Peak shaving and cost minimization using model predictive control for uni- and bi-directional charging of electric vehicles

Abstract

Uni- and bi-directional electric vehicle charging schedulers are key enablers in the mitigation of the negative impacts, such as peak powers, of electric vehicles and renewable energy sources on the grid. This paper presents two electric vehicle charging schedulers based on model predictive control algorithms for four different charging strategies to minimize electricity bills and peak powers for a local energy system. The schedulers use forecast values for photovoltaic production, load demand and electric vehicle state, from an existing database, a deep recurrent neural network and an electric vehicle battery model, respectively. The charging strategies are tested in a simulator that uses real and recent charging events from a charging location in Brussels. The results show the possibility to drastically reduce peak powers and with them, electricity bills as well, near real-time. Bi-directional charging shows the best results compared to uni-directional charging. These good performances are amplified by oversized PV systems but reduced with higher minimum state-of-charge.

Keywords

• Electric vehicle
• Model predictive control
• Peak shaving
• Smart charging
• Vehicle-to-grid