Journal of Energy Research and Reviews

This paper presents a Data-Augmented Model Predictive Control strategy that is optimized with Multi-Objective Particle Swarm Optimization for Active Power Filters operating in Electric Vehicle charging networks with random and unbalanced load profiles. The approach integrates physical modeling with data-driven forecasting based on historical charging logs, environmental parameters, and tariff signals to predict harmonic-rich current and a specific amount of demand based on previous events. These predictions form the basis of an adaptive MPC cost function that is optimized using MOPSO to minimize Total Harmonic Distortion, reactive power, and switching, while maintaining DC-link envelope and IEEE-519 compliance. Simulation results show that the proposed controller greatly improves the power quality and energy efficiency of the system. Under unbalanced nonlinear load profiles, THD is minimized to 2.7%, leading to an 88% reduction compared to the uncompensated case and 35% cost savings compared to conventional MPC. The power factor increased from 0.93 to 0.998, and DC-link ripple was confined to ±1.5%. Switching frequency decreased by 8%, lowering overall system efficiency by 2%. Resilience was verified in Monte Carlo simulations with N = 100 using Gaussian noise with σ = 3%, random grid impedance variation, and dynamic EV charging events under multiple random scenarios. The THD was 2.6% to 2.9%, far below IEEE limits, demonstrating robust disturbance rejection and dynamic adaptability. So, by combining optimal MOPSO with predictive intelligence from data, the MOPSO–MPC methodology guarantees near-unity power factor, low distortion, and enhanced stability under realistic grid variabilities. This makes it an appealing and scalable solution for improving power quality in smart-grid applications, as well as coordinating and deploying renewable energy facilities and Electric Vehicle infrastructure for electro-mobility.