With the aging of the electric grid and growth in renewable energy generation, the need for ancillary services to maintain grid stability will rise. The Renewable Portfolio Standard in California alone requires that by 2020, 33% of the state’s energy is derived from renewable resources. The variability that these renewables add to the grid will necessitate an increase in demand of ancillary services (for example, area regulation and frequency regulation beyond what is currently available with dispatching and spinning reserves). Many authors have proposed that some form of energy storage augment the current spinning reserve to provide the ancillary services needed to keep the grid stable. This paper explores the effects of adding energy storage to the grid. It does so by using a simple grid model to determine the attributes of an ideal energy storage medium best suited to provide ancillary services. These attributes are then compared to available energy storage technologies such as ultracapacitors, flywheels, and batteries.
For any self-contained grid, be it a large grid covering most of North America or a small island grid with a single generator, the grid operator must insure that the generating capacity and the power demand are balanced at all times. In the simplest grid scenario consisting of a single generator and an equally sized load as shown in Figure 1, balancing is achieved by running the generator at capacity which ensures that there is adequate power available for whenever the load it turned on.
The disadvantage of this simple solution is that the generator consumes fuel regardless of whether it is producing power or not. If the load is used infrequently, then the amount of fuel to keep the generator running on standby could greatly outweigh the amount of fuel used to power the load. Ideally, if the load demand is known in advance, then the generator can be switched on just prior to load, and switched off just after to dramatically reduce the fuel consumed. The amount of time required for the generator to be switched on prior to the load being applied, is set by the generator’s ramp rate, or time needed to transition from fully off to fully on. Unfortunately, in most real power grids the exact timing of load demand is not known ahead of time.For larger power grids, where no single load represents a substantial portion of the total capacity, the solution to this problem is typically to set generation just slightly higher than current load demand, leaving a buffer window to accommodate future loads being switched on. As loads are then switched on or off, the generation capacity is then ramped up or down to maintain the buffer window. In North American power grids, this process of constantly adjusting generation capacity (to maintain a small buffer window above the current load demand) is highly stratified as shown in Figure 2 from the Energy Storage Market Assessment Guide published by Sandia National Laboratories. 
The first layer is the baseload generation and consists of generators that cannot easily adjust their power output (such as nuclear plants or large steam turbines). Energy purchases from other grid operators and intermediate duty generation, sold on a day-ahead market, is added to the baseload generation to meet the predicted daily load demand based on historical projections. The final adjustment of generation capacity throughout the day is done at three finer levels consisting of load following, area regulation, and frequency regulation. As a group, these levels of control are called ancillary services by the industry.
Load following generation is typically dispatched by the grid controller, based on a 15-minute look-ahead window, so the generators that provide this service need not be highly responsive. Area regulation, on the other hand, responds to a continuous reg up/reg down signal from the grid controller and is expected to respond within five minutes. Finally, frequency regulation is a small buffer window of generation capacity above the current load demand that this entire control scheme is intended to provide. Often called spinning reserve in the industry, the minimum size of this buffer window is determined by historical load variability and the level of control that the grid has experienced. For example on a typical summer day California ISO delivers about 40 GW of power to its customers during peak demand, and will maintains a regulation buffer of around 350 MW to ensure demand balance. Hydro and combined cycle turbines supply most of this 8.7% spinning reserve buffer. Each of these kinds of generators can dedicate up to 16% of their rated output to regulation. The balance comes from steam turbines, which can only dedicate about 9% of their rated output.
To improve efficiency and reduce fuel consumption, grid operators are under constant pressure to reduce the size of the buffer window, but this may result reduced grid reliability and increases in brownouts and power outages. This places grid operators under great strain as total load variability is on the rise in North America due to both increased power demand and the introduction of renewable energy sources.Renewables such as wind and solar place a particular stress on the grid because their output cannot be fully controlled. When the sun is shining, it is advantageous for the grid operator to use the available solar power to reduce total fuel consumption. However, when a cloud comes and reduces the power output of the large solar array, such as the one shown in Figure 3, other generation must be brought online to replace the loss of solar power.
Because of their uncontrolled output, renewables can be thought of as large negative loads on the grid, contributing greatly to the total variability. To accommodate this variability, more ancillary services are required by the grid operator. The grid operator has a choice at this point to either increase spinning reserve for more frequency regulation, or to increase area regulation. Increased area regulation is the preferred method to deal with increased variability because of the fuel cost associated with increasing spinning reserve.
To explore how much area regulation is required to support renewables on the grid, as well as how energy storage can effect this required regulation level, a simplified grid model was developed. In the model, generation consists of a single power plant dedicated entirely to area regulation and a single photovoltaic plant as shown in Figure 4. The output of the power plant and photovoltaic plant combine to supply a constant power load. The power plant has equal generation capacity as the load and has a five-minute ramp rate, the minimum ramp rate required to provide area regulation under most U.S. grid operators.
The purpose of this model is to explore the effect that various levels of solar power have on the power quality and grid stability in this simple system. In the grid, minor mismatches in generation and demand manifest themselves as frequency disturbances and voltage excursions. If the voltage excursions are large enough, distribution relays will trip off, shedding load from the grid and cause brownouts or blackouts in the areas they serve. Southern California Edison publishes a power quality handbook that outlines the limits of acceptable voltage excursion levels and durations within their grid during normal operation (as shown in Figure 5). Within this model, the power quality levels will be used as the limits of normal grid operation, and any excursion outside of these levels will cause the load to trip off.Much as in a real grid, once the load has tripped off, the relay will wait five seconds before attempting to reconnect the load. The solar power variability in this model is based on one-second solar irradiance data collected in a temperate northern hemisphere location with moderate cloud cover. For each run of this model, the name plate power rating of the solar plant is increased, representing a higher level of solar penetration within a grid with fixed area regulation resources (as represented by the power plant with a five minute ramp rate), and number of trip events are recorded as a measure of grid instability. Figure 6 is illustrative of the power quality produced in this model with 50% solar penetration. As a note, the number of trip events predicted in Figure 6 was well over 200 in a signal day.
Next, energy storage will be introduced to the simple grid model. As shown in Figure 7 the energy storage is assumed to have the same power capacity as the load, and the only parameter that is changed in each consecutive run of the model is the number of reserve minutes of backup power that can be stored.
The energy storage is assumed to charge and discharge to the grid and to maintain a 50% state of charge when not in use. With this method, the energy storage acts like a form of spinning reserve, providing the buffer window to allow the area regulation of the power plant time to respond. Even a small amount of energy storage can have a dramatic effect on power quality. Referring to Figure 8, 30 seconds of reserve energy storage were enough to eliminate all system trip events for the entire day.
As described previously, the simple grid model has two input variables consisting of the level of solar penetration and the number of minutes of reserve energy storage, and a single output, trip events, representing power quality. Figure 9 shows the results of the model over one month of solar data, when solar penetration is swept from 0% to 100%, and energy storage reserves are incrementally increased from 0 to 90 seconds.
Two key features are immediately apparent when inspecting Figure 9. The first is that without energy storage, as represented by the first black curve in the figure, the grid is stable until about 20% solar penetration. Based on area regulation, with a five-minute ramp rate, a complete loss 20% solar power would take just one minute to compensate for.
The model’s stability below 20% solar penetration indicates that solar power output is rarely fully lost in less than one minute. This is in agreement with large-scale photovoltaic power plant statistics which have been shown to rarely exceed 40% in a one-minute interval, as shown in Figure 10 reproduced from Carl Lenox’s paper “Variability in a Large-Scale PV Installation” .
The next feature that is apparent is that each incremental increase in energy storage reserves provides diminishing returns in power quality. For example, 20 reserve seconds of energy storage allows the grid to remain stable with up to 40% solar penetration, doubling the amount the grid can absorb through area regulation alone. Thought of another way, 20 reserve seconds of energy storage allows the same amount of solar power variability to be absorbed with only half as much generation dedicated to area regulation. To achieve another doubling in solar penetration, the energy storage reserves would need to be increased to well beyond 90 seconds, indicating that the majority of the energy storage benefit is derived from the first few tens of seconds of reserve capacity.Next, the throughput for the energy storage was calculated to determine the life expectancy for various storage technologies in this application. For a solar penetration level of 50% and an energy storage reserve capacity of 30 seconds, the energy throughput equivalent is 8.6 complete charge/discharge cycles in a single day. At this rate, an energy storage medium with a 10 year rated life would need to last the equivalent of 31,000 charge/discharge cycles by its end of life.
Based on this, it can be generally concluded that for optimum performance providing ancillary services, the energy storage system needs a reserve capacity of a few tens of seconds and a cycle life in the tens of thousands.
Choosing The Best Energy Storage System
Having identified the key feature of an ideal energy storage system, the next question of interest is which of the available storage technologies today is a best fit for providing ancillary services to the grid. Because of all the factors involved, making direct cost comparisons between different storage technologies is extremely difficult. However, a general indication of how well-suited a particular storage technology is for an application can be determined by comparing the specific energy of the technologies over the reserve times relevant for the application. In Figure 11, the specific energies of electrolytic capacitors, ultracapacitors, flywheels, lithium-ion batteries, and lead acid batteries are compared over a range of 10 ms to 2.8 hours of reserve capacity. In the range of interest for ancillary services, (10 to 100 seconds), it is apparent that, ultracapacitors, flywheels, and lithium-ion batteries compete for the highest specific energy.
To further narrow the technologies of interest, the nominal cycle life for each technology is compared in Figure 12. From the simplified grid model it is clear that to have a reasonably long service life in an ancillary service application, the technology of interest needs to have a cycle life of 10,000 cycles or higher.
Figure 12 indicates that of the three technologies identified to be good candidates based on specific energy, only ultracapacitors and flywheels meet the cycle life requirements needed to have a greater than 10 year service life in an ancillary service application.
When comparing these two technologies with an eye for grid compatibility however, two key points stand out. The first is that for safety reasons flywheels are typically installed underground requiring significant site prep and excavation, while ultracapacitors typically can be installed above ground in skid-mounted containers. This difference can greatly impact installation time and cost for end user. The second advantage that ultracapacitors present that they are an inherently solid state technology. The means that there are no moving parts to wear our if neglected, and so not regular maintenance to done. In contrast flywheels typically require periodic maintenance of their vacuum pumps and main shaft bearings. This maintenance adds up over the system life, and can become extra burdensome if the energy storage system are distributed over many remote location as is the case when something solar farms.
Summary And Conclusions
This paper has explored the impact of increasing levels of solar penetration on grid reliability with the use of a highly simplified grid model. Even with this model’s simplifications it is clear that unmitigated solar power will adversely affect grid reliability at penetration levels above 20%. In addition this model was used to show that only a small amount of reserve power would be needed to smooth a solar plant’s output to the point that is could easily be absorbed by the grid without any adverse effects. This model also shows that the ideal characteristics of such an energy storage system would be a calendar life of greater than 10 years, a cycle life of greater than 10,000 full cycles, and a reserve energy of between 10 and 100 seconds. Of all the available energy storage technologies that could be used in this application, ultracapacitors stand out at being the ideal match for the challenges of this application. Based on this work it is the authors opinion that solar energy and ultracapacitors together can build the base for the resilient, reliable and renewable future grid.
By: R. Shaw Lynds Sr. R&D Systems Engineer Maxwell Technologies
1) Eyer, Jim, and Garth Corey. Energy Storage for the Electricity Grid: Benefits and Market Potential Assessment Guide. Sandia National Laboratories, February 2010, Report SAND2010-0815.
2) Mills, Andrew. Short-Term Variability of Solar Power. Lawrence Berkeley National Laboratory, October 2010, Presentation at Solar Power International.
3) Power Quality Handbook. Southern California Edison Power Quality Department, 2013.
4) Lenox, Carl. Variability In A Large-Scale PV Installation. SunPower, October 2009.
We spoke with Shaw Lynds to find out more about ultracapacitors and energy storage. Listen to the full interview here.
This is very useful information