UNI-MB - logo
UMNIK - logo
 
E-resources
Full text
Peer reviewed
  • A novel high impedance faul...
    Biswal, Tapaswini; Parida, S.K.

    Electric power systems research, August 2022, 2022-08-00, 20220801, Volume: 209
    Journal Article

    •The summation of the accumulated difference of residual voltage is used for discriminating HIF from any other no-fault events.•DWT analysis is used to compute the wavelet features.•The wavelet features are fed to the decision tree to classify the fault event and the faulty zone.•The proposed method is also validated for unbalanced loading, varying the DG parameter and distribution line length, and the addition of noise.•The performance indicators such as accuracy, precision, recall, F-measure of the method are compared with other classifiers like SVM, KNN, and ensemble classifiers. Conventional overcurrent protective relay is unable to detect high impedance faults (HIFs) in the micro-grid system owing to the reduced levels of fault magnitude. In this paper, numerous fault detection methods are discussed for developing a technique to identify HIFs, and also three steps approach of summation of accumulated difference of residual voltage, discrete wavelet transform (DWT), and decision tree(DT) approach has been used to detect and classify the fault. It also differentiates the faulty events from the various non-faulty ones such as switching of generators, capacitors, loads, etc. The residual voltage is used for the extraction of wavelet coefficients by DWT and fed to a DT classifier for fault event classification and confusion matrix created to view the accuracy. After classification, the predicted fault and the actual fault are obtained subsequently by training the data is 99.5% true positive rate and 0.5% false-negative rate. Similarly, this approach is also applied for faulty zone and non-faulty zone classification. The accuracy of the technique is analyzed with other classifiers like support vector machine(SVM), k-nearest neighbor(KNN), and ensemble classifiers. The DT classifier yields more accurate results compared to other classifiers. The proposed method performance is further evaluated by taking different performance indicators such as precision, recall, and F-measure. The proposed technique gives an overall accuracy of 99.95%, precision index of 100%, recall index of 99.9%, and F-measure index of 99.94% as compared to other classifiers under normal operation. The proposed method is also validated with unbalanced loading, varying DG parameters, varying distribution line length, and the addition of noise in the signal. To implement and demonstrate this method, a 5 bus micro-grid system integrated with a wind turbine-based system(WTS) generator is simulated using Power System Computer-Aided Design (PSCAD) software.