Electric Vehicle Presence Prediction for Model Predictive Control in Home Energy Management Systems

Home Energy Management Systems (HEMS) are pivotal to reduce energy consumption and improve energetic self-sufficiency of households equipped with photovoltaic (PV) and energy storage systems. However, the majority of such devices considers electric vehicles (EV) as additional loads, without exploiting the energy storage provided by the vehicle battery. This is mainly due to the inherent variability in the vehicle’s presence at home, posing a challenge for accurate forecasting. To embrace the EV potential in the HEMS, in this paper it is proposed a predictive algorithm for EV presence that can be integrated into HEMS based on Model Predictive Control (MPC). The realized algorithm estimates the EV presence at home and its battery state of charge (SoC) when it returns home. This algorithm is composed of two parts: a frequency matrix and a Semi-Markov model. The former is trained to identify the most common behavior associated with the house’s occupancy, while the latter assists the algorithm in adjusting predictions when they deviate from the actual observations. The time step, optimizer, and constraints considered in the algorithm are described. Through the proposed algorithm, the HEMS can prepare the EV before leaving home, with the optimal SoC.