Research Of The Detection And Control Strategy Based On Neural Networks

Along with the rapid development of power and electron technique, an increasing number of power and electron equipments are used diffusely in the course of industry-agriculture production and human daily life. The power and electron equipments have the characteristic of non-line load, though they can improve the efficiency of industry-agriculture production and the quality of human life, they also make much harmonics and reaction power current into power grid and brought a lot of power quality problems. Active power filter has a strong point that it can achieve the function of harmonics and reaction power compensation. And it can also well compensate the objects of which frequency and amplitude changes in the same time, this prosthesis the LC passive power filters'shortcoming very much.Firstly, the work principle and technical development of active power filter is introduced in this paper. The shortcomings and advantages of all kinds of harmonic detection and current control is also summed up. The neural networks is proposed make up for the faults of the existing technology.A novel harmonic detection method based on adaptive neural networks is proposed in this paper. This method has high accuracy, adaptability to load currents. On the basis of traditional hysteresis band control, a BP neural network is trained to achieve the function of hysteresis band controller and improve the performance, carrying out the dynamic trace and control of command current.The system is robustness and insensitivity to the parameter changes due to the BP controller.The output LCL filter of active power filter determines the compensation performance and stability,it is a key problem of the desigh of the APF system.It is blind and imprecise to determine the parameter of the output filter by engineering experience. So the parameters of LCL filter of 200A experimental prototype are optimized by genetic algorithm in this paper. This method can meet the various technical requirements. In addition, solving the problem of excessive energy consumption produced by damping resistor, we use virtual resistance made by control strategy, improving efficiency of the APF system and avhieve great performance.MATLAB models corresponding to the detection methods and control strategies proposed in this paper are given, their validity and feasibility are validated by simulation.