Overload Mitigation in Electric Vehicle Smart Charging Algorithms Using Photovoltaic Generation Forecasting

Recently, the growing use of electric vehicles (EVs) and distributed energy resources (DERs) has brought about the emergence of new stakeholders in the public EV charging market. This has resulted in a need for innovative techniques in electric vehicle smart charging (EVSC) to lower charging costs and optimize renewable energy sources (RES) usage. However, an unsupervised EVSC system may experience overloads or violations of the maximum power defined by the charging point operator (CPO) due to fluctuations in photovoltaic (PV) generation. To address this issue, this study proposes a heuristic method to prevent overloads in the electric vehicle aggregated system (EVAS), considering various 15-minute interval PV generation prediction methods. Specifically, this research compares the effectiveness of different prediction models, such as ARIMA, LSTM, and CNN-LSTM, in minimizing overloads during a 4-week evaluation period. The study's outcomes are expected to offer valuable insights into creating more efficient and reliable EVSC systems. Both Machine Learning (ML) methods succeeded in reducing the occurrences of overloads by approximately 50%. However, ARIMA method yielded even more impressive results, reducing the occurrence of violations by approximately 87%.