Degradation Classification of Low-Voltage Zinc Oxide Varistors Using K-NN Algorithm

In this paper, the degradation of low voltage varistors-based arrestors exposed to slow-front overvoltage surges is analyzed. MOV samples are degraded under different set number of surges. The reference voltages of varistor samples are measured before and after the application of switching surges. The KNN algorithm is then applied to analyze the percentage change in reference voltage values and classify the degradation condition. Additionally, the ANOVA statistical test is used to determine the correlation between the number of surges and the percentage change in reference voltages. The KNN classification accuracy is reported to be 96.4%, with precision and recall values of 96.9% and 96.4% respectively. The F-1 score is calculated to be 96.4%. Results indicate that the KNN algorithm effectively classifies the degradation condition of the MOVs. In contrast, the ANOVA results show that there is a significant mean reference voltage difference observed between the groups of varistors that were exposed to different sets of switching surges. This suggests that the number of switching surges applied has a notable effect on the reference voltage values of the varistors.