Cascaded Multilevel Inverter with Reduced Switch Count for Total Harmonic Distortion Reduction using Optimization with Neural Approach

Multilevel inverters (MLIs) are highly efficient and produce a high-quality output voltage, making them useful in traction systems, flexible AC transmission systems (FACTS), electric drives, and renewable energy (RE) systems. However, traditional methods for reducing harmonics often incur significant computational overhead and rely heavily on the initial value selection, resulting in suboptimal performance. This manuscript proposes a hybrid technique for minimizing total harmonic distortion (THD) using a cascaded H-bridge multilevel inverter (CHBMLI). The proposed method combines the Snow Ablation Optimizer (SAO) and Macro Graph Neural Networks (MGNN). Hence, it is named the SAO-MGNN method. The primary objective of the manuscript is to minimize THD and obtain optimal switching angles. The SAO is employed to optimize the switching angles without maximizing the size of the inverter. The MGNN is used to predict the parameters of the multilevel inverter. The SAO-MGNN technique is utilized to find the appropriate switching angles for an MLI that generates an output voltage of seven levels. The proposed technique's performance is tested on MATLAB and compared with existing approaches. The proposed approach demonstrates better results than existing methods, are Particle Swarm Optimization (PSO), Grey Wolf Optimization (GWO), and Artificial Neural Networks (ANN). The proposed SAO-MGNN technique achieves minimal harmonic distortion in a seven-level cascaded H-bridge inverter, delivering a phase voltage THD of 3.22% and a line-to-line voltage THD of 1.22% under resistive load conditions.