Development of a Machine Learning System for Power and Internal Consumption Prediction and Degradation Assessment in Solar-Vanadium Redox Flow Battery (VRFB) Hybrid Energy Systems

Kurzfassung

The increasing demand for sustainable energy has driven the adoption of renewable energy systems, particularly solar power, but its intermittent nature creates challenges in energy storage and system stability. Vanadium Redox Flow Batteries (VRFBs) offer a promising solution due to their long cycle life and scalability, while machine learning enables accurate modeling of complex system behavior. This project develops a machine learning-based predictive framework for VRFB degradation assessment, consisting of three models: solar power prediction, internal energy consumption prediction, and VRFB degradation prediction. Due to limited real experimental data, a simulation-based Approach was adopted using solar data from PVGIS and MATLAB/Simulink to generate datasets for model training. The generated data were integrated into an Energy Management System (EMS) to manage energy distribution among the load, Li-ion battery, and VRFB, and the system outputs were used to predict degradation under different operating conditions. The proposed framework reduces reliance on costly and time-consuming experiments, supports maintenance planning through early detection of degradation, and improves energy allocation strategies, while also enhancing system efficiency and stability, with the potential for further development using real-world data and advanced control strategies to extend VRFB lifetime and optimize overall performance.