An Improved SVM-Based Cognitive Diagnosis Algorithm for Operation States of Distribution Grid

• Published in: Cognitive computation Vol. 7; no. 5; pp. 582 - 593
• Main Authors: Yang, Jun, Gong, Lingyun, Tang, Yufei, Yan, Jun, He, Haibo, Zhang, Leiqi, Li, Gang
• Format: Journal Article
• Language: English
• Published: New York Springer US 01.10.2015; Springer Nature B.V
• Subjects: Algorithms; Artificial Intelligence; Biomedical and Life Sciences; Biomedicine; Classification; Cognition & reasoning; Computation by Abstract Devices; Computational Biology/Bioinformatics; Entropy; Fault diagnosis; Machine learning; Neural networks; Neurosciences; Smart grid; Support vector machines; Wavelet transforms; Wavelet-packet entropy; SVM; Smart distribution network; Operation states; Cognitive diagnosis algorithm
• ISSN: 1866-9956; 1866-9964; 1866-9964
• DOI: 10.1007/s12559-015-9323-2

Intelligent diagnosis of operation states of distribution grid is a prerequisite to the self-healing ability of a smart grid. In this paper, an improved support vector machine (SVM)-based cognitive diagnosis algorithm is proposed to cognize the current operation state of distribution grid by classifying the disturbance into different operation states. Based on the current measurement in distribution grid, wavelet-packet time entropy is developed to extract features of the operation states. Considering the rejection recognition of multi-class classification, an improved SVM multi-class classifier based on a kernel metric is constructed. To investigate the performance of the proposed cognitive diagnosis algorithm, simulations of real distribution grid cases are carried out in PSCAD–EMTDC. Compared with wavelet-packet energy and Fuzzy C-means, the simulation results demonstrate that the proposed cognitive diagnosis algorithm can achieve higher accuracy and more robust performance on different grids and fault conditions.