针对传统PI控制存在的动、静态控制性能较差的缺点,提出一种基于神经网络的PI控制方法。将神经网络与传统PI控制结合,构建神经网络PI控制系统,建立三层BP神经网络,并通过梯度下降法对各项参数进行修正,从而实现k_p、k_i参数的在线调节。仿真及试验证明,与传统PI控制方法相比,使用神经网络的PI控制系统在不同外部条件下都具有更快的响应速度和更小的超调量,可明显提高系统的动、静态性能。

Aiming at the disadvantages of traditional PI control such as poor dynamic and static control performance, a PI control method based on neural network was proposed. Combining neural network and traditional PI control, we constructed the neural network PI control system. Threelayer BP neural network was established, and through the gradient descent method the parameters were modified to achieve on line adjustment of k_p and k_i parameters. Through simulation and experimental verification, it was proved that compared with the traditional PI control method, the PI control system using neural network had faster response and smaller overshoot under different external conditions, which could significantly improve the dynamic and static performance of the system.

TM 33

国家自然科学基金项目(51275137)