Fault Prediction Method Research of the Power Plant Fan
To solve the problem of the power plant fan fault prediction,proposed that combining with neural network method and nonparametric density function estimation methods based on parzen window the estimation to achieve fault detection.To improve the prediction performance of neural network,used PSO method,which can realize weights optimization of the neural network prediction,avoid falling into local optimum.Using sliding time window achieve the multi-step prediction of the neural network,and ensure the prediction accuracy.Then,fault is predicted by prediction residuals through density function estimation and hypothesis test.Finally,by using the vibration fault prediction of the air feeder of a power plant in Shanxi as research object to test this method,the simulation result illustrate this fault prediction algorithm can predict the fault of fan timely and effectively.
power plant fan fault prediction neural network PSO non-parametric density estimation
Jianping Sun Ming Gao Yalun Li
Automation Department, North Electric Power University, Baoding, Hebei, 071003, China
国际会议
太原
英文
99-104
2012-08-17（万方平台首次上网日期，不代表论文的发表时间）