基于神经网络控制的有源滤波器的研究

Research on active power filter with a neural network controller

随着电力系统中负荷的多样化,使得电网的谐波干扰和无功冲击越来越严重,对电力系统的安全可靠运行和用电设备的正常稳定工作构成了威胁,因此谐波的实时检测和补偿也就变得越来越重要.基于此背景,提出了一种基于神经网络控制器的新型有源滤波器,通过神经网络控制器对电网谐波进行分析预测,并通过开关决策来控制补偿电流发生电路从而实现滤除谐波的功能.其中,神经网络的训练是采用遗传算法和小脑模型来实现的.仿真结果证实了神经网络控制器的有效性.

The harmonic interference and reactive power impulsion in power systems become worse because the kinds of loads in electric power systems are more.The harmonic analysis and detection in real time are important for the safety and reliability of electrical equipment.A novel active power filter (APF) with a neural network (NN) controller is proposed,which is used to analyze current harmonics in power system.The APF with a NN controller controls the current compensation generator by switch decision-making and injecting compensation current into the system.The NN controller is trained by evolutionary computation and cerebella model articulation controller (CMAC).The results of simulation confirm the validity of the proposed NN controller.