By Manling Hu and Jie Ying
High voltage electricity towers. Image by AdobeStock
The research is supervised by Arthur van Benthem, Professor of Business Economics and Public Policy at Wharton, focusing on understanding how battery energy storage systems (BESS) operate in electricity markets, specifically within PJM.
Power markets and battery storage
To most people, electricity is invisible – it is simply available at the flip of a switch. Behind the scenes, however, the grid requires constant coordination to keep supply and demand balanced. In the U.S., this work is handled by Regional Transmission Organizations (RTOs) and Independent System Operators (ISOs), which run wholesale electricity markets and maintain system reliability.
This research focuses on PJM Interconnection, which serves more than 65 million people across 13 states and Washington, D.C. Within PJM, generators and storage resources participate in multiple markets. Two of the most important for battery storage are:
- Energy markets, where electricity is bought and sold in advance through the day-ahead market and then adjusted closer to real time in the real-time energy market.
- Ancillary service markets, which pay resources to provide reliability services that keep the grid operating smoothly.
Jie Ying is a Master of Environmental Studies Candidate at Penn.
One key ancillary service is frequency regulation. The power grid operates at a constant frequency of 60 hertz, and even small imbalances between supply and demand can cause that frequency to drift. Frequency regulation resources respond within seconds to correct these imbalances. Batteries are especially well-suited for this service because they can change output almost instantly. They submit regulation offers, specifying how much fast-response capacity they can provide and how accurately they can follow the grid operator’s rapid instructions, called regulation signals, which are updated every 2-4 seconds.
Manling Hu is a Master of Environmental Studies Candidate at Penn
As a result, battery energy storage systems can earn revenue in both markets—buying and selling electricity in energy markets (energy arbitrage) while also providing fast reliability services like frequency regulation. But how a battery participates is shaped by real physical and economic limits. State of charge (SOC) is the battery’s fuel gauge, indicating how much energy is available at any time; power limits determine how quickly it can charge or discharge; and battery degradation captures gradual wear from repeated cycling and calendar aging, which affects both performance and project economics.
Turning classroom questions into a research project
Hu and Ying’s collaboration began in Spring 2025 in Professor van Benthem’s Energy Markets and Policy course. Through the group discussions and team assignments, they discovered a shared curiosity about how power markets actually function and how renewable energy projects translate market signals into revenue.
Those questions grew into a research idea: could the tradeoffs between revenue capture and operational constraints be modeled systematically to better understand how batteries should participate in PJM’s markets? When EII released a call for proposals through its PPA Renewable Energy Research Program, they seized the opportunity to turn that idea into an applied research project.
Modeling battery participation with data and Python
Like real power market participants, Hu and Ying began by working through PJM’s market manuals, the rulebooks that define how resources, such as power plants or batteries, submit offers into markets, how performance (how quickly and accurately a resource responds to PJM signals) is scored, and how payments are settled. This step was essential for translating market rules into model inputs.
They then conducted a targeted literature review, reading how other researchers have modeled battery bidding and dispatch under uncertainty in market signals, operational constraints, and battery degradation. They also spoke with industry experts to sanity-check assumptions about battery operations.
Day-ahead optimization results showing the baseline charging schedule, regulation capacity offers, and the resulting state-of-charge trajectory for the battery energy storage system under PJM market conditions.
With that foundation in place, Hu and Ying built their framework in Python, combining data processing and optimization into a single workflow. The model follows a two-timescale structure:
- In the day-ahead stage, a two-stage stochastic optimization decides both the battery’s energy market offer (how much electricity to sell at what price) and its regulation market offer (how much fast-response capacity to commit to follow PJM’s regulation signals)
- In the hour-ahead stage, a rolling-window method updates decisions as real-time conditions evolve, ensuring feasibility as the battery’s SOC and market signal changes.
Throughout, the model enforces real-world constraints such as power limits, SOC boundaries, battery degradation, and PJM’s fast regulation (RegD) dynamics.
They then tested the framework using PJM market data, simulating a full day of dispatch and revenue across both energy and ancillary services. The results show that the battery earned most of its revenue by providing fast frequency regulation, responding to grid signals in seconds, while traditional energy arbitrage played a smaller supporting role. They also found that updating decisions hour-by-hour with a rolling look-ahead can achieve performance close to an unrealistic “perfect foresight” strategy without requiring knowledge of future market conditions.
As battery storage continues to scale across regional grids, Hu and Ying hope this work can serve as a practical reference for students interested in energy research and for power market participants seeking to better understand how storage resources interact with electricity markets.