Abstract: Aiming at the problem that it was not easy to deduce the transfer function of the controlled object when three parameters were adjusted in the traditional PID (prop
orationalintegraldifferential) controller, and these parameters were difficult to be adjusted manually, we proposed a new algorithm to adjust the PID controller parameters. The algorithm combined neural network with genetic algorithm. First, the algorithm used the simulation function of the neural network to assist the genetic algorithm calculating the fitness, and the trained neural network was used to simulate the controlled object. Then three parameters of the PID controller were constantly optimized in the evolution of genetic algorithm. Compared to the traditional parameter tuning, the simulation results show that the proposed algorithm has strong robustness and fast response speed.

Key words: neural network, PID parameter selftuning, computer application, genetic algorithm