Identification of faulted line section in microgrids using data mining method based on feature discretisation

摘要

Protection is the main challenge for the operation of microgrids. This paper presents a new data mining method for identification of faulted line section for the protection of microgrids. This method uses wavelet packet transform (WPT) to extract a set of features from the fault voltage and current waveforms. The features are then pre-processed and used for identifying the faulted line section through classification. Three classifiers are examined in this paper. In order to improve the classifiers performance, two pre-processing steps are applied on the features. First, using three different feature selection methods, the irrelevant and redundant features are removed from the feature set. Secondly, using a supervised discretisation technique, the continuous features are converted into finite interval features. The performance of the proposed method is investigated and compared with previous methods using extensive simulation study on a complex microgrid. The results indicate that with only six discretised features, the proposed method has a fast and accurate performance by using the three classifiers examined, whereas one is more effective. Moreover, the proposed method only uses the measurement at the main substation, obviating the need for communication links to exchange data as used by most previous methods.