I’ve been writing about grid-scale energy storage since August 2008 when I published a Seeking Alpha article titled “Grid-Based Energy Storage: Birth of a Giant,” a simple primer on energy storage for the grid that’s surprisingly relevant today. A subsequent Seeking Alpha article from February 2010 titled “Grid-Based Energy Storage Represents A $200 Billion Opportunity” was my first attempt to estimate the market value of various grid-scale energy storage applications. The key takeaway from both of these early articles remains unchanged: A few high-value applications offer very attractive economics for battery-based systems but a host of low-value applications can’t offer decent returns under any circumstances.
While this article mentions Tesla (TSLA), the poster child for magical thinking in the renewable energy sector, it’s not about any particular company or subsector. My conclusions are bullish for traditional energy producers as a group and bearish for renewable energy promoters as a group. For some diversified companies like General Electric (GE), they’re bearish for one business unit, like wind turbines, and bullish for another business unit, like gas turbines. In the final analysis, this article is a thematic piece where each investor will need to decide how my conclusions should impact his personal investment strategy.
The fundamental problem of renewable power is intermittency.
I’ve always been amazed at a strange mental disconnect that’s common among renewable power advocates. On one hand, they freely acknowledge that industrial societies can’t function without stable power grids to supply electricity on demand, 24 hours a day, seven days a week, 365 days a year, and with 99.999% reliability. On the other hand, they insist that we have a moral duty to use non-dispatchable, intermittent, and generally unreliable power from renewables despite the fact that intermittency is the mortal enemy of a stable electric grid. The electrons from wind turbines and solar panels may be green and squeaky clean, but their intermittent electric current is the grid equivalent of sewage in a mountain stream.
According to the California Independent System Operator, or CAISO, the biggest challenge of managing a greener grid is maintaining a precise balance between supply and demand as the percentage of intermittent power from renewables increases. To meet the challenge, CAISO is working overtime to develop a fleet of flexible power resources with the capacity to:
- Sustain upward and downward ramp;
- Respond for a defined period of time;
- Change ramp directions quickly;
- Store energy or modify energy use;
- React quickly to meet expected operating levels;
- Start with short notice from a zero or low-electricity operating level;
- Start and stop multiple times per day; and
- Accurately forecast operating capability.
While the contract price for green electricity from a wind- or solar-farm may be cheaper than the contract price for electricity from a conventional power plant, the downstream cost of making green power stable, reliable, and useful in the electric grid can be immense, which is why electricity in states that have implemented renewable portfolio standards is often more costly than it is in states that haven’t implemented RPS programs.
Grid-scale batteries, an overview.
In the last decade, a series of megawatt-scale demonstrations proved that batteries are well suited to work smoothing the second-to-second, minute-to-minute and hour-to-hour variability of power from wind and solar systems, an application known as “short-duration renewables integration.” The demonstrations also proved that batteries can respond more quickly and accurately than conventional resources like spinning reserves and peaking plants. As a result, large battery arrays are becoming the stabilization technology of choice for short -duration renewables integration. While large battery arrays are an excellent fit for short-duration renewables integration and a few other high-value utility applications, the potential ubiquity of battery technology in grid applications and the size of the business opportunity are invariably overestimated by advocates, politicians, dreamers, and promoters who give their imaginations free rein without critically examining the details.
In January 2019, researchers from the University of California San Diego, or UCSD, published a paper titled “Combined economic and technological evaluation of battery energy storage for grid applications” in Nature Energy. While their analysis is too technical for most investors, this graphic summary of the four applications the researchers modeled is both relevant and instructive.
While grid-scale energy storage has been a research priority for over a decade, the transition from pilot scale demonstrations to commercial scale implementations is just beginning. In the last two years, Tesla has won two large contracts.
- In June 2017, it was awarded a contract to build a 100MW/129MWh grid-based battery array in South Australia. Under the original plan, the batteries would be opportunistically charged with excess wind power and the stored power would then be fed back into the grid during periods of peak demand. That application, known as “energy time shift,” is shown in the upper left hand corner of the UCSD graphic. After the battery array was installed, the operator decided it also could use the battery array for “frequency control and ancillary services,” an application that’s shown in the lower right-hand corner of the UCSD graphic. This use of one facility to provide two classes of service is called “aggregation,” and it does wonderful things for end-user economics.
- In July 2018, PG&E asked the California Public Utilities Commission to approve its plan to buy a 182.5MW/730MWh battery array from Tesla that will provide “congestion relief services,” an application that’s shown in the upper right hand corner of the UCSD graphic, together with frequency control and ancillary services. Since PG&E’s planned facility also will aggregate a pair of high-value services, it's likely to offer acceptable system performance and solid end-user economics.
While both of these projects are well suited to the performance capabilities of batteries, they’re unusual because they offer opportunities to aggregate high-value services that need rapid cycling and create several revenue events in an average day. Most installations can’t capitalize on aggregation opportunities, and without it, economics degrade rapidly. The key takeaway here is that for grid-based batteries to be a paying proposition, they must be used frequently and provide high-value services to the grid operator. Storing electricity for its power value is usually a paying proposition, but storing electricity for its energy value is usually not.
Three timeframes for evaluating renewables intermittency.
A couple of weeks ago, I downloaded data for the following column graph from CAISO’s Renewables Watch website. It shows average hourly power production from utility-scale solar during the 30-day period that began on March 17, 2019, and ended on April 15, 2019. I think the graph does a fine job of highlighting three different timeframes for evaluating renewables intermittency.
If you look at the tops of the daily column clusters, you’ll see they’re not always smooth curves like you see in solar power discussions. That’s because real-world conditions like cloud cover change from minute-to-minute and every change can impact power output. So, it’s not unusual to see situations like the zoomed box for March 24th where hourly power production climbed from zero to 9,700 MW in the early morning, fell by 1,500 MW (-15%) around mid day, climbed by 900 MW (+10%) in the afternoon, and then fell back to zero by early evening. The intra-day variability would be even more striking if I could access and graph 1-, 5- or even 10-minute intervals, but the hourly data clearly shows why CAISO needs a fleet of flexible resources to maintain grid stability as intermittent power from renewables becomes more prevalent. The ugly truth is that raw electric current from renewables is just too intermittent and too dirty for power grids in industrial societies.
Now that you understand the intra-day variability that’s the primary focus of CAISO’s flexible resource initiatives, I invite you to consider the relative heights of the daily column clusters and note the way they bounce up and down from day to day with occasional peaks above 10,000 MW, frequent peaks in the 7,000 to 10,000 MW range and one peak below 6,000 MW. For this 30-day period, the average daily solar power production was 80,800 MWh, the maximum was 104,800 MWh (130% of average), and the minimum was 40,300 MWh (50% of average). The key takeaway here is that day-to-day variability in solar power production, an issue that’s not particularly important to CAISO, is a full order of magnitude greater than intra-day variability.
Understanding the third critical timeframe requires you to focus on the first six column clusters. Power production was high on March 17th and 18th, plummeted by 60% on the 19th and 20th, and then recovered into the normal range on the 21stand 22nd. The reason for this deep multi-day trough was a simple winter storm moving through the state. The takeaway is that multi-day intermittency, which can be several times greater than day-to-day intermittency, will occur with predictable regularity and unpredictable timing.
When I downloaded my data there was nothing special about the dates. I just wanted to confirm that a one-day graph I used in an article was a typical spring day. I did not expect the data to provide clear examples of intra-day, day-to-day and multi-day intermittency. While this 30-day period is highly instructive, it would be dangerous to assume that the major swings in my small data set are as bad as it gets. I suspect that a multi-year analysis would uncover extreme examples for all three classes of intermittency. Since 99.999% reliability is a critical operational goal for electric grids and un-remediated intermittency will invariably result in power outages, capital equipment planning must focus on extremes rather than averages.
Cumulative impact of uncorrelated intermittencies.
While focusing on power output from utility-scale solar was useful for exploring the nature of intra-day, day-to-day, and multi-day intermittency, it can’t give you a good feel for the true magnitude of the problem because it excludes incremental intermittency from wind farms and residential solar. My last graph is an effort to size the magnitude of the intermittency problem that combines:
- CAISO data for utility-scale wind projects;
- CAISO data for utility-scale solar projects; and
- My estimate for small-scale solar.
To estimate average hourly power production from small-scale solar, I divided hourly power production from utility-scale solar by the installed nameplate capacity of 11,800 MW and then multiplied the product by the installed residential nameplate capacity of 7,900 MW. While my calculations probably overstate the variability of residential solar which is not as efficient as utility-scale solar, I think the error is small enough to provide a reasonable overview.
When you combine utility-scale wind, utility-scale solar and residential solar, the maximum intra-day intermittency was about 2,500 MW. For the 30-day period, the average daily power production from intermittent resources was 182,200 MWh, the maximum was 249,500 MWh (137% of average), and the minimum was 110,500 MWh (61% of average). For the two-day trough on March 19th and 20th, the cumulative shortfall was 126,684 MWh. For a grid that must be kept in balance 24 hours a day, seven days a week, 365 days a year, with 99.999% reliability, intermittency of this magnitude in an enormous challenge.
Pulling it all together
Battery-based energy storage systems for the grid are quite profitable in situations that require rapid cycling and can aggregate multiple functions in a single facility. Unfortunately, the total demand for high-value services like frequency regulation is limited. In its Q4 2018 Report on Market Issues and Performance, CAISO sized its peak frequency regulation demand for 2018 at roughly 2,000 MW. While battery arrays that are installed to support short-duration renewables integration may be able to underbid current service providers and win a share of the frequency regulation revenue, additional battery capacity that exceeds total demand is likely to drive frequency regulation prices down to levels that are currently unthinkable. It may ultimately be good for electricity consumers, but it will be very bad for facility owners.
If one looks solely at intra-day effects, the peak intermittency during my 30-day sample period was about 2,500 MW. Since Tesla’s Powerpack for commercial customers and the large PG&E project discussed in this article both have power to energy ratios of 0.25, it seems reasonable to assume that remediating a peak intra-day intermittency of 2,500 MW would require battery arrays with a combined capacity of 2,500 MW / 10,000 MWh that cost about $4 billion. If battery arrays were used to remediate 100% of the intra-day intermittency, some of the systems would be in a position to aggregate short duration renewables integration with other grid services and the rest would have to settle for a single revenue stream. The systems with multiple revenue streams would probably offer solid economic returns for their owners, but the systems that could only garner a single revenue stream would probably offer marginal returns. Therefore, in the real world, it’s unlikely that battery systems will be deployed to serve 100% of the theoretical demand for intra-day intermittency remediation.
When the analysis progresses from intra-day intermittency to day-to-day intermittency, power becomes less of an issue and total energy capacity takes center stage. In my 30-day example, the maximum power production was 67,300 MWh greater than average and the minimum was 71,700 MWh less than average. The average daily deviation from the mean was 24,200 MWh. While remediating the day-to-day intermittency during my sample period would require battery arrays with an incremental capacity of 61,700 MWh that cost about $25 billion and weigh 500,000 tonnes, the batteries would only generate the equivalent of 10 revenue events per month as opposed to several revenue events per day for an optimized intra-day system. Those economics can’t work without a couple of billion dollars a year in standby charges for reserve capacity. While it’s a meaningless gee-whiz statistic, a 61,700 MWh wall of Tesla Powerpacks would extend for 158 miles, more than enough to fence the 140-mile California-Mexico border.
Where using 61,700 MWh of batteries to remediate day-to-day intermittency would be prohibitively expensive, using an additional 55,000 MWh of batteries to remediate predictable multi-day intermittency would be preposterous because the incremental batteries would only generate a few revenue events per year. While some might argue that doubling the height of the border wall could be a good thing, it’s no way to provide plentiful, cheap and reliable electricity for all.
It’s easy to read about the wonders of grid-based storage and imagine a world where a combination of renewables plus storage makes fossil fuels a thing of the past. However, it’s impossible to map a path to that world that doesn’t drive electricity costs to stratospheric levels. Since the law of economic gravity says that the cheapest solution wins, investors are well advised to check their imaginations at the door when discussing the future of energy storage on the electric grid. There simply is no credible scenario where renewables plus storage will replace fossil fuels and relegate conventional power systems to the ash-heap of history.
Disclosure: I am/we are short TSLA TROUGH LONG DATED PUT OPTIONS. I wrote this article myself, and it expresses my own opinions. I am not receiving compensation for it (other than from Seeking Alpha). I have no business relationship with any company whose stock is mentioned in this article.