Journal of Beijing University of Posts and Telecommunications

  • EI核心期刊

JOURNAL OF BEIJING UNIVERSITY OF POSTS AND TELECOM ›› 2017, Vol. 40 ›› Issue (6): 43-49.doi: 10.13190/j.jbupt.2017-100

• Papers • Previous Articles     Next Articles

Prediction of Power System Security Based on Stacked Sparse Auto-Enconder Neural Network

LI Xiao-yu1, LI Shu-fang1, LI Wen-qi2, TIAN Shi-ming3   

  1. 1. Beijing Key Laboratory of Network System Architecture and Convergence, Beijing University of Posts and Telecommunications, Beijing 100876, China;
    2. State Grid Henan Electric Power Company, Zhengzhou 450052, China;
    3. China Electric Power Research Institute, Beijing 100192, China
  • Received:2017-10-06 Online:2017-12-28 Published:2017-12-28
  • Supported by:
     

Abstract: In order to satisfy the accuracy rate and timeliness performance for transition stability assessment(TSA) with a safety margin, a prediction model of stacked sparse auto-encoder network (SSAEN) is proposed based on deep learning. Firstly, the matrix obtained from the bus voltage can be treated as a pattern diagram with the TSA operating mechanism. Then, the dominant property of the pattern diagram hierarchically is mined by adopting SSAEN. And, the domination features and their evolution are described by analyzing the connection weights of the layers. Next, employing the logistic classifier can identify the dominant property. Finally, the system stability in the future time is successfully predicted. Competitive prediction speed and accuracy of the proposed SSAEN can be achieved from the simulation results in the IEEE-39-bus system, which is important to provide a sufficient safety margin when the system suffers from a temporary instability.

Key words: transient stability assessment, stacked sparse auto-encoder, dominant feature, classifier

CLC Number: