Research On Predictive Function Control Methods Based On Neural Networks

It is a new trend combining the predictive control and intelligent control to improve the qualities in industry process control. Neuron-control is an important branch in intelligent control fields, which is widely used for modeling and controlling. The predictive control based neural networks is becoming a significant research issue. The theory and applications of predictive function control (PFC) based on neural networks are studied in this thesis. The main contents are as follows:1. Considering the neuron adaptive ability and the PFC robustness, a neuron predictive function control method for plants with uncertainties is proposed. With an example of the basis weight control of a paper machine, the experiments are made. The simulation results show this algorithm has good performance.2. To simplify PFC control law, the simplified multi-variable PFC method is proposed. In which, the neuron is used to produce the control signal for reducing calculation. This method is used to control the unit power plant of 125MW and 300MW, and simulation results show that the satisfactory performance is reached.3. The multiple model PFC method for nonlinear plants is proposed, which is based on modeling of fuzzy recurrent neural network (FRNN). The genetic algorithm is used to optimize the parameters of the RFNN model and every sub-model has its own PFC control law. then the control signal is integrated according to Gauss functions in the multiple models. A pH neutralization process is taken as a plant to implement this method. The simulation shows the efficiency of this control method.