Convolutional Neural Network-Based Fault Detection in Distribution Networks Using Voltage Magnitude Measurements

Fault detection is a challenging task in power distribution networks, particularly with limited coverage of measurement equipment. When a fault occurs, accurate and fast fault detection technology is crucial for maintaining the reliability of the power supply system. To address the issue, this paper introduces a data-driven fault detection method for distribution networks, targeting two independent tasks: fault type classification and fault location. The approach relies solely on voltage magnitude measurements from a limited number of buses, eliminating the need for phase angle information and the high costs associated with phasor measurement units. The one-dimensional convolutional neural network is employed to extract features from the voltage data. Various fault scenarios are employed to train and test the model, covering different load conditions, fault types, fault resistances, and fault locations. Simulations are conducted on the modified IEEE 34-bus system, and the experimental results highlight the high accuracy of the proposed model in both fault type classification and fault location.