This paper mainly sets up a nonlinear steam temperature model under different loads for control purpose by using a sliding window nonlinear identification technique based on BP neural network (BPNN) for single input and single output nonlinear plant modeling. Because of some obvious advantages in computation and simulation, a discrete intelligent simulation prediction (DISP) model based on BPNN for power plant thermal process nonlinear superheat steam temperature is created. The proposed procedure applies to the process that has linear model around each operating point. The main feature of nonlinear modeling based on BPNN is to cover a number of operating points within one global model, to follow the dynamic and static change of the process over a wide range of operating conditions for even large variation of the load. The technique developed here achieves this objective, it starts from the input-output description of the process and ends with a BPNN nonlinear model. The simulation and prediction results show that DISP model has good robustness and noise filtering ability with small and medium noise and also has merits of high accuracy and fast speed.
展开▼
