Solar Load Cycles and Storage Requirements
By Doug Houseman and Vadim Zheglov
Most discussion of how using solar photovoltaic cells to account for most electricity generation has focused on the daily cycle of load versus the daily generation cycle. That cycle has been heavily analyzed, and there are a large number of companies working on local and neighborhood storage to support the daily cycle. But there is a much longer cycle that gets little attention and has a much larger impact on the cost of investment and on the efficient use of the solar generation. Simulations based on data from four major U.S. grid systems show that long-term storage needs will be large. This is but one of many scenarios that need to be explored.
Some years ago the question was raised at a conference of how much long-term energy storage would be required in a grid that depended entirely on renewable resources. Since roughly that time, our team at EnerNex has been collecting and analyzing data on wind, solar and other resources, while at the same time looking at a wide variety of storage devices and applications. This work led us to classify storage as long-term, short-term, daily cycle and smoothing. Here, we consider only long-term storage requirements.
In our study of long-term storage, we gathered data from four of the major U.S. grid systems, ERCOT, PJM, MISO and CAISO, to provide the empirical basis of a simple model. That enabled us to do a simulation in which solar resources would generate exactly the total energy needed over the course of a year to match total demand, and then we looked at the differences between demand and supply over the course of an “averaged” year. To err on the side of conservatism, this scenario put an extreme upper bound on the long-term storage requirements—no other scenario exceeds this one for the total storage requirement. (In the real world, because a mix of generation would exist, storage requirements would almost certainly be significantly smaller.)
Reviewing two years of data from 2011 and 2012 from the four ISOs/RTOs, we found that there were no significant differences between the years for the purposes of long-term storage estimation, though there were significant differences for short-term and daily-cycle storage requirements. In other words while there were strong month-to-month and year-to-year differences in the patterns, for purposes of estimating long-term storage size, the variations were not significant.
Once the hourly data for solar were scaled to match the annual demand, simple mathematical models were run to determine for each period whether the energy supply was net positive or negative. As presented here, monthly net intervals are used, while the shorter, hourly intervals are ignored. Our assumption is that storage would exist to handle the hourly, weekly and monthly variations, so that those will net out in long-term storage requirements. The pattern of solar generation turned out to be similar for the four ISOs. Demand for energy heavily exceeds the net-zero type generation from solar in the winter months, which comes as no surprise. ERCOT (Texas) and PJM (the Mid-Atlantic States) both show demand exceeding supply in the summer months, because of the heavy use of air conditioning in both regions.
In the four grid systems, net charging versus discharging ranges from a peak positive of 9 Terawatt-hours (December) to negative 5.3 TWh (April and May) in CAISO (California): + 6.5 TWh (August) and - 7.3 TWh (March) in ERCOT; + 19.3 TWh (December) and - 13.3 Twh (March) in MISO (the Middle West); and + 24.2 TWh (December) and - 17.5 TWh (April) in PJM.
In California, the storage filling period starts in March and finishes in October. From November through February storage has to discharge to cover the demand that exists. The maximum amount of energy that has to be stored is 21 TWh, which represents about 7 percent of total annual demand for energy in that market.
In Texas, the cycle has two parts: There is a summer discharge period from June through September to cover the heavy demand for air conditioning, and a second in December and January. In the other months of the year, there is a need to fill the long-cycle storage. The total need based on the ERCOT data is 16 TWh or 5 percent of annual demand, a quantity that is relatively low primarily because storage can be cycled twice a year rather than once.
In the midwestern service territory as it existed in 2011-12, MISO would see heavy discharge from November through February. It has a relatively large demand for storage, over 54 TWh, or just under 11 percent of the total energy that the MISO customers demand.
PJM has an interesting pattern that is somewhat similar to ERCOT in that it has a dual cycle annual storage use. In July and August and in November through January the system discharges from storage to support demand. In the rest of the year, the storage system will need to fill to support this demand. Because of the twice annual cycle, as in Texas the system can be smaller and more capital efficient because of this difference. Storage demands for the long-cycle needs are 47 TWh or approximately 7 percent of annual energy demand.
Considering those patterns in their totality, it is clear that long-cycle storage will be a key component of any generation plan that is designed from an economic generation point of view and uses a very large portion of variable renewables. Without the buffer that long term storage offers, the grid would have to contain many more generation units and manage the flow of power from those units actively; short-term, daily and smoothing storage would be critical components in systems that rely mainly on variable generation, with little or no long-term storage available.
To be sure, some inter-regional transmission may help reduce the storage demand, and allow balancing between regions. But in general the patterns are the same from region to region across the country. There is no region in a theoretical net-zero scenario where solar over-generates in December and January. This should not be surprising since this is the period of the lowest solar radiation during the year. So there will be rather sharp limits on the ability of solar generation in one region to fill in for the deficiencies in another region.
Altogether, the four ISOs/RTOs considered here account for roughly 50 percent of all U.S. power consumption and therefore can used as a proxy for the whole country. Extrapolated from the four, total long term storage needs for an all-solar United States would be approximately 275 TWh, roughly 7 percent of total delivered electricity in 2012.
As stated at the outset, our estimates represent a theoretical maximum limit of storage requirements. In practice, those requirements can be significantly reduced by:
- Mixing photovoltaic systems with wind systems, as the lowest annual wind generation periods often match some of the better solar generation periods.
- Completing transmission improvements between different regions, so that solar energy can be better exchanged among regions.
- Adding solar generation to go above the net-zero scenario here. That would mean that some solar generation would be off-line for months at some time each year.
- Pushing the envelope on energy efficiency and demand response to help not only with the daily cycle issues, but also for the long-cycle needs. Focusing on air conditioning and other summer devices should be the top priority.
Finally let us note that electric vehicles may grow to be a significant part of the transportation industry, raising overall demand. They would have a major potential impact on the daily cycle as well as potentially impacting the long-term cycle, based on the annual and monthly driving pattern for U.S. drivers.
Doug Houseman, an IEEE senior Member, is Vice President for Technical Innovation at EnerNex and has served as Chief Technology Officer at Capgemini. With extensive experience in the energy and utility industry, he has been involved in projects in more than 30 countries. He was designated part of the World Generation Class of 2007, one of 30 people in the global utility and energy industry so named by World-Generation Magazine and the World Generation Forum. He was the lead investigator on one of the largest studies on the future of distribution companies published by CEATI, and for the last five years has been working with more than 100 utilities and manufacturers, 50 governments, and five international agencies/NGOs on a wide range of industry issues. He was one of the primary authors of the IEEE Power & Energy Society’s GridVision 2050. He obtained his bachelor’s degree at the U.S. Naval Academy and did graduate work at the University of Michigan.
Vadim Zheglov, an IEEE member, is a Power Systems Consultant at EnerNex. He received his M.S. degree from the Tennessee Technological University in 2010 and his engineering diploma from the University of Applied Sciences, Wurzburg-Schweinfurt (Germany) in 2007. At Tennessee Tech, he was part of a student group that planned, installed and maintained photovoltaic systems and small wind turbines at various east Tennessee high schools. Prior to that, he worked on development of wireless power transmission devises, integration of automation tools and design of printed circuit boards at Philips Technology and BASF.