Deep Learning Techniques for Transmission Line Fault Diagnosis: A Comparative Evaluation

Early fault detection and adjustment are critical in power systems for efficient and cost-effective power transmission and distribution, as well as their viability as available renewable energy deployment strategies. This study demonstrates and contrasts three separate diagnostic methods for high-voltage transmission lines and microgrid systems that deal with fault detection and classification. The first diagnostic system utilizes a multi-level with a restricted Boltzmann machine (RBM) for the diagnosis of faults in the microgrid system, allowing the framework to retrieve probabilities across its parameters. The second technique employs a self-attention convolutional neural network (SAT-CNN) framework for detecting and classifying faults to enhance the capability of rapid restoration in high voltage transmission lines. A self-attentive weight-sharing capsule network (WSCN) is suggested as the third method to obtain high quality and high classifier accuracy with small sample size. This paper depicts an adequate discussion about these algorithms and analyzes the effect of data types. The immunity of the methods is also evaluated using multivariate inconsistencies and noisy operational characteristics. A comparative analysis is further provided, emphasizing the benefits and drawbacks of each method, to find the most effective method among these methods.