Research On Harmonic Sources Identification Based On Harmonic Monitoring Data

With the construction and development of the ultra-high voltage(UHV)and smart grid,power quality(PQ)problems caused by non-linear loads and equipment have been highlighted.Harmonic pollution will make energy and equipment consumption,lead to grid operation safety problems,and bring huge economic losses.In the harmonic comprehensive elimination process,harmonic source identification is the primary problem to be solved.Generally,harmonic source identification based on simulation grid model has low accuracy.The State Grid has built a harmonic monitoring and analysis system,which collects a large amount of PQ data and lays a foundation for data-driven harmonic analysis.In this paper,aiming at the harmonic source identification problem,combined with the harmonic monitoring data,the related research will be carried out around the harmonic source identification and harmonic responsibility division.The main work of this paper is as follows.1.Aiming at the accurate identification of harmonic sources,this paper proposes a classification method of harmonic sources driven by PQ data.This method takes into account the diversity of indicators,the non-linear and sequential characteristics of data.Firstly,based on the distribution characteristics of PQ data,the massive data are sifted and sampled.Then the feature subset is extracted by the sequence backward selection algorithm after evaluating the feature importance based on random forest(RF).Finally,the data are aggregated at fixed intervals,smoothed by the sliding average method,and put into Long Short-Term Memory(LSTM)network for model learning and prediction.Experimental results demonstrate that the proposed method is more accurate than the existing RF method,and has better generalization performance.2.To reasonably divide the harmonic responsibility in the case of multiple harmonic sources,this paper proposes a harmonic contribution evaluation method.Firstly,extract the exceeding standard PQ data based on the operation monitoringpoint and the exceeding standard data analysis results.Then,use Pearson correlation coefficient to calculate the correlation between any two monitoring points.Next,according to the correlation results,and combining the Pearson correlation coefficient threshold and box diagram analysis rules,we get the strong correlation branch set of all main lines.Finally,evaluate the contribution of each strong correlation branch set to the main line according to the DTW algorithm.Experimental results show that compared with the known partial correlation coefficient measurement,this method can better distinguish the relevant monitoring points,and shows a higher accuracy in identifying the dominant interference sources in the regional power grid.At the same time,the harmonic contribution evaluation result is more precise,and more suitable for the timing characteristics of the PQ data.3.Based on the above research results,this paper designs and implements harmonic source identification prototype system,and realizes the automatic application of harmonic source identification and harmonic responsibility division method.