Data Analysis In UHF PD Monitoring For Power Transformers

During the operation of power transformers, insulation deterioration is still the major fault. Partial discharge (PD) is one of the prime reasons to cause insulation deterioration occurring inside power transformers. The PD on-line monitoring can be used to judge the insulation situation of power transformers accurately and timely. It's valuable to prevent accident faults in power transformers and insure stable performance of power system. Ultra-high-frequency (UHF) PD monitoring approach is focused on recent years because of its effectiveness to avoid noises.In order to solve the difficulties in pulses extraction and classification of the on-line monitoring system in power transformers, a research on data analysis and knowledge discovery of UHF PD monitoring system in power transformers was conducted using the raw data obtained from the on-line monitoring system of 251B, 1B, 14B, 18B transformers in Gezhouba Hydro Power Plant.Firstly the theory of UHF PD on-line monitoring in transformers and signal characteristics were studied, and an overview of anti-interference measures and data analysis methods in UHF PD monitoring system were given. Then referring to the process of knowledge discovery, the target datasets were established on the basis of the structure and principle of the monitoring system. After filtering and de-noising of the datasets, a method based on sliding window and energy threshold was used to extract the information of a single pulse. In the end, the general characteristics of pulses under different conditions of the transformer unit were summed up on the platform of the Optimal Maintenance System for Hydro power plants(HOMS).Through the analysis of the typical pulses, an algorithm for pulses classification and recognition based on pulse width, polarity and distribution characteristics was designed and implemented. The results showed that this algorithm had the characteristics of less calculation and high efficiency, and was able to classify and identify kinds of typical pulses effectively.