Research On Model Method Based On Data For The Complex Thermal System

In this paper,the founditon of model in complex thermal system was the main line, the partial least-squares regression and support vector machines regression method which based on data-driven model were studied,an improved method on gross error detection and correction was put forward;Aimed at the problems occurred in the operation of Power generating units,separately found the main steam flow on-line prediction model based on PLS regression and the mid-point temperature prediction model based on LS-SVM.To resolve the disadvantages of parameter selection method based on grid search,particle swarm optimization algorithm was presented to select the parameters.The simulation results indicated that:the PLS method can deal with the problem of multiplicity effectively,but weakly deal with the problem of strong non-linear,and the PSO-LS-SVM method can solve the model of non-linear,and faster Fast,high-precision.