The application of machine algorithms to the early warning and diagnosis of electrical equipment failures has become more and more extensive. Because it can effectively prevent equipment damage and further damage to the power grid and cause irreparable consequences, it has a significant effect on the maintenance of stable operation of the power system. At present, the main application of machine algorithms are: BP neural network, SVM, deep learning (including: RNN, CNN, DBN). This paper first gives an overview of the development and basic concepts of machine algorithms. Secondly, the basic principles of various machine algorithms and their applications in early warning and diagnosis of electrical equipment failure are introduced. Finally, this paper looks forward to the development trend of deep learning in fault warning and diagnosis.