Optimizing BESS Operations with Hybrid Modeling and Machine Learning

Stationary Battery Energy Storage Systems (BESS) play a crucial role in providing reliable and cost-saving use of grid power, while also normalising integration of cheaper and greener renewable energy sources. Current industrial energy management systems (EMS) for a BESS rely on rule-based control or traditional optimization methods, such as linear programming. A model based control may be able to provide a better optimal solution. However, the complexities of fluctuating energy prices, diurnal variation of renewable energy sources, unpredictable load demands, and the dynamic performance characteristics of batteries makes deriving a model a difficult task to discover and implement. This leads to suboptimal BESS utilization, resulting in increased energy costs and reduced battery lifespan. This project aims to develop a hybrid approach, involving physics based battery models, with data-driven machine learning techniques to optimise the BESS charging and discharging schedules. Historical data on the load demand, energy prices, weather, production schedules, etc. will be used to accurately forecast short-term demand and tune the BESS parameters. The overall optimisation objective being to minimise costs, maximise energy efficiency, and ensure operational safety and reliability.