Dr. Allison Weis directs Ascend’s battery storage team and provides technical guidance on a variety of optimization analytics products for renewable integration, flexible grid resources, and system planning. She manages valuation and real-time dispatch optimization for storage and renewable systems operating in wholesale markets. She also leads work on hydro system optimization for clients facing different system constraints and goals and development of planning analytics for optimal resource selection across transmission connected, distribution connected, and behind-the-meter resources. She brings her understanding of flexible grid resources from her work modeling the integration of wind and solar with batteries and electric vehicle charging at Tesla, and in her PhD from Carnegie Mellon prior to joining Ascend.See All Posts
Battery storage is entering the electricity scene at an extraordinary pace. It seems the entire energy industry is looking closely at the technologies, economics, and financing hurdles to stand-alone storage options and renewable energy pairing opportunities.
There are four “S” factors critical to battery profitability: Siting, Sizing, Stacking, and (Bid) Strategy. In this piece, we break down those four elements and offer the development and investment community a starting point for due diligence analysis.
When bankrobber Willie Sutton was asked why he robbed banks, he famously responded, “that’s where the money is.” For battery storage, the money is in price spikes. Batteries operate as physical arbitrage machines to the volatile real-time energy market. Buy low, sell high. Transmission congestion and high load requirements have always been important contributors to price volatility. Now, variable generation adds a further multiplying effect, increasing the impacts of forecast error, creating a perfect environment for highly volatile market prices.
As the California Energy Imbalance Market (EIM) expands to a large portion of the western U.S., nodal volatility has become increasingly more important for those new regions trading into the CAISO system. Of course, the myriad of market forces (e.g., transmission expansion, behind-the-meter generation, vehicle electrification and new loads) will cause continual change in supply and demand relationships. But as renewables, with their difficult-to-forecast ramping behavior, become the dominant supply resources serving load, we believe price volatility, reflecting the fundamental need for flexible generation, will be with us for a long time.
Figure 2 is a “heat map” of the California nodal system based on average 2015-2017 congestion persistence, a measure of price volatility. We see congestion-related volatility surrounding the urban areas, as expected, but also in interior sections of the state with significant renewable resource supplies and reduced access to loads. Nodal price analysis within supply or demand-constrained areas represents a valuable initial screen, but still needs to be combined with land cost, siting, and other considerations to find an optimal location.
Battery sizing can be confusing and often misinterpreted. Power output and duration are generally multiplied together to calculate the energy capacity. However, a 5 MW battery capable of 4-hour duration (representing 20 MWhs) will have very different economics than a battery sized to 20 MW, 1-hour duration or a 40 MW, 30-minute duration asset.
The critical factor when targeting the volatile real-time market is in the quantity and duration of price spikes, both positive and negative. Most price spikes only persist for short bursts as real-time markets find equilibrium relatively quickly.
Regional market rules are also a primary input. California has two real time markets, a 5- and 15-minute. In California, the two RT markets are only modestly correlated. Volatility in one does not portend volatility in the second. In 2018, for example, spikes in the 5 and 15 minutes RT markets at the SP15 node overlapped only 22% of the time.
There are 4 primary sandboxes a stand-alone or paired battery system can play in, represented by a 2×2 matrix: Day-Ahead and Real-Time scheduling cross-referenced with Energy and Ancillary Service markets. Ancillaries are further broken down into smaller categories. In the CAISO, for example, four ancillary products markets are established: regulation up, regulation down, spinning reserve and non-spinning reserve. Battery investment should be based on strategies to maximize revenue per increment of battery capacity, representing an analytic integration of battery location, sizing and bid strategy.
However, optimization is complex – the reason for the arbitrage opportunity, forecast error, translates to lack of perfect foresight. Nonetheless, advancements in machine learning have led to improved prediction in price spike probability. Meaning: good analytics and data capture can increase your odds of selecting the most profitable market to participate in.
Figure 4 offers insight to an analysis we conducted for a client in the California market. We found, in this case, a shorter duration battery maximizes revenue in both regulation and energy markets (including both very low and negative prices for charging and very high prices for discharging) per unit of capacity installed.
While each battery investment is unique, our fundamental analysis indicates that RT energy market revenues will be the most valuable component of the revenue stack. While ancillary service prices are expected by some to increase across markets as renewable energy penetration progresses, as recently predicted by Lawrence Berkeley Labs, the volume represented in the ancillary service market is very thin relative to the energy market. Ancillary markets can become quickly saturated, leaving little opportunity for new flexible resources. Instead, our models show battery value will be maximized by capturing negative (or very low) prices to charge and erratic price spikes in RT markets to discharge, creating a powerful arbitrage opportunity that will be sustained over time due to the much larger volume of forecast error driving RT volatility.
Once the battery investment is complete, it’s time to monetize the available revenue sources. Bid strategy will change seasonally and even week to week. For example, in spring and fall months, battery profit may be tied to soaking up negative or very low prices due to forecast error in the quantity of realized solar production. In contrast, July might see very few low or negative prices but far more upside volatility during the complex evening ramp as solar production declines and residential load obligations kick in.
Profit maximization will be tied to two primary factors: weather-derived, short-term (1-2 day) forecasting of price volatility, and continual evaluation of real-time market conditions and current battery state of charge to optimize next-day and next-hour energy and ancillary bids.
Batteries are the perfect tool to capture price volatility sure to come from our increasing reliance on variable energy resources. But good analytics will be the instruction set to use these tools to maximum effectiveness.
Mike Mendelsohn serves as manager of marketing & analytics at Ascend Analytics. Dr. Alison Weis is the company’s manager of optimization analysis.
Ascend Analytics is an innovative software service company focused on energy analytics. Energy portfolios and markets have increased in complexity, making decision analysis more difficult. Ascend's solutions provide the core analytic infrastructure to streamline processes, enhance understanding, and support decisions.