基于在线参数辨识补偿的直线感应电机低开关频率模型预测控制策略

Low Switching Frequency Model Predictive Control Strategy Based on Online Parameter Identification Compensation of Linear Induction Motor for Urban Rail Application

为降低变换器硬件成本及开关损耗,强电流大功率的轨道交通用直线感应电机及驱动系统通常需运行在低开关频率模式下。但是,传统的低开关频率控制方法经常会在引入较为严重的谐波电流和推力波动的同时放大微处理器离散化带来的影响,从而恶化观测器性能。针对该问题,该文提出一种带在线参数辨识的多步长有限集模型预测控制策略,可有效降低低开关频率下的电流谐波,同时利用高精度参数增强系统的可靠性。首先,采用目标函数带开关项的模型预测控制算法,实现了高采样频率下的低开关频率控制,并借助多步长模式成功抑制了因开关频率降低所引起的谐波含量上升等问题。其次,详细对比分析了两种方法控制下的低开关频率模式对观测器性能的影响。基于分析结果,同时考虑到轨道交通用直线感应电机中励磁电感受边端效应影响程度较深,进一步综合分析了该模型预测方法的参数敏感性并引入参数在线辨识算法,实现了低开关频率下的励磁电感的精确估计,提高了模型预测控制算法的预测精度和参数鲁棒性,降低了驱动系统的谐波含量。大量仿真和实验充分表明,所提出的方法可很好地实现低开关频率下直线感应电机励磁电感的准确跟踪,配合多步长模型预测控制算法,可对驱动系统电流谐波进行有效抑制。

To reduce the hardware cost of converter and switching loss, the linear induction motor(LIM) and drive system for urban rail application with strong current and high capacity usually need to operate in a low-switching-frequency mode. However, the conventional low-switching-frequency control method often amplify the influence of discretization error in micro-processor while introducing severe harmonic current and thrust fluctuations, thereby deteriorating the observer performance.Therefore, a multistep finite control set model predictive control(FCS-MPC) strategy with online parameter identification is proposed, which can reduce the current harmonic under low-switchingfrequencies and enhance the reliability of system by using high precise parameters. Firstly, the MPC algorithm with switching item in cost function is adopted to realize low switching frequency control at high sampling frequency. Furthermore, the multistep mode is used to suppress the increase of harmonic content caused by the reduction of switching frequency. Then, the influence of the two different low-switching-frequency realization methods on performance is compared. Considering that the magnetizing inductance in LIM for urban rail is deeply affected by the end effect, the parameter sensitivity of the MPC method is further analyzed and the parameter identification is introduced, which realizes the accurate estimation of magnetizing inductance at low-switching-frequency, improves the precision of prediction accuracy and parameter robustness of MPC, and reduces the harmonic content in the drive system. Simulation and experimental results show that the proposed method can well realize accurate tracking of the magnetizing inductance in LIM at low-switching frequency, and effectively suppress the current harmonic of the drive system combined with the multistep FCS-MPC algorithm.