Optimized Algorithm Research Of Power Transformer Fault Diagnosis Based On The Analysis Of Oil And Gas

Due to the increasing of power users,along with tremendously promote and advocate of smart grid construction, we put forward higher requirements to the security management and stable operation of power system. Power transformer is one of the most important transmission installation in power system, it ensures the stable operation of power transformer and act as an important part to maintain the normal operation of power system. Therefore, studying of the health condition monitoring and fault diagnosis of power transformer has always been a hot topic in this field. In recent years, a variety of solutions to fault diagnosis of transformer has been proposed by home and abroad scholars. Previous study showed that the intelligent diagnostic method, which based on the Dissolved Gas Analysis, is the most representative way. In this paper, we optimize and improve the DGA based on the thorough study of it, also combine it with intelligent algorithms, thus we propose a new fault diagnosis algorithm for power transformer.There are two main criteria for fault diagnosis of power transformers: diagnosis accuracy and diagnosis speed. In order to make further improvement of diagnosis accuracy and diagnosis speed of fault diagnosis, we optimize and integrate current diagnostic algorithms based on the study of them: first, use quantum behavior of Support Vector Machine fault diagnosis algorithm, namely using SVM to classify large power transformer and find optimization of SVM parameters by using improved genetic algorithm of quantum behavior. Second, using K-nearest neighbor clustering analysis fault diagnosis algorithm, namely using K-nearest neighbor clustering analysis to classify the suspicious area based on the first step. Finally, compare and analyze the performance of improved quantum genetic algorithm and common genetic algorithms by simulation experiments.Simulation experiment shows that the improved quantum genetic algorithm only needs 50 generations to get optimal classification model parameters, while the ordinary genetic algorithm needs 170 generations; at the same time, the combination of clustering analysis and SVM reduces misclassification samples of single support vector machine(SVM). Therefore, we can see that this kind of power transformer fault diagnosis algorithm can effectively improve diagnosis speed and accuracy of the fault diagnosis, which achieved the desired purpose.Finally, we sum up the above study and propose prospect of the further work of our system.