Performance comparison of three deep learning techniques applied to the fault detection of a transmission network

This study highlights the development of intelligent techniques (decision trees DT, support vector machines SVM and deep learning DL) to solve the issue of fault detection and location in transmission networks. These techniques are developed using both MATLAB and Python programming languages and tested considering an IEEE 39-bus standard test network. thirteen cases of faults are investigated. Further examination of performances (both accuracy and rapidity) of these methods are carried out to identify the most beneficial approach. The assessment of the studied methods achievements highlighted the effectiveness of the DL approach, which achieved an impressive accuracy rate of 100%. However, the convergence time of the DL algorithm (0.9 seconds), as obtained from various tests, surpassed other algorithms. Nevertheless, when considering different results acquired through testing on other standard networks (IEEE 14-bus and IEEE 09-bus), it becomes evident that, unlike the DT and SVM algorithms, the convergence time of the DL algorithm decreases as the volume of data increases.