A hybrid technique for grid-tied photovoltaic (PV) systems utilizing a modular multilevel inverter (MMI) topology

Abstract

This paper proposed a hybrid strategy for grid-tied photovoltaic systems utilizing a modular multilevel inverter (MMI) topology. The novel control strategy is named the Mexican Axolotl Optimization (MAO) and Recalling-Enhanced Recurrent Neural Network (RERNN) technique (MAO-RERNN). The major purpose of this study is to develop a hybrid MAO-RERNN-based control strategy for a 31-level Modular Multilevel Inverter in grid-tied photovoltaic systems to minimize Total Harmonic Distortion (THD) and minimize the switching losses to enhance power quality through optimized switching state prediction and adaptive gate pulse generation. The controller generates gate pulses for level generator switches to produce staircase waveforms of unipolar signals and converts these signals into bipolar using polarity changer switches. MAO adjusts the gate pulses for level generator switches, producing staircase waveforms of unipolar signals, which are then converted into bipolar signals using polarity changer switches. RERNN is employed to predict and optimize the switching states of the inverter, ensuring efficient and accurate control. Implementation in MATLAB validates the performance, demonstrating superior results compared to existing methods. The proposed MAO-RERNN technique achieves a THD of 1.25 % with a 31-level inverter, showcasing its effectiveness in enhancing output quality and system robustness.