| PID control is widely adopted in many fields because of its simple structure, highreliability and easily implementation. The traditional PID has good control effect if theparameters of system model have not big variation. However, with the development of themodernindustry,theprocess ofindustrial productionis sometimes nonlinear,uncertainandisdifficult to establish accurate mathematic model, so using traditional PID controller isimpossibletoachieveperfectcontrollingeffect.Artificial neural network(ANN) is an important branch of intelligent control, it has theability of self-study, self-adapting and self-organizing. So combining the ANN and the PID,form the compound ANN-PID controller which need not set the accurate mathematic model.It identifies and sets the controlled process parameter automatically, and can solve theproblem of conventional PID controller with difficult parameter setting, unforcementrobustnessandnot real-timeparametertuning.The work of this paper mainlyconsists of two parts. At first, analysis the traditional PIDalgorithm deeply. For the system's request of the integral: when the error is biggish, theintegral should be reduced, contrarily, the integrate should be strengthened. Based on thisideal, put forward the elastic-integral PID algorithm based on the shifted integral. Thesimulation shows that the algorithm can make the control system more steady. In the nextplace, BP neural network PID controller is researched primarily in the paper. With the BPalgorithm, the choice of initial weight influences on the appearance of local minimum pointand improvement of network constringency. If the choice of scope is not right, at thebeginningof the learning process, may arise the"pseudo-saturation"phenomenon, even enterthe local minimum point, and the network is not converged. Therefore, use the global-searchability of particle swarm optimization (PSO) to optimize the initial weights of BPNN, avoidthe possibility of slow constringency, the existence of local minimum, and so on. For theprecocious phenomenon of PSO, the idea of mutation will be introduced to the PSO, andmaketheimprovedIPSO-BPoptimizetheweightsofthePIDcontroller. The paper carries out the simulation experiments about the elastic-integral PID and theshiftedintegral PID algorithms. Theresults indicatethat elastic-integral algoritm can improvethe stability of the system. And the paper also does the simulation experiments of theimproved PSO modifying the parameters of intelligent BP-PID controller.Both the stepresponse curve and the error tracking curve reflect that the BP-PID control systemoptimizated by IPSO algorithm, has the advantage of quick convergence speed and goodsearching ability, can go into the steady state as soon as possible, reform the controlperformance,andgetthesatisfiedcontroleffect. |
