基于BP神经网络和分数阶PI^(λ)D^(μ)的VIENNA整流器控制策略

Control Strategy of Vienna Rectifier Based on BP Neural Network and Fractional Order PI^(λ)D^(μ)

为了解决传统比例积分(proportional integral,PI)控制器调节速度慢、追踪性能差等问题,提出了误差逆传播(back propagation,BP)神经网络和分数阶PI^(λ)D^(μ)(BP-FOPID)相结合的控制方法。首先根据维也纳(Vienna)整流器的拓扑结构推导出数学模型。接着根据数学模型设计双闭环控制,其中外环采用BP神经网络对参考电流进行非线性拟合,内环采用分数阶PI^(λ)D^(μ)控制器对外环输出的参考电流进行跟踪。另外,由于直流侧上下桥臂电容电压会存在不平衡的问题,本文采用了计及小矢量的改进空间矢量脉宽调制(space vector pulse width modulation,SVPWM)调制策略。最后,在MATLAB/Simulink中建立相应的仿真模型,仿真结果表明,本文所提的控制策略能够达到控制目标且控制性能上优于比例积分控制器。

For the problems of slow regulation and poor tracking performance of proportional integral controllers,a control method combining BP neural network and fractional-order PI^(λ)D^(μ)(BP-FOPID)was proposed.First,the mathematical model based on the topology of the Vienna rectifier was introduced.Then,a double closed-loop control was designed according to the mathematical model.The outer loop used a BP neural network to perform nonlinear fitting of the reference current,and the inner loop used a fractional order PI^(λ)D^(μ)controller to track the reference current output by the outer loop.In addition,due to the problem of imbalance in the voltage of the upper and lower bridge arms on the DC side,this article adopted an improved SVPWM modulation strategy that takes into account small vectors.Finally,the corresponding simulation model was built in MATLAB/Simulink.The simulation results show the effectiveness and superiority of the proposed BP neural network control and fractional-order PI^(λ)D^(μ)controller.