The Research Of Neural Network Control System With Disturbance

Based upon the deficiencies of the robustness in neural network control system, take example by neural network model reference adaptive control system, show the bad result, analyze the effecting mechanisms, point two potential reasons. And based on these, this paper presents a novel robust neural network model reference adaptive control structure and a novel robust neural network adaptive control algorithm. In the end, we apply them to system control, and have a good result. Compared with the classical neural network model reference adaptive control system, the algorithm which this paper present possesses some advantages as following: 1 On the initial stages of control, this novel structure, which integrate with traditional control strategy by robust gene, improve the robustness and stability of closed system. 2 A neural network is used as compensator to compensate the influence result from robust controller whose action become weaken. It can improve the dynamic performance and steady accuracy of control system. 3 Based on control algorithm of the classical neural network model reference adaptive control system, this paper present a novel. It adopts amendatory data of input and output as supervised signal. So it can eliminate the effects of unmodelled dynamics and disturbance. A good deal of simulation results show the novel control algorithm has the advantage over basic algorithm in the convergent precision, especially on resisting to the noise effects.