The increasing importance of intermittent renewable energy sources suggests a growing importance for energy storage as a way of smoothing the variable output. In this paper I investigate factors affecting the amount of energy storage needed, including the degree of intermittency and the correlations between wind and solar power outputs at different locations.
This paper has explored some basic issues in the economics of energy storage. There are two different functions that storage has to perform: one is to shift solar power produced in the daytime to the night, assuming that there are not sufficient other sources of power available at night. The second is to smooth out fluctuations in the output of renewable energy.
It might be possible to design a system in which either or both of these functions are redundant. There could be sufficient carbon-free generating sources to meet demand out of daytime hours - nuclear, geothermal, hydro and others - in which case no time-shifting of solar power would be needed.
Construction of the Salt Tanks which provide efficient thermal energy storage so that output can be provided after the sun goes down, and output can be scheduled to meet demand requirements. The 280 MW Solana Generating Station is designed to provide six hours of energy storage. This allows the plant to generate about 38 percent of its rated capacity over the course of a year
These same sources could also be used to smooth the output of a stochastic renewable source. But it could also be possible to avoid the need for storage to smooth output by spatial diversification of renewable energy sources, so lowering the correlation between different sources, and by building large amounts of capacity. And in both cases the need for storage could be further reduced by demand management, giving consumers incentives to reduce demand on the grid at times of shortage and to shift loads to times of surplus. Consumers have many ways of storing energy at their disposal - they can store hot or cold water in tanks, make ice when power is available and use it to cool air when it is not, and store energy in batteries, in particular in the batteries of electric vehicles. Denholm and Margolis (2016) consider this last option in detail.
The bottom line is that the question we are focussing on - how much energy storage would be needed if an economy such as that of the US were to move to much heavier dependence on renewable energy, as is implied by the goal of substantial reduction of greenhouse gas emissions - is probably not well-posed. As de Sisternes et al. (2016) note, “In general, while energy storage appears essential to enable decarbonization strategies dependent on very high shares of wind and solar energy, storage is not a requisite if a diverse mix of flexible, low-carbon power sources is employed, including flexible nuclear power.” There are some routes to low GHG emissions that travel via extensive use of storage, and others that make little if any use of these technologies and use other ways of managing intermittent power supplies. Which is best seems to be a matter of costs. If storage costs continue to fall, storage will feature prominently in the ultimate solution: otherwise we will work with a range of alternatives.
In my earlier paper Heal (2016) I assumed that the ability to store about two days of energy consumption would be needed in a world with 66% renewable energy, divided equally between wind and solar PV. Where does this discussion leave that assumption? It is convenient to discuss solar and wind separately. One third of total consumption would come from solar PV, and several studies (for example Denholm and Margolis (2016)) suggest that this level of solar penetration could be accommodated without storage: that it could all be used during the day. If we were to seek to shift some of this to other times of day, we would probably want to shift less than one half of solar output, which is less than one sixth of daily consumption.
The output from the one third of capacity that is wind could be smoothed by spatial diversification, as analyzed in section 4.1 above, but not completely, and at the cost of improving the connectivity of the grid (in my earlier paper I included the cost of an increase in miles of high voltage lines by 25%). We could adopt the approach of overbuilding wind capacity to reduce the probability of a power shortage to some acceptable level, combining this with demand management to cope with the low-probability eventualities. The illustrative calculations in section 4.1 suggest that in this case we might need to install six times the demand we want to meet, which would mean doubling the amount of wind capacity in my earlier paper. This would increase cost by $0.78 trillion. As I noted in my earlier paper, a battery large enough to store one day’s output from a wind turbine would cost twice as much as the turbine itself. It follows the extra capacity, with possible curtailment in the event of strong and persistent winds, is probably less costly than storage. In the earlier paper I allowed between $2.2 and $5.1 trillion for storage capacity, and for less than the lower limit here we could both build extra wind capacity for smoothing and construct some storage. The bottom line is that the storage figures I used in Heal (2016) are probably too high: less than one day of storage capacity might be adequate.
by Geoffrey Heal
National Bureau of Economic Research (NBER) www.NBER.org
NBER Working Paper No. 22752; Issued in October 2016