Battery Optimization in Microgrids using Markov Decision Process Integrated with Load and Solar forecasting

Rising climatic concerns call for unconventiona/renewable energy sources which reduce the carbon footprint. Microgrids with battery systems that integrate a variety of renewable energy resources play a key role in utilizing these energies in a more efficient and environmentally friendly manner. This paper presents a framework based on Markov Decision Process (MDP) integrated with load and solar forecasting to derive an optimal charging/discharging action of battery with rolling horizon implementation. The load forecasting is implemented using time series model and PV output forecasting is implemented using regression model. The control algorithm is developed to reduce the monthly billing cost by reducing the peak load demand while also maintaining the state of charge of the battery. The presented work simulates the control algorithm for one month based on historic load and solar data. A simulation covering one month yielded results showing that a microgrid with battery bank controlled by the MDP algorithm reduces the maximum load demand by 23.3%, leading to a cost saving of 33.1%.