Model for Predicting the Major Faults of a Regenerative System Based on an Artificial Neural Network
Yan Li, Junhui Yu · Journal of Engineering for Thermal Energy and Power · 2011
To effectively predict the faults of a regenerative system,established were three error BP(back propagation) neural network models for predicting the fault signs and phenomena of a regenerative system based on the Traingda,Traincgf and Trainrp algorithm respectively.In such a case,the input layer was the fault signs and the output one was the fault phenomena.The data actually measured in a power plant were used to conduct a training and testing of the three prediction models.The training and testing results show that the model based on the Traincgf algorithm has the smallest testing error and a relatively quick converging speed.Its network was of a 9-7-9 structure with its momentum factor being 0.6 and the learning speed being 0.8.The error BP neural network model based on the Traincgf algorithm can effectively predict the fault phenomena of a regenerative system by using the fault signs,thus providing a certain reference value for testing the faults of a regenerative system.