Research on Fault Early Warning of Wind Turbine Based on IPSO-DBN

Abstract

Aiming at the problem of wind turbine generator fault early warning, a wind turbine fault early warning method based on nonlinear decreasing inertia weight and exponential change learning factor particle swarm optimization is proposed to optimize the deep belief network (DBN). With the data of wind farm supervisory control and data acquisition (SCADA) as input, the weights and biases of the network are pre-trained layer by layer. Then the BP neural network is used to fine-tune the parameters of the whole network. The improved particle swarm optimization algorithm (IPSO) is used to determine the number of neurons in the hidden layer of the model, pre-training learning rate, reverse fine-tuning learning rate, pre-training times and reverse fine-tuning training times and other parameters, and the DBN predictive regression model is established. The experimental results show that the proposed model has better performance in accuracy, training time and nonlinear fitting ability than the DBN model and PSO-DBN model.