The traditional fault diagnosis method of dissolved gas analysis for oil immersed transformers has the problem of slow fault diagnosis speed.This paper proposes a multi-scale fusion stacked Auto-encoder fault diagnosis method for oil immersed transformers.First,obtain the infrared detection atlas of oil immersed transformer high-voltage bushing,then cut and process the atlas into grayscale images,flatten these grayscale images into one-dimensional feature vectors,and input them into SAE,and extract features from the data by setting the number of different hidden layers to obtain the cod-ing part taken from the encoder.These features are accumulated to obtain hidden layer features at different scales.Then these features are input into the kernel extreme learning machine classification model optimized by the sparrow algorithm for fault diagnosis.The fault diagnosis method proposed in this paper has a high accuracy in fault diagnosis.
维普
万方数据
