Dynamic Spiking Graph Neural Networks

Electric Vehicle (EV) charging navigation in Smart Grids (SG) optimizes route and charging station selection for EV, ensuring efficient energy use, reduced waiting times, and balanced load distribution. The integrated management system improves transportation efficiency and energy system stability respectively. Accurate determination of EV charging station status represents a significant obstacle in navigation as it affects route planning through information on availability and operational conditions and current queue lengths. To overcome these drawbacks, this paper proposes a hybrid approach for enhancing EV charging navigation in SG. The process begins by gathering data from EV charging navigation dataset, which is then passed through a pre-processing phase. Maximum Correntropy Unbiased Minimum-Variance Filter (MCUMVF) is employed to clean and remove the missing value in the input data. The pre-processed output was fed to Dynamic Spiking Graph Neural Network (DSGNN) the data enters the classification phase, to enhance the accuracy of classification. The busy, available and out-of-service status of charging station is successfully classified by using DSGNN. The weight parameter of DSGNN is optimized using Kookaburra Optimization Algorithm (KOA). The DSGNN-KOA technique is implemented in MATLAB and evaluated using various performance metrics, including accuracy, precision, recall, F1-score, specificity and Root Mean Squared Error (RMSE). The results show that the DSGNN-KOA method outperforms existing approaches, such as Deep Reinforcement Learning (DRL), Machine Learning (ML), Enhanced Multi-Agent Neural Network (EMANN), Deep Learning (DL) and Multi-Agent Deep Neural Network (MADNN). DSGNN-KOA demonstrates exceptional performance through its 2.72 MAE and 0.85 RMSE values combined with 98.6% prediction accuracy and 98.4% precision for predicting available busy and out-of-service EV charger categories.