ANN-Optimised Space Vector PWM for a Single-Phase 5 kW Grid-Connected Solar PV Inverter: Harmonic Reduction, Efficiency Enhancement, and Thermal Performance Under Himalayan Irradiance Conditions

Grid-connected solar photovoltaic (PV) inverters constitute the critical power conversion interface between PV arrays and the utility distribution grid, and their performance in terms of output power quality, conversion efficiency, and thermal reliability directly determines the techno-economic viability of distributed solar generation installations. The Himalayan states of Himachal Pradesh and Uttarakhand, with annual global horizontal irradiance (GHI) of 4.8–6.2 kWh/m²/day in the mid-altitude (800–2000 m) zone and significant irradiance variability due to cloud shadow effects, mountain terrain masking, and seasonal snow cover, present a demanding operating environment for inverter control systems that must maintain output voltage quality and grid synchronisation across rapidly varying DC input conditions. Conventional sinusoidal PWM (SPWM) and space vector PWM (SVPWM) strategies apply fixed switching patterns that are suboptimal under variable irradiance, producing elevated total harmonic distortion (THD) and reduced conversion efficiency at partial load — the predominant operating condition under Himalayan weather patterns. This study proposes an Artificial Neural Network (ANN)-optimised SVPWM control strategy for a 5 kW single-phase H-bridge inverter in which a three-layer feedforward ANN trained on simulated PV array operating points dynamically adjusts the space vector dwell times and switching sequence to minimise output voltage THD across the full modulation index range (0.5–1.0). The proposed ANN-SVPWM achieves THD of 3.2% at M=0.9 (versus 18.4% for conventional SPWM and 12.1% for standard SVPWM), peak conversion efficiency of 97.4% at 3 kW output (versus 95.1% for conventional SVPWM), DC link voltage ripple of 4.8 V peak-to-peak (versus 18.4 V conventional), and dynamic voltage recovery time of 8 ms under 50% step load change (versus 82 ms conventional PI control). Heat sink temperature rise at full 5 kW load is reduced from 48°C to 31°C, extending IGBT junction life by an estimated factor of 2.8. Hardware-in-the-loop (HIL) validation using dSPACE DS1104 confirms simulation results within 2.1% deviation. These findings demonstrate the practical viability of ANN-based adaptive PWM control for improving solar inverter performance in mountain irradiance environments representative of north Indian hill-state distributed generation deployments.