Active and Reactive Power Control in Three-Phase Grid-Connected Electric Vehicles using Zebra Optimization Algorithm and Multimodal Adaptive Spatio-Temporal Graph Neural Network

Three-phase grid-connected Electric Vehicles (EVs) are critical for optimizing energy flow, managing Active Power (AP) for charging and discharging, and controlling Reactive Power (RP) to ensure voltage regulation. These features enhance grid reliability and support the seamless integration of large-scale EVs into power grids. However, the unpredictable frequency of charging sessions creates challenges such as voltage fluctuations and grid imbalances, adversely affecting power quality (PQ) and stability. To address these issues, this study proposes a hybrid approach for AP and RP control in three-phase grid-connected EVs. The novel ZOA-MASTGNN technique integrates the Zebra Optimization Algorithm (ZOA) with the Multimodal Adaptive Spatio-Temporal Graph Neural Network (MASTGNN). The ZOA dynamically optimizes system parameters, improving power management, reducing Total Harmonic Distortion (THD), and enhancing grid stability. Meanwhile, MASTGNN predicts optimal control actions, mitigating harmonics, regulating voltage dynamically, and adapting to changing operational conditions in grid-interactive EV systems. The suggested method was implemented on the MATLAB platform and evaluated with existing approaches, including Resiliency-Guided Physics-Informed Neural Networks (RPINN), Elman Neural Networks (ENN), Multilayer Feed Forward Neural Networks (ML-FFNN), Deep Neural Networks (DNN), and Particle Swarm Optimization-Artificial Neural Networks (PSO-ANN). Results showed significant improvements, achieving 19.36% load current THD and 3.52% source current THD, while outperforming other approaches in efficiency and effectiveness. This framework addresses key challenges in large-scale EV integration, offering scalable and practical solutions for sustainable power grid operations.