Analytical processing of online monitored dissipation factor based on morphological filter

Analyzing the trend of the on-line monitored dielectric dissipation factor (tan δ) series is a very effective method to diagnose the insulation fault of high voltage apparatus. However, due to the influence of environmental temperature, humidity and field noise interference, etc., the series is too nonstationary to observe its trend directly. In this paper, the problem of a trend-extracting method through relative comparison is briefly pointed out. Based on the analysis of the regularity of the practical tan δ series, a trend model of the series is then built. Since the variation regularity of the real insulation condition is different from that of all influential factors, another trend-extracting method by using the filter is put forward. The morphological filter is introduced to extract the trend of the tan δ series. To verify the effectiveness of the morphological filter, the median filter and wavelet transform are employed to process a simulated tan δ series. The results by the morphological filter are more stationary and smoother than those obtained by other methods. Finally, morphological filters with the structure element in different width are employed to analyze a practical tan δ series. The results show that the influences of different factors can be effectively removed.