Multi-Harmonic Responsibility Division Based on DBSCAN Clustering and Correlation Analysis

Traditional methods for multi-harmonic responsibility division necessitate the aid of physical models. This paper introduces a multi-harmonic responsibility division method, grounded solely on existing monitoring data, and based on DBSCAN clustering and correlation analysis. First, the collected sample set of monitoring data is subjected to the DBSCAN clustering method, grouping data with a relatively stable correlation between feeder harmonic current and bus harmonic voltage into the same scenario. Secondly, the correlation analysis method is employed to analyze the correlation between feeder harmonic current and bus harmonic voltage for multiple harmonic sources within the same scenario, yielding the correlation coefficient for harmonic responsibility division. Conclusions are drawn from the data, eliminating the need for the aid of physical models. Finally, the method proposed in this paper is validated and analyzed on real-world data. Considered the differences in the correlation between multi-feeder harmonic currents and bus harmonic voltages under different operating scenarios. In scenario 1, feeder 1 has a higher correlation coefficient with feeder 2 and is the main contributor to harmonic pollution. In scenario 2, feeder 2 has a negative correlation coefficient, making it the victim of harmonic pollution. However, feeder 1 and feeder 3 have a higher correlation coefficient and should be held responsible for the main harmonic pollution. The method's principle is straightforward and easily implemented, offering a foundation for rapid harmonic responsibility division in diverse scenarios.