AI-Based Adaptive Power Factor Correction for Reactive Power Reduction in Distribution Systems

Abstract

The widespread application of inductive loads in today ‘s power distribution systems where power factor (PF) correction has become an integral requirement, has led to power factor degradation as a serious problem and requires correction for more efficient power distribution. In this paper, an innovative AI based Adaptive PF correction scheme is proposed to integrate conventional PF correction capacitor bank compensation with an Artificial Neural Network (ANN) based predictor-controller for dynamic reactive power compensation. A detailed MATLAB/Simulink 3–phase distribution system-based model has been evolved which yields operational data sets for ANN training and testing with different loading conditions. Proposed ANN predictor-controllers are trained with back-propagation algorithm applying MSE (Mean Squared Error) function as a network performance measure. The network prediction accuracy 96.9 to 97.8% and the generalization capability has been established through comprehensive results analysis. The proposed architecture significantly decreases the reactive power demand by approximately 77%, apparent power consumption by 28.1% and the dynamic response by 3 times compared to fixed capacitor compensation method. Moreover, the (PF) is maintained between 0.95–0.96, confirming its effectiveness for real-time adaptive power factor correction in modern distribution systems.