Fault Diagnosis Of Oil Immersed Transformer Based On Improved Neural Network

Oil immersed transformer is one of the most important equipment in power system.Once it fails,it will have a negative impact on social security and economic growth.When the transformer fails,how to quickly and accurately identify different types of faults is the first priority in power system fault diagnosis.At this stage,DGA has been determined to analyze the dissolved gas in transformer oil in the following ways: 1.Three ratio method,obtaining the content and type of five gases for ratio,and judging by content ratio;2.The voiceprint vibration method judges the transformer by the sound and vibration when the transformer is running;3.The characteristic gas method is used to judge whether the transformer has a fault through the characteristic gas.Among the three methods,the three ratio method can be used to judge and analyze whether the transformer has failed only when the transformer has failed.If the transformer operates under normal conditions,we use the three ratio method for calculation and analysis,which will make technicians misjudge when judging the transformer fault,thus causing unnecessary waste of manpower and material resources.The voiceprint vibration method is usually used to diagnose equipment defects and faults,so there is a lag in diagnosis.The special gas method is of little significance for transformer fault diagnosis due to its real-time nature.Therefore,how to quickly carry out transformer fault diagnosis is of great significance to both the stability of power system and the prevention of transformer faults.Moreover,these methods still have the following shortcomings in transformer fault diagnosis: 1.A large number of transformer fault data are required to make accurate judgments,and the accuracy rate drops;2.The transformer fault cannot be detected online,which takes a long time;3.Poor anti-interference ability;4.The robustness is poor when judging transformer faults.In view of the low accuracy of traditional methods and the instability of discrimination methods,a fault diagnosis method for oil immersed transformer based on improved neural network is proposed.The transformer DGA database is taken as the research object,and based on this,the transformer fault diagnosis is carried out by KNN-BP model constructed by optimizing the weights and thresholds of the BP neural network with KNN.It only takes a few seconds to complete the troubleshooting.The main research results and contents of this paper are as follows: Based on the background,significance and research status of transformer fault diagnosis,the characteristics of gas production under different dissolved gases and fault conditions are described.Secondly,through the deficiency of traditional methods,an online detection method is proposed,which is trained and simulated by BP neural network,and the recognition rate is 65%.It not only converges slowly,but also vibrates in the training process,and the recognition rate is low.It needs to further improve the recognition rate,so it is used to improve the accuracy by optimizing the BP neural network.Because of the shortage of the original data,it can not meet the transformer fault prediction very well,so this paper proposes a variational self encoder to expand the original data.A certain amount of data can be expanded to obtain an equal amount of reliable data,and the problem of insufficient data can be solved through the variational self encoder.The fault identification rate and accuracy are improved.According to the different fault types corresponding to different characteristic gases in transformer oil,a transformer fault diagnosis method based on KNN optimized BP network model is proposed.In order to further improve the fault identification rate and accuracy,the concept of KNN network is introduced.Through BP neural network model and KNN neural network model,the KNN-BP model is established,and the modified data of variational self coding is used for simulation.The identification rate is90%.From the simulation results,it can be seen that compared with the other two algorithms,KNN-BP neural network has the lowest error in fault diagnosis,the best stability,and greatly improved accuracy.It is an effective method for transformer fault diagnosis.Comprehensively using the technical paper of this research,the KNN-BP neural network fast identification transformer fault diagnosis model is constructed.Finally,the BP neural network and KNN-BP network are simulated and compared,which makes the convergence algebra and diagnosis accuracy of its transformer fault diagnosis significantly improved.Through this way,the transformer fault is judged in a timely manner,ensuring the safe operation of the transformer.Figure [23] Table [15] Reference [70]...