Research On Intelligent Predictive Control And Its Simulation Application In Circulating Fluidized Bed Boiler Combustion Process Control

Predictive control is a kind of control strategy which can meet the requirements of industrial production control. Through the means of predictive model to describe the relationship between input and output of the system, Predictive control compares the output of the predictive model and the reference trajectory, meanwhile it obtains the optimal control in the rolling optimization. Based on the improvement of Vortex Search algorithm, this paper argues that a predictive control method of Fast Learning Network can be applied in the field of circular fluidized bed boiler main steam pressure and bed temperature control. The results show that the predictive control method can achieve the expected control effect. The specific work of this paper is as follows:In this paper, an optimization algorithm, the Vortex Search(VS) optimization algorithm is introduced and improved. It analyzes the theory and disadvantages of VS optimization algorithm. Still based on the theory Chaos Vortex Search(CVS), optimization algorithm and Opposition Vortex Search(OVS) optimization algorithm are proposed. CVS optimization algorithm applies chaos theory to the generation of alternative solutions to improve the accuracy and convergence speed of the algorithm. OVS optimization algorithm applies opposition theory in the generation of alternative solutions, while taking into account the speed of the algorithm to improve the optimization performance of the algorithm late iteration at the same time.Aiming at the problem of random question in the process the Fast Learning Network does the input weights and hidden layer node threshold, this paper declares a Fast Learning Network modeling method based on CVS algorithm. The method applies the input weights and hidden layer nodes threshold range and number as CVS optimization algorithm to find optimal space and dimension, the network output and the training data output are root mean square error as the fitness value optimization. In this way, the training results of the Fast Learning Network can be determined to achieve the optimal input weights and hidden layer node thresholds. By comparing the common Fast Learning Network and Extreme Learning Machine, the thesis proves that the Fast Learning Network be optimized by CVS has better accuracy and stability on the modeling practice. To refer the OVS algorithm, a predictive control method of Fast Learning Network based on OVS is proposed in this article. It identifies the prediction model by CVS and calculate the optimal control in rolling optimization procedure by OVS, it also makes this predictive control method to have good control quality.The paper applies the Fast Learning Network which based on OVS-optimized in the field of circular fluidized bed boiler main steam pressure and bed temperature control; it also practice the test of simulation in the process of comparing ordinary PID control and single neuron PID in many cases. The experimental result shows the Fast Learning Network predictive control, which based on OVS, has good speed of response, anti-noise and anti-interference capability. In the case of main steam pressure and the bed temperature controlled object model changing, the output of the system can still be quickly recovered to the set value.