Hybrid AI–GA–DQN Optimization for Harmonic Mitigation in Smart Inverters

arief budi laksono (Universitas Islam Lamongan)
Zainal Abidin (Universitas Islam Lamongan)
Abdur Rohman Wakhid (Universitas Islam Lamongan)
affan bachri (Universitas Islam Lamongan)
Wahri Sunanda (Universitas Bangka Belitung)
Wahyu Mulyo utomo (Universitas Tun Hussein Onn Malaysia)

Nucleus Journal
Vol. 5 No. 2 (2026)

This paper proposes a hybrid Artificial Intelligence–Genetic Algorithm–Deep Q-Network (AI–GA–DQN) control framework for inverter systems to minimize harmonic distortion and enhance dynamic voltage regulation. A mathematical state-space model incorporating harmonic dynamics is formulated, and a hybrid optimization approach is applied to adaptively tune modulation indices and switching parameters. The GA provides global parameter optimization, while DQN enables reinforcement learning-based adaptation under variable load conditions. MATLAB/Simulink simulations show that the proposed controller achieves a 75.6% reduction in Total Harmonic Distortion (THD), improves power factor to 0.995, and enhances system efficiency to 97.8%, compared with conventional PWM and GA-only controllers. The results confirm the effectiveness of the AI–GA–DQN strategy in achieving real-time harmonic suppression and superior transient performance in intelligent power converters.