A New Partitioned Algorithm of Innovation Graph Technique

• Published in: 2007 IEEE Power Engineering Society Conference and Exposition in Africa - PowerAfrica pp. 1 - 7
• Main Authors: Yanjun Zhang, Yibin Shi, Suquan Zhou, Yongjie Leng, Tianyu An
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01.07.2007
• Subjects: Artificial neural networks; Electric variables measurement; Large-scale systems; Network topology; Partitioning algorithms; Power system reliability; State estimation; Switches; Technological innovation; Velocity measurement
• ISBN: 1424414776; 9781424414772
• DOI: 10.1109/PESAFR.2007.4498042

This paper presents an innovation graph technique partitioned algorithm to identify multiple interacted bad data, topology changes and topology errors for the real-time large-scale electric network. First, the innovation graph technique is introduced and the reason that the innovation graph technique cannot adapt to large-scale electric network is analyzed. Second, an innovation graph technique partitioned algorithm is proposed. At last, the merits and demerits of the innovation graph technique partitioned algorithm are analyzed. Based on innovation graph technique, a simplified network which equals to the original network is constructed in the innovation graph technique partitioned algorithm and then the simplified network is used to calculate reckoned innovation of boundary branches of each sub-network. The reckoned innovations of sub-network boundary branches are used to correct the reckoned innovations of sub-network inner branches. The algorithm reduces model error superposition in the innovation graph technique, improves the accuracy of branch reckoned innovations, and blocks the error diffusion between sub-networks. Therefore the innovation graph technique partitioned algorithm can identify multiple interacted bad data, topology errors and topology changes in large-scale electric network quickly and accurately. The method can be applied on real-time large-scale electric network.. In this paper we propose a methodology to identify multiple interacted bad data, topology changes and topology errors for the power system state estimation.