Based on LM algorithm of an artificial neural network

a model for fault warning diagnosis of wind power units has been established.The LM algorithm of neural network is an improved BP algorithm

but the problem of easily falling into local minimum is still existing.For this

a momentum method has been applied to bounce out from the local minimum

achieving good effectiveness.By using Matcom tool

the hybrid programming method between VC++ and Matlab softwares has been realized

the problem of complex matrix calculation encountered in the process of algorithm programming being resolved

enhancing the performance of algorithm.The improved algorithm has been used in training and inspection of the fault warning diagnosis model for a wind power unit with structure of 15-22-4

result shows that the converge performance of the network is good.The model of said algorithm has already been embedded into software of fault warning diagnosis for wind power units.