Fault Diagnosis of Power Transformer Based on Stacked Sparse Auto-Encoders and Broad Learning System

The power transformer is an important part of the power system, and the diagnosis of the power transformer plays an important role in its safe operation. In order to improve the reliability and accuracy of power transformer fault diagnosis, this paper uses the dissolved gas in power transformer oil to propose a fault diagnosis method based on stacked sparse auto-encoders(SSAE) and broad learning system(BLS). The sparse auto-encoder has a powerful data reconstruction ability, which can extract the essential characteristics of the fault data by reconstructing the original data, and improve the diagnostic accuracy. The broad learning system reconstructs the network through incremental learning, and uses the pseudo-inverse calculation method to quickly solve the hidden layer-output layer weight, avoiding the use of gradient update method, improving training speed and preventing local optimization. Using KNN classifier to realize the feature clustering and label classification of the target domain samples. The simulation results show that the proposed method can effectively identify the fault type of power transformers with a satisfactory accuracy.