Application of Data Stream Outlier Mining Techniques in Steam Generator Safety Early Warning System of Nuclear Power Plant
Liu Ding-ping, Zheng Kaitao, Yan Qiqi · 2013
Mining outliers in data streams is a popular research issue in data mining field, which can help to find outliers under abnormal condition and then corresponding measures can be taken. The security guarantee of nuclear power plant is the center topic for discussion of the development of nuclear power plant (NPP). As an important equipment of NPP, steam generator (SG) will produce large amounts of real-time data streams in the process of running every day, such as temperature streams of U-shaped tubes. Mining temperature stream outliers (TSO) of U-shaped tubes is helpful to the implementation of early warning behavior, which contributes to the safety of NPP. In this paper, the authors propose a novel algorithm for mining TSO of U-shaped tubes. By using data stream labels to mark the abnormal frequent items, the data stream is considered to be abnormal if it is marked three times continuously. The proposed algorithm is tested by experiments. And experimental results show that the algorithm has higher accuracy and better scalability.