A Hybrid CNN-BiLSTM Framework for Residential Electric Vehicle Charging Demand Forecasting

Abstract

The growing number of electric vehicles (EVs) has brought new changes to residential electricity consumption patterns. Therefore, predicting EV charging demand accurately is important for the operation and management of distribution networks. This paper presents a forecasting method that combines a Convolutional Neural Network (CNN) and a Bidirectional Long Short-Term Memory (BiLSTM) network for residential EV charging demand prediction. The model uses several types of information, including time-related features, ambient temperature, electricity prices, and previous charging data. The CNN part is used to identify short-period changes in charging sequences, while the BiLSTM part is applied to describe the time-dependent relationships of load variations. An Artificial Neural Network (ANN) model is selected for comparison. The simulation results show that the proposed CNN-BiLSTM method obtains better prediction results than the ANN method according to MAE, RMSE, MAPE, and coefficient of determination. The proposed approach can be used as a useful tool for residential charging control and distribution network operation.