Fault segment location method for distribution network based on random forest algorithm

Abstract

The invention discloses a fault segment location method for distribution network based on random forest algorithm, and belongs to the technical field of the distribution network and automation thereof. The method defines a feeder switch state of the distribution network and a feeder segment code, and determines an original fault sample set. Sample preprocessing is performed by oversampling and a feature selection method based on a learning model by considering the high dimensional imbalance characteristics of the original fault sample set. The method is based on the preprocessed sample set, proposes a random forest algorithm to mine the relationship between a fault attribute and a fault section, and constructs a fault location prediction model. The method achieves segment positioning of single and multiple faults in the distribution network, and has strong fault tolerance to distortion information.

Description

technical field [0001] The invention relates to a distribution network fault section location method based on a random forest algorithm, and belongs to the technical field of distribution network and automation thereof. Background technique [0002] In recent years, with the access of distributed generation (DG) to the distribution network, the traditional distribution network has changed from a single power source radiation network to a multi-power source complex structure with bidirectional power flow. At the same time, due to the continuous expansion of the distribution network scale and the increasingly complex network topology, the number of feeder switches and sections has also increased significantly, and the relationship between the fault current limit of the feeder switch and the fault section becomes complicated when a fault occurs. In addition, monitoring terminals are generally installed outdoors in harsh environments. Under extreme weather, the fault information...

AI Technical Summary

Problems solved by technology

However, due to the large number of feeder sections and the inconsistent occurrence probability of fault locations, the characteristic number and sample number of historical fault data are high-dimensional and unbalanced, which greatly reduces the prediction accuracy of the model. Therefore, it is necessary to preprocess the original data before modeling

As an implementation method of data mining technology, machine learning algorithm can be used to locate the fault section, but the speed and accuracy of the location are affected by the complexity and convergence of the algorithm

Images