Multi-parameter prediction modeling for analyzing the trend of turbine oil online monitoring parameters

With increasing of the parameters in online monitoring of turbine oil, the effective and appropriate trend prediction and forecast of main operation parameters become feasible. And it is very important to determine the working condition of turbine system. Several methods used for forecasting the development of online monitoring parameters of the steam turbine oil were researched. After analyzing the merits and defects of the methods, the BP Neural Network model were selected to predict development of the main monitoring parameters, account for the variety and relevance of online monitoring data. And the results turn out to be satisfied. The research work in this paper includes the selection of BP Neural Network, designing of forecast plan, predict strategy, error analysis. The nonlinear multi-parameter prediction modal based on BP neural network was simulated by using MATLAB. Through analysis of experiment result, it is proved that this method is reliable. And experiment result shows that the predicted data are well correspond with verification data. Therefore, the modal is considered to be used in prediction the trend of particle concentration. It is furthermore used to predict and diagnosis the working condition of turbine system.