基于M估计稳健回归的多谐波源责任区分

Quantifying Harmonic Responsibilities of Multiple Harmonic Sources Based on M-estimation Robust Regression

为对多谐波源系统中各谐波源的谐波责任进行正确区分,提出一种基于M估计稳健回归的评估方法。首先根据公共连接点处各谐波电压相量之间的关系定义了评价谐波源责任的指标因子。利用时间序列分割法选择出满足分析要求的谐波电压和谐波电流历史数据。利用反复加权的最小二乘迭代法求解回归系数,根据回归系数计算出谐波责任指标因子,进而评估多谐波源系统中各谐波源的谐波责任。该方法大大削弱了数据异常值对回归系数的影响,克服了传统最小二乘法对奇异数据敏感、评估精度低的缺点。仿真分析和实际算例均验证了该方法的有效性以及相对于传统最小二乘法的优越性。

Based on the M-estimation robust regression, a method to determine the harmonic contributions of multiple harmonic sources was proposed in this paper. Firstly, the harmonic impact index factor was defined based on the phasor relationships among all harmonic voltages on the Point of Common Coupling. Then the useful history data of the harmonic voltages and harmonic currents were selected out by using the time segmentation method. After that, the regression coefficients were estimated by adopting an iteratively reweighted least squares algorithm. Finally, the harmonic impact index factor was calculated based on the regression coefficients and then the individual harmonic contribution of multiple harmonic sources was obtained. The proposed M-estimation robust regression method weakens the bad effect of the singular data on coefficient estimation, so it has higher estimation accuracy than the traditional least square method. Both simulation and case study results both verify the validity of the proposed method and its superiority over the least square method.