AI-Driven Optimization of a Cascaded H-Bridge 11-Level Converter Using Reinforcement Learning

In real time, the performance of a cascaded H-bridge 11-level converter is optimized using reinforcement learning (RL) in this paper. Traditional methods generally fail to maintain power quality and adapt to changing load conditions, leading to increased total harmonic distortion (THD) and inefficiency. The RL agent continuously learns from the system's statesuch as output voltage and current and adjusts the switching of H-bridge modules dynamically to reduce switching losses, minimize THD, and maintain balanced direct current DC-link voltages, with Deep Q-Network (DQN) algorithms. The simulation results show that RL-based controllers significantly improve efficiency to 98%, reduce THD 2.45%, and enhance fault tolerance compared to conventional approaches, demonstrating the effectiveness of artificial intelligence (AI)-driven control when optimizing multilevel converters in complex, real-time environments in MATLAB/Simulink. This research work is aligned with Sustainable Development Goal (SDG) 7. The improved fault tolerance and reduced THD support reliable and sustainable energy systems also contribute to cleaner and more efficient energy distribution.