Research On Transformer Fault Diagnosis Method Based On Deep Learning

With the improvement of voltage level and the increasing demand for electricity,the security of power equipment in power system is facing serious challenges.Due to the important role of power transformer in all power equipment,the operation status of transformer will directly affect the operation status of power system,so it is of great significance to diagnose potential faults of transformer.Dissolved gas analysis(DGA)in oil is the primary method for transformer fault analysis.Early transformer fault diagnosis methods are based on DGA theory.The main diagnostic methods include ratio method and shallow machine learning method.These methods have achieved good results in a period of time.With the improvement of voltage level and the development of sensor technology,a large number of unlabeled data are generated,which can be used to diagnose transformer faults.As a result,higher accuracy is required.Thanks to the popular application of deep learning,this paper applies deep learning algorithm to transformer fault diagnosis.A transformer fault diagnosis model based on stack sparse automatic encoder network and a transformer fault diagnosis model based on stack sparse noise reduction automatic encoder network are constructed.The research contents of this paper are as follows:Firstly,the background and significance of this study are introduced,the current methods of transformer fault diagnosis and their problems are described,and the current research status of in-depth learning is introduced in detail.Secondly,the source of dissolved gas in power transformer oil and the relationship between dissolved gas in oil and fault types of powertransformer are introduced.The feasibility of fault diagnosis of power transformer based on the type and content of dissolved gas in oil is illustrated.Finally,a data set of dissolved gases in oil is set up.A transformer fault diagnosis model based on stack sparse automatic encoder network and a transformer fault diagnosis model based on stack sparse denoising automatic encoder network are constructed.Its number of hidden layers,hidden layer nodes,sparse parameters,noise figure and iteration are studied.The optimal network model parameters can be determined by adjusting the parameters under the ratio of training and testing samples.Testing on the test set shows that the correct rate of diagnosis is about 97%,and compared with traditional transformer fault diagnosis methods,the correct rate of diagnosis is increased by about 15%.The results show that the diagnosis model constructed in this paper is feasible and effective in transformer fault diagnosis scheme.