Partial Discharge Pattern-Recognition Method Based on Embedded Artificial Intelligence

This paper proposes a method for detecting and recognizing partial discharges in high-voltage (HV) equipment. The aim is to address issues commonly found traditional systems, including complex operations, high computational demands, significant power consumption, elevated costs. Various types of were investigated an HV laboratory environment. Discharge data collected using high-frequency current sensor microcontroller. Subsequently, this underwent processing transformation into feature sets the phase-resolved discharge analysis technique. These features then converted grayscale map samples PNG format. To achieve classification, convolutional neural network (CNN) was trained on these samples. After successful training, model adapted deployment microcontroller, facilitated by STM32Cube.AI ecosystem, enabling real-time recognition. study also examined storage requirements across different CNN layers their impact recognition efficacy. assess algorithm’s robustness, accuracy tested under varying voltages, insulation media thicknesses, noise levels. test results demonstrated that algorithm could be effectively implemented achieving exceeding 98%.