State monitoring and fault prediction of power grid equipment are carried out by using deep learning technology

With the increasing demand for power in modern society and the increasing complexity of power grid, traditional power grid monitoring methods are facing severe challenges in real-time and accuracy, and it is urgent to find new technical solutions. This study is based on deep learning technology and takes power transformer, a key equipment of power grid, as the research object, to establish transformer fault diagnosis model and prediction model of dissolved gas concentration in transformer oil, so as to provide reference for transformer condition monitoring and fault prediction. A fault diagnosis method of transformer based on Focal loss stack sparse noise reduction autoencoder was established by deep learning framework, and the fault diagnosis results were output by Softmax classifier. In addition, this paper establishes transformer fault diagnosis model based on actual case data, and the diagnosis accuracy of the trained model reaches 93.3%. On this basis, the variation trend of dissolved gas concentration in transformer oil is forecasted in order to realize transformer fault warning effect. The prediction model of dissolved gas concentration in oil based on TPE algorithm was established through deep learning framework, and the single step prediction result was finally given. By harnessing the full potential of deep learning, we will support the construction of a smart grid and drive the entire power industry towards smarter and more reliable solutions to ensure the safe and stable operation of the grid.