Fault Classification in an IEEE 30 Bus System using Convolutional Neural Network

Fault detection and classification are crucial task in the power system. It should he done immediately or in the least time to isolate the faulty phase rapidly from the system. Due to the increment in transmission lines of a power system network, the conventional methods become inappropriate because of their dependency on fault currents. Various machine learning-based approaches have been carried out in the past for fault classification, but it requires pre-processing in data and manual feature extraction & selection process. Therefore, it is time-consuming and is remarkably feature dependent. However, Convolutional Neural Network (CNN) has exhibited as a more reliable classifier with automatic feature extraction intended to classify the fault accurately. In the proposed scheme, the raw data of voltage and current signals sampled by the system are given to CNN directly without any pre-processing steps. The proposed model extracts the reliable features automatically from raw data and thoroughly performs the multi-classification of fault, including the no-fault case. The proposed model is tested in an IEEE-30 bus system. A 10-fold cross-validation technique is used to demonstrate the performance efficiency of the model in terms of various metrics such as accuracy, precision, recall and F1 score. The average accuracy of 99.27% has been achieved for the proposed work.