Predictive Thermal Management of an Industrial Battery Energy Storage System

The paper deals with the thermal management problem of an industrial battery energy storage system (BESS). To meet the demands of maintaining battery temperature in a suitable thermal range and ensure economical operation, we formulate the model predictive controller (MPC) using a linear model of BESS obtained from real-time data. Since the BESS heating, ventilation, and air conditioning (HVAC) system’s actuators operate only in two stages (on/off), we are dealing with a non-convex optimization problem. Additionally, the control actions associated with HVAC are physically not allowed to be simultaneously active, leading to nonlinear constraints. We employ the MPC with binary inputs to overcome this issue and compare it to relaxed convex MPC with a conditional switcher. The results show that both methods can keep the temperature in predefined intervals considering an appropriate choice of the prediction horizon. However, binary formulation produces less energy-demanding operation.