An Analytical Method for Fault Detecting, Classifying, and Section Locating Based on Principal Component Analysis in Distribution Network

Distribution networks are exposed to various faults due to their varied structure. Detecting and locating faults in distribution networks can prevent widespread outages and heavy damage to the network. Most methods of detecting, classifying, and locating faults in power transmission lines are based on the current measurements at the sending and receiving ends of the lines. Using these methods in radial distribution networks where terminals are connected to more than one branch is challenging and requires more current sensors. For this purpose, this paper proposes an analysis method based on the principal component analysis (PCA) technique using the voltage signals measured on the terminal. First, the voltage of all terminals of the distribution system is measured. Next, the phase-decomposed signals are applied to the PCA, and the $T^{2}$ statistic corresponding to each phase is calculated. In addition to fault detection and determining the time of fault occurrence, this statistic is used to classify the faults. By determining the time window containing three cycles on the $T^{2}$ statistics signals, two indices, peak-to-peak index (PTPI) and phase angle difference index (PADI), are defined to classify the faults. Then, the Q-contribution diagrams obtained from PCA are used for fault section locating in the distribution network. The proposed method is implemented on an IEEE 13 node test feeder. The simulation results in MATLAB software confirm the effectiveness and high accuracy of the proposed method.