Hierarchical Classification of Grid Event Signatures Using a Public Data Repository

• Published in: IEEE Power & Energy Society General Meeting pp. 1 - 5
• Main Authors: Annalicia, Christabella, Joo, Jhi-Young
• Format: Conference Proceeding
• Language: English
• Published: IEEE 21.07.2024
• Subjects: Accuracy; data repository; Discrete wavelet transforms; disturbance signatures; event classification; hierarchical classification; Kernel; Libraries; machine learning; Measurement; Power systems; Random forests; Subject matter experts; Support vector machines; Transforms
• ISSN: 1944-9933
• DOI: 10.1109/PESGM51994.2024.10689036

This paper provides an overview of the development of the Signature Matching Tool (SMT) for the Grid Event Signature Library (GESL), a publicly available repository of disturbance event signatures observed in electric power systems. The GESL contains signatures that contain multiple labels, e.g., phase, status, equipment, in nature, where the labels of each signature are assigned and arranged in a hierarchical, tree-like structure by subject matter experts (SMEs). The SMT thus adopts a local classifier per node (LCN) approach, which is a type of hierarchical classification method, in order to help GESL users identify the natures of unlabeled signatures. The SMT achieved an average accuracy of 83%, with up to 100% accuracy for certain labels, across five root (Primary) labels using a Random Forest binary classifier as the local node classifier. Hierarchical validation metrics, such as precision, recall, and F-1 score, are also calculated to validate the performance of the SMT.