Analysis of Electric Vehicle Charging Demand Forecasting Model based on Monte Carlo Simulation and EMD-BO-LSTM

Abstract

The stochastic charging behaviors of Electric Vehicle (EV) users illustrate the negative effects of bulk charging during peak hours on the grid. To overcome this problem, the bulk EV charging demand forecasting approach is investigated using historical EV charge demand dataset and EV driver mobility statictics in this paper. In this model, a Monte Carlo Simulation (MCS) is perfomed that considers the charging behavior of EV users for the generation of EV charging times. Moreover, the EV charging times are combined with the bulk EV demand hybrid forecasting model using decomposition and deep learning time series method. In first stage, the EV demand time series dataset are divided to improve the model performance by empirical mode decomposition (EMD). Then, all decomposed signals are forecasted separately using the Bayesian optimized Long Short-Term Memory LSTM network (BO-LSTM). Finally, to evaluate the model perfomance, the power system analysis using IEEE 33 busbar test system is performed in terms of distribution network power losses, busbar voltage drops and transformer loading conditions.