Assessing partial discharge intensity of electrical equipment based on UV detection and the ANFIS algorithm.

UV detection has been generally used in external insulation detection of equipment. Currently used detection methods have many shortcomings, such as incomprehensive consideration of influencing factors, poor model adaption, and low accuracy. We propose a method that detects the ultraviolet pulse number (P), temperature (T), and humidity (H). Meanwhile, discharge intensity (J') was preliminarily estimated by UV detection using a fuzzy algorithm. Next, the ANFIS algorithm with a self‐study mechanism and an artificial neural network was introduced, which assessed that the discharge intensity is J. A UV detection circuit was established while carrying out a field test. The preliminary evaluation data of the fuzzy were used as training and testing data of ANFIS. The optimized model was obtained after ANFIS multiple trainings. Experimental results show that the relationships between P, T, H, and J in the optimized model are consistent with those of theoretical deduction. The assessment results remain highly stable with the occurrence of changes in external conditions, which indicates the feasibility of the ANFIS algorithm in UV detection of discharge intensity of electric equipment. This study can provide effective suggestions for online insulation monitoring and equipment maintenance. © 2018 Institute of Electrical Engineers of Japan. Published by John Wiley & Sons, Inc. [ABSTRACT FROM AUTHOR]