Data Driven Modeling and Simulation for Energy Storage Systems

A widely used choice for delivering and storing energy on demand in many modern electrical systems is that of rechargeable batteries. When assembled in packs, application of such batteries takes the form of energy storage systems in a variety of configurations, such as a microgrid, which play a key-enabling role in future energy solutions by integrating distributed renewable energy sources and storage systems. In the interest of using batteries more effectively, their behavior under various operating conditions must be understood. This paper describes a framework for estimating the terminal voltage of a lead-acid battery at different working temperatures using feed forward and recurrent neural network techniques for use in large-scale microgrid applications. The methodology is developed for Li-ion batteries and then extended to lead-acid batteries, where simulation results are compared against extensive laboratory tests performed at varied conditions found in photovoltaic and electric vehicle charging applications. Utilization of the resulting energy storage system model is illustrated in a large-scale simulation for a microgrid application.