capital cost of hydro without pumped capacity is stated as $5,110/kW, indicating that adding PHS would make the figure considerably higher. Let us imagine not equipping CSP with storage, thereby reducing construction cost by c. $1500/kw of capacity (see Lenzen et al. Table 1), and transferring the storage task to PHS. To deliver 1`kWh after storage would then involve a capital cost of ($6,256 - $1,500) to generate it plus $5,110+ to store and regenerate it via PHS, i.e., = $9,866+. Thus to provide the 19.4 GW required in Fig. 6 would involve a capital cost of $191+ billion. If the life of both components was 30 years the annual $6.3 billion p.a. would indicate an additional 3.1c./KWh on the production cost due to capital, O and M and transmission, and an additional 7+c/kWh after applying the above 2.3 multiplier.

            Conclusions on electricity.

The quantities arrived at above are not intended to be regarded as precise or confident but as indicating the kind of analysis needed to derive implications for the likely retail price of electricity from the studies such as that by Lenzen et al. which deal only with an understandably limited set of production cost factors.  However the assumptions and derivations set out above would seem to support the general impression that the retail price resulting from the pattern of generating technologies arrived at by Lenzen et al. would be much higher than the production cost they arrive at.

The mix of technologies arrived at by Lenzen et al. involves a considerable use of biomass, which few other countries could adopt.  Although the Australian biomass resource would seem quite adequate to enable this quantity, applications of the approach to other national supply situations should recognize that Australia has one of the highest per capita biomass resource bases of any country, around five times the global average. Thus for some other regions, notably Europe with its relatively limited biomass resource, resort to this option for back up purposes is likely to be quite limited.

This exploration suggests that on the global scale the quest for 100% renewable power supply will involve retail prices that might be tolerable in renewables-rich regions but again even there are likely to be at least significantly economically disruptive.

OTHER PROBLEMS ASSOCIATED WITH RENEWABLE ELECTRICITY SUPPLY.

The European situation re power?

It would be of great value if the Lenzen et al. approach could be applied to the European situation.  I assume that the model is available for application to climate data from other regions, in this case ideally a large area extending to North Africa, the Middle East and Siberia. My attempt to analyse the European situation (2013a) in the light of the limited mount of information then available indicated that there would be impossible problems, set mainly by their much less favourable solar conditions, the “anti-cyclonic” periods in European winters that can be long, cloudy, cold, stable and windless for weeks, and the relatively low biomass per capita available.  Their best strategy for dealing with intermittency and storage issues would seem to be “turkey nest” pumped hydro storage, although there are significant difficulties with this (discussed below).

             The EROI issue.

This topic seems to remain confused and unsatisfactory.  The commonly quoted EROI for PV has been around 8+ (ranging to 14) but when Prieto and Hall (2013) attempted to take in all the energy costs and losses involved in the Spanish PV system they concluded that it was around a surprisingly low 3. Palmer (2013) came to a similar conclusion for rooftop PV in Melbourne. Weisbach et al. (2013) conclude that if the energy cost of providing storage for PV is taken into account the EROI is again around 3. PV industry sources dispute these figures, especially over definitions and boundaries.

What we need for all energy technologies are breakdowns of all energy costs and losses according to several agreed/standard categories. These would include energy going into a) upstream factors, such as producing the materials and factories that make the turbines, b) producing the turbines in the factories, c) installing them, d) operating and maintaining them, including dealing with breakdowns, and the energy costs of all vehicles, clothing, accidents etc., e) decommissioning including subtracting energy values for recycled materials.  Either within these categories or added on should be all the easily and usually overlooked items Prieto and Hall tried to take in. This way of organizing the data would enable meaningful comparisons between technologies and anyone who thought particular items should not be included could use the relevant categories to derive an EROI value. Different indices of EROI would be recognized, e.g., upstream, purely within factory, all-inclusive…

The main EROI concern I have is that among the few studies of wind or solar thermal embodied energy none seem to take in all relevant factors as Prieto and Hall have attempted to do for the Spanish PV system.  Such studies would be very likely to arrive at values well below those commonly assumed. (The issue also indicates the need to look carefully at embodied energy overlooked in O and M cost tables.)

             A PV limit?

There is an indication from simulations that the contribution from PV might not be able to exceed 20%. (Riesz and Elliston, 2016), although as Blakers argues extensive use of PHS would increase the percentage. Output from fixed or non-tracking modules is very low in the morning and late afternoon and peaks at midday, while summer demand peaks in the afternoon and evening. Modules could be oriented to maximize input in late afternoon, but that would further reduce their morning contribution. Aligning some for early and some for late contributions would lower overall system capacity significantly.

             Geothermal?

It seems fairly clear that geothermal is not going to make a significant contribution in Australia in the near future, if ever. There are very substantial amounts of energy involved but the recent efforts to tap them in Australia have failed.  The main company exploring the field has gone bankrupt.  Evidently the technical difficulties are much greater than was expected. Wood et al. (2012) say the flow rates required are 10 times those achieved in the oil industry.

The rapid depletion of gas yield from fracking might be significant here. Geothermal involves forcing water down a drill hole then some horizontal distance through rock to an up pipe, and the question is what path will it take between the holes and how effective will this be in harvesting the heat in the rock volume that does not lie directly between holes?  Gas flows from fracked fields deplete fast and this suggests that in geothermal systems water might quickly take out heat directly between the holes but not harvest much of the total amount in the field.

The optimism expressed by the AEMO re 100% renewable electricity supply was significantly due to the assumed geothermal contribution. The following discussion does not consider geothermal.

             Intermittency, storage and back up issues and options.

Where there is sufficient biomass available it provides the best storage option (…except for those few regions where hydro or geothermal is abundant.) In those many situations where biomass is far from sufficient, notably Europe, pumped hydro storage would seem to be the best option.  Following is a brief discussion of the main options (…except Compressed Air Energy Storage), finding that all are problematic. The most promising, apart from biomass, which is pumped hydro storage, is likely to add significantly to costs when applied to intermittent sources (see below.)

First we should be cautious re the amount of storage it is assumed will be needed.  Some advocates of pumped hydro option proceed as if 24 hours would be sufficient (e.g., Blakers, 2016, Hearps et al., 2016) but Abdulla, et al. (2014) say 58 hours, and Pickard (2014) says for Australia the goal should be 7 days. Oswald (2008) says 6 days for Europe. Aghahosseini, A., D. Bogdanov, and C. Breyer, (2016), say 14 days for the US. Fig. 6 in Lenzen for the difficult five day period show that the backup sources CSP and biomass must meet half to three quarters of demand on all of those days. It was noted above that Riesz and Elliston found CSP could not store for more than 5 hours supply in winter.

            Biomass.

Both the Elliston et al. and the Lenzen et al. analyses depend considerably on the ability to use a biomass capacity for back up purposes.  There would be little doubt that this is the right strategy for Australia, because Australia has one of the highest biomass resource potentials in the world, perhaps around 5 times the world per capita average.

The AEMO (Crawford, et al., 2013) estimate of the amount of biomass available for energy production in Australia, 96 Mt/y, is quite sufficient to deal with intermittency. This estimate will be used here. Others indicate much higher potential yields, e.g., Foran (2015), but the conclusion arrived at by Farine et al., (2012) is only 65% of the AEMO figure. Even if the AEMO estimate is taken there is too little to back up electricity supply as well as to make the major contribution to the need for liquid fuel for transport. (Discussed below.) Lenzen et al. say the amounts at the upper end of the biomass quantity range they consider “…are unlikely to be available given competing food and biodiversity objectives.”  It is noteworthy in their Fig. 3 that what might be regarded as the most acceptable 100% RE mix, that is one which keeps wind down to 30%, requires 18% of electricity to come from CSP and about the same amount from biomass.  That would require about 38 TWh, 137 PJe, or 526 PJ of biomass, which is 29 M tonnes. This seems feasible, being only 27% of the maximum quantity they consider, but it would reduce the amount left for transport to around 66 Mt, corresponding to about one-third of present liquid fuel energy used.

World biomass energy potential is limited. The IPCC, 2011, estimates total plantation plus waste potential at c. 420 EJ/y (see also European Biomass Industry Association, undated). The 2014 IPCC Working Group 3 report states a much lower global potential, up to 270 EJ/y of primary energy.  

If 450 EJ/y was put into liquid fuel production it might yield 180 EJ/y of ethanol, which if shared among the 9.7 billion people expected by 2050 would provide them with18 GJ per person; European consumption for road transport alone is almost 40 GJ per person. Australian non-electrical inal energy donswumption is about 120 GJ per person p.a.

The Europeans seem to be in an especially difficult situation. Estimates of future European biomass potential put it at sufficient to meet at most a mere 3% of transport fuel demand, and possibly only a quarter of that amount. (European Commission, 2013, Wetterlund et al., 2012, de Witt and Faaij, 2010, Bentsen and Felby, 2012.) European maximum 2020 potential biomass production has been estimated at 13% of present total primary energy consumption.  (Europa, 2012.)

So biomass is not likely to make a big contribution to enabling backup of electricity production and provision of liquid fuels for a world intent on having rich world “living standards”.

Batteries.

For battery storage on a very large scale the cost would have to fall a long way. It is commonly argued that this will happen as it was the case with PV, and this might be so but it is not clear whether the two are comparable. What factors were responsible for the PV decline, and are these relevant to the battery case?  Mass production is obviously a common factor, but PV module production in China seems to have been heavily subsidised, and based on low-cost labour. Palmer (2016) believes the present embodied energy cost of production can only be halved.

Batteries are being used in relatively small scale utility level electricity supply systems.  EPRI says the cost is between $500 and $2,500/kWh (Patel, 2012.)  The biggest grid-level system is in Fairbanks, Alaska.  It weighs 1,300 tonnes but can only store and deliver 46 MW for 5 minutes.  This corresponds to $1,600/kWh.  (Nelder, 2010.) To store the output of a power station for 24 hours would require storage of 24 million kWh, and at the Fairbanks cost it would cost would $39 billion … which is approaching 20 times the cost of the power station.

Another indicator might be the cost of the Tesla Powerwall domestic battery.

This is $7,000 for a set that stores 5.8 kWh/d, i.e., $1,206/kW of storage capacity.  (Shahan, 2015.) At this rate to store output from a 1 GW power station for one day, i.e., 24 GWh, would cost 24M kWhx $1,206 = $29 billion, which is about the cost of 8 - 10 coal-fired power stations. In addition the battery set might have to be replaced after about 13 years, so the cost for a system that replaces 50 year coal fired plant would be c. $104 b.

Others who doubt that batteries can back up grid level supply systems include Istvan (2015), and Morgan, (2015). Palmer (2016) says batteries “ … could only usefully contribute a short term role to buffering variable renewable energy.”

Optimistic battery expectations depend primarily on use of Lithium, and although resources are at times regarded as large, they are not if the task is defined as enabling rich world “living standards” for all the world’s people. Wister (undated) says up to 0.3 kg of Lithium metal are needed to store 1kWh, and a vehicle might need 40 kWh of storage.  Thus if electrified the world’s present c. 1 billion vehicles would need over 50 Mt. Gruba and Medina (2010) report a global resource of 38 Mt. Wister (undated) reports 28 Mt and points out that reserves are much smaller than estimated/predicted resources, around 13 Mt. However it is likely that the coming demand will see reserves increase significantly.

Far more problematic would seem to be the amount likely to be required for grid level power storage. To store the output from a 1GW solar power plant to enable supply for three cloudy days, 3x24hx1GWh = 72 GWh would have to be stored. If 0.3 kg is needed to store 1 kWh this would require 23,000 tonnes of Litiium. If the world’s present 3,000 power stations were renewable equipped with 3 day storage 72 Mt would be needed. And if 9.7 b people were to rise to present rich world per capita power consumption the figure would be multiplied by about 6.

Possibly the biggest problem is to do with battery lifetime.  According to Smil (2010, p. 29), Lithium batteries deteriorate quickly in the field. Conger (undated) states c. 5 years for EV batteries.  Others say that in future they will be able to last the 15 year lifetime of a vehicle. The Lithium can be recycled but evidently it is often spread through electronic components in ways which make recycling difficult. Consider the capital cost of equipping solar or wind power stations with 3 – 7 day storage capacity that has to be paid for perhaps 3 or more times over the period in which a coal-fired power station could guarantee against intermittency and storage problems at no storage cost. Performance also deteriorates significantly if temperature rises above 30 degrees Celsius, making Lithium batteries problematic for Australian conditions.

These numbers cannot be taken very confidently because significant differences might be made by greater Lithium discovery effort, technical advance and price changes but they do indicate the magnitude of the improvements that would be required before Lithium storage could make a major contribution to a renewable world.

Store in Electric Vehicle batteries?

The common assumption that EV batteries can make a major contribution involves significant problems, in addition to the above battery limitations. The main one is that vehicle batteries need to be fully charged when the vehicle is to be used, which is typically twice a day and is not always predictable. This sets difficulties regarding the time available to carry out the various processes involved. It would take time to recharge the batteries from the drive to work, then there would be the storage time until the energy is needed somewhere in the economy, and the time which that use takes, then it would take time to recharge the battery again to be ready for the drive home from work. (Rapid charging could become the norm, but at a significant capital cost.) Many car users could not predict confidently when they were likely to want to use the car.

Then this storage and output arena has to align with the times when surplus renewable input is available.  The capacity of EVs to charge up, store, supply to the grid, and recharge before use is greatest at night, but there is low need for power then. It is therefore difficult to see how an electric vehicle battery could be available for a sufficient length of time to make a major contribution to meeting the need for electrical storage.

However it could be that each vehicle in a large fleet might be programmed to enable a small drawdown on its battery under specified conditions and times, thereby accumulating a significant amount of useable capacity for the grid to draw on at any time.

The main need for stored energy does not occur within a single day but involves coping with demand through sequences of cloudy and calm days, and it is not evident how vehicle batteries could make any contribution to solving this problem.

The limitation set by a lengthy charging time could be eliminated by battery swap systems at “refuelling” stations. However this would significantly increase the quantity of batteries required, with consequences for system energy and materials costs, including Lithium demand.

Finally there are the economic implications for the car owner. If he pays a retail price of 25 c/kWh for the electricity to charge his battery, loses some of this in charging and sells the remainder back to the grid at the c. 7 c/KWh which other electricity suppliers are paid he is not likely to want to be involved in the scheme. In addition his contribution to the storage task would be shortening his (already short) battery life due to the additional charging and discharging cycling.

These considerations align with the impression Lenzen's review (2009) expressed, i.e., that not much excess wind energy is likely to be storable in vehicle batteries.

Pumped hydro storage, (PHS).

This is by far the most promising storage option, involving electricity generated by water that has been pumped up into dams. Almost all large scale electricity storage schemes that have been built are of this kind, indicating that the rest are much less attractive. 

However it is a mistake to take the figures from existing PHS systems as indicating its viability and cost for storing renewable energy.  It will be explained that the intermittency of input makes a major difference to the quantities and costs involved.

The energy efficiency of PHS is sometimes claimed to be 80 – 90%, but for the c. 20 existing schemes reported in the ROAM Report (2012) efficiency is on average around 75%, with one at 50%+.  This is misleading because existing projects do not involve highly variable input energy from renewables, which reduces pumping efficiency (…large volumes of water in the pipes accelerating and decelerating), and involves major pumping capacity implications (below.)

In view of the magnitude of the storage problem retrofitting existing dams can’t be the answer. World hydro-electric generation meets only about 15% of electricity demand, and the present 10.7 EJ/y output is not likely to be doubled.   Hydro electricity has contributed c. 9% of supply in Australia but seems to be heading for 5% in recent dry years.  It can provide 18% of demand for a short period.  So a large scale PHS solution would have to involve construction of many new systems.

There are not enough sites for construction of a large amount of normal dam capacity. Mackay’s analysis (2008, p. 232) of the UK situation with high rainfall concludes that the potential capacity is only a small fraction of what would be needed. NREL (2012, 12 – 22) states a surprisingly low PHS potential for the US, 35 GW (electricity consumption in the US is c. 500 GW), at a cost of $5,595/kW.

The most promising possibility would seem to be the “turkey nest” arrangement whereby the same water can be circulated between relatively small high and low ponds, or the sea is used as the low “dam” and a small pond is constructed on a high nearby cliff. The Melbourne Energy Institute (Hearps et al., 2012) says there are 15 sites in Australia with “technical potential”.  (This term is not very meaningful; there are many considerations that typically greatly reduce what’s technically possible to what is feasible after all social and ecological cost factors have been taken into account.)  They say ecological problems would rule out some of these sites, especially seepage of salt water into soils, although proposals usually assume plastic sealing of the ponds. The capital cost MEI arrives at is not high, to provide the equivalent of the 25 GW average national demand for 20 hours. The ROAM study (2012) also found that Australian storage capacity could be large, c. 516 GWh, sufficient to meet total Australian demand for 20 hours. However the Report says that environmental considerations would rule out many of these sites. (Several have estimated a need for much longer storage; see below.)

            But how applicable is PHS to intermittent input?

The literature on storage of renewable energy seems to give no attention to the fact that analyses of existing PHS systems refer to a very different situation and task, and these do not indicate the feasibility or cost when the task is to store highly intermittent input. This is because the output from sun and wind varies considerably almost all the time, whereas the information on PHS performance and costs etc. derives from situations where there are long periods of pumping at constant rates, typically to enable coal-fired generators to tick over through the night.  Large volumes of water in tunnels that might be 13 metres diameter and many kilometres long take a lot of energy to get moving and if the energy available to pump is waxing and waning the energy lost in frequent pump rate changes would be considerable; prohibitive according to Lang who suggests that for the Australian Snowy scheme the pump rate would have to be constant for at least three hour periods.  Existing PHS schemes have not been designed or costed to cope with variation in pumping rates due to variation in power coming in to be stored. (However turkey nest schemes would largely overcome this particular problem.)

Input intermittency creates significantly increased costs in three areas.

1.    Additional generating capacity.

If PHS is to be added to a renewable system in order to make a large contribution then the generating capacity to do the storing will have to be additional to what was there.  The initial capacity would have been only sufficient to meet demand (i.e., more or less, most of the time). Yes at times the initial system would have generated more than was needed and some energy would have been spilled, but a well-designed system relying on for instance biomass for back up would have little or no wastage of generated power.

This means that every kWh to be delivered from PHS will involve/require the addition of separate expensive plant to generate it.

2.  The pump rate.

Those spiky wind plots mean that much of the energy available for storage would only be available for those usually short periods when the winds were very high.  This means we would need a lot of pumping capacity ready to turn on and run at full bore for relatively short periods. This is not the situation with those fossil fuelled systems ticking over through the night sending power to the pumps at nice stable rates.

3. The storage volume.

The proposals by Blakers and by Hearps et al. assume 15-20 GWH would be sufficient for Australia, enabling demand to be met for about 20 hours.  But as has been noted above some people see a need for much greater storage; 58 hours in the case of Abdulla, et al. (2014), and 7 days according to Pickard (2014). Lenzen’s Fig 6. Indicates a need for storage to supply half to three quarters of demand for every one of at least 5 days. Oswald (2008) and Aghahosseini, Bogdanov, and Breyer, (2016) state several days for overseas sites.

So it would be a serious mistake to take the costs evident in existing PHS systems attached to fossil-fuelled systems as indicating costs for systems intended to solve intermittency problems set by intermittent inputs. 

Note that intermittency also sets this problem for most other storage options, such as hydrogen, compressed air or heat.  All would either involve very large and infrequently used capacity to convert the energy into a storable form at times of high surplus availability, and large capacity to hold the stored energy, or would have to dump a lot of energy that was being generated at those peak times. Palmer (2016) recognizes this, saying storage “…exhibits marked diminishing returns.”

The intermittency situation has surprising implications for plant sizes and costs.  This is evident in the study by Connolly et al. (2012, also reported on p. 36 of the Melbourne Energy Institute study.) on the conditions to enable c. 100% electricity supply from wind in Ireland. Irish demand is on average 3.5 GW. Fig. 22 shows that 2 GW of wind capacity would enable wind to contribute 20% of demand without storage.  To be able to contribute 60%, 6 GW of generating capacity and 10 GWh of storage would be needed.  But for the contribution to be 96% capacity would have to rise to 9 GW and storage to a remarkable 500 GWh.  In other words as wind penetration rises above 50% generating capacity and storage required rise at an accelerating rate, and if wind was to supply almost all demand then wind generating capacity and pumping capacity would have to be 2.7 times national demand, and there would have to be enough storage to meet demand for 6 days.  And Ireland is possibly the best wind region in the inhabited world.


These considerations cast doubt on the figures given in the ROAM report, especially the ease with which a 20 hour supply could be provided. The report sets out the 38 sites identified in order of increasing cost per kWh up to a $1,000/kWh cut off (at which point cost is accelerating steeply.) So to add sufficient sites to extend storage from 20 hours to several days would require use of many of those far more expensive sites. Even at $1,000/kWh an Irish PHS with 500 GWh storage would cost $500 billion … which is enough to build about 155 coal-fired power stations…when Ireland’s demand can be met by less than 4.

Because Budischak et al. (2012) found storage for a 100% renewable electricity supply system to be so costly they recommended reducing the need by greatly increasing the generating capacity, to about 8 times as much as would meet demand if in the form of coal-fired plant, and accepting considerable power dumping.

Finally it would seem that the wind and solar inputs would have to be dealt with separately, with an additive impact on costs.  If wind surpluses tended to occur when solar surpluses were low then much less pumping capacity would be needed than if they often coincided.  But they are in fact often likely to peak at the same time (good wind in the middle of a sunny day) so if we were to collect all the surplus solar and all the surplus wind energy at these times we would need the great deal of pumping capacity required to harvest the combined/coinciding quantities (…or, again, settle for less than that but let a lot of power go at these times, meaning wasted costly generating capacity.)

Hydrogen.

In my uncertain view hydrogen is likely to be the best option for meeting liquid fuel demand in a future energy system, but not for an industrialised, affluent energy-intensive consumer society. It is costly and inefficient but the basics are technically simple. (Bob Lloyd has had experience with remote settlements and does not regard it as a viable solution. Personal communication.) However in a Simpler Way society there would be very little need for energy storage, and the limited biomass and pumped hydro capacities would be able to meet much of the demand.

If we assume 0.7 efficiency for production of hydrogen from electricity, an optimistic 0.8 for storage and distribution by compression, pumping or tanking, fuel tank filling, and 0.4 for fuel-cell operation, the overall efficiency would be 22%. Bossel, (2003, 2004), confirms this general impression. Pellow et al. (2015) estimate 30%, and do not find hydrogen to be a viable option for storing surplus wind energy. In fact plausible assumptions can make the final figure closer to 10%. (North, 2005.)  Mackay says “…hydrogen powered vehicles are a disaster.” They use more than three times as much energy as a petrol driven car. (2008, p. 7.)  To these factors must be added to large embodied energy costs of compressors, high pressure tanks and pipes.

It is not clear that fuel cell efficiency can be raised to c. 0.5 - 0.6 as some assume.  In addition, platinum resources are insufficient for large scale use of PEM fuel cells, and for electrodes in hydrogen electrolysis (Gordon, Bertram and Graedel, 2006). Other forms of fuel cell might become viable, but the prospect of alternative electrodes seems to be debatable.  Because the hydrogen atom is very small and light it leaks through valves and seals easily.  It also reacts with other elements, and makes metals brittle.  How often would pipes etc. have to be replaced, or fitted with plastic liners?  How much energy and resources would it take to put in a plastic pipe distribution system (inside steel pipes to take the pressure)?  Consider the extent of the existing gas supply infrastructure; another more expensive system about as big would have to be installed for a hydrogen distribution system (if only because the gas system will still be in use.) 

Bossell (2004) details these and other difficulties.  For instance he points out that a standard road tanker can deliver 20 tonnes of petrol, but it would only deliver 320 kg of compressed hydrogen. To pump hydrogen to Europe from the Sahara would take 65% of the energy going into the pipe line at the start. It is therefore not likely that energy-intensive societies could be run on hydrogen shipped around the world in tankers from sites such as the Antarctic where winds are strong. Pressurised tanks in vehicles would add weight, reducing the efficiency of vehicles, and constitute a greater explosive crash risk. 

Consider the capital and embodied energy costs of a system to deliver 1000 MW via storage.  This would have to include the capital cost of the wind turbines, the transmission lines, the hydrogen generating plant, the compressing, pumping and storage equipment capable of handling very large volumes of gas, and on top of all this the cost of the “power station” required to produce electricity from the stored hydrogen.  The last item would be equivalent to the 1000 MW coal or nuclear power plant that would have avoided the need for all this plant on the hydrogen path.  To deliver the initial 1000 MW electricity input we would need 3,000 MW of wind capacity, even at an ideal wind site, and 4,340 MW at the world average site where capacity is 0.23 (IPCC, 2007, Section 4.3.3.2.)

The AEMO (2012, 12 - 7) study of renewable potential assumed that hydrogen storage for power generation was too costly to consider. It noted that efficiency might only be 28%, and regeneration of electricity after storage as hydrogen would be very costly.

However it will be argued below that a 100% renewable energy supply system for an industrialised-affluent society would have to involve large scale use of hydrogen, for purposes that cannot be run on electricity.

            Storage conclusions?

The above notes indicate that the storage and back up problems set by the goal of 100% renewable electricity are not likely to be easily solved. Biomass seems to be by far the best option, where it is available in large quantity, but it will also be in demand for liquid fuel production. It seems clear that it cannot be anything like a solution to both problems on the global scale.

“Turkey nest” PHS would seem to be the best general storage option but it is likely to add at least a significant amount to the already considerable cost of 100% renewable electricity indicated above.  However it is not likely to help with the much larger task of meeting total energy demand, which the following section suggests would require large amounts of hydrogen.

                                    MEETING TOTAL ENERGY DEMAND.

Meeting electricity demand from renewable sources is a quite different task to meeting total energy demand.  At present electricity makes up less than 20% of total energy demand in a rich country, and it is the form most easily provided by renewables. Biomass is the only renewable form that does not directly produce electricity. Providing all forms of energy needed from renewables is a much more difficult task than simply scaling up the power supply system by a factor of 5. This is mainly because most of the remaining 80% of energy used is not presently in the form of electricity and therefore sets problems to do with a) the nature and number of these other forms, and b) costs and losses in switching them to electricity, c) the amount that cannot conveniently be switched, and d) the energy and dollar costs of converting electricity or biomass into these more difficult forms.

To analyse the situation well we would need a complete list of the different kinds of energy in the present total energy budget, such as how much liquid fuel is needed for what purposes and we would need to ask about the extent to which electricity might be able to replace each of these. How might we run trucks, ships, aircraft, remote mines etc.? What might the conversion losses and efficiencies be, and what might be the ultimate total renewable system cost? Unfortunately there is little information on these issues, and it will be seen that differing assumptions make a considerable difference to conclusions. Therefore the following exploration must be regarded as uncertain and indicative only.  However it does illustrate the kind of analysis that is needed, and it provides a strong case that achieving a 100% renewable system will at best be very difficult and costly.

            Estimating a 2050 Australian total energy budget.

The approach will be to begin with an attempt to work out how much energy in what forms might be needed in Australia by 2050, given the goal of converting as much as possible to renewable forms. Following is a summary of the assumptions and approach taken.

Present (2015) final energy consumption 4,130 PJ. 

Electricity 810 PJ, 20% of final.

Transport 1603 PJ, 39% of final.

Population will multiply x 1.82. (ABS, 2012.)

It is quite difficult to estimate what “business as usual” (BAU) demand is likely to be. Australian power demanded from the grid has declined in the 2010-2014 period, but this has been partly due to the post GFC recession, the closure of big users such as smelters, and to the rapid uptake of rooftop PV. More recently demand has recently begun to increase again. The approach here has been to consider what 2050 demand would be given BAU, then add the effect of for instance shifting to electric vehicles.  It will be assumed that BAU demand will increase in proportional to population, and thus be 7,520 PJ. There is reason to think this could be an underestimate because since 1974 energy consumption has shown a smooth increase that has been faster than population growth. Yet this assumption is quite uncertain and the implications of taking a much lower 2050 BAU target will be discussed later.

Also highly uncertain is the likely effect of energy conservation effort. There is considerable scope for this and quite optimistic possible reductions are often claimed. However there seem to be few if any numerically based technical estimates of whole system savings, (as distinct from discussions of specific areas in which spectacular achievements are likely to be made; Amory Lovins’ works provide many of these.) 

Thus 2050 final energy demand will be taken as,

Electricity,   1,472 PJ

Transport,   2,917 PJ

Remainder  3,131 PJ

Total            7,520 PJ

1.    Electricity provision:

Assume 94% of electricity provided by wind, solar and hydro, plus 6% from biomass used for back up (from Elliston, Diesendorf and MacGill, 2012), requiring 340 PJ/y biomass assuming AEMO 26% generation efficiency. (This is a very optimistic assumption for wind and solar, and the biomass quantity is about one-third that of the “real-world probable” path discussed by Lenzen et al., noted above, and thus leaves considerably more biomass for liquid fuels than the Lenzen et al. analysis would leave.)

AEMO (Crawford et al., 2013) estimate Australia could harvest c. 96 million tonnes p.a. for biomass energy, i.e., 1,728 PJ/y, including municipal wastes. These figures are likely to be considerably too high as AEMO notes that they do not include energy costs for biomass production and transport, nor any embodied energy costs in plant, trucks etc. Farine et al. 2012, arrived at a figure around half as big. However Foran assumes much greater amounts. I will assume about 1,600 PJ can be provided. Thus after backing up power 1,260 PJ of biomass would be left.

2.    Transport energy provision:

Assume all passenger vehicles can be EVs, doubling energy efficiency (not trebling, in view of the high embodied energy cost of EVs; see TSW: Note on EV efficiency), one third of light trucks each run on electricity, ethanol and hydrogen, half heavy trucks run on ethanol and half on hydrogen (transfer of freight to rail not accounted, but that would greatly increase light truck use for distribution from rail heads), air transport fuelled by ethanol, shipping energy use not included.

      Amount required in 2050, 2917 PJ.

      Amounts and proportions, (from A.B.S. 9208.)

                  Road. 73% of transport energy, i.e.,               2,129 PJ

                  Passenger           58% of road energy  1,235 PJ

Light trucks    17%  “                          362 PJ

Heavy trucks  22%                             468 PJ

Other                3%                               64 PJ

                  Rail. 3% of transport energy, i.e.,                         87 PJ

                  Air. 19% of transport energy, i.e.,                       579 PJ

                  NEI. 4% of transport energy.                              120 PJ

No figures for shipping are included. The amount involved in the Australian economy would be significant but almost all vessels would be foreign owned so fuel quantities would not be recorded in Australian accounts.  Fuel would have to be liquid or hydrogen etc., not electrical, so production from renewable sources would be problematic, given scarcity of biomass; see below.

                  Amounts and forms of transport energy required:

Passenger vehicles.      1,235 PJ needed if BAU.  Assume all are electric vehicles, but at only double present energy efficiency, not treble (after taking into account their high embodied energy cost), therefore need 617 PJ of electricity.  (NB. Thus EV cars save only 617/7520 = 6.2% of final energy.)

Light trucks.   362 PJ needed if BAU. Some trucks for very light deliveries can be EV’s, some for heavy distribution would have to be hydrogen or biomass-ethanol. So assume one third each electric, hydrogen and biomass ethanol, so 120 each hydrogen and biomass and 60 PJ electricity for EV light trucks. (This unrealistically assumes light trucks use no more energy than passenger vehicles.)

Heavy trucks.   468 PJ needed if BAU.  These can’t be EVs (Friedmann, 2016), so assume half hydrogen and half biomass ethanol, so 234 PJ hydrogen and 234 PJ biomass ethanol would be needed. (The hydrogen component is probably quite implausible, given the weight of the hydrogen storage tanks that a truck would need; see Bossel, 2004, and Friedmann, 2016.)

Rail. Assume all electrical.   87 PJ

Air.  579 PJ of liquid fuel (assume ethanol.)

Other transport. 87 PJ needed. Assume 43 PJ of ethanol and 43 PJ of hydrogen.

(A shift of much freight from heavy trucks to rail would reduce the heavy truck total, but it would increase medium trucks for distribution from the rail head. Neither is taken into account here.)

      So total needed for transport :   

Electricity  for EVs      765  PJ

Hydrogen                     398  PJ

 Biomass  ethanol       977 PJ

     Total.     2,140 PJ

3.    Remaining energy provision:

This is a large amount, 3,131 PJ and 43% of total demand.  It is quite difficult to find information on its composition, or to work out how much of it could be provided via non-electrical paths, and therefore how much additional electricity this category would require to be generated.

The issue is complicated by the fact that some of the electricity consumption accounted above would presently be going into low temperature heating. Let us attempt to take low out temperature heat energy by assuming (unrealistically) that it can all be provided without electricity, e.g. from simple solar thermal panels. The two most relevant figures available are, firstly residential heating and cooling makes up about 5.6% of total energy use (although the quite small cooling fraction of this would involve electricity.) Industrial plus commercial energy use add to about 32% of the total. If one quarter of this second quantity is low temperature heat that need not be provided via electricity, then residential + industrial + commercial low temperature heating might add to 12% of total energy. This would mean that 12% of the BAU target 7,530 EJ, i.e., 904 PJ, can come from solar thermal panels, reducing the remaining category to 2,227 PJ. (In the real world much heating and cooling will be carried out by heat pumps, using electricity, but the above assumption is that all is provided by solar thermal panels, unrealistically reducing the total electricity demand to be met. Also note that if solar radiation in winter was 6 kWh/m2/d and could be collected at 100% efficiency, to collect 904 PJ/y or 2.5 PJ/d would require 416 million m2, adding a significant embodied and dollar energy cost that is ignored here.)

Let us make the simplifying assumption that all of this remainder can be provided by electricity. (This assumption is likely to be far from correct.)

                                    Summary of totals:

Energy forms to be generated.

                                                 Electricity               Biomass                Hydrogen.

            Electricity demand.     1,384 PJ             340 PJ biomass

Transport  demand.       765 PJ             977 PJ as ethanol,         977 PJ                                                                         = 2,443 PJ of biomass

Remaining 45%.         2,227 PJ

                                                            -------------------------------------------------

Totals.   4,376 PJ            2,783 PJ of biomass        977 PJ

Complications.

Crawford et al. (2013) estimate that 1,728 PJ of biomass will be available for energy purposes in Australia. If 340 PJ is used to back up electricity production this would leave 1,388 PJ, but from above 2,783 PJ is needed. The shortfall of 1,395 PJ would have produced 558 PJ of ethanol, if conversion efficiency is 40%.  This will now have to met by hydrogen, bringing that total to 1,535 PJ.

But to have 1 unit in the form of hydrogen about 1.7 units must be generated in the form of electricity (… even ignoring the large energy cost embodied in hydrogen production, storage, and pumping equipment and losses. (Efficiency of conversion estimates vary; Honnery and Moriarty, 2009, state 55% which would mean the multiple is 1.82.) Thus generating the hydrogen would require 2,610 PJ of electricity.   The electricity total would then become 4,376 + 2,610 = 6,986 PJ. The present hydro contribution should be subtracted, but it is relatively small so the electricity total will be taken as 6,940 PJ.

There are other factors that would tend to increase these numbers significantly. No provision has been made in this budget for any embodied energy costs and for hydrogen these would be substantial. No provision has been made for shipping energy, said to be 7% of world transport energy. Australia is heavily involved in trade, via almost entirely foreign owned ships. No provision has been made to deal with peak demand.  The figures are only for quantities needed to meet average demand, i.e., for the annual total quantity to be supplied. The difference between average and peak is large; the Elliston, Diesendorf and MacGill study took peak demand as 41% above average demand, and Lenzen et al. took it as 61% higher.

More optimistic assumptions?

Š      From here on BAU energy demand might not increase as fast as population. However a much lower plausible assumption for 2050 BAU demand would not make a major difference to the cost conclusions arrived at below.

Š      The electricity generating task would be significantly reduced if the biomass contribution could be greatly increased beyond the AEMO estimate, which some think is quite possible. (Foran, 2008.) But the difference would not be huge; if 40 million ha could be planted yielding 720 PJ/y this would provide about 290 PJ of ethanol, which is only about 30% of the amount needed by transport in the above table.

Š      There will probably be significant efficiency improvements other than those assumed here, which were only for EVs. As noted, optimistic claims about possible demand reduction are common but these tend to be for “technical potential” and there are few detailed and impressive attempts to estimate what might be achieved in the real world where many limiting factors operate.  The most common estimate seems to be in the region of 30% on BAU.

Š      Some believe the conversion of biomass to methanol is the preferred option, and efficiency might be raised to 50%.

On the other hand it is likely that in 2050 many functions will require greatly increased energy inputs, such as mining poorer ores, processing poorer ores, water desalination, denser settlements involving much high rise construction and living, recycling higher proportions of waste. The quest to improve productivity will probably increase energy demand, as it has recently been realized that productivity growth is significantly due to adoption of more energy intensive ways. Global freight and especially tourism and air traffic are expected to increase faster than population. Energy will be needed to cope with climatic challenges to agriculture and increasing effort will be needed to deal with generally accelerating ecological deterioration. Above all dealing with the many effects of climate change will add very large energy costs that do not exist at resent, including defensive works such as sea walls, settlement relocation, salt water incursions into agricultural land, remedying storm damage, dealing with refugees, adapting to altered rainfall patterns e.g. making dams redundant, dealing with pest surges and algal blooms, developing new crops for altered conditions. Warming weather will significantly increase air conditioning use, adding greatly to the task of providing for peak supply. Probably the greatest problem will be set by the fact that the IPCC greenhouse targets assume taking very large amounts of carbon out of the atmosphere after 2050 and this would involve enormous energy costs.

            Conclusions.


The above discussion of the Lenzen et al. study indicated that the retail price of electricity would be well above 46c/kWh, and could be over 70c or around three times the preset cost. To provide 6,940 PJ/y, i.e., 1,927 billion kWh, at 46 c/kWh could cost $887 billion p.a., which is around 55% of 2015 GDP, or 28% of 2050 GDP assuming 2% economic growth. The present rich country total expenditure on energy is usually well under 10%. (After recent significant rises the Australian figure has been estimated at 8.2%. Pears, 2017.)  But this this includes taxes added on after all production and distribution costs. (For example 40% of the price paid for petrol in Australia today is a tax added by government to the retail supply price.) If we could take out taxes we would probably find that the retail cost paid for energy in Australia today was closer to 6% of GDP. Note that if we had confident numbers for all the10 factors noted above as likely to increase retail price it could well be more or less double the 46 c /kWh minimum used above in this paragraph.

It would obviously be impossible to pay out anything like 28% of GDP for energy. Hall and Klitgaard (2014) find that when energy expenditure remains above about 5.5% of US GDP for some time recession occurs.

Even if the 2050 demand was no greater than the present demand the total cost would not be affordable.  Nor would it be affordable if conservation effort cut 30% off BAU demand. In other words far lower assumptions than have been made in this exercise would have to be true before the costs arrived at could be affordable. And all this is for Australia, which has possibly the most favourable renewable energy conditions in the inhabited world

The belief that 100% renewable energy supply is probably the main element in the tech-fix faith held by most people, including green and left people. They think there is no need to shift from something like present energy and resource intensive lifestyles and systems, or from an economy driven by growth. If the position arrived at in this reassessment is sound then the big global problems cannot be solved unless there is dramatic reduction in rich world per capita levels of consumption, the present economy is abandoned, there is immense cultural change away from individualistic, competitive acquisitiveness, and transition to some kind of radically Simpler Way. (That this would be workable and attractive is argued at thesimplerway.info/).

Again it should be said that these derivations and figures are anything but clear and confident, much depends on assumptions and those used here are open to debate, but they do indicate how very different the assumptions would have to be to arrive at tolerable costs for a 100% renewable energy supply system.

                                                ----------

Australian Bureau of Statistics, (2012), 3222.0 - Population Projections, Australia, 2012 (base) to 2101 - Series A http://www.abs.gov.au/ausstats/abs@.nsf/Lookup/3222.0main+features52012%20(base)%20to%20210

Australian Bureau of Statistics. (2014) 9208.0 - Survey of Motor Vehicle Use, Australia, 12 months ended 31 October,  http://www.abs.gov.au/ausstats/abs@.nsf/mf/9208.0

Abdulla, K. A. Wirth, S. K. Halgamuge, and K. Cl. Steer, (2014), “Selecting an optimal combination of storage and transmission assets with a non-despatchable electricity supply, in Information and Automation for Sustainability, 7th International Conference, pp.  1-6.

Abengoa, (2016), PS 10 Solar Power Plant, Wikipedia. https://en.wikipedia.org/wiki/PS10_solar_power_plant

AETA, (2012),   Australian Energy Technology Assessment, ABARE, Canberra.

Aghahosseini, A., D. Bogdanov, and C. Breyer, (2016), “100% Renewable Energy in North America and the role of solar photovoltaics, in EU-PVSEC conference, June 20-24, 7DV.4.8, Munich, Germany.

Ahmed,N., (2017), Fake News, Failing States, Dortrecht, Springer.

Blakers, A., (2016), “Pumped hydro, solar and wind can deliver a 100% renewable NEM”. Ecogeneration, 25th Nov.

Bolinger, M., and J. Seel, (2014), Utility Scale Solar: An Empirical Analysis of Project Cost, Performance and Pricing Trends in the United States. US Dept. of Energy, Berkeley Laboratory. Sept. https://emp.lbl.gov/sites/all/files/lbnl-1000917.pdf

Budischak, C., Sewell D., Thompson H., Mach, L., Vernon, D. E., Kempton W., (2012), “Cost minimised combinations of wind power, solar power and electrochemical storage, powering the grid up to 99.9% of the time”, Journal of Power Sources, 225, 60 – 74.

California Energy Commission, (2014), Estimated Cost of New Renewable and Fossil Generation in California, May.

http://www.energy.ca.gov/2014publications/CEC-200-2014-003/CEC-200-2014-003-SD.pdf

Chhatbar, K., and R. Meyer, (2011), The influence of meteorological parameters on the energy yield of solar thermal power plants, MS in Renewable Energy, University of Oldenburg, Germany http://epic.awi.de/24776/1/chp2011.pdf

Concentrating Solar Power, (undated), https://en.wikipedia.org/wiki/Concentrated_solar_power

Conger, (undated), How Stuff Works Auto. http://auto.howstuffworks.com/fuel-efficiency/vehicles/lithium-ion-battery-car2.htm

Crawford, D., T. Janovic, M. O”Connor, A. Herr, J. Raison and T. Baynes, (2013), Potential for electricity generation in Australia from Biomass in 2010, 2030, and 2050,  AEMO 100% Renewable Energy Study, 4 Sept. CSIRO Report EP – 126969.

Danko, P., (2015), Ivanpah Solar Plant Picking Up Steam,  Breaking Energy January 28, http://breakingenergy.com/2015/01/28/ivanpah-solar-plant-picking-up-steam/

De Castro, C., (2017), Personal communications.

Diederen, A. M., (2009), Metal minerals scarcity: A call for managed austerity and the elements of hope, TNO Defence, Security and Safety, P.O. Box 45, 2280 AA Rijswijk, The Netherlands.

Dietrich, C., (2016), Ivanpah CO2 Emissions: More Than 10 Times Solar PV, Pete Danko, Clean Energy News and Commentary,March 29. Wikimedia Commons, http://petedanko.net/ivanpah-co2-emissions-more-than-10-times-solar-pv/

Electric Power Research Institute, (2010), Program on Technology Innovation: Integrated Generation Technology Options, 25-Nov.,112 pp.

http://www.epri.com/abstracts/Pages/ProductAbstract.aspx?ProductId=000000000001019539

Elliston B., Diesendorf, M., and I. MacGill, (2012). “Simulations of scenarios with 100% renewable electricity in the Australian National Electricity Market”,  Energy Policy, 45, 606 – 613.

Elliston, B., I. MacGill, and M. Diesendorf, “Least cost 100% renewable electricity scenarios in the Australian National Electricity Market, Energy Policy, 59, 270 –

Farine, D. et al., (2012), “An assessment of biomass for bioelectricty and biofuel and for greenhouse gas emission reduction in Australia”, Bioenergy, 4,148) -175.  doi:  10.111/j.1757-1707.2011.o1115/x

Feldman, D. K. et. al., (2014), Photovoltaic System Pricing Trends; Historic, Recent and Near Term Projections,  SunShot, US Dept. of Energy. http://www.nrel.gov/docs/fy14osti/62558.pdf

Foran, B., (2008), Australian forestry for a better future Challenges, opportunities and strategies.  Institute for Land Water and Society, Charles Sturt Univ. Albury.

Geyer, M ., (2008), Potential of Solar Trough Installations in Europe, http://ec.europa.eu/energy/res/events/doc/geyer_andersol/pdf.

Gruber, P and P. Medina, (2010), Global Lithium Availability: A Constraint for Electric Vehicles? MSc Thesis, Univ. of Michigan.

Hall, C. A. S and K. A. Klitgaard, (2014), Energy and the Wealth of Nations, Dordrecht, Springer.

Hearps, P., R. Dargaville, D. McConnell, M. Sandiford, T. Forcey, P. Seligman, (2014), Opportunities for Pumped Hydro Energy  Storage in Australia, Melbourne Energy Institute, 27the Feb.

.

Hinkley, J., et al., (2012), Concentrating Solar Power Drivers and Opportunities For Cost Competitive Electricity, CSIRO, Canberra.

Honnery, D., and P. Moriarty, (2009), “Estimating global hydrogen production from wind”, Int. J. Hydrogen Energy, 34: 727-736.

Istvan, R, (2015), “Intermittent grid storage”, Climate Etc.  July 1.

Jones,S. A., N. J. Blair, R. Pitz-Paal, P. Schewarzboezl and R. Cable, (2001),  “TRNSYS modeling of the SEGS VI parabolic trough solar electric generating system”, Proceedings of Solar Forum 2001, Solar Energy: The Power to Choose, April 21-25, Washington DC.

Lenzen , M., B.  McBain, T. Trainer, S. Jutte, O. Rey-Lescure, J. Huang, (2016), “Simulating low-carbon electricity supply for Australia”, Applied Energy

179, pp. 553–564.

Lovegrove, K., A. Zawadsky and J. Coventy,  (2007), “Parabolic dish solar concentrators for multi-megawatt power generation”, Solar World Congress, Beijing, Sept .18- 223.

Lovegrove, K., M. Watt, R. Passy, G. Pollock, J. Wyder, and J. Dowse, 2012.  Realising the Potential of Solar Power in Australia. IT Power, for the Australian Solar Institute.

Marin, O.C., (2015), (Personal Communication), Corporate Communication, Calle Servoi, Ochoa 4, FORAN, www.sener.es

Miskelly, P., (2012), “Wind farms in Eastern Australia – Recent lessons”, Energy & Environment, Vol. 23, No. 8, 1233 – 1260.

Morgan, J., (2015) “The Catch-22 of energy storage”, On Line Opinion, 10th March.

NREL, (undated) System Advisor Model (SAM) ,Case Study: Gemasolar Fuentes De Andalucía https://sam.nrel.gov/sites/sam.nrel.gov/files/content/case_studies/sam_case_csp_salt_tower_gemasolar_2013-1-15.pdf

Odeh, S.D., Behnia M and Morrison, G.L., (2003), Performance Evaluation of Solar Thermal Electric Generation Systems, Energy Conversion and Management, V44, 2425-2443.

Palmer, G, (2013), “Household Solar Photovoltaics: Supplier of Marginal Abatement, or Primary Source of Low-Emission Power?”  Sustainability, 5, 1406-1442; doi:10.3390/su5041406.

Palmer, G., (2016), “A framework for incorporating EROI into electrical storage”, Australian-German College of Energy and Climate, University of Melbourne.

Pears, A., (2017), “2017 will be a big year for Australia’s energy system: here’s what to look out for”, The Conversation, January 25.

Pellow, M. A., C. J. M. Emmott, C. J. Barnhart and S. M. Benson, (2015), “Hydrogen or batteries for grid storage? A net energy analysis”  Energy Environ. Sci., 8, 1938-1952 DOI:10.1039/C4EE04041D

Pickard, W. F., (2014), Smart grids vs the Achilles heel of renewable energy; Can the needed storage infrastructure be constructed before the fossil fuel runs out, Proceedings of the IEEE, 102, 7, July.

Pietro, P. and C. Hall, (2013), Spain's Photovoltaic Revolution, Dordrecht, Springer.

Raman, P. and N. K. Ram, (2013), “Design improvements and performance testing of a biomass gasifier based electric power generation system”, Biomass and Bioenergy, Volume 56, September, Pages 555–571.

Renewable Energy Focus, (2010), “2025 solar and wind energy costs charted”, Renewable Energy Focus, 22 February.

Riesz, J., and B. Elliston, (2016), “Prioritising research investment and policy mechanisms for

deployment of renewable technologies to achieve low cost, high renewable

electricity systems”, Energy Policy, Volume 98, November, 298–308.

ROAM, (2012), ROAM Report on Pumped Storage Modeling for AEMO 100% Renewables Project, Dept. of Climate Change and Energy Efficiency, 24 Sept.

Shahan, Z., 2015), How Tesla battery storage compares with rivals on prices, 11 May. file:///Users/sandranorris/Desktop/MANUSCRIPT%201/Storage%20battery%20Tesla

Sharma, R., C. Manzie, M. Bessede, R.H. Crawford, M.J. Brear  (2013), Conventional, hybrid and electric vehicles for Australian driving conditions. Part 2: Life cycle CO2-e emissionsTransportation Research Part C: Emerging Technologies, Volume 28, March 2013, Pages 63–73.

Sveredrup, H., and V. Ragnarsdottir,  (2014), Natural Resources in a Planetary Perspective, Geochemical Perspectives, 3, 2., Oct.

Symons, J., (2016), ”Are the Greens really the climate radicals we need?”, The Conversation, June 22,

Tahil, W., (2010), “How much Lion does an Electric Vehicle really need”, Meridian Research. http://www.meridian-int-res.com/Projects/How_Much_Lithium_Per_Battery.pdf

Thodey, J., (2013), personal communications.

Timensen,M. J. A., A. P. C. Faaij, C. N. Hamelinck, M. R. M. van Hardeveld, (2002), ‘”Exploration of the possibilities for production of Fischer Tropsch liquids and power via biomass gasification”, Biomass and Bioenergy, Volume 23, Issue 2, August,129–152.

Trainer, T., (2013a), "Can Europe run on renewable energy? A negative case", Energy Policy, (63) 845 – 850.

Trainer, T., (2013b), "Limits to solar thermal energy set by intermittency and low DNI: Implications from meteorological data", Energy Policy, http://www.deepdyve.com/lp/elsevier/limits-to-solar-thermal-energy-set-by-intermittency-and-low-dni-3d0eZOnztX

Trainer, T., (2014) Extract on Biomass-gas-electricity, from Comments on the proposal for 100% renewable energy electricity supply in Australia by Elliston, Diesendorf and MacGill.

Trembath, A., (2015), “Is there an upper limit to intermittent renewables?”, The Breakthrough, May.

http://thebreakthrough.org/index.php/voices/energetics/a-look-at-wind-and-solar-part-2

TSW: Note on EV efficiency. thesimplerway.info/RE-EVefficiency.htm

Unnamed, (Undated), Solar Parabolic Trough. US DOE, Office of Energy Efficiency and Renewable Energy. http://Solarpaces.org/soilar_trough.pdf

Viebahn, P., S. Kronschage and F. Trieb, (2004), New Energy Developments for Sustainability, Project 502687.

Wister, W., (undated), “ Is there enough lithium in the world to replace all petroleum-cars with battery-electric vehicles?  https://www.quora.com/Is-there-enough-lithium-in-the-world-to-replace-all-petroleum-cars-with-battery-electric-vehicles

Weisback, D., G. Ruprecht, A. Huke, K. Cserski, S. Gottlleib and A. Hussein, (2013), “Energy intensities, EROIs and energy payback times of electricity generating power plants”, Energy, 52, 210- 221.

Wilson, N., (2011) “Gemasolar Plant; Elegant, but Pricey”, The Energy Collective, June 6th. http://www.theenergycollective.com/nathan-wilson/58791/20mw-gemasolar-plant-elegant-pricey

Wood, T., T. Edis, D. Mullerworth, H. Morrow, (2012), No Easy Choices: Which Way to Australia’s Energy Future? Technology Analysis. The Grattan Institute.

M. Worley, M., and J. Yale, (2012), Biomass Gasification Technology Assessment, NREL Consolidated Report, NREL.

Wright, M. and P. Hearps, (2010), Australian Sustainable Energy

Zero Carbon Australia Stationary Energy Plan, Energy Research Institute,