Wide Area Backup Protection System Based On Multi-information Fusion Research

With the power system is becoming more flexible and complex, the development of modern power proposes new requirements of backup protection. Traditional backup protection can not accurately locate the actual fault location, and relay tuning is complex, exposing some problems. In recent years, wide area measurement technology, computer technology and digital substation communication technology provides a broad space for the development of relay protection. In promoting the above factors, based on WAMS measurement information, considering design and configuration from a systems perspective the wide-area backup protection has been more and more attention.The paper start from the status of wide area backup protection, main content involved wide area backup protection based on faulty component recognition principle were reviewed. On the basis of analysis and comparison of existing algorithms wide area backup protection, multi-information fusion theory is introduced.The paper adopt the regional centralized structure as wide area backup protection structure, using wide area backup protection algorithm based on multi-information fusion, fuse the various circuit breakers and protective action status informations, emulate by rough sets-probability neural network model. By mining multiple wide area information and their relationship, rough set use the redundancy and complementarity of information to reduction and optimization. The negative impact of wide area backup protection fault identification caused by the acquisition, transmission of information can effectively be inhibit by protection using probabilistic neural network. The example analysis shows that, based on multi-information fusion, rough sets-probability neural network model can effectively use redundancy and complementarity multi-information, in the case of information missing or incorrect can still guarantee higher accuracy rate of fault recognition. By comparing with the BP neural network, radial basis function neural network, generalized regression eural network, the method has obvious advantages of accuracy and fault tolerance.The research of this paper is funded by special fund of Key laboratory of Guangdong Regular Higher Education Institutions (ZDSYSZOO701).