Improving Algorithmic Trading for BESS on Intraday Electricity Markets through Distributional Price Forecasts
summary
The gist
Grid-scale battery energy storage systems (BESS) are becoming increasingly important in short-term energy markets…
In short
The research extends a standard battery trading strategy by using advanced price forecasts based on the entire distribution of mid-prices, not just averages. This novel approach improves profits by up to 42% compared to traditional methods by allowing batteries to anticipate future profitable trading opportunities.
Key concepts
- Rolling Intrinsic (RI) Policy
- This is a standard trading rule that re-solves the optimal battery action after every market update. It focuses on maximizing immediate revenue based only on the current state of charge and existing contracts, often leading to short-sighted decisions.
- Empirical Cumulative Density Function (ECDF)
- Instead of using simple average prices, this method predicts the ECDF of mid-prices for the rest of the trading day. This captures the full range and dynamics of order book prices, providing a richer view for planning future trades.
- Lazy Clairvoyant Concept
- This is a strategy where batteries plan their actions based on anticipated future price movements derived from forecasts. It involves extending tradable limit orders with virtual orders priced using these forecasts to guide trading decisions forward.
Terminology used across episodes
This episode discusses
- Improving Algorithmic Trading for BESS on Intraday Electricity Markets through Distributional Price Forecasts · Paper Radio
- Battery valuation on electricity intraday markets with liquidity costs
- Coordinated Trading Strategies for Battery Storage in Reserve and Spot Markets
- The Value of Battery Energy Storage in the Continuous Intraday Market: Forecast vs. Perfect Foresight Strategies
- Rolling intrinsic for battery valuation in day-ahead and intraday markets
- Data-driven Calibration Sample Selection and Forecast Combination in Electricity Price Forecasting: An Application of the ARHNN Method
- Deep Learning for Electricity Price Forecasting: A Review of Day-Ahead, Intraday, and Balancing Electricity Markets
- The Evolution of Probabilistic Price Forecasting Techniques: A Review of the Day-Ahead, Intra-Day, and Balancing Markets
- Corrected Support Vector Regression for intraday point forecasting of prices in the continuous power market
- OrderFusion: Encoding Orderbook for End-to-End Probabilistic Intraday Electricity Price Forecasting
- Orderbook Feature Learning and Asymmetric Generalization in Intraday Electricity Markets
- Probabilistic intraday electricity price forecasting using generative machine learning
- Optimal Order Execution in Intraday Markets: Minimizing Costs in Trade Trajectories
- Online Distributional Regression
The paper
Improving Algorithmic Trading for BESS on Intraday Electricity Markets through Distributional Price Forecasts · Read on arXiv
Simon Hirsch, David Schaurecker, David Wozabal
House of Energy Markets & Finance (HEMF) · Statkraft Trading GmbH · Eidgenössische Technische Hochschule Zürich · Operations Analytics Department, Vrije Universiteit Amsterdam
Transcript
Introduction to the show: ident: Quantitative Finance. The day in quantitative finance: models of markets, risk and prices.
Ruth: We're looking at the paper: "Improving Algorithmic Trading for BESS on Intraday Electricity Markets through Distributional Price Forecasts".
Warren: Here's the gist of the paper. Grid-scale battery energy storage systems (BESS) are becoming increasingly important in short-term energy markets, where they help balance fluctuations in variable renewable generation In this paper, we propose an extension to the myopic rolling intrinsic policy that incorporates intraday price forecasts tailored specifically to the BESS trading problem Our approach addresses this gap by introducing a novel target variable derived from the limit order book and its dynamics The resulting forecasts are designed to inform the optimization about future profit opportunities while preserving computational efficiency and avoiding the burden of full stochastic optimization In a realistic high-frequency simulation setting covering 2023 and 2024, our method improves profits by up to 42% over the naive rolling intrinsic formulation We find evidence that BESS size and duration jointly influence both overall profitability and the marginal benefit of incorporating forecasts.
Ruth: Market Structure and Optimization Framework The continuous intraday market (IDC) in the European internal electricity market uses continuous trading as its settlement mechanism, which allows participants to react instantaneously to new information without waiting for auction clearing The optimization problem is formulated as a Mixed-Integer Linear Program (MILP) that defines the intrinsic value based on an initial state of charge and pre-existing contractual positions This formulation maximizes revenue by considering trading volumes subject to order fills, minimum bid-size constraints, and power constraints.
Warren: The standard rolling intrinsic (RI) policy repeatedly solves this intrinsic problem after every market update or a fixed delay.
Ruth: Forecast Integration and Target Variable Selection To correct the RI's greediness, the paper introduces a "lazy clairvoyant" concept, which plans injection and withdrawal based on future price developments The proposed method augments the RI by extending the tradable set of limit orders with virtual forecast orders assigned prices derived from empirical distribution forecasts The target variable for forecasting is the empirical cumulative density function (ECDF) of mid prices for the rest of the trading session, which is preferred over volume-weighted average prices because it captures order-level dynamics Forecasting Models and Evaluation Metrics Six different forecast models are employed to predict quantiles of the empirical distribution of midprices F, including benchmark models like Naive Day-Ahead and Naive ECDF Error metrics such as L1 and L2 errors are used, which are equivalent to the Wasserstein distance between the true and predicted ECDF Forecast stability is also evaluated by calculating the average directional change and absolute change of forecasts The analysis shows that forecast errors are negatively correlated with profits, indicating that better forecasts lead to better trading decisions.
Title and authors: Ruth: Let's talk about the title and who wrote it. It’s "Improving Algorithmic Trading for BESS on Intraday Electricity Markets through Distributional Price Forecasts" >
Warren: That sounds very technical, but honestly, what it tells me is that this paper is focused on making battery storage systems more profitable in those intraday electricity markets >
Ruth: Right. And the authors are Hirsch and his colleagues from ETH Zurich and Statkraft Trading GmbH, which gives them a strong background in both energy markets and quantitative trading >
Iris: That sounds like they're bringing expertise from a solid place on how these systems actually operate, which is good because the intraday market is where most of the action happens >
Warren: It's about taking existing strategies, like myopic or rolling intrinsic optimization and making them better for BESS trading specifically >
Ruth: Exactly. And they are trying to fix a known weakness in those older methods, which is that they don't capture what matters for BESS flexibility >
Iris: So what does "distributional price forecasts" mean in practical terms? Is it more than just guessing the average price for the next hour >
Warren: It means they are looking at the whole picture of possible prices, not just one single number, which is why they are interested in capturing order-level dynamics >
Ruth: Precisely. They're trying to model how uncertainty plays into their trading decisions when you have a storage asset like a battery involved >
Iris: That sounds like they're moving from a deterministic view to a probabilistic one, which is often necessary when dealing with variable renewable generation >
Warren: It’s about having better tools to plan injections and withdrawals based on future price developments, which is the goal here >
Ruth: So it's not just about predicting the next price; it's about understanding the shape of the uncertainty around that price prediction >
Iris: That sounds like a deeper dive into forecasting models than what we usually see in this kind of paper >
Warren: It’s an attempt to make those forecasts more actionable for real trading systems, which is where I think this work fits in >
The paper's summary: Ruth: So, to summarize the paper "Improving Algorithmic Trading for BESS on Intraday Electricity Markets through Distributional Price Forecasts," they are extending a myopic rolling intrinsic policy with intraday price forecasts tailored for the BESS trading problem >
Warren: They are essentially taking an existing optimization strategy and augmenting it with better, more specific price information derived from the limit order book dynamics >
Ruth: That information is what makes their approach distinct, as they're moving away from just using volume-weighted prices during the last hours of trading >
Iris: So, the paper essentially introduces a novel target variable based on the empirical cumulative density function of mid prices for the rest of the trading session >
Warren: That ECDF is preferred over volume-weighted average prices because it captures order-level dynamics, which is what they need for BESS trading >
Ruth: That's a key technical detail. They are planning how to use that distribution to inform their optimization about future profit opportunities while keeping computational efficiency high >
Iris: So, the main goal is to get better trading decisions without the full burden of a full stochastic optimization setup >
Warren: They are using this distributional information to guide their trading plan toward those better opportunities in a way that respects the constraints of the market structure >
Ruth: It’s about adding intelligence to an existing policy, making it more reactive to real-time price action within the order book >
The paper's improvements: Ruth: Let's talk about what they actually improved. They suggested augmenting the myopic rolling intrinsic policy by incorporating these intraday price forecasts tailored for the BESS trading problem >
Warren: So, the improvement comes from that specific integration of those forecasts derived from the limit order book dynamics >
Iris: They are introducing a novel target variable derived from this limit order book and its dynamics, which is what makes their method different >
Ruth: That variable allows the optimization to consider future profit opportunities while preserving computational efficiency and avoiding the burden of full stochastic optimization >
Warren: And they’re using this new information to guide their trading plan in a way that respects the market structure >
Iris: So, it's about making sure that we are optimizing for things like order fills and minimum bid-size constraints while still getting good profit signals >
Ruth: They also found that by incorporating these distributional forecasts, the economic benefit of the battery size and duration actually increases with both factors >
Warren: That means bigger batteries or longer durations give you a bigger advantage from this forecasting approach >
Iris: And they developed an adaptive strategy that selects optimal parameters alpha and a stop-loss parameter gamma based on cumulative profitability over a rolling calibration window >
Ruth: So, the system can dynamically adjust its risk management settings as it sees how profitable things are accumulating over time >
Warren: That's smart. They also showed that this adaptive strategy outperforms the benchmark for two and four hour batteries by up to twenty-seven percent and forty-two percent respectively >
Conclusion: Ruth: So, to wrap up, the paper "Improving Algorithmic Trading for BESS on Intraday Electricity Markets through Distributional Price Forecasts" shows that incorporating these distributional price forecasts can improve profits by up to forty-two percent over the naive rolling intrinsic formulation >
Warren: That improvement is substantial. It comes from using a novel target variable derived from the limit order book dynamics instead of just volume-weighted prices >
Ruth: And it shows that BESS size and duration jointly influence both overall profitability and the marginal benefit of incorporating those forecasts >
Iris: So, this is about getting smarter about how we plan injections and withdrawals by planning for future price developments using the empirical cumulative density function >
Warren: It’s a concrete method for improving trading decisions while keeping things computationally manageable within the market structure >
Ruth: It gives us a solid framework to consider for making our algorithms more reactive to real-time price action in those markets >
Iris: So, we've seen that distributional forecasting provides actionable insights into the shape of uncertainty around future prices >
Warren: And it’s a concrete step forward in how we approach this kind of problem computationally, moving towards incorporating these probabilistic forecasts directly into the optimization >
More episodes
- The day in conversation for 2026-11-19
- Daily Summary for 2026-11-19
- The day in conversation for 2026-11-18
- Daily Summary for 2026-11-18
- 2601.05274-On the use of case estimate and transactional payment data in neural networks for individual loss reserving
- 2009.00868-Loss-Given-Default Modeling by Post-Last Passage Time Process
- 2606.30070-Financial Resilience Evaluation: From Conditional Expectations to Dynamic Convex Risk Measures
- 2212.12687-Measuring price impact and information content of trades in a time-varying setting
- Daily Summary for 2026-11-17
- 2401.08323-Dynamic portfolio selection under generalized disappointment aversion