Enhancing Charging Station Power Profiles: A Deep Learning Approach to Predicting Electric Vehicle Charging Demand

The transportation sector is a primary driver of rising fuel consumption and greenhouse gas (GHG) emissions. Electric vehicles (EVs) are considered a promising solution to these environmental issues. However, due to variances in charging demands, widespread EV adoption may pose problems to the distribution network's reliability. Numerous methods are employed to forecast EV charging demand to overcome these difficulties. This study evaluates the performance of four well-known deep learning models-artificial neural networks (ANN), recurrent neural networks (RNNs), long short-term memory (LSTM), and gated recurrent units (GRUs)-in forecasting the charging demand for EV customers once a charging session begins. Additionally, the paper proposes a two-layer charging station energy management system aimed at smoothing the power profile of a charging station with high power demands by integrating solar energy from photovoltaic (PV) panels. According to the findings, the GRU regression method demonstrates a slight advantage over the remaining three models in predicting power charging requirements. Notably, the GRU regression model exhibits the lowest Mean Absolute Error (MAE) of 2.6391. These results hold the potential to aid Moroccan authorities in enhancing the dependability of the grid utility in the near term and providing guidance for the strategic expansion of charging infrastructure in the long term.