From Prediction to Profit: Integrating ML-Based EPEX SPOT Price Forecasts with BESS Optimization

Abstract

The transition towards a decarbonized energy system is reshaping the structure and dynamics of electricity markets. Renewable energy sources such as wind and solar now account for a growing share of supply, but their variability intensifies price volatility and uncertainty. This, in turn, increases the demand for flexible resources that can provide system stability and enable participation in short-term markets. Large-scale Battery Energy Storage Systems (BESS) are central to this development: they are no longer purely technical assets but active market participants, whose intelligent operation can simultaneously maximize economic returns and contribute to system reliability. Their effective deployment, however, depends on accurate forecasting, optimization frameworks that embed both technical and market constraints, and algorithms that translate these capabilities into actionable trading strategies. This thesis presents a modular Python-based prototype that combines machine learning–based electricity price forecasting with optimisation-based scheduling of BESS under realistic technical and regulatory constraints. By linking a forecasting engine with advanced optimization routines and an interactive user interface, the system supports both methodological experimentation and practice-oriented decision-making. Applied to historical and predicted EPEX SPOT data, the prototype enables scenario-driven analyses such as capacity scaling and degradation sensitivity. These experiments demonstrate how forecast accuracy, operational limits, and degradation costs shape the economic value of storage assets. The prototype thus serves both as a research testbed for methodological development and as a decision-support tool for storage operators and electricity traders evaluating storagebased trading strategies under realistic market constraints.

Date of Award	2025
Original language	English
Supervisor	Rainhard Dieter Findling (Supervisor)