Research On Line Loss Analysis Of Low-voltage Distribution Network Based On Data Mining Technology

In recent years,with the continuous growth of power load,the line loss of lowvoltage distribution network is also increasing,and the number of low-voltage stations is large,the line loss analysis is still lack of effective technical means,which seriously affects the benefit of power grid enterprises.Line loss analysis methods such as theoretical calculation of line loss,abnormal detection of line loss and prediction of line loss have guiding significance for improving the lean management level of power grid and formulating targeted loss reduction measures.The main contents of line loss analysis research based on data mining technology are as follows:(1)Several theoretical line loss calculation methods for low-voltage distribution network are compared and analyzed,and their advantages and disadvantages and application scenarios are expounded.In view of the three-phase load unbalance that often occurs in the low-voltage station area,the equivalent resistance method and the split-phase equivalent resistance method are used for calculation,and the effectiveness of the split-phase equivalent resistance method for line loss theoretical calculation is verified.(2)In order to obtain a large amount of power consumption data of distribution network,a modeling scheme was proposed to build 10 k V distribution network based on Pandapower power system analysis tool and simulate the characteristic index data of the station area: a 10 k V distribution network simulation model was established jointly by Pandapower and Python.The distribution network model built by Pandapower was used to collect the data of the station area,and the influence of each index on the line loss and the difficulty of obtaining the data were comprehensively considered.10 variables such as the central distribution variable capacity and power supply radius were selected as the research indicators to provide data reserve for subsequent research.(3)For the current distribution network line loss abnormal analysis methods,the difference between theoretical line loss and actual line loss is mostly used to judge,which has great limitations in timeliness and accuracy.By comparing different outlier detection algorithms,isolated forest algorithm is selected as the basic algorithm,and aiming at the shortcomings of isolated forest algorithm,a combination model based on improved K-means clustering algorithm and isolated forest anomaly detection algorithm is proposed to detect line loss anomalies.Finally,the validity of the proposed line loss anomaly detection model is verified by a numerical example.(4)To solve the problem of low accuracy in line loss prediction of support vector machine(SVM)model,an improved Grey Wolf optimization algorithm(IGWO)based on dimensional learning hunting and search strategy was introduced to optimize the penalty parameters and kernel parameters in the SVM model,and an improved grey Wolf optimization SVM model(IGWO-SVM)was constructed.Numerical examples show that the addition of IGWO optimization algorithm can reduce the prediction error of SVM.Compared with the unoptimized SVM model,the prediction accuracy of the proposed IGWO-SVM model is significantly improved.