Integrating NARX Neural Network with K-S Test for Accurate Partial Discharge Detection in Transformers

Condition monitoring of transformers is crucial in preventing unexpected outages and equipment damage due to the ageing and deterioration processes. However, the traditional methods of setting alarm thresholds are problematic due to the dynamic changes in the operational and environmental conditions of transformers. This paper proposes a novel approach to partial discharge (PD) detection by integrating the Nonlinear Auto Regressive eXogenous (NARX) neural network with the Two-Sample Kolmogorov-Smirnov (K-S) test for online monitoring of a partial discharge index, called Peak to Average Power Ratio (PAPR). The proposed model aims to enhance the accuracy of PD detection by increasing the chances of identifying normal process behavior and reducing false positive alerts. The results of the study demonstrate the effectiveness of the NARX-K-S integration in improving the accuracy of partial discharge detection and provide a new direction for the condition monitoring of transformers.