A hybrid Bayesian network-based deep learning approach combining climatic and reliability factors to forecast electric vehicle charging capacity

The increasing adoption of electric vehicles (EVs) necessitates advanced predictive models to accurately forecast charging demand and ensure reliable infrastructure planning. This study introduces a novel analytical framework that integrates queuing network and Bayesian network models to enhance the prediction accuracy and reliability of EV charging demand. The objective is to develop a comprehensive system that accounts for various influencing factors, such as meteorological conditions and charging pile failure rates, to optimize EV infrastructure. The methodology involves creating a hybrid Bayesian Network-based deep learning (HBNDL) system architecture. This architecture uses extensive transaction data and climate analysis to build a detailed model of EV charging pile reliability. Additionally, two algorithms are designed to assess the usage and reliability of charging stations. The framework's effectiveness is tested through a series of experiments evaluating its performance in short-, medium-, and long-term prediction scenarios. The results demonstrate that the HBNDL framework significantly improves prediction accuracy and infrastructure reliability. The integration of queuing theory and Bayesian network models with deep learning techniques results in a robust system adaptable to various conditions. Experimental validation shows that the proposed framework outperforms existing models in forecasting EV charging demand, particularly under varying environmental influences.