Support Vector Machine-Integrated High-Definition MCPWM Technique for Improving Power Quality in Multilevel Inverters

What if machine learning could predict inverter harmonics before prototyping? Conventional pulse width modulation (PWM) techniques in cascaded H-bridge (CHB) multilevel inverters (MLIs) struggle with low-frequency odd harmonics, high switching losses, and uneven power distribution, limiting efficiency and adaptability in industrial applications. For the first time, this article introduces a support vector machine (SVM)-based harmonic prediction algorithm integrated with a high-definition multicarrier PWM (HD-MCPWM) technique for MLIs. The SVM algorithm forecasts lower order odd harmonics (5th, 7th, 11th, and 13th) and determines optimal switching angles. Meanwhile, HD-MCPWM ensures uniform switching losses and stable inverter output currents. Together, this approach achieves 4.47% harmonic distortion with a 33.3% reduction in switching losses. Validated through theoretical models, simulations, and prototypes, the proposed technique delivers 97.8% peak efficiency, demonstrating its superiority over conventional methods as a reliable, efficient solution for MLIs in industrial applications.