•  
  •  
 

Keywords

low-voltage system;leakage detection;random forest algorithm;residual current;

Abstract

With the rapid increase in the scale of low-voltage distribution system and user demand, leakage faults of user lines and household electrical equipment occur frequently, which increase the risk of electric shock and electrical fire accidents. Residual current protector is a common method to detect low-voltage leakage fault. In recent years, due to the existence of leakage current to ground of lines (or equipment), it also frequently false operates, which greatly reduces the operation rate and reliability of protective equipment. To overcome these issues, this paper proposes a leakage detection technology for low-voltage distribution system based on random forest （RF） algorithm. In order to closely simulate the real leakage fault scenario, the original residual current data close to the real fault scenario can be obtained by fully considering the interference factors such as excessive normal leakage current and frequent switching of load in the adjacent branch of the fault scenario. Through data preprocessing of the original residual current data, the frequency domain and time domain characteristics of the residual current are analyzed, and the time-frequency characteristics are extracted using the Fourier transform algorithm to complete the establishment and training of the low-voltage system leakage detection model. The leakage detection model is tested under the condition of multiple interference factors, and the results show that the detection accuracy of the leakage fault can reach 99.98%, realizing the leakage fault detection of low-voltage distribution system under the condition of multiple interference factors. Finally, support vector machine (SVM) algorithm, K‑nearest neighbor (KNN) algorithm and the leakage fault detection accuracy based on random forest algorithm are compared to verify the accuracy and feasibility of the proposed leakage fault detection model of low-voltage system.

DOI

10.19781/j.issn.1673-9140.2024.03.005

First Page

38

Last Page

47,115

Download

Share

COinS