神经网络自适应PI控制器在有源滤波器中的应用

Application of neural network adaptive PI controller in active power filter

针对有源电力滤波器的电流跟踪控制问题,设计一种神经网络自适应PI控制器。该控制器将神经网络技术和PI控制器设计方法相结合,控制器采用PI控制器结构,具有结构简单、计算时间短、易于实现等优点。同时利用神经网络技术,使其输出作为最优控制规律下的PI控制器的参数值,并根据误差大小对控制器参数进行在线实时自适应整定,从而满足大工况、全工作条件和最优性的要求。仿真实验表明,神经网络自适应PI控制器较一般的PI控制器有更快的响应速度和更高的补偿精度,而且经过神经网络自适应PI控制器作用后,其电网电流的谐波畸变率和电流跟踪误差均降低到PI控制器的55%左右。

For current tracking control problems in active power filter （APF）, a neural network adaptive PI （NNAPI） controller is designed. It combines the neural network technology with PI controller structure. It adopts the PI controller structure, which has the advantages ofsimple structure, short computing time and easy realization. Meanwhile, the neural network technology is used to make the output of APF be the PI controller parameters values under an optimal control law, and conduct an online real-time adaptive setting of the parameter values according to the error, thus to meet the requirements of full range working conditions and optimality. Simulation experiments show that NNAPI controller has quicker response speed and higher accuracy of compensation comparing with general PI controller. When NNAPI controller works, its total harmonic distortion rate of electric current and the current tracking error are reduced to around 55 percent of that using PI controller. This work is supported by the National High Technology Research and Development Program of China（No.2008AA04Z129）. National Natural Science Foundation of China （No. 60504010）.