An Intelligent PID Controller Based On Improved DE Algorithm

PID control mode is convenient to use for its simple structure and working manner, but its basing on strict and exact mathematics model and its outdated parameters tuning styles make it work ineffectively and fall demands of complex systems such as non-line、 time-varying system. Intelligent PID control mode emerges with the development of swarm intelligence optimization and intelligent control mode. An intelligent PID controller based on improved differential evolutional algorithm is proposed in this paper.In recently DE algorithms, mutation rate and crossover rate tuning styles are blind and non-self-adapting, so convergence speed is not improved apparently; group diversity is another aspect which is not paid enough attention, and the deficiency of population diversity can lead premature convergence. An improved DE (KDE) algorithm is proposed in this paper to improve the convergence speed and avoid premature convergence. KDE algorithm adapts multi-group parallel to make sure group diversity and feedback to tune parameters in self-adaption mode. And the investigation of KDE algorithm with a set of nineteen benchmark problems shows KDE algorithm outperforms, or at least comparable to the DE algorithm and some other adaptive and self-adaptive DE algorithm in terms of convergence speed, average fitness and number of iterations.Based on KDE algorithm, the intelligent PID controller presented in this paper is a compound controller integrated by fuzzy control technology and neural network control technology which executes "fuzzy control" with multi-layer structure of neural network and enhance the control accuracy by approaching function characteristics of neural network. In order to make the controller work more efficiently, KDE algorithm is used to optimize the initial value of membership function parameters and control rules, then it is the hybrid algorithm combined by KDE algorithm and BP algorithm solving the self-adaptive adjustment of membership function parameters and network weights on line, which accelerates the convergence speed and improves control accuracy. The experiment on three-order system verifies the PID controller makes the system response much more quickly while guarantees better control performance on system overshoot. At the last of this paper, the intelligent PID controller is employed in servo system. Design and simulate the three-loop control system of servo system, and the simulation results shows that, in contrast to conventional PID controller, the control system based on the intelligent PID controller proposed in this paper has faster response speed and better following performance.