三相四线有源电力滤波器新型神经预测控制

Novel Neural Network Predictive Control Scheme for Three-phase Four-wire Active Power Filter

为补偿有源电力滤波器的控制延时,提出了一种新型三相并联型有源电力滤波器的RBF神经网络预测控制方案。建立三相四线制并联型有源电力滤波器的数学模型及电流预测控制的离散化模型,设计神经预测控制器,通过在线训练权值提高控制精度,控制有源滤波器产生用以抵消非线性负载的谐波电流。应用Matlab对该方法进行了仿真,并在以DSP为核心控制器的基础上对其进行了实验研究,仿真和实验结果表明该预测电流控制方法实时快速性好、动态特性好,验证了所提方案的可行性和正确性。

In order to compensate the control delay of active power filter（APF）,a novel predictive control strategy based on radial basis function neural network is proposed.The mathematic model of three-phase four-wire APF and the discrete model of predictive current control scheme are established.Then a neural network based predictive controller is designed.All the connecting weighting coefficients are trained on-line to improve the control accuracy,which makes the active power filter working properly to cancel load＇s line harmonic current.The simulation studies are carried out using Matlab,and an experimental study system has been conducted based on DSP.Simulations are performed to verify the feasibility and accuracy of the proposed method,and experimental results show that the predictive current control method has a good real-time quality and good dynamic characteristics.