Mexican axolotl optimization algorithm with a recalling enhanced recurrent neural network for modular multilevel inverter fed photovoltaic system

This manuscript presents a grid-connected photovoltaic (PV) system employing a modular multilevel inverter (MMI) topology with an advanced hybrid control technique. The proposed MAO-RERNN control method integrates the Mexican Axolotl Optimization (MAO) algorithm with a Recalling-Enhanced Recurrent Neural Network (RERNN) to achieve optimal power conversion, improved stability, and reduced total harmonic distortion (THD). Unlike traditional multilevel inverters (MLI), the MMI structure utilized in this work requires fewer power electronic components, reducing cost and complexity. The MAO algorithm optimally tunes control gain parameters offline, creating a dataset of optimal control signals, while the RERNN method predicts and applies real-time control signals, ensuring an efficient and stable power supply. The control strategy generates a smooth sinusoidal output by employing a staircase waveform approach and effectively mitigates external disturbances to maintain grid compliance. The proposed method is implemented in MATLAB/Simulink, and its performance is analyzed through comparisons with existing control techniques. Results demonstrate that the MAO-RERNN control strategy achieves superior power regulation, reduced THD, and enhanced robustness, making it a promising solution for next-generation PV-based inverter systems.