Harmonic Impedance Estimation Method Based on Data Filtering and Improved Rank Regression

In modern power systems, the integration of numerous power electronic devices leads to random fluctuations in harmonic impedance. To improve the accuracy of harmonic impedance estimation, this paper utilizes the similarity secondary measurement principle to improve the DBSCAN clustering algorithm. This improvement filters out measurement data where the harmonic voltage and harmonic current have an approximately linear relationship and categorizes data with different impedance values into distinct groups. Additionally, a robust weighting function is incorporated into the rank regression model to expedite calculations, enhancing the robustness and accuracy of estimation results. The effectiveness of the improved clustering algorithm is validated by analyzing the third harmonic data obtained from measurements and comparing the harmonic impedance magnitude curves before and after filtering. Furthermore, the accuracy of the improved rank regression algorithm is confirmed by estimating harmonic impedance using several algorithms under the premise of non-time-varying background harmonics and comparing the errors.