A Day-Ahead Forecasting Model for Probabilistic EV Charging Loads at Business Premises

Abstract

Focusing on every individual electric vehicle (EV) while optimally charging a significant number of EV units at the workplace is normally computationally burdensome. Such charging optimization requires not only a long runtime but also a large CPU memory due to numerous decision variables involved. This paper develops a new combined state of charge (SOC) based methodology to calculate day-ahead combined probabilistic charging loads for a large number of EV units. Here, several models are proposed to estimate different combined statistical parameters based on historical data. The proposed methodology determines the transition of the combined SOC distribution of EV units from one timeslot to the next using these estimated parameters. Various strategies of SOC-based charging (e.g., unfair and fair modes) are investigated to control EV loads. Numerical results show that the proposed SOC-based charging can reduce the number of decision variables significantly, and require less computational time and memory accordingly.