Abstract The increasing integration of renewable energy sources necessitates advanced power conversion systems capable of maintaining high power quality in compliance with grid standards. Multilevel inverters (MLIs) have emerged as a promising solution; however, their nonlinear switching characteristics introduce harmonic distortions that challenge compliance with IEEE 519 limits. To address this issue, this paper proposes a novel hybrid artificial intelligence-driven harmonic optimization framework for selective harmonic elimination (SHE) in MLIs. The proposed approach integrates Accelerated Particle Swarm Optimization (APSO) and Adaptive Spiral Dynamic Algorithm (ASDA) into a unified hybrid scheme that enhances global exploration and local exploitation, thereby improving convergence reliability and harmonic minimization performance. To enable real-time implementation, an Artificial Neural Network (ANN) is developed and trained using optimal switching angle datasets generated by the hybrid optimizer, allowing instantaneous prediction of switching angles without the need for iterative online computation. The effectiveness of the proposed framework is validated across 7-level, 13-level, and 21-level cascaded H-bridge MLIs, demonstrating scalability and robustness under varying operating conditions. Simulation results show that conventional high-frequency PWM techniques fail to consistently satisfy IEEE 519 requirements, whereas the proposed hybrid ANN-based approach achieves significantly reduced total harmonic distortion (THD) and ensures standard compliance, particularly in higher-level inverter configurations. Furthermore, the integration of hybrid optimization with learning-based control provides a computationally efficient and scalable solution for harmonic mitigation in grid-connected photovoltaic systems. Future work will focus on hardware implementation and real-time validation.
Deshi et al. (Mon,) studied this question.