Below is a new, detailed critique by Dr Ted Trainer of the simulation studies by Elliston, Diesendorf and MacGill on how eastern Australia might be run off 100% renewable energy. The summary:
Three recent papers by Elliston, Diesdendorf and MacGill (2012, 2013a, 2013b) elaborate on a proposal whereby it is claimed that 100% of present Australian electricity demand could be provided by renewable energy. The following notes add considerations arising from the last two papers to those discussed in my initial assessment of the first paper. My general view is that it would be technically possible to meet total Australian electricity demand from renewables but this would be very costly and probably unaffordable, mainly due to the amount of redundant plant needed to cope with intermittency. This draft analysis attempts to show why the cost conclusions EDM arrive at are probably much too low.
Ted has also updated his critique of the Zero Carbon Australia’s report on 100% renewable energy by 2020. The original BNC post is here, and the updated PDF here.
Ted notes the following:
These efforts have taken a huge amount of time and I am still not clear and confident about my take, mainly because neither party will cooperate or correspond. Thus I have not been able to deal with any misunderstandings etc. I have made. Both critiques are strengthened by information I have come across since circulating previous commentaries, but they are essentially elaborations on the general line of argument taken in earlier attempts.
I find this unwillingness to engage on these criticisms by the primary authors disappointing, but typical.
Introduction
I think these three papers are valuable contributions to the considerable advance that has occurred in the discussion of the potential of renewables in the last few years. My understanding of the situation is much improved on what it was three or four years ago and I now think some of my earlier conclusions were unsatisfactory. EDM take the appropriate general approach, which is to look at how renewable technologies might be combined at each point in time to meet demand, or more accurately, to estimate how much capacity of each technology would be required, especially to get through the times when solar and wind input is minimal. EDM put forward a potentially effective way of coping with the problem of gaps in their availability via biomass derived gas for use in gas turbines. My earlier analyses did not consider this.
It is not difficult for an approach of this kind to show that electricity demand can be met, and many impressive 100% renewable energy proposals have been published. (For critical analyses of about a dozen of these see Trainer, 2014), but a great deal of redundant capacity would be needed, and the key questions are, how much, and what would it cost? My present uncertain impression is that Australia might be able to afford to do it, but if it could it would be with significant difficulty, i.e., with major impacts on lifestyles, national systems and priorities, and on society in general.
A major disappointment with the EDM analyses is that for some crucial elements no data, evidence or derivations are given and as a result the proposal can only be taken as a statement of claims. We need to be able to work through the derivations in proposals such as this to see if they are sound or what questionable assumptions might have been made etc. Consequently I have had to spend a lot of time trying to guestimate my way to an assessment of the cost conclusions and it is not possible to confident about the results.
Required capacity?
A merit of the EDM approach is to take as the target the present demand. This avoids the uncertainty introduced when attempting to estimate both future demand and the reduction in demand that conservation effort etc. might make.
Two of the plots given in EDM 1 (Elliston, Diesendorf and MacGill, 2012), set out the contributions that might be combined to meet daily demand over about 8 days in 2010, in summer and winter. In my first discussion of the proposal it seemed to me that when these contributions were added the total capacity needed would be much more than the paper stated. The total amount of plant required to supply an average 31 GW was stated as 75.5 GW of peak capacity. (In his response to Peter Lang, Mark Diesendorf said their total requirement is 84.9 GW.) However in EDM 2 (Elliston, Diesendorf and MacGill, 2013a) the total has been significantly raised, to 104 GW. (The proposal from Hart and Jacobson 2011, for 100% renewable supply in California, also involves a large multiple of average demand, greater than four, and in the proposal by Budischak , et al. 2013 for California it is around 8.) Unfortunately this is one of the crucial numbers and claims in the proposal which it is not possible to assess because the derivation and the necessary background data on weather patterns are not provided.
The following discussion is in terms of present capital costs, as stated by AETA. The figure that seems to me to be of most use is the capital cost of sufficient capacity to deliver I kW, at distance, in winter, net of embodied and other costs and losses. This is much higher than the usual figure for capacity to generate 1 kW in peak conditions.
Wind
EDM assume that wind will meet 58% of demand in one of their scenarios, but there is no discussion about whether this is possible (although there is a sentence recognising that 50% would set significant problems.). Lenzen’s review (2009, p. 19) concluded that the general limit before problems arise is likely to be around 20%, and possibly under 20%. More recently Kuchinski (2013) states 20% as the extreme upper limit and 10% as the norm. It is difficult to generalise as there are studies where higher penetrations have been achieved, but these are typically for unusually favourable regions such as Denmark which is able to export large surpluses easily. (This export capacity also enables Germany to have such a large PV component.) The complicating factor is that higher penetrations bring problems such as grid instability, back-up needed, and power dumping, which “technically” can be solved, but at a significant cost in equipment such as the interconnectors that have to be built to make exporting possible. (South Australia is contemplating the construction of two, at a possible cost of $4 – 10 billion…which would pay for up to 3 coal-fired power stations and these would deliver several times as much power as the whole wind sector in SA. Miskelly, p. 1249.)
This assumption of large dependence on this relatively low cost technology among the renewable options is a major determinant of the questionably low final capital cost sum arrived at by EDM. If the quantity limit is only half that assumed then much more of the more expensive PV and ST plant would have to be used.
It is not possible to assess how adequately EDM deal with the intermittency of wind. The recent CSIRO study of intermittency (Sayeef, 2012) makes clear the magnitude of the integration problem, and the many difficulties it sets, even for penetration levels around 25%. Miskelly’s analysis of Australian wind power around 2010 and 2011 makes the magnitude of the integration problem clear, even for the present Australian situation where penetration is low. In one year output from the whole 1.9 GW peak capacity network was negligible 156 times, for a total of 6.5 days. In May 2010 there was virtually no wind anywhere in Australia for three days.
It is generally assumed that adding widely distributed farms reduces intermittency of aggregate supply but when Miskelly compared the situation before and after the installation of much capacity he found that this did not “smooth” aggregate input. These findings mean that the entire wind capacity has to be backed up by fossil fuelled generation capacity, and almost none of the previously existing fossil fuelled capacity can be retired. As Palmer points out the German PV sector has added some 30% to power generating capacity without adding to power consumed, so this is added capacity that can be substituted for fossil fuelled capacity from time to time, but it is not capacity that adds to power provision.
Even more problematic are the implications of the associated ramp rate problem. Miskelly documents the frequent precipitous rises and falls in total output. He says this means that for a large scale wind sector the back up can’t be the existing coal-fired plant but would have to be newly built gas-fired plant. (The EDM proposal aligns with this conclusion as backup would be provided by the very large biomass-gas-electricity sector they assume.)
As noted above, the estimate of system capital cost offered below will focus on the present capital cost of sufficient plant to send out to the grid (or deliver to users) an average kW in winter conditions. It is not clear what an appropriate winter wind capacity factor would be. AETA (p. 46) assumes 38% as the annual average but inspection of the NEM wind data shows that in July 2010 the average capacity factor was in fact well under 28%. The world annual average is around 25%. The examination of data from the whole Australian wind sector carried out by Miskelly shows that in one winter month in 2010 capacity averaged 24%. Surprisingly for South Australia where wind resources are unusually good, Miskelly and AEMO (2012) both document winter capacity lower than summer, and in some months it is 60% lower. However for working purposes the winter figure assumed below will be a very generous 38%. This would make the capital cost of the capacity to send out 1kW in winter $6,660, based on the AETA present capital cost figure of $2,250/kW(p).. (2012, p. 46.)
Next we need to take into account the assumption that the lifetime for wind turbines is only 20 years compared with 30 assumed for the other renewable technologies involved here, (and 50 for coal, AETA, 2012 p. 27.). Given that the estimation below considers a 30 year period, the assumed lifetime for PV and ST, this makes the 30 year capital cost figure $9,987 for the amount of plant needed to send out a 1 kW flow in winter. (Wind capital cost is not expected to fall much if at all in future; see AETA, 2012, p. 76.)
A thorough analysis would have to increase this figure to take into account the embodied energy costs of wind plant and the “downstream” energy costs and losses. Within the latter would be Jefferson’s recent finding that wind systems operate for fewer hours than has previously been assumed. (Jefferson, 2012.) (Hughes found that UK turbines are lasting only about 12 years, but this could be due to retirement of older units, so will be disregarded here.) Weisbach et al. (2013) and others state an ER of around 16 for wind but do not seem to have taken into account all upstream embodied energy factors or downstream losses (as Pietro and Hall have attempted for PV, below.) However Lenzen and Munksgard (2001) did take at least some of them into account and arrived at an ER of 6. These findings leave the issue unsettled but suggest an all-in embodied energy cost of 6% – 17%. No account of wind embodied energy cost will be included in the estimations below.
The problematic PV component
PV is assumed to contribute c. 17 – 20% of electricity. This seems to be an appropriate assumption. Denholm and Margolis (2007) see 10 – 20% as the limit and believe it is not likely to be more than 15%. This seems to be valid because if in each day PV was to contribute 20% of the average 23 GW then it would have to generate 0.2 x 23 GW x 24 h = 110 GWh/day. But it would have to do this during the c. 7 hours in which there is significant solar energy, so its output then would have to average 16 GW, meaning that the wind, solar thermal and hydro sectors would have to be cut back to only a combined 30% of total output in that period. (Also grids would have to be strengthened to enable 70% of demand to come from PV sites for some hours, then to draw around 100% from other sites. This suggests that PV would not be allowed to contribute more than about 15%.
It is not generally realised that PV can’t make much difference (unless the storage problem is solved.). If it provides 15% of electricity that’s less than 3% of present rich world energy consumption. A related problem is that supply does not align well with demand, indicating that its contribution would be limited without extensive storage capacity. In summer air conditioning use often makes electricity demand peak late in the afternoon. Palmer reports a study in NSW finding PV input to be zero at this time. AEMO reports Victorian PV contributing at a rate of 1% of capacity when power demand peaks in winter. (Below.) Palmer says, “… PV is not suited to taking on a primary network role or delivering sufficient surpluses of energy when a fuller account of embodied energy is included.” ( 2013, p. 1407.)
EDM assumes PV modules on rooftops in Brisbane, Sydney, Canberra and Adelaide, eliminating distribution losses. Their total capacity corresponds to around 2 kW(p) per household, or 15 modules, which would not seem to be plausible. However winter global radiation in these four latitudes averages around only 2 kWh/m2/d in some months. Presumably wind and inland solar thermal sources are to make up for rooftop PV in winter, but we are not able to see how this is to be done or whether there is sufficient capacity assumed to do it, especially through difficult periods.
Embodied energy costs
There is no discussion of embodied energy costs or “downstream” losses or the resulting EROI for PV (or any other of the renewables), yet these are crucial considerations. For PV EROI has been commonly assumed to be around 10 (Fthenakis, undated, says up to 60), but only recently has attention been given to attempting to estimate an all-inclusive value. It is now clear that two major and previously neglected categories of energy cost and loss have to be taken into account. The first is to do with the “upstream” factors, such as the relevant fraction of the energy it took to construct the aluminium plant that made the material for the module frames. A few studies have attempted to do this seem to have arrived at a figure of around 33% for the total energy cost of production for PV. (Crawford, 2011, Crawford, Treloar, and Fuller, 2006, Lenzen, 1999, Palmer, 2013, Prieto and Hall, 2013, and Lenzen, Dey, Hardy and Bilek, 2006.) A recent and detailed study, by Weisbach, et al., (2013) arrives at an ER of 3 to 4.8, and they say that the studies by Battisti et al., (2005), Ito et al., (2008), Meije et al, (2003),Raugeri et al., (2010), and Alsema (2000) are in good agreement. (I’m not sure about this.)
Inquiries to AEMO in 2014 found that it does not have evidence on embodied energy costs of PV and although it discusses the “payback period” this is calculated only in dollar terms, i.e., time to repay the monetary investment. In other words it is surprising and disappointing that AEMO’s statements and statistics do not take embodied energy costs into account.
The “downstream” factors include costs and losses after the plant has been installed and which accumulate over its lifetime, such as losses in wires, connections and inverters, invertor failure and replacement, accidents that damage equipment (e.g., rats or possums that cause shorts), badly aligned panels (which seems to me must be a significant factor, from observation of the many less than ideal rooftop mountings), shading of panels (a tiny amount of shading can cut output dramatically), dust and bird droppings on panels, corrosion at terminals and connections, heating of panels (which can cut midday input by more than 25%), poor maintenance by householders, and deterioration of panel efficiency due to age. Prieto and Hall (2013) list 15 such factors and estimate that these cut output to 65% of rated capacity.
Previous discussion has recognised some of these loss factors and have used the term “performance factor” to refer to the reduction they cause. For PV this has been thought of as around 0.75 – 0.8. What Palmer, Crawford, and Prieto and Hall are saying is that the previously assumed models for ER calculation have not included all relevant factors and have used “boundary” assumptions that have been too narrow, and have therefore produced a performance factor that is too generous,
Power dumping must be accounted as a different kind of energy loss affecting EROI. (My stand alone homestead PV system dumps most of the power it produces in summer, yet is not big enough to meet winter demand reliably, despite 800 ah of battery capacity.) Palmer’s Figure 3 shows that about 10% of German PV output in a summer period was exported. This was possible because neighbouring countries did not have as much PV as Germany and were not experiencing output gluts when Germany wanted to export. If they all had as much PV the Germans would have to dump in summer. Australia could not export.
There is another category of “downstream” costs/losses attributable to PV and most other renewables but not previously subtracted from embodied energy and “downstream” accounts. The recent OECD report (2013) discusses the significant costs of “grid adjustment” to accommodate renewables, including the cost of strengthening grids to be able to cope with variation in amounts of power coming from differing regions. When centralised coal fired sources provide all power relatively simple networks can be built, e.g., to deal with a fairly regular and stable voltage drop from power station to distant consumers. But with renewable sources much power might be coming from one region today and from another tomorrow. This seems to be a factor additional to those Prieto and Hall deal with. It would become more problematic if the contribution of wind to a system is assumed to be high, and again EDM assume up to 58%.
There is not yet agreement about definitions and boundaries in the discussion of EROI (e.g., do you include the energy cost of security services at the PV farm, or of workers travelling to the module factory, or those associated with financing? These are necessary costs; there could be no PV output if they were not paid.) Palmer (2013) and others use the term “extended EROI” when all these upstream and downstream factors are taken into account. For PV systems Palmer arrives at a remarkable 2.1 – 3+ for the extended EROI of rooftop PV systems in Melbourne. This means payback time might be taking 15 of the 30 year lifetime of the system (p. 1434), and in its lifetime a module could only produce enough energy to pay for its own production plus that of one more.
Even more surprisingly a similarly very low figure has also been arrived at by Prieto and Hall (2012) for actually functioning Spanish utility scale systems, where one would assume many of the factors associated with roof top systems would not be problematic, such as quality of maintenance.
This recently opened cluster of issues will take time before confident conclusions emerge. It must be remembered that ER values depend significantly on a particular set of conditions especially the location of the modules and its typical level of solar radiation, and the size of system assumed (e.g., small penetration might involve no backup provision or dumping loss.)
It is very important to recognise that we seem to have no clear idea how rooftop PV is performing, or what contribution to the grid is being made by each kW of installed capacity. The first point that is usually overlooked is that if you install a 3.5 kW capacity system you will almost certainly never get more than about 2.5 kW out of it; even before taking into account all the above “downstream” loss factors. The peak rating refers to ideal conditions and a radiation level few if any house roofs ever get. (See for instance Damn The Matrix, 2013.) My efforts (e.g., via communications with AEMO) to get information on the actual amount being generated by PV in Australia or anywhere else, and the proportion of this that is fed into the grid, have found that no data is available. The relevant data probably cannot be collected because neither the amount of power a household produces nor the amount it uses before sending the surplus to the grid are generally recorded. These would be very difficult factors to get good data on. AEMO admits that it does not know how much power rooftop PV in Australia generates, and that its estimate that 50% of what is generated is fed in is a guess. (Personal communications, Jan. 2014.)
It seems therefore that we actually have little idea what a given amount of rooftop PV capacity will contribute to the grid. AEMO’s lengthy and detailed discussions (2013a, 2013b) are based on assumptions and theoretical models involving highly uncertain estimates for some of the key factors for which there seems to be no evidence, and they completely ignore embodied energy costs and “downstream” losses. AEMO does however state a figure for the amount PV contributes when demand peaks. They say that in winter the 426 MW installed PV capacity in Victoria contributes a mere 4 MW (…and this too seems to be an estimate, not a measured observation.). However this is late in the day and it would be more meaningful to know what PV contributes during the whole of a typical winter day. Unfortunately personal communications found that AEMO does not have this data.
The important point from all this is that we seem to be in no position to make assumptions about what contribution rooftop PV can make at what cost, or how much plant would be needed to achieve a chosen contribution. EDM like AEMO and the rest of us can only deal in terms of peak ratings and thus theoretical outputs in ideal conditions and these are likely to be quite misleading indicators of actual achievements in the field, especially at less than ideal sites in winter. As Palmer, and Hall and Prieto stress, many factors then detract from those theoretical figures, leaving us highly uncertain about how much PV capacity we would have to build and pay for in a proposal designed to deliver a certain net amount to the grid. What we can be sure about is that if we calculate in terms of the commonly stated dollar cost per kW(p) of capacity, and commonly quoted ER figures, we will seriously underestimate the cost.
These considerations make it very difficult to assess the EDM proposal as it is not possible to work out confidently how much PV energy can be sent to the grid net of the above costs and losses, at what capital cost. We won’t know until the new uncertainty over PV embodied costs and other losses is resolved, and until actual input data is available, but given that these factors have previously been ignored it is plausible that the general net ER for PV will be half or less of the value previously assumed, effectively at least doubling previously assumed sector capital costs.
Estimating costs
AETA reports the present capital cost for fully installed PV at $3,800/kW(p) (p. 43, but $3,380 on p. 87?). (However much higher 2010 figures are given by Black and Vetch, 2012. NREL states a much higher figure, $(US)7,690/kW(p) for rooftop and $4,790 for commercial scale.) EDM use the AETA future cost of $1,677, which is one-fifth of the NREL figure for present cost (and it is present cost that matters for affordability; see below.)
Because costs have recently fallen rapidly a confident assumption cannot be made here but the AETA figure will be used below. It is stating what it would cost to have a sufficient area of PV panels etc. to produce 1kW in peak radiation at (I assume) 15% efficiency. If that area received peak radiation for 24 hours, i.e. 24 kWh/m2/d, it would produce a constant 1 kW, but in winter radiation in the Sydney region only totals about 5 kWh/m2. (Melbourne or Tasmanian winter radiation is around half this level.) So we would need 24/5 = 4.8 times as big an area to produce an average 1kW in winter. Therefore the capital cost of sufficient PV panels etc. to produce the equivalent of a constant average 1 kW in winter (i.e., ignoring the storage issue) would be around $18,240. (Wilson, 2011, states the annual average as $20,000, so his winter figure would be considerably higher than $18,240.)
This is a gross output figure and to it we should add the cost added by all the embodied energy costs and the downstream loss factors referred to above. If Hall and Prieto, and Palmer are right in claiming an ER of 2 – 3 these costs and losses add to around half the lifetime energy produced, so would more or less double the above cost figure. It is not possible to make a confident ER assumption so I will assume an ER two to three times as high, i.e., 6. This would seem to be generous compared with the commonly stated ER of about 8 – 10 which does not take in any of the upstream embodied costs or downstream loss factors discussed above. The capital cost figures then become, $21,900 for the capacity to deliver a constant 1 kW in winter. Of course none of these figures take in the cost of the storage (or back up) to enable the 1 kW to be sent constantly, but that is not relevant to the EDM proposal.
The solar thermal component
Output issues
The solar thermal capacity is assumed to be 9.4 to 13.3 GW in the two EDM 2 scenarios, contributing 8.3 – 12.5% of power. Again it is not possible to assess the acceptability of these low percentages (for the most costly sector) but they seem to be much too low given the above points made re wind and PV. If the wind limit is 20% (not 50+%), the PV limit is 15% (not 20%), biogas 7% (as assumed), and hydro 5% (as assumed, plausibly) then the costly solar thermal sector would have to contribute around 53%, which is 4 – 6 times the amount EDM assume. Again this would greatly increase total system capital costs.
Fig 2 in EDM 1 representing combined inputs for each of a number of days enabled some visual confirmation of capacities needed. The ability to see these combinations is crucial in assessing the plausibility of such proposals, but this is not possible with the second and third papers. Thus we can’t check whether the following apparent problem evident in that figure has been resolved. It is stated in the paper that 15.6 GW of solar thermal capacity would be needed, but Fig. 2 seems to show that at times the solar thermal component will be sending out perhaps 27 GW. There would be much worse days than the one represented in Fig. 2, suggesting that on those days the amount would be greater still.
Whether the 9.4 – 13.3 GW of solar thermal capacity stated in EDM 2 would be sufficient would depend on long-term weather patterns, and again we are not given evidence on these. How often would there be no wind or sun for several days in a row, requiring electricity to come mainly from hydro and biomass-gas-generation? (See below.)
My recent attempt (Trainer 2013a) to analyse the AEMO weather data for its implications for solar thermal power has increased my concerns about the potential of solar thermal in less than ideal conditions. I took six widely distributed good sites from central Australia to Mildura and found that in the 92 day period at the end of 2010 there were 12 (non-overlapping) periods each lasting 4 days or more, including 48 days in all, in which DNI averaged across the sites did not reach 500 W/m2 at any time during the day. Reference to power curves (mostly available for dish-Stirling systems but also somewhat evident for central receivers…and apparently much worse for troughs) indicates that almost no power would have been generated on these days. It seems that if daily radiation received is under 6 kWh/m2 day solar thermal output will be poor, and possibly very poor. This is also stated for troughs by Odeh, Behamia and Morrison, 2003, Fig. 14, and is evident in the plot by Elliston et al. (undated), and other sources discussed in Trainer (2013a). In the slides from Elliston (2012) there is considerable evidence on this issue, and it is not clear why this does not seem to have been integrated into the papers. As Elliston says the slides document “Some very long low irradiance events.” Some of these exceed 5 days of negligible radiation. A 6 day event is noted for Roma in 2000. A simulated solar thermal plant output for Cobar indicates no output for almost 4 consecutive days. It is said that in the south of the continent these events are most frequent in winter, which is the time of highest demand.
It would seem that energy production from dish-Stirling units in the rather high latitudes ZCA assumes (Wright and Hearps 2010), e.g., Mildura, could be around only 10% of what might be expected given an annual average daily DNI/m2 of 4.3 W/m2. (Central receivers would do somewhat better than dishes, but I cannot get much power curve data for them so can’t say how much better.)
The above study (Trainer 2013a) also explored the possible effects of low DNI on the efficiency of solar thermal power generation. It seemed that as average DNI goes down towards 500 W/m2 or the daily total goes towards 6 kWh/m2/day, efficiency is markedly reduced. For some dish-Stirling systems it is clearly reported that when DNI is down to 400 W/m2 efficiency and output fall to zero. This has profound implications. It seems to mean that although solar thermal systems are excellent contributors in summer they will make little contribution in winter even in ideal regions, and that at other times they will be subject to frequent periods of no output for periods of 3 or more days. It also means that calculations of the embodied energy cost and pay back time for solar thermal need to be rethought, because these have been based on the mistaken assumption that annual average DNI for a site can be multiplied by a single, constant plant efficiency figure (which refers to performance in peak radiation) to yield annual or winter output, when in fact much of the DNI making up that average will be too weak to generate any power. In other words it is a mistake to simply take a DNI/m2 figure for a site and multiply it by a constant efficiency figure (ZCA assumes 17%) to arrive at a figure for power produced, because if the DNI figure is middling to low plant efficiency will be disproportionately lower, and it is possible that in winter in some locations which are satisfactory in summer no power at all will be produced.
For instance ZCA assume solar thermal located at three sites where winter DNI averages under 5 kWh/m2/d. This means that at the Mildura site where DNI is 4.5 kWh/m2/d, about half the time it will range down to well under that value. At 4 kWh/m2 DNI is not likely to rise above 400 W/m2 for more than 2 hours a day, meaning that power generated on that day would be negligible.
More recently I have found that total output from the Australian wind system was close to zero on two occasions within the periods I identified above in which there was negligible solar radiation (i.e., 5th October and 20th November 2010.). Thus for the EDM proposal there would very likely be times when total demand has to be met by biomass-gas-electricity plus hydro, so if EDM are intending to provide for a 37 GW peak demand, the biomass-gas-electricity capacity would have to be much greater than the 23 GW they assume (see below.)
It is stated in EDM 1 but not explained that to use more than 6 hour solar thermal storage would not be viable. However EDM 2 assumes 15 hour storage and why this is now regarded as viable is not explained.
Embodied energy costs for solar thermal do not seem to have been taken into account. I am not aware of any satisfactory studies of this factor to date, dealing with all upstream embodied energy costs, as for PV above. (Dale’s review, 2013, lists only 8 studies, with no reference to upstream or downstream costs and losses.) The few figures in the literature indicate a 6 – 11% cost, and none seem to have included upstream embodied energy costs or possible downstream losses. In my analyses I assume a 10% embodied energy cost but when thorough analyses are available the figure is likely to be higher.
Note that higher Solar Multiples and storage capacities (discussed below) add to embodied costs and reduce ERs, because they do not increase the amount of energy produced, they only increase the delivery period possible. The evidence below suggests that this factor might increase embodied energy costs by more than 50%.
Similarly there seems to have been no study of downstream costs and losses for solar thermal. Many of the 15 factors Prieto and Hall take into account for PV would be relevant, such as losses in connections to the grid, energy used by workers and vehicles, and in plant security and in construction financing.
Capital costs
The Solar Multiple issue
This seems to me to be confused in both the EDM and Zero Carbon Australia (Wright and Hearps, 2010) analyses. EDM assume 15 hour storage and refer to this as a SM of 2.5. ZCA assume 17 hour storage and refer to this as involving a SM of 2.5. However it seems that these figures should be 3.5 and 3.83.
If the SM is 1 there is no storage and the field is just big enough to drive the turbine at more or less full power for the approximately average 6 hours equivalent that the sun is shining at full strength. If the SM is 2 the field is twice as big to enable collection of twice as much energy and thus to store for 6 hours operation, if there is 12 hours storage the SM is 3, and if there is to be 15 hours storage the SM will have to be c. 3.5. This is the usage explicitly discussed in Lovegrove et al., (2012), James and Hayward (2012, p. 12), and Trieb, et al. (2009) who state, correctly it seems to me, 18 hours storage means the SM is 4. The IRENA review of solar thermal power (2012) makes this use at a number of points, e.g., pp. 8, 14, 18 and 30. The US DOE (2012) more or less corresponds, saying that for 11 hour storage the SM would be 2.5; i.e., a little larger than the above references would suggest.
In EDM 2 on p. 8 the apparently correct figure is quoted from AETA: “The AETA provides cost data for CST plants with six hours of thermal storage and a solar multiple of 2.” But the next sentence says, “As the simulations are based on CST plants with 15 hours storage and a solar multiple of 2.5 …the capital cost of the simulated CST plant was derived by scaling the solar multiple by 1.25 and the storage by 2.5.” The ratio seems to have been arrived at by dividing 2.5 by 2 but the ratio of total plant cost is not the same as the ratio of the SMs. It is produced by the costs of the additional field and storage added to the other costs, such as for the tower.
This is not simply a matter of definition; it affects field and storage cost estimates. AETA (2012 p.37) gives cost estimates for central receivers without storage and with 6 hour storage, but not for longer periods. They say that adding field and storage to enable 6 hour operation from storage increases plant cost from $5,900/kW(p) to $8,308, i.e., by 41%. For a plant with 15 hour storage this amount of field and storage would have to be added to the cost of a plant with 6 hour storage another 1.5 times. Thus the total cost would be 60% higher than the cost of the plant with 6 hour storage, a multiple of 1.6 not 1.25.
Lovegrove et al., (2012) give figures for LCOE which seems to imply a higher multiple. They indicate that 15 hour storage would involve raise the LCOE to 1.8 times that for 6 hour storage. Trieb (2010) indicates a similar figure.
It will be assumed below that a SM of 3.5 adds 50% to plant cost.
Capital cost estimation
Following is an imprecise derivation of a figure for the capital cost of sufficient plant to deliver at distance a net 1kW in winter. It will be based on the information given in the NREL (2010, 2011) Solar Advisory Model for a (theoretical/modelled) example 100 MW central receiver. This is located at Blytheside River, Southern California USA, the estimate assumes 6 hour storage and air cooling, and provides for interest charges. The cost estimate is $658 million and therefore is $6,580/kW(e)(peak). (However NREL says this figure has been set at lower than actual present construction costs; Turchi, 2014.) The average annual rate at which power would be sent out is estimated at 40% of peak capacity, and the winter rate is 28 % of peak, indicating that the capital cost for the capacity to send out each 1 kW on average would be $16,250, and to do so in winter it would be $23,214.
A number of considerations would greatly increase this figure.
• The energy loss of 10 – 15% for long distance transmission, e.g., from the Sahara to Northern Europe, or from the distant sites that are ideal in Australia.
• The capacity to collect and store for 15 hours might multiply the cost of the NREL example by 1.5+.
• The embodied energy cost of the plant might take 10% of energy produced.
• The increased cost of construction in remote areas. The best situation for solar thermal plant is in desert areas, especially North Africa for European supply. According to Lovegrove et al. (2012) this might multiply total costs by a factor of 1.3+.
Taking these factors into account (assuming 10% for transmission loss and a factor of only 1.5 for the SM item) would multiply the above figure by 2.4 to $55,900/kW. The figure does not include the energy costs of plant O and M. (Nor does it take in the dollar or embodied energy costs of building, operating and maintaining the long distance transmission lines but these might best be kept separate, as EDM do.).
This cost figure is likely to appear to be extreme, and mistaken. It is not put forward here with great confidence but the above discussion makes the derivation clear, along with the assumptions etc. that would need to be revised if a significantly lower figure is to be established.
The cost of locally sited coal-fired plant capable of delivering 1 kW in mid-winter, and requiring no long distance transmission lines, assuming a .8 capacity factor, would be c. $3,700. (AETA, 2012). Adding fuel costs might bring the total to $5.7 billion, which is around one tenth of the above solar thermal cost. (This does not include an embodied energy cost for coal-fired plant but this would be relatively low as the 50 year life AETA assumes makes the ER for coal-fired power generators high.) Gas fuelled costs would be significantly lower.
Gemasolar
However the capital cost reported for the recently completed Gemasolar plant in Spain are far higher than the commonly stated solar thermal cost estimates, such as those given by NREL’s SAM. This 20 MW plant has 15 hour storage capacity, which the EDM proposal assumes. It has been reported as having a capital cost of $(A)548 million, which is $(A)27,400 /kW(p). (Solar Australia, 2011.) Some sources state 200+ million euro, including AETA. The plant is claimed to send out 110 million kWh/y, so the capital cost per annual average kWh sent out (as distinct from peak) could be around $41,000, whereas for coal-fired capacity it would be $3,100 according to AETA. If a 50 year plant life for coal fired power is assumed, as AETA does, p. 27, the annual capital cost per kW sent out for Gemasolar would be around 20 times that for coal. (Figures on what Gemasolar actually does send out are not made public. NREL publishes a graph but it does not enable monthly figures to be derived, and personal communications with NREL confirm that data is not released.) Taking into account embodied energy costs would raise these numbers further. Above all, these are average annual figures –what it sends out in winter, and the capital cost of sufficient plant to send 1 kW then, would be significantly lower.
Of course Gemasolar is the first of its kind with 15 hour storage, so the cost is likely to fall significantly eventually. However estimates of future capital costs generally assume only a 33% reduction for solar thermal units built in Australia (the expectation in AETA, (2012, p. 71), and AETA expects Gemasolar capital cost to only fall by 20% (p. 37.)
Also relevant is the fact that 15 hour storage does not mean approximately 24 hour supply or 100% capacity. Gemasolar’s capacity factor has been reported as 63%. (Wilson, 2011.)
Solar Thermal Conclusions?
It would not seem to be possible to state confident conclusions regarding capital costs for the actual amount of power delivered by the solar thermal sector. No large commercial scale central receivers have been built. NREL and all other sources approached confirm that operators will not release performance data. We do not know whether real world costs will be close to the theoretical estimates such as those given by NREL. More importantly we do not know what delivered figures would be, net of embodied, upstream and downstream costs and losses. And we do not know what these values would be in winter when it seems that for solar thermal power annual average values for efficiency and output do not apply.
Note also that not all cost estimates are as low as those from AETA and it should not be confidently assumed that costs will fall. The most quoted of the (few) estimates ventured expect falls, but some do not. (Guielen, 2011, Hayward, 2012.) Indeed it is likely that energy and materials costs will rise steeply in future. Lovegrove et al. note the recent jump in the price of steel. (2012, p. 190.) (The NREL cost is lower than the AETA cost but NREL points out that the SAM statements deliberately under-state present costs. Turchi, 2014.)
Another cost concern is that the value of the Australian dollar has been falling since the minerals boom. When the main capital cost estimates quoted at present were made it was around equal to the US dollar but early in 2014 it had fallen to around 80 US cents. If this is a return towards the long term level of around 70+ US cents, the cost of imported elements (all wind turbines, PV modules, and solar thermal turbines) might have to be multiplied by 1.4.
The efficiency of plant with SMs above 2 also needs to be considered. Some believe a solar multiple of 2.5 is about the limit because of the distance to outer reflectors (increasing the “cosine effect” loss as the average angle between mirrors and sun increases, and increasing heliostat spacing due to increased shading.) It would be interesting to know how the efficiency of Gemasolar is affected by these factors.
Again unfortunately the picture is obscure. The $55,900/kW figure above is at first site difficult to believe, and critical revisions of the derivation are invited, but there would seem to be reason to assume that present capital cost of delivering a net 1kW in winter would be a significant multiple of the figure EDM assume for future cost.
The biomass-gas-electricity sector
One of the significant recent advances in the discussion of renewable energy and the storage problem seems to have been the realisation that this might best be done by use of biomass to produce gas that can be delivered and stored via the existing gas supply system, and used to generate electricity when needed. Discussion of this option is a major merit of the EDM proposal, but again the quantities assumed cannot be verified, and sector costs assumed seem to be much too low. In addition it is not clear that the technology is viable in view of the difficulties it involved.
The proposal is intended to cope with a 37 GW peak demand. Given the above discussion of big gaps there would be periods in which the biomass-gas-electricity sector would have to provide total demand minus the hydro component which is assumed to be a maximum of c. 7 GW. Therefore it seems there would have to be at least 30 GW of biomass-gas-electricity generating capacity on hand, not the 23 GW stated.
Very few biomass-gas-electricity systems have been built and the viability of the technology has not been established. Significant uncertainties are expressed in the literature.
The Grattan Report (Wood et al., 2012) on renewables notes the problems. 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. One source says “…no viable technology has been available to produce refinery grade syngas from biomass.” (Syngas Technology, 2013.) EDM use the conclusions on various renewable technologies from the AETA review but that source reinforces doubts about the viability of the biomass-gas-electricity path and does not attempt to estimate capital costs. It gives only seven lines to the technology, 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.)
It seems clear that the proposal 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 the section on gas generation. This is not a section on the complete biomass-gas-electricity generation process or its costs.) But biomass-gas-electricity capital costs would add to the turbine cost the costs for the gas producing plant, including removing impurities and compressing and pumping, and all the machinery going into biomass production, harvesting, drying, and transportation. Kendry (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. It is not possible to state a confident figure here.
Biogas backup
Depending on how often how much gas would be required for back up, it could be that relatively little gas producing plant would be needed because it could be continually restocking storage at a constant rate. Again the amount can’t be determined without detailed analysis of weather patterns and gaps. However the EDM proposal does involve a considerable use of gas, generating up to 7.1% of power supplied. This suggests that the gas production rate would have to be sufficient to average an electrical output of 7.1% x 23 GW = 1.63 GW and therefore that the relevant capital cost figure would be for plant and other equipment capable of providing biomass at a rate of greater than 3.26 GW (if conversion efficiency via gas turbines is 50%, but more than twice as much if steam via biomass burning has to be used). This means the energy content of the biomass input flow would have to be at least 103 PJ/y corresponding to c. 5.7 million tonnes p.a. (None of this includes embodied energy etc. costs.)
Estimates of biomass potential vary greatly but Australia would have no great difficulty providing the amount needed according to this estimate. AETA’s figures on available landfill, sugar and other woody waste biomass material indicate that these could generate around 5% of present electrical energy used, i.e., 34 PJ/y (pp. 50 – 51), corresponding to a biomass input flow of c. 100 PJ/y. Milibrandt and Overend (2008) state a higher figure; all collectable waste might produce 293 PJ/y in fuel corresponding to 36% of Australian petrol consumption. If necessary a relatively small amount of plantation biomass could be used, but it must be kept in mind that to meet transport demand via ethanol or methanol would draw heavily on his potential. ABARE and Geoscience Australia say our total biomass potential is only c. 480 PJ/y. Transport now takes over c. 1,800 PJ/y of fuel, and if all this was to be provided in the form of ethanol c. 4,000 – 5,500 PJ/y of biomass would have to be going into its production. The lower figure would require a daunting 250 million tonnes p.a., from perhaps 35 million ha.
(These numbers rule out meeting global transport demand via biomass; around 16 billion ha of plantations would be needed…on a planet with only about 10 billion ha of productive land. The IPCC, 2011, estimates total plantation plus waste potential at c. 420 EJ/y, which would give 9 billion people an average liquid fuel budged around 20% of the present Australian figure…and there are several reasons for thinking the IPCC figure is too high; see Trainer 2012a. The 2014 IPCC Working Group 3 report states a much lower global potential, up to 270 EJ/y of primary energy. Lang points out that the availability of biomass year after year in a land prone to severe and protracted droughts, and floods, is highly variable and uncertain.)
Given AETA’s above doubts about producing gas by pyrolysis the system might have to be based on biomass burning to generate power via steam, which is the only path AETA discusses. Crawford (undated) reports the capital cost of biomass CHP at $5,500/kW, and AETA (p. 52) states $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 and would therefore be less adept at plugging sudden gaps.
Then there are energy efficiency considerations. Gas turbines can be quite efficient but 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%. He also says that because of the difficulty of removing the tars etc. mixed in the gas it can be preferable to produce gas by fermentation at even lower yield.
Thodey also raises the issue of plant size. He says generators running on biogas have to be large to be efficient. If they are in the 2 – 5 MW range efficiency is only c. 20 – 25%. Because of the problem of trucking distance it is generally assumed that biomass fuelled operations would have to be small and numerous, distributed throughout biomass producing areas.
These figures probably refer to within-plant efficiencies and are very unlikely to have subtracted embodied figures, or other energy losses and costs such as for producing and trucking the biomass. It seems therefore that the overall sector energy efficiency would probably be no higher than around 20%. This low figure would mean that a lot of plant and equipment would be required to produce the amount of biomass and gas needed. The low efficiency would also increase the amount of biomass required. All this would also have a cost in embodied energy. It might be preferable to store via hydrogen, which has low overall efficiency..
Again confident conclusions are elusive, mainly because we can’t be clear about the amount of gap-filling power that would be needed, nor about the efficiency and cost, or indeed the viability, of plugging gaps by biomass-gas-electricity. It would seem however that the sector’s capital cost would be much greater than is assumed by EDM 2.
Hydroelectricity
It should not be assumed that there are no problems or limits regarding use of hydroelectric generating capacity and pumped storage capacity to plug gaps. Lang (2009) discusses these, including the limit that might exist on the amount of water that can be released into a river over a short period. EDM assume c.12 TWh, 5.9% of demand, from 7.1 GW of capacity. Using surplus wind and solar energy for pumped hydro storage involves the problem of whether the surplus rates are sufficient, continuous and lasting. Lang points out that pumps getting large volumes to start moving long distances might have to run for three hours to be efficient.
The dumping issue
It is puzzling that the amounts of plant stated in EDM Table 5 would generate only and no more than (approximately) the amount of power stated in column 3, which adds to the actual 200 TWh annual Australian demand. That is there seems to be no need to dump any surplus generated because when normal capacity assumptions are applied to those amounts of plant we arrive at the quantities of power listed in the relevant column of the table.
For instance Table 5 says that 34.1 GW of wind plant is going to contribute 94.8 TWh/y. But for that much plant to provide that much useable power over a year it would have to achieve an average capacity factor of 31.7% (a plausible annual average figure), provided that every Watt produced was used.
Similarly the 29.6 GW PV sector is assumed to provide 41 TWh of useable power. It could produce this amount if its capacity factor was 15.8%. This is about the generally accepted figure for PV, meaning that in a year that much PV plant would indeed have an accumulated output of 41 TWh, but it is very likely that much of it would be produced when it was not needed.
Thus the assumption seems to be that almost all power generated can be used when it is being generated. The gaps left by wind and solar will be filled by the large biogas component. (It has capacity equal to total average power demand, 23 GW.) This could well be a sound strategy, but the lack of dumped wind energy in particular is confusing.
Whether these claims are sound cannot be seen without access to the weather patterns assumed. However I one scenario it is claimed that the biomass-gas-electricity sector will only need to contribute 2.6% of total power demanded. This means that the four sources are being claimed to be capable of combination throughout varying weather patterns and demand patterns to almost always provide almost exactly the amount of power needed.
Table 5 EDM 2, (p. 23) states that only 4% of energy produced would have to be dumped. This is a surprisingly low figure. Huva, Dargaville and Caine’s exploration of renewable supply for Victoria (2013) found that for almost half of the time shown in their Fig. 7 the amount of wind capacity needed for wind to contribute its share would be putting out around two and a half times the total amount of power needed at those times. Only about 65% of power produced by wind could be used. (This is only an exploratory study indicating the approach being taken by a more detailed analysis that is underway, so its findings can’t be taken as conclusive. However the weather assumptions it is based on are highly favourable to renewables, involving a 20 day average, not winter, period containing only one difficult half day.) In the ZCA proposal it appears that almost half of the solar thermal output is to be dumped.
Remember that one of the EDM scenarios assumes that more than 50% of power is assumed to come from the highly variable wind sector. If at times this 34.1GW of capacity was running at 80% of peak it would be generating 27.3 GW… which is considerably more than total demand. (In the 10% discount scenario wind capacity is 47.1 GW and at 80% this would be generating 38 GW.) Clearly much of the time 34.1GW of wind capacity could be expected to be generating much more power than is needed given that the other 42.9 GW of solar would also be contributing.
To summarise, the claim seems to be that almost all power generated by wind, sun and hydro can be used, and almost never will their combined output fall short of demand. From the numbers given this does not seem to be possible, but the information given does not seem to enable the situation to be understood. Perhaps judgment should be reserved until the derivation from underlying weather data is made available.
Transmission losses and costs
EDM 2 advances on EDM 1 by including an estimate for transmission line costs, but again this is presented to the reader as a bald statement and despite lengthy and obscure discussion (of “genetic algorithms”) it is not shown how the figure is derived or that it is plausible. It is said that a simplified system is assumed and actual costs would probably be higher. Again this would require detailed access to weather patterns and the occurrence of low inputs and gaps, enabling assessment of the sufficiency of the transmission infrastructure claimed.
The total system cost?
Following is a rough attempt to compare only the capital cost sum (for the EDM 2 “least cost” and 5% discount rate scenario), with alternative conclusions derived from the above figures. The exercise is not precise or confident but it is a transparent indication that the capital cost sum for the amount of plant to deliver the power EDM assume would probably be several times that arrived at by EDM.
Table 1a sets out the capital costs for the amount of plant EDM 2 says is required in GW.
This seems to align with EDM estimates (they do not give separate figures for capital costs.) Their lowest cost total, $19 billion p.a., adds O and M costs and a 5% discount rate.
Because we cannot see whether the EDM conclusions re the amount of plant needed are sound, the alternative capital cost estimate in Table 1b is derived from the percentages of power contributed, as stated by EDM 2. It sets out the costs of these amounts of plant derived from the above conclusions re the amount of plant and capital required to deliver I kW at distance in winter.
N.B. EDM use estimated future capital costs whereas the alternative cost estimates in Table 1b use AETA’s estimated present costs. As will be discussed the crucial issue for renewables is affordability and in the early decades of the build-up period present costs will have to be paid, not the costs to which plant are predicted to fall by 2030 or 2050.
The difference is considerable. For Wind the ratio is 1.4, for PV it is 2, and for Solar thermal 1.5.
The above EDM sum arrived at in Table 1a does not include the transmission cost. The alternative table 1b below (unrealistically) assumes a transmission cost only for solar thermal.
This annual sum is around 2+% of Australian GDP. Present world investment in electricity supply, including distribution, is around .28% of GDP. (McCollum et al., 2012, , Rahai, et al., 2012, IEA, 2011, IPCC, 2014,Ch. 16.)
There are several reasons why the real world multiple would be much higher than this $32 billion figure.
As discussed re dumping above, Table 1 b assumes that the power can be produced when it is needed and there will be no dumping. But to deliver the quantities in column 1 of the table there would probably have to be much more generating capacity than would produce these amounts, because much of the output would not be being produced when it was needed. If significant dumping would occur this would affect sector ER and thus raise the capital cost per delivered kW assumed in Table 1b.
The PV, wind and solar thermal capital costs assumed in the above alternative derivation are likely to be much lower than those that will eventually be arrived at following thorough accounting of embodied and downstream costs and losses, and of the solar multiple and low DNI/efficiency issues. EDM do not include or discuss these factors. (Note again the high Gemasolar cost from the real world, as distinct from predictions from theoretical modelling, and the fact that the NREL rooftop PV cost estimate is double AETA’s figure, which is double the figure EDM use.)
We cannot assess whether the amount of plant assumed is sufficient to cope with gaps in the wind and solar resource, and there are reasons for thinking that it is not.
The winter wind capacity factor assumed, 38%, could be 60% higher than it should be.
In the alternative costing the embodied energy costs of the long distance transmission systems have not been included. These must be deducted from system output.
A realistic cost for the biomass-gas-electricity sector would include all the elements in addition to the turbines, and none of these have been included in the alternative costing. If the biomass has to be burned the capital cost for generation would probably be 6 – 7 times as high as EDM assume.
No account has been taken of the energy costs of O and M within the various sectors, including for the transmission system. These reduce power delivered and thus raise the capital cost pe