We all know that concentrating solar thermal technology in California has been delivering ‘no fuel’ electricity for two decades. Now advanced solar thermal electric options are dropping in price and some companies are introducing thermal storage to match power demand. Here, David R. Mills and Robert G. Morgan explain how this technology can deliver very much more. Their modeling shows that solar thermal power could not only replace most fossil-fueled electricity generation in the US, but could replace petroleum-based transportation. They argue it’s not only technically, but economically feasible – and not just for the US but for China and India as well.The sun is a much larger practical energy resource than any non-direct solar resource. Consequently, solar electricity is the most likely means to nearly eliminate contributions to global warming from electricity generation by mid-century. Furthermore, with thermal storage much cheaper than electrical, mechanical or hydrogen storage, solar electricity will probably be predominantly in the form of solar thermal electricity (STE) with thermal storage, rather than photovoltaic solar electricity with electrical or mechanical storage.
STE (often referred to as CSP, or concentrating solar power – though this, strictly speaking, includes concentrating photovoltaics) uses a field of solar reflectors to create a hot fluid to run a heat engine such as a Rankine or Brayton cycle. This is a proven concept using Rankine cycle turbines. It has, for instance, been successfully demonstrated in the Californian desert for two decades using commercial parabolic trough technology and steam turbines, achieving an annual field availability of 99%.
The US National Renewable Energy Laboratory (NREL) uses a conservative future total plant availability of 94%, due primarily to operations and maintenance (O&M) requirements of the conventional steam turbine used. Central receiver (CR) technology, in which a small receiver on a high tower is illuminated by a field of mirrors below, has also been developed using two-axis tracking heliostat reflectors and a commercial plant, PS10, has begun operation in Spain. A third option recently developed is the linear Fresnel reflector (LFR) system in which long steam pipe receivers on towers are illuminated by long heliostats (see REW March–April 2008, page 65–67). A CLFR (compact LFR) system is the basis of a recently announced 177 MW project by the PG&E utility in California. Both CRs and LFRs currently generate steam directly with low parasitic pumping losses and could be used in gigawatt-scale fields.
STE can also use low-cost energy storage in artificial thermal reservoirs. Oil storage was successfully demonstrated commercially in the mid-1980s and molten salt is being commercialized in parabolic trough plants in Spain. In addition, very low cost water-based thermal storage is expected to be commercialized within two years using technology that is currently under development. Thermal storage can actually lower the cost per kWh because it reduces the turbine size required for a given thermal output. In STE designs using storage and no fuel, there is also immunity from fuel cost rises over the long term.
Currently, the cost per kWh of STE is reaching the cost of natural gas generation in California and is expected to be near US new plant coal-fired generation cost when plants get to the 500–1000 MW scale in a few years. Any technology that can displace coal and gas generation could also potentially eliminate vehicle emissions using plug-in electric vehicles.
Modelling and analysis
A technical examination of both electrical supply markets and automotive applications, without detailed reference to economics, has been undertaken using 2006 data to reveal correlations between solar output and grid load requirements in California and Texas.
The data on grid usage is based on hour-by-hour grid load data from the California Independent System Operator (CAISO) and the Electric Reliability Council of Texas (ERCOT). Meanwhile, a simple vehicle usage model was developed from energy use data from the same sources together with national load figures for the US.
The model for the solar thermal aspects is part of the project model developed commercially for the CLFR system. However, any solar system with the same number of hours of storage will exhibit broadly similar correlations. The model uses ray trace results to form a map of optical performance against sun position in the sky. Two models were used, an internally developed one and, for reference, SolTrace from NREL. The maps were incorporated into a TRNSYS model – a transient systems simulation program with a modular structure – of the collector and power system, which can incorporate storage modules.
A simpler project model with project financial modules was also examined with a collector and storage performance modules that are cross-correlated with the TRNSYS model to ensure accuracy. The project model was run for every hour of the year and, where possible, load data was entered on an hourly time step. Both the TRNSYS and the project models develop a value for collected solar radiation from archived solar radiation data, as available from NREL.
The individual power block peak thermal efficiency was assumed to be 33%, although this does not affect the basic conclusions that would apply for any turbine size and efficiency. The plant fleet size was arbitrarily made to equal the peak load requirem ent of the state or country being modelled. In this case, based on 2006 data, this implies 50 GW for California, 63 GW for Texas, and 1067 GW installed and 789 GW non-coincident peak load for the US overall.
The solar multiple (SM) is the ratio of actual solar array size to the minimum size required to run a turbine at full capacity at solar noon in midsummer. Solar multiples greater than one are required when delivering power outside daylight hours using storage. The storage used is only enough to carry load for one or two days, and is used to match hourly output fluctuations in solar input with hourly load. These storage levels do not provide seasonal or even weekly storage, so are subject to local weather events, especially sustained cloudy periods.
However, with the CLFR, solar multiples of up to two are possible, even when not using storage. This causes overproduction of thermal energy at peak solar periods in summer, which is discarded by turning some of the reflector field off-focus, but also allows better utilization of the turbine at other times, increasing plant capacity factor. Because our models currently show improved economics using a solar multiple of two in fields without storage, we use this figure as the non-storage configuration with the best correlation with grid-load.
Replacing fossil generation
The capacity factor is the ratio of actual energy supplied to the maximum possible supply by the installed turbines over that period. In Figure 1, modelled monthly capacity factor (CF) is given for a 50 GW installation using the 2006 the Californian ISO grid load. The collector model uses an array with a solar multiple of two, with the array being assumed to have storage only in the thermal mass of the array pipes, fluid, and steam drums. It can be seen that, partially due to the SM2 strategy, the CF is reasonable and the array covers about 40% of the annual California load. This is excellent for a non-storage technology, but not enough to allow the technology to generate the majority of power on the grid.
In Figure 2, the same turbine fleet is now provided with arrays in the SM2, SM3 and SM4 sizes, all with 16 hours of storage. The chart shows the SM3 case to exceed the grid load requirement at all times except in winter, using a peak turbine capacity equal to the peak load of 50 GW. The 16-hour figure was chosen because it was financially optimal for the SM3 case; many other storage levels were attempted. The correlation with annual load is 92%, without the application of any peaking plant, with only 3% of energy having to be dumped by turning excess collector capacity off-focus. At SM2, the monthly load is never carried, but zero energy is dumped. At SM4, the entire grid load is carried, but 22% of energy is dumped. The lowest kWh cost case is therefore near SM3, because the turbine operates at close to the capacity factor required by the grid, while little energy is dumped.
In Figure 3, the model results for the Texas ERCOT grid are given for SM2, SM3, and SM4. Again, 16 hours of storage was assumed. The chart shows the least cost SM3 case to fall short in summer, using a peak turbine capacity equal to the peak load hours of the year. This was 63 GW, recorded in May 2006. Again, SM3 is the best factor, with a 91% correlation, without needing peaking plant.
Supplying the entire US?
While the high supply fractions are compelling from a regional viewpoint, a more ambitious thought experiment addresses the supply of the entire national grid from the modelled Texas and California solar arrays.
Of course, supply of the US would take place from many southern and western states, but using two distant states like California and Texas is illustrative.
In Figure 4, the dashed line indicates the 2005 national grid profile scaled to the 108 GW coincident peak of both the CAISO and ERCOT. The result – surprisingly – is even closer to the two-state blended solar generation correlation, with 96% of the national annual grid supply accessible to least-cost SM3 STE. However, this chart was prepared by using monthly national data, not the hourly data available through CAISO and ERCOT. Nevertheless, there is a close match between the forms of load patterns of Texas, California, and the national grid, suggesting that similar amounts of storage could be used to the same effect.
Furthermore, there would be a tendency for extreme local weather events to be averaged out, and there would be hundreds of solar plants available with flexible storage and considerable geographic diversity. For this reason, a result close to or better than that in the California case is not unreasonable.
This close correlation in a country having a severe winter in the northern regions might seem not to be intuitively correct, but the excellent seasonal match at the national level can be better understood if one realizes that winter home heating loads are largely carried out by non-electrical energy, typically gas and oil, and that air-conditioning is mostly electrical. This produces a close national load correlation with solar seasonal availability similar to that previously calculated for the warmer states. In 2005–2006, the US national grid had a generating capacity of 1067 GW and non-coincident peak load of 789 GW. Based on the current technology, a CLFR with SM3 and storage would require 3.9 km2 for 177 MW, translating a national land requirement equal to 23,418 km2 or a square with 153 km sides.
Transportation – replacing oil
Recent developments in lithium ion batteries and supercapacitor technologies may provide the possibility of rapidly recharging electric vehicles which would use zero fossil fuel. The electricity for such vehicles would come from the national and state grids, and therefore can potentially be supplied by grid-connected renewable energy with low climate impact.
The annual US figure for 2006 vehicle emissions has been calculated by the Department of Energy at 2 billion tonnes of CO2 equivalent (CO2e). This is close to the annual US electricity generation emissions of 2.3 billion tonnes CO2e.
A Socolow Wedge – a concept that illustrates the scale of emissions cuts needed in the future, and provides a common unit for comparing the carbon mitigating capacities of various energy and storage technologies – is a saving of 1 billion tonnes of carbon emissions per year reduction. Converted to tonnes of CO2e, a single wedge is 3.7 billion tonnes of CO2e per year. Seven wedges are required to stabilize the atmosphere at 550 ppm of CO2 over 50 years, so the potential of removing emissions from the US generation and vehicle fleets is 17% of the entire global reductions required. The potential in other markets such as China, Europe and India is also large.
The US national vehicle fleet travelled 10 trillion miles in 2005–2006. Battery electric vehicles typically use between 0.17 and 0.37 kWhe per mile, so for 1 x 1013 miles of vehicular travel the US would need 1.7–3.7 x 106 GWh to eliminate fully vehicle emissions from fuel use. National solar generation would consequently have to climb by 42%–91% to accommodate an entirely electrified transport sector. The land area requirement for the supporting CLFR generation plant would thus climb to 182–211 km on a side, about 1509–2029 GW.
Superimposed on our electricity load, this would have some implications. Although fast charging will be available, it is likely that much of charging will take place in the home garage, leading to a stronger night load.
Because we do not have hourly data for the entire US grid, nor a typical charging pattern, we can look at a simple model using the more extreme effect of placing 91% more generation into one state, California, spreading the charging period over the period between 9 pm and 9 am. It is likely that technical improvements would drive vehicle efficiency toward the lower end of the range after a decade of manufacture, although this is not considered in this model. Nor does this model benefit from time zone displacement, as would occur in a national model. For both reasons, it can be regarded as a worst-case scenario.
Figure 5 shows a calculation for California, such that peak generation is now 95.5 GW rather than the existing 50 GW. It can be seen that the effect on the model correlation is marginal, with the SM3 configuration continuing to be preferred and the correlation slightly improved over the 50 GW California model shown in Figure 2 at 93%. This suggests that, on a national basis, the correlation will also remain high with a grid load that totally includes the vehicle sector. For more efficient vehicles, the added grid load would be smaller, but the correlation similar.
Discounting future commodity inflation, the authors believe that, without additional improvements in end use efficiency, future systems with storage will require $4.5–6 trillion in capital investment at today’s prices to provide generation capacity which supplies the great majority of both static and transport usage. The current cost of imported oil to the USA at $100 per barrel at an import rate of 13.2 million barrels a day – in 2005–2006 – is $482 billion per year.
The simple payback time in balance of payments of petroleum savings alone is approximately 9–12 years even without any consideration of substantial coal and gas fuel savings for the non-transportation sector. This very simple economic argument neglects very large additional benefits to the local environment, that, in addition to global environmental benefits, would include a much cleaner atmosphere in urban areas and the avoidance of associated health costs.
The net benefit to the economy over 40 years is thus clearly very positive, and far from the net cost that emissions reduction is commnly portrayed to be. Of course, should significant energy efficienty be implemented in the period before 2050, the additional generation required would be substantially reduced.
Solar thermal as a future energy source
Although it is often said that ‘solar cannot produce baseload electricity,’ STE is probably the only currently available technology that can be considered for a globally dominant role in the electricity sector over the next 40 years.
Humankind evolved to be most active when the sun was up, and this is why human activity and energy usage correlate significantly with the energy delivery from direct solar systems. Additional seasonal correlations detected result from the influence of the US national building air-conditioning load, which is greater toward summer months when the sun delivers more direct solar energy to the earth’s surface. We have up to now largely neglected these advantageous correlations when designing power systems technology. Such hourly and seasonal natural correlations with energy output from a solar system are substantially enhanced using storage. An immediate advantage is that load-following solar plant does not need expensive peaking plant back-up and it is clear that natural correlations can be used to economic advantage in solar power system design.
The relevance of baseload generation as a technical strategy needs to be carefully re-examined. Human activity does not correlate well with baseload coal or nuclear output and it should be recognized that baseload is what coal and nuclear technologies produce, not what is required by society and the environment.
Solar power with storage can take up as much of the grid generation load or vehicle energy load as is desired, and can host other clean energy options by treating them as a negative grid load. A mixture of storage and non-storage renewable options thus appears to be fully self-consistent as an alternative to the present generation mix, with the main co-contributors to STE probably being hydro and wind.
Not only is STE an energy option of great significance, but with only 16 hours of storage it has sufficient diurnal and seasonal natural correlation with electricity load to supply the great majority of the US national grid (and by logical extension, those of China and India) over the year, with the hourly solar radiation data including typical cloudy weather patterns. Furthermore, STE can supply much of an electrified transportation market without destroying these natural correlations.
An almost complete elimination of both fossil-fuelled generation and oil usage for transportation in the US appears to be technically feasible and will cost less than continuing to import oil.
Zero emissions technology is required to replace most of current generation by mid-century to meet stringent climate goals. What is now needed to facilitate such a vision is a rethink of the function and form of electricity grid networks, and the inclusion of high capacity factor solar electricity technology in the design of continental electricity systems.
David R. Mills is the chairman and chief R&D officer of California-based Ausra Inc. of California and Robert G. Morgan is the chief development officer.
Figure 1. Solar contribution to grid load in California assuming no storage and a solar multiple SM2. the annual contribution if 40%
Figure 2. Solar contribution to CAISO annual loads (16 hours storage)
Figure 3. Solar contribution to ERCOT annual loads (16 hours storage)
Figure 4. The effect of blending the grids from Texas and California and using SM3 arrays
Figure 5. Solar contribution to CAISO annual load with oil use onverted to electric load