Modeling And Optimal Control Of Double Coal Based On Intelligent Algorithm

Double input and double output steel ball coal mill(hereinafter referred to coal mill) is widely used in coal-fired power plants in China.Coal pulverizing system is the main equipment,which is characterized by high efficiency grinding,equipment running stable, low rate of repair,and other efforts.It has many advantages,at the same time,there are a lot of questions:First,consumption of high milling,up to the plant about 20%of electricity consumption;Second,the low level of automation and control mill.This dissertation is for the above two issues are analyzed.This dissertation first introduces the overall structure,characteristics and the control tasks of coal mill,and gives the existing control method an in-depth study.Precise modeling of coal mill is great significance for its performance monitoring and control.In the structure and running mechanism in depth analysis on the basis of the relevant parameters,in the GUI of Matlab7.0 environment,set up three input,three output neural network model of the coal mill,using particle swarm optimization to improve the initial weights of network and threshold value to optimize and improve the speed of network convergence.Application of Fuxin Power Plant Unit BBD-4360-type 80 group training samples to train the model,20 groups of simulation results of the test samples show that the neural network model with higher accuracy.After the model set up,through the existing control method to conduct in-depth research,the use of predictive control modeling convenience,robustness,and other advantages,based on the idea of multi-variable nonlinear inverse program.In order to overcome the insufficient of traditional methods in nonlinear modeling and control algorithm to realize,using neural network nonlinear approximation ability to set up prediction model,using another feed-forward neural network to achieve the inverse system, and as controller.Application real-time monitoring data,the neural network inverse controller is trained.Finally,in the Matlab7.0 environment,with the traditional mathematical model of negative pressure at the entry,import and export pressure,and outlet temperature simulation compared of three indicators.Simulation results show that the dissertation proposed combining predictive control and dynamic inverse control can overcome the time-varying coal mill model mismatch caused by the impact of system time-varying, uncertainties and environmental interference,realize decoupling control,consistent with the design specifications and safety requirements,has a certain practical application value.