Oil-filled electrical equipment fault diagnosis method based on graph attention neural network

The invention relates to the field of artificial intelligence and power systems, in particular to an oil-filled electrical equipment fault diagnosis method based on a graph attention neural network. According to the method, five kinds of characteristic gas generated when the oil-filled electrical equipment breaks down and correlation possibly existing among the gas are used as graph data input, six kinds of fault types are used as output, the graph attention neural network is built, and fault diagnosis of the oil-filled electrical equipment is achieved. The network not only can learn the nonlinear relationship between the characteristic gas and the fault type, but also can learn the mutual relationship between the characteristic gases. According to the method, an attention mechanism is adopted, more implicit relations are fully excavated through transverse and longitudinal deep excavation, the problem that some fault types of the oil-filled electrical equipment cannot be diagnosed is solved, and the fault dia