Hybrid CNN-LSTM Forecasting Model for Electric Vehicle Charging Demand in Smart Buildings
Nikolaos Tsalikidis, Paraskevas Koukaras, Dimosthenis Ioannidis, Dimitrios Tzovaras
Abstract
The accelerated shift towards renewable energy sources has signalled the widespread adoption of Electric Vehicles (EVs) as the primary mode of transportation. Concurrently, smart building technologies are becoming essential for achieving sustainability goals and promoting energy-efficient practices. However, the anticipated integration of private EV Charging Stations (EVCS) in today’s smart buildings and the stochastic nature of EV charging patterns present challenges in efficiently managing residential EV charging demand. Recognizing the importance of accurate early knowledge of future EV load demand, this research emphasizes developing an advanced Deep Learning (DL) step-by-step approach for forecasting residential EV charging demand. The proposed hybrid CNN-LSTM model extracts temporal features from the input data. Then, these features are passed on to LSTM layers for sequential learning, contributing to an increased accuracy in EV charging demand forecasts compared with other traditional DL approaches.