Abstract
Dissolved gas in oil (DGA) is a common means of monitoring the condition of an oil-immersed transformer. The concentration of dissolved gas and the ratio of different gases are important indexes to judge the condition of power transformers. Monitoring devices for dissolved gas in oil are widely installed in main transformers, but there are few recorded fault data of main transformers. The special operation and maintenance modes of main transformers leads to the fault modes particularity of main transformers. In order to solve the problem of insufficient samples and the feature uncertainty, this paper puts forward an unsupervised mutual information method to select the feature verified by the optimized support vector machine (SVM) model of particle swarm optimization (PSO) method and tries to find the feature sequence with better performance. The methos is validated by data from nuclear power transformers.
Highlights
Framework of the Feature Selection Method Based on the support vector machine (SVM) Model for MainThe gas concentration values measured in the continuous operation process of nuclear power transformer are obtained, and there is almost no-fault data
The feature sequences selected by the supervised mutual information feature selection algorithm and the unsupervised mutual information feature selection algorithm are applied, respectively
This paper proposes an unsupervised mutual information feature selection method to calculate Dissolved gas in oil (DGA) monitoring data of main transformer and output feature selection sequence
Summary
The gas concentration values measured in the continuous operation process of nuclear power transformer are obtained, and there is almost no-fault data. The features used in various power transformer condition diagnosis methods based on dissolved gas in oil are extensively studied. On this basis, the initial feature set is formed. Different number of feature sets are selected sequentially and verified by optimized SVM model for transformer fault diagnosis. In order to reduce the contingency of the experiment, the 5-fold verification method is used to process the training samples and test samples to verify the validity of the selection feature in the diagnosis of the nuclear power transformer condition diagnosis. Based on the accuracy of diagnosis, the feasibility of different feature extraction algorithms in the condition diagnosis of main transformers is analyzed
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have