Fault classification in distribution system utilizing imaging time-series, convolutional neural network and adaptive relay protection

This paper presents a fault classification model in the transmission lines and classify faults while keeping the coordination between the primary and the backup relays by adaptively changing the relay's parameters accordingly. The problem to be addressed through this paper is the need for a protection system that can dynamically adjust the relay's settings and operation to enhance their response to the fault. This model is based on convolutional neural network (CNN), by implementing Gramian Angular Field (GAF) to transform voltage and current signals into images for extracting temporal features. The coordination between primary and backup relays is optimized to minimize primary relay operating time. The proposed model was evaluated using a 9-bus test system to determine optimal relay coordination based on fault's type. The proposed fault classifier's achieves 100% accuracy in classifying the faults while achieving the optimal solution in 0.047 s.