Comparative Study of HMM and BPNN in Detecting Corona Discharge on 20 kV Cubicle Based on Voltage and Sound

Abstract

Corona discharge is one of the most common disturbances in 20 kV cubicles, potentially reducing efficiency and damaging equipment. Early detection is essential to improve system reliability and prevent failures. This study compares the performance of the Hidden Markov Model (HMM) and Backpropagation Neural Network (BPNN) in detecting corona discharge based on acoustic signals and voltage measurements. The recorded corona discharge sounds were processed using Linear Predictive Coding (LPC) for feature extraction, followed by clustering and classification. HMM achieved an accuracy of 84.44% in voltage-based clustering and 100% in noise-based clustering, while BPNN demonstrated more consistent and reliable results, reducing detection errors. The proposed system enables early identification of corona discharge symptoms, provides acoustic-based clustering at specific thresholds, and offers an early warning mechanism for technicians. Overall, the findings indicate that BPNN is more effective than HMM in identifying corona discharge in 20 kV cubicles, thereby supporting preventive maintenance efforts and enhancing the reliability of power distribution networks.