无刷直流电机输出转矩的非线性偏最小二乘回归估计

摘要

针对无刷直流电机直接转矩控制系统中,电机输出转矩估计或观测存在需检测电机状态信号难以精确地检测、观测或计算等不可避免的难题,提出一种利用电机线电流和转速两个易检测信号进行电机输出转矩估计的方法.该方法首先基于拉丁超立方抽样方法确定建模数据,确保离散数据的可靠性;然后通过核函数将建模数据中的自变量变换到特征空间,将自变量与因变量之间的非线性关系在特征空间中线性地展开;最后在特征空间中利用偏最小二乘回归方法建立电机输出转矩的估计模型.通过对转矩估计精度和控制系统性能进行分析及实验验证,证明了所提转矩估计方法能准确地估计电机输出转矩,确保电机直接转矩控制系统的性能,该方法具有较强的实用性.%For the direct torque control (DTC) system of a brushless DC motor (BLDCM), there are some conundrums cannot be avoided such as motor output torque estimation or observation can not be observed or calculated precisely.A new motor output torque estimation method is proposed in this paper, which only use two easily measured state signals like the motor speed and line current to output torque estimation.The method first to determine the modeling data and to ensure the reliability of the discrete data based on Latin hypercube sampling method.A kernel function is then used to transfer the independent variables of the modeling data to the feature space, by which the nonlinear relationship between the independent variables and the dependent variables are linearly expended in the feature space.Finally, the motor output torque estimation model is established in the feature space based on the partial least squares regression method.By analyzing the accuracy and performance of the control system, it proves that the proposed method can accurately estimate the motor output torque and ensure the control performance when it is applied in a motor direct torque control system.Because this method has good practicability, it can provide a new method for torque estimation in the motor control system.