Deep learning-based power quality enhancement in microgrids with hybrid energy storage and photovoltaic EV charging stations

The rapid growth of Electric Vehicles(EVs) and Photovoltaic(PV) systems in modern microgrids has intensified power quality issues, including voltage fluctuations, harmonics, and frequency deviations. Hybrid Energy Storage Systems (HESS), integrating batteries and supercapacitors, provide stability support but require intelligent coordination for optimal performance. Conventional control strategies struggle to manage fast PV variations and sudden EV charging surges,resulting in unstable voltage profiles and degraded power quality. Existing Deep Learning approaches lack effective attention mechanisms and adaptive optimisation for accurate short-term forecasting and energy allocation. This study proposes an Intelligent Animal Migration–optimised Attention-Long Short-Term Memory (IAM-Att-LSTM) framework for microgrid power quality enhancement. A real-time microgrid dataset of 2500 records, including voltage, current, frequency, harmonics, PV generation, EV load, Battery SOC, and SuperCap SOC, was preprocessed using missing-value imputation and Z-score normalisation. The Attention-LSTM captures temporal variations in PV output and EV demand, while Intelligent Animal Migration Optimisation adaptively manages energy sharing between battery and super capacitor units. Simulation results demonstrate superior predictive accuracy with a MAPE of 0.95% and MARPE of 1.85, reducing voltage deviations and improving dynamic response during PV fluctuations and EV charging spikes. IAM-Att-LSTM improves microgrid power quality with intelligent prediction and optimized HESS control for EV-integrated systems.