Research On Reverse Modeling Of Tube Wall Temperature Of High-temperature Heating Surface In Power Plant Boiler

The model of tube wall temperature of high-temperature heating surface in boiler is built based on reversed modeling theory using of the real-time data generated by power plant operation.While the data is preprocessed, the outliers of multipoint parameter with measurement redundancy are mined based on grubbs criterion, according to the continuty of measured values and the improved lauta criterion, in single, double point conditions, meanwhile, apply the relation of two points (related parameters) to judge outliers.The most dangerous tube wall temperature of heating surface,which is the underside of right side tube of final reheater, serve as the output variable of demanded temperature model from a practical view and according to mechanism analysis.The input variables are preliminarily picked form the steam side and gas side by thermal analysis of tube wall for the characteristic of reverse modeling,then them are filtered by grey correlation analysis which verified through the actual situation.The modeling nonlinearity is tested by F-test. The nonlinear model of wall temperature is achieved with least squares support vector machines, using hybrid genetic algorithm to optimize model parameters. The fitness function and the model input variables are changed to prove the the correctness of the modeling initial set. The predictive ability of moedling in the other load range indicated the applicability of variable conditions. The comparison with BP neural network model reflects the advantages of least squares support vector machine in the nonlinear modeling of wall temperature.The results show that the reverse modeling method can solve the nonlinear modeling in large load range of tube wall temperature of high-temperature heating surface, contribute to wall temperature monitoring and prediction,which is important to the working life of heating surface metal materials and the safe operation of boiler.