摘要：AbstractWhile the use of supervised learning algorithms for fault detection and classification is well-studied in Artificial Intelligence, the use of unsupervised learning for fault detection receives less attention. Additionally, work on substation fault detection has emphasized the use of unsupervised learning in conjunction with neural network modeling to detect and diagnose unknown fault states. It is accomplished by combining two techniques: 1) an incremental one-class method for detecting anomalies and 2) a dynamic shallow neural network for the fault state. Moreover, the proposed work was used to data samples to detect faults, and the results were much superior to those obtained in prior research. Additionally, experimental research is performed on online process-based substation equipment to determine the validity of the technique. The findings indicate that the suggested framework is an effective tool for detecting and categorizing known and unknown process problems.
