Unraveling the complexity of merchant energy storage projects
This blog post is part 1 in a 2-part series on merchant storage.
You know an industry is maturing when developers discuss the merchant market not as a source of upside revenue, but as the only source of revenue. Power plants have been designed, financed, built, and operated, significantly relying on wholesale revenue since the late 90’s; so why is it so difficult to replicate this strategy for energy storage?[1] Typical power plants can turn a profit if their average costs of fuel plus operations and maintenance (O&M) are lower than the average price at which they can sell electricity. Simply put, higher electricity prices tend to result in higher profits. However, average electricity prices matter little to energy storage project operators. The volatility of prices is what matters for storage profitability,[2] and volatility is expected to increase alongside the growing penetration of wind and solar capacity in many grids around the world.[3]
At DNV we are increasingly being asked to model and evaluate the 20 year revenue potential of storage assets in wholesale markets. Our experts discuss approaches to optimization modeling, forecasting, managing degradation, planning augmentation, implementing scenario analysis, dealing with uncertainty, warrantying throughput, and the inimitable value stacking. Inevitably, we are asked if this revenue stream can be financed. This is when it gets complicated.
While we only focus on a few in this post, there are many dimensions we can use to compare and contrast typical fossil fuel power plants to energy storage projects operating in merchant markets, as shown in Table 1.
Table 1 – Comparison between power plants and energy storage operating in merchant markets*
* Note that the table is highly generalized for each dimension to illustrate differences. There are exceptions and outliers that may contrast with our assumptions listed here. This table focuses on the dimensions that impact profitability and intentionally avoids qualitative comparisons by ignoring aspects such as environmental benefits or future grid needs. We do not dive into detail on each dimension in this post, but leave open the opportunity to discuss the other topics in future posts. Please provide feedback in the comments below on topics you would like covered. + Focus of blog post
Energy Sources and Flows
Despite the lengthy list above, the most impactful difference to understand storage financing compared to fossil fuel power plants lies in the energy source (fuel) for each asset type. Power plants generate electricity by converting natural resources into electricity while storage recycles electricity, taking a low value product at one time of day and turning it into a higher value product at another (i.e., arbitrage), incurring losses along the way. The energy flow of power plants is linear, whereas the flow for storage is U-shaped (see Figure 1). Typical fossil fuel power plants purchase an energy-dense fuel from one commodity market and convert it into electricity for sale in another commodity market (the electricity market). The energy source for these power plants is functionally limitless.[4] Storage projects buy and sell the same commodity (electricity) in the same market. Storage must therefore leverage intra-hour and intra-day price variations.[5] Energy storage is further limited by its capacity. Even though its electrical energy source is functionally limitless, storage can charge only until full and discharge until empty – and the electricity flows only one way at a time.
Figure 1 – Energy flows for power plants and energy storage projects
Forecast Everything – Believe Nothing
Lenders have become comfortable financing new power plants because owners can often sign contracts or hedge long-term fuel supplies which allows them to lock in their fuel costs for an extended period. A financial model may evaluate cash flows based on 20 years of monthly natural gas prices compared to the forecast of 20 years of average monthly electricity prices. Once assumptions for construction and O&M costs, as well as efficiency losses, are incorporated, the model can churn out the project’s expected return. In this highly simplified example, the financial analyst only needs 240 forecasted inputs each for fuel and electricity (20 years x 12 months/year), see Figure 2. This is no easy feat, but it pales in comparison to the hundreds of thousands of prices needed to model a 20 year storage project with hourly price forecasts.
Figure 2 – Monthly future natural gas prices (left) for power plant revenue modeling and hourly forecasted wholesale electricity prices (right) for storage revenue modeling, 2020 – 2032
At the most basic level, to maximize revenue, storage projects must charge during the lowest priced hours and sell during the highest priced hours each day. But storage project operators do not know prices with certainty in advance of committing to provide electricity each day. Prices must be forecasted, and any reliable forecast has uncertainty, also described as forecast error. Typical power plants plan dispatch by bidding the price at which they expect to sell electricity based on known costs (of fuel and O&M). Storage projects must plan dispatch based on forecasted costs (of electricity) for charging and forecasted prices (also of electricity) for discharging. Typical power plant operators know and control the cost side of the equation and forecast the revenue, while storage operators must forecast both.
Forecast error for storage comes in two flavors, both of which need consideration in financial modeling. There is the obvious uncertainty associated with forecasting hourly prices over a 20 year period – no one can get this exactly right. But the subtle uncertainty lies in understanding the ability of storage to dispatch each day to capture the highest and lowest prices. For example, a storage operator may forecast the lowest priced hours tomorrow to be from 2 to 6 am and the highest to be from 4 to 8 pm, and set the storage dispatch schedule (or bid prices) accordingly (i.e., charge from 2 to 6 am, discharge from 4 to 8 pm). Yet, the lowest prices may actually occur from 1 to 5 am and the highest may occur from 6 to 10 pm. The operator’s dispatch will be less than that which would have maximized revenue due to day-ahead forecast error. Power plants are not impacted by this uncertainty because they do not have to forecast their daily fuel costs and they are not energy limited and can therefore operate during all high price hours rather than attempting to target only the top few.
Operational Strategies
The recursive nature of storage charging and discharging, coupled with its energy capacity limitation, results in a dramatically different operational strategy compared to fuel-powered generation. Typical power plants can operate for long periods of time by generating and selling electricity if the prices are high enough – profiting more in some hours than others, but nonetheless profiting. Storage operators need to decide, for each time interval, when to charge and discharge, at what rate, and for how long. There is an opportunity cost for each decision to charge or discharge, and an opportunity cost for each decision not to charge or discharge. Storage operators need to constantly choose between making $X in this hour with 90% probability, or holding out for a little while longer to make perhaps $1.5X in a later hour with a 65% probability. In practice, energy storage projects often seek to maximize revenue from the higher volatility that tends to appear in sub-hourly (5 minute and 15 minute) electricity prices rather than hourly.
Let’s recap: Storage earns revenue from opportunistically buying and selling electricity at hourly or sub-hourly intervals while fossil fuel power plants sell electricity at a price higher than the costs to generate it. Power plants use average price forecasts while storage prices need to be forecasted at every interval over 20 years for a financial model. Power plant revenue can often be calculated in a spreadsheet. Storage price forecasts need to be input into software that can simulate storage dispatch taking into account imperfect daily dispatch (see Figure 3). This is what makes merchant storage complicated.
Figure 3: Illustrative inputs into cash flow models for power plants and storage projects
Read part 2 of this series: 7 Lessons Learned from Merchant Energy Storage Projects. You can also explore more detail about our modeling capabilities by viewing our recent webinar, Merchant storage projects - Balancing bidding strategy with storage health.
To learn more about merchant storage, listen to our Norton Rose Fulbright Currents Podcast: Merchant storage: A new frontier.
If you have any questions or comments, please feel free to reach out to us here. Read Part II of the blog post here.
[3] https://emp.lbl.gov/publications/impacts-high-variable-renewable