A Neural Network-Based Model Predictive Control for a Grid-Connected Photovoltaic-Battery System with Vehicle-to-Grid and Grid-to-Vehicle Operations

The growing integration of photovoltaic (PV) energy systems and electric vehicles (EVs) introduces new challenges in managing energy flow within smart grid environments. The intermittent nature of solar energy and the variable charging demands of EVs complicate reliable and efficient power management. Existing strategies for grid-connected PV-battery systems often fail to effectively handle bidirectional power flow between EVs and the grid, particularly in scenarios requiring seamless transitions between vehicle-to-grid (V2G) and grid-to-vehicle (G2V) operations. This paper presents a novel neural network-based model predictive control (NN-MPC) approach for optimizing energy management in a grid-connected PV-battery-EV system. The proposed method combines neural networks for forecasting PV generation, EV load demand, and grid conditions with a model predictive control framework that optimizes real-time power flow under various constraints. This integration enables intelligent, adaptive, and dynamic decision making across multiple objectives, including maximizing renewable energy usage, minimizing grid dependency, reducing transient responses, and extending battery life. Unlike conventional methods that treat V2G and G2V separately, the NN-MPC framework supports seamless mode switching based on real-time system status and user requirements. Simulation results demonstrate a 12.9% improvement in V2G power delivery, an 8% increase in renewable energy utilization, and a 50% reduction in total harmonic distortion (THD) compared to PI control. The results highlight the practical effectiveness and robustness of NN-MPC, making it an effective solution for future smart grids that require bidirectional energy management between distributed energy resources and electric vehicles.