Abstract: The paper introduced a PID self-tuning control system which was established with neural network technology,gave structure of the system and analyzed structures and learning algorithms of BP neural network and RBF neural network in details.The system adopts three-layer BP neural network and its outputs are parameters of PID controller.It identifies controlled objects through variable-structure RBF neural network,provides gotten gradient information of output to input to BP neural network,and BP neural network optimizes parameters of PID controller according to the information.The simulation result showed that to nonlinear system whose parameter perturbation is larger,the system has fast convergence speed,strong ability of dynamic response,good stability and stronger robustness and adaptability.