Harnessing Artificial Intelligence for Improved Harmonic Reduction in Rectifier Systems: A Hybrid Power Filter Approach

Power electronics-based electrical equipment is widely used across modern industrial sectors, offering advancements in energy conservation, efficiency, performance, and industrial needs. However, these devices-such as rectifiers, converters, inverters, Variable Frequency Drives (VFDs), Uninterruptible Power Supplies (UPS), furnaces, and other equipment-are categorized as non-linear loads, leading to waveform distortion in the electric power supply. This waveform distortion is a significant issue, causing Power Quality (PQ) problems in both power systems and local distribution networks. Total Harmonic Distortion (%THD) is a key metric for assessing the extent of harmonic pollution in an electrical system. The IEEE 519-2022 standard provides clear guidelines on voltage and current harmonic limits (%THDV and %THDI) based on system voltage levels. This study focuses on reducing harmonics produced by a three-phase rectifier through the implementation of a hybrid harmonic filter that integrates both passive and active components. To improve the effectiveness of the active filter, an Artificial Neural Network (ANN) is employed. Real-time measured data are used to train the ANN, resulting in better %THDI reduction. The article presents a performance analysis of the proposed active harmonic filter, hybrid harmonic filter, and ANN-trained hybrid harmonic filter. Testing and validation of the proposed hybrid harmonic filters are conducted using the MATLAB simulation platform.