Adaptive Harmonic Mitigation in Synchronous Machines Using a Hybrid Active Filter and Artificial Neural Network Technique

The increasing use of nonlinear loads introduces harmonic distortion into power systems, potentially compromising the efficiency, stability, and operational reliability of synchronous machines. This study evaluates an artificial neural network (ANN)-assisted active filtering approach for mitigating harmonic distortion under varying load conditions. A mathematical model of a synchronous machine was developed and implemented in a Simulink environment. The active filter was integrated with an ANN comprising five input neurons, five hidden neurons, and five output neurons to support adaptive compensation of harmonic components. Simulations were conducted for the Sapele power station model under no-load (0 MW) and loaded (20 MW) conditions. Performance was evaluated using current, voltage, and frequency signals, with unfiltered waveforms providing the reference for assessing compensation. Waveform distortions were observed under the no-load condition and increased at 20 MW. In the unfiltered 20 MW case, current and voltage total harmonic distortion (THD) reached 18.88% and 17.56%, respectively, while the frequency signal exhibited a reported distortion value of 17.54%. Integrating the active filter with the ANN reduced current THD to 4.71%, representing an approximately 75.1% relative reduction. Improvements were also observed in the simulated voltage and frequency profiles. These findings indicate that the proposed approach reduced current harmonic distortion and improved waveform characteristics under the simulated operating conditions. The study supports the potential application of ANN-assisted active filtering to power quality improvement in synchronous machines and provides a basis for developing adaptive harmonic compensation strategies. Experimental validation and evaluation across additional load levels are needed to establish the approach’s robustness, real-time performance, and practical suitability for power station applications.