At the Trust, yet verify blog, Michel has written a couple of posts about the Hornsdale Power Reserve. I had intended to do a post on this energy storage facility for a while and commented that I was planning to do a post but hadn’t gotten around to it. When I said would not have to produce a post Michel said his was only one way to look at it and there are other possible views. After reading the second post I decided to make a point about this system as it relates to New York State energy policy.
According to the Hornsdale Power Reserve website “At 100MW/129MWh, the Hornsdale Power Reserve is the largest lithium-ion battery in the world, and provides network security services to South Australian electricity consumers in concert with the South Australian Government and the Australian Energy Market Operator (AEMO). The Hornsdale Power Reserve is a facility comprising of a 100MW/129MWh Tesla Powerpack system located approximately 15km north of Jamestown in South Australia.”
In the first post Michel addressed the claim made by renewable energy advocates that batteries can replace natural gas for peaking and gap filling. The advocates point to Hornsdale profits as proof of this claim. However, he showed that the facility was making most of its money providing Frequency Control Ancillary Services. There is great value to the electric grid having a way to quickly address variations in grid frequency and this system does that well. However, peaking requirements are different and there is no sign that this small a system can provide much value for that need.
In the second post he responded to a heated discussion on a reblog of the first post on the blog “Utopia, you are standing in it!“ An advocate for renewable energy claimed that battery systems like Hornsdale can compensate for intermittency problems. In this particular case, the intermittency from break downs at fossil-fired power plants was the cited example. Michel concluded that “A power plant with a capacity of 1,480 to 2,200 MW breaking down is by no means a “minor” event and if a battery system with a capacity of 55 MW / 80 MWh (or even a 100 MW / 129 MWh in case of the Hornsdale Power Reserve) is able to compensate for such an event, then it most probably wasn’t a “break down””.
In this post I want to look at the use of energy storage batteries for short-term fluctuations of renewable energy resources at an example solar facility. On March 17, 2020 North Park Energy submitted their Public Involvement Program Plan as the first step of the New York State electric generation permitting process. Their proposed Declaration Energy Center Project in Seneca County is proposed to have a generating capacity of 450 MW covering a project area of ~4,400 acres of land of which ~ 2,500 to 3,000 acres will be used for the solar energy center.
In this post I want to show that there is another aspect of the intermittency problem. To date I have focused on intermittency energy storage when wind and solar resources are low and must be replaced. In particular, those times when the wind does not blow at night. The problem I want to address here is short-term intermittency when solar is affected by variable clouds.
In order to determine how solar electric generation could vary over short periods I used short-term meteorological data from the NYS Mesonet meteorological system. The NYS mesonet is a network of 126 weather observing sites across New York State. The official website of the Mesonet includes a tab for live data that brings up station information for the 126 operating individual sites that shows that available data include wind direction and speed, temperature at two levels, relative humidity, precipitation, pressure, solar radiation, snow depth, and camera images. For this analysis I obtained 5-minute archived meteorological data for the Rush and York Mesonet stations near Caledonia, NY. I previously used these data to evaluate the output from a proposed 180 MW solar facility nearby but will assume that these data also represent conditions that could be expected at the Declaration Solar 450 MW project.
I analyzed the five-minute solar insolation (watts per meter squared) to determine how the electric generation output from photovoltaic solar panels would vary. I only looked at five-minute periods when the solar insolation was greater than zero for a fifteen-day period (12 July through July 26 2017). The Summary Solar Insolation Statistics for Rush and York page lists statistics for different parameters in three tables. I list the standard descriptive statistics for both stations in the first table. There are over 2,700 five-minute periods when there was solar insolation greater than zero in this period. The average insolation was around 300 watts per meter squared and the maximum insolation was just over 1,000.
In order to assess solar variability effects, I calculated the absolute difference between successive observations. In the second table I list the summary statistics. While the average difference is around 50 watts per meter squared the standard deviations are 1.7 times greater suggesting that there is a lot of variability. Importantly, the maximum difference between successive periods is 702 watts per meter squared at Rush and 675 at York. The maximum solar output difference divided by the maximum solar radiation represents the maximum variability. Any solar facility near Rush could generate up to a change of 68.9% of the maximum output in successive five-minute periods and at York the maximum change would be 62.4%. The last table shows the frequency distribution of the five-minute absolute difference percentage change. Note that 1% of the time solar generation output varies 40% or more.
The solar variability analysis provides an overview of the data for one example period. The archive for the Mesonet includes daily meteograms or graphs of different weather parameter which provides an overall description of the day. The reports for the July 18, 2017 solar radiation example are available at these links: Rush and York but I have also provided the Rush and York 18 July 2017 meteograms which combines the two. (Note that the hours listed are UTC times so you need to subtract five hours to get to Eastern Daylight Time.) During the proceeding night temperatures dropped enough that in the morning there was very high humidity and possibly fog with light winds so I don’t think there were clouds. Once the sun came out the temperature rose quickly but note the variability throughout the day.
The Solar Variation in Five-Minute Meteorological Data at the Rush and York NYS Mesonet Stations table lists all the meteorological data available from the Mesonet five-minute data archive and the calculated absolute solar insolation difference between five-minute periods. This table highlights cells where the difference between successive 5-minute periods is greater than half the maximum observed solar radiation in the Insolation % of Max Delta column. Note that during this 8-hour period that value was exceeded seven times at Rush and six times at York.
The Solar Electric Output Variation in Five-Minute Meteorological Data at the Rush and York NYS Mesonet Stations table lists estimated power output based on two testing conditions used to rate the output of solar photovoltaic modules for a more limited period than the previous table. Although there is a temperature factor that should be included to increase accuracy, I am only going to consider the effect of insolation. In this instance ambient temperatures were close to the testing temperature of 20 deg C so it probably is not much or a factor. The first test condition is Photovoltaics for Utility Scale Applications (PV-USA) and that determines maximum output when the insolation equals 1,000 watts per square meter. The second condition is normal operating cell temperature (NOCT) which rates maximum output when the insolation equals 800 watts per square meter. The Declaration Energy Center Project is proposed to have a generating capacity of 450 MW. My naïve formula for solar output was simply the observed input solar insolation times solar PV capacity (450 MW) divided by the test condition insolation.
Using Rush meteorological conditions, the largest 5-minute change occurs from 1245 to 1250 EDT when the output changes 66% or 301 MW using PV-USA or 376 MW using NOCT. It might be more of problem for the York data at 1325 EDT when the output went down 59% but then back in the next five-minute period 52%. That is a PV-USA load swing of 287 MW and 256 MW or a NOCT load swing of 359 MW and 320 MW.
That is concerning enough but the fact is that the partly cloudy weather on July 18, 2017 commonly covers a large area and it would not surprise me in the least that most of New York was in these conditions. As of March 17, 2020 there are 36 solar facilities totaling 5,759 MW and another four solar plus storage facilities totaling 740 MW currently in the New York State active permitting queue. Let’s say that half of the solar-only facilities were in the same weather pattern. In that case I expect that the solar output would be jumping around by hundreds of MW every five minutes and in the worst case by a couple of thousand MW. This would be large short-term variability for the grid to handle and remember that I showed earlier that 1% of the time solar generation output varies 40% or more between five-minute periods.
The obvious solution is to require solar facilities to have an energy storage battery that could buffer the output from the solar panels to the grid. Instead of tying the solar facility directly to the grid the output would go to a battery that should be able to provide power with less fluctuations.
This post adds another caveat to the claims of renewable energy advocates. The all-renewable electric grid has to rely on energy storage. At first glance energy storage only appears necessary to provide power when renewable resources are not available for hours or days. However, the requirements are more complicated and nuanced than that.
In previous posts at the Trust, yet verify blog, it was shown that energy storage systems such as the Hornsdale Power Reserve cannot provide peaking support or compensate for intermittency problems at larger power plants unless they are much larger. On the other hand, Hornsdale is making money providing Frequency Control Ancillary Services.
In my previous energy storage work I considered energy storage with respect to renewable energy resources to determine costs. In order to rely on renewable energy, stored energy has to be provided at night when the wind is not blowing. This article shows that in addition to the longer-term storage problem, there is a short-term renewable output fluctuation problem that we may have to rely on battery energy storage to resolve.
There is no question that energy storage systems like the Hornsdale Power Reserve can provide value to the electric grid. However, we have shown that increased renewable resources will cause multiple problems that have to be resolved. On one hand, it is clear that much larger battery systems will be needed for the more obvious requirements but it is not clear to me that those systems could also be used to simultaneously resolve the other problems described. I suspect that batteries operated to provide Frequency Control Ancillary Services or to address the five-minute fluctuations issue have to be operated differently because in both instances they have to react to increases and decreases in power. Therefore, the batteries would not be charged to the maximum level. On the other hand, if the primary purpose is energy storage then you need to keep them charged as much as possible.
Renewable advocates are fond of saying that wind and solar development costs are cheaper than fossil-fired alternatives. However, when you start adding the costs to make those intermittent sources available as needed that argument fails because of the energy storage issues described here.