A Data-driven-based Single-phase-to-ground Fault Section Location Method for Port New-type Distribution System

The increasing integration of renewable energy generation in the port distribution system poses significant challenges to grid stability and reliability. Single-phase-to-ground faults (SPGFs) have a high probability of occurrence, but the fault signals are often weak, and the diverse fault output characteristics of distributed generations further interfere with the weak fault characteristics. In this paper, a novel data-driven SPGF section location method for port distribution networks is proposed. The characteristics of phase current transient components under SPGFs in distribution networks are first analyzed in order to identify the differences between faulty phase current fault transients and non-faulty phase current fault transients. In order to enhance the predictive accuracy of deep learning models, it is essential to obtain a large amount of fault sample data as the foundational dataset. However, in actual operations, the quantity and diversity of fault data are insufficient to meet the sample learning and training needs of deep learning algorithms. Consequently, a model of a novel 10 kV distribution system for a green port was constructed in MATLAB/Simulink. The parameters of the modules affecting the fault current were modified using the sim function, and the model operation was controlled accordingly. Batch simulations were employed to obtain a diverse set of fault samples, and these samples were utilized as inputs for a convolutional neural network (CNN), which facilitated feature extraction and classification prediction. This approach enabled the classification of phase currents from feeder terminal units (FTUs) at the distribution automation station and, in conjunction with the topology of the distribution network, achieved the localization of fault sections. The analysis of the resulting location data indicated that this method was both viable and offered a certain level of accuracy for fault location.