Transformer Fault Identification Method Based On Point Density Weighted FCM Clustering And Quantum Genetic Neural Network

In the transmission of power grids,transformers are important nodes that cannot be replaced.Ensuring the stable operation of transformers is the prerequisite for ensuring the economy of our country and the quality of life of the people.With the passage of time,the components of the transformer will inevitably age,and even lead to failures and loss of the transformer after serious.At the same time,the transformer will be inevitably failed due to the influence of the manufacturing process and external environment during commissioning.Has a huge impact on the economy and life.Therefore,it is imperative to prospectively repair and improve the level of fault detection for the transformer.With the advancement and popularization of artificial intelligence technology,the relationship between dissolved gas in oil and faults has been clarified,and the identification of transformer faults based on dissolved gas in oil data has become a research hotspot in recent years.The main research contents of this article are as follows:First,taking the DGA data of dissolved gas in oil as the research object,by analyzing the correlation between the gas production and the fault type of the transformer during the cracking of insulating oil,it is clear that the five gas content data is used as the basis for fault identification.At the same time,by analyzing the fault The redundancy of the characteristic gas at the time of occurrence,using the principal component analysis method to extract features from the five-dimensional data,simplify the data structure,and improve the calculation efficiency of the model.Secondly,on the basis of DGA characteristic data,the transformer fault type data and fault type are divided,and at the same time,the measured fault and the division result are compared to filter out the wrong division data in the fault type division.In this paper,the FCM method of point density weighted optimization is used in data partitioning.The sample data is weighted by point density calculation to enhance the spatial recognition effect of missing item data.The results show that the FCM method based on feature data and point density weighting The accuracy rate of the type division is improved by 13.54% compared with the basic FCM.The accuracy of the failure type division is better.In the data screening of wrong division,on the one hand,this paper compares the above division results with the measured transformer failure records,Select the wrong sample for further fault type identification and diagnosis.On the other hand,in order to solve the problem of no comparison of the actual measurement records in the actual fault diagnosis after the initial division,this paper uses the comparison of cluster membership to outliers.Filter by point to extract the sample data that may be misclassified to the greatest extent.Finally,a transformer fault classification model based on quantum genetic neural network is built for the fault classification and fault identification of the extracted fault data and outlier data.Based on the DGA data after feature extraction,the genetic algorithm is optimized through the qubit coding and quantum revolving gate method in quantum computing,and the weights of neural network neurons are optimized in the model to reduce local optimization problems and improve the entire The accuracy of the model.Experimental analysis shows that the accuracy of transformer fault classification optimized by quantum genetic algorithm is improved by 5.36%,and the accuracy rate is 91.70%.