Use the BP neural network to evaluate the risk of Thermal Project Investment.An evaluation model is established first.Then the connection between weights and threshold value,which are restrainted by the error functions in the model,are calculated.And the output layer is exported as a network result.The sample sets of evaluate network are evaluation results of high reliability and authoritativeness,which can be summarized by experts from previous data of similar Thermal Project Investment processes.Last,store the connection weights and threshold value of the evaluations,and forecast projects to be evaluated.