Analysis of influencing factors of Dissolved Gas in oil

摘要

As an important means of insulation testing, Dissolved Gas Analysis (DGA) in oil reflect the insulation condition of transformer. However, a lot of factors influence the dissolved gas of transformer. Some of them should be taken into consideration during the analysis, especially when it comes to the typical gas concentration. In this paper, an evaluate method of influencing factors is proposed. With this method, the effect degree of influence is quantified and the typical gas concentration could be modified. Firstly, collect and process the DGA data, removing invalid data and filling missing data. Then, classify DGA data according to each influence factor. With the different classification, each cluster center can be figure out by K-means. Calculate the average Euclidean distance of cluster centers as the influence evaluating indicator. Each kind of classification corresponding to a Euclidean distance value, regard the Euclidean distance value as the influence level. The further distance means the greater influence. Finally, classify the data according to the influencing factor and calculate the typical value with the statistic method. As an example, DGA data from two grid has been collected and tested. It comes to a conclusion that voltage grade, operation years, oil type have much stronger influence than the installation location. According to the conclusion, the classification of DGA has been done and typical gas concentration which corresponds to voltage grade, operation years, oil type has been calculated. The field test shows that, with the modified typical gas concentration, the defect was identified correctly, which means the great value of this method.