Artificial neural network-based power management for hybrid microgrid with SoC-supervised storage and dual-mode grid operation

Hybrid microgrids (HMGs) in grid‑interactive settings historically relied on single‑source generation or single‑storage control that faltered under renewable intermittency, fast transients, and storage aging, constraining reliability and autonomy. This study developed a lightweight power‑management system that coordinated solar photovoltaics (PV), wind, battery–supercapacitor (SC) storage, the utility grid, and AC load and electric‑vehicle (EV) charging to increase local use and resilience. The Hybrid Artificial Neural Network‑Integrated Synergistic State‑of‑Charge Supervised Power Management System (HANNI‑S3‑PMS) operated in both grid‑connected and islanded modes on an AC/DC MG. Control was hierarchical: primary proportional–integral loops regulated converter currents and voltages; an artificial neural network (ANN) provided maximum‑power‑point tracking (MPPT) for PV and wind and decentralized power sharing; and a tertiary state‑of‑charge (SoC)‑supervised layer coordinated power flows and charging via ANN optimization with phase‑locked‑loop (PLL) synchronization for seamless transitions. MATLAB/Simulink studies spanning varied loads, renewable ramps, and grid contingencies showed near‑complete renewable capture, exceptionally fast transient recovery, tightly regulated DC‑bus voltage with minimal overshoot and quick settling, and resilient load balancing, while reducing reliance on the utility grid and curbing storage stress. Frequency-domain and time‑domain analyses verified stability, and comparisons with representative MG controllers indicated higher efficiency with low computational overhead. HANNI‑S3‑PMS provided a data‑driven alternative to computationally heavy optimizers, significantly removed dependence on pre‑tuned parameters, scaled to real‑time multi‑node deployment, and advanced autonomy, reliability, and seamless grid/island operation in HMG.