Enhancing Electric Vehicle Charging Predictions: A Physics-Informed Neural Network Approach

Electric vehicle (EV) charging stations have been evolving to offer better and more efficient power delivery methods. A key area of research in this field involves predicting the power consumption of EV charging stations. Many researchers have addressed this issue using machine learning and deep learning methods, however, forecasting models struggle to predict individual charging sessions with high resolutions. In this paper, a deep neural network (DNN) is constructed to forecast the EV charging current profile and state of charge (SoC). Charging current and SoC are mathematically formalized and embedded into the DNN’s loss calculation to build a physics-informed neural network (PINN). The performance of the models is assessed through analysis of their prediction error, utilizing actual data from real EV charging sessions.