Study on Improved Neural Network PID Control of APF DC Voltage

According to the active power balance principle, the paper analyzed the approximate mathematical model of APF. In order to optimize the control effect of DC bus voltage in APF, PID control method based on improved BP neural network is adopted to do closed-loop control to the system. The two strategies, adding momentum method and adaptive learning rate adjustment, are combined to improve BP network, which can not only effectively suppress the network appearing local minimum but also good to shorten learning time and improve stability of the network furthermore. The improved BP network adjusted the parameters such as KP and KI of PID controller according to the operation state of the system and realized optimum PID control. The experiment studies show that on condition of load power and harmonic content changing, APF system, controlled by PID control method based on improved BP network, can assure the harmonic distortion keeps in an allowed range and the dc side voltage becomes stable in a short time.