Key Features Selection of Power System Operation Via Improved Clustering Algorithm

Based on the advanced information technology, it has become an important topic to develop smart grid with big data playing a leading role. However, with increase of data dimension, the problems of dimensional disaster and sparse information become increasingly prominent. In this paper, we proposed an improved the clustering algorithm, which combines the partial priority clustering and clustering ensemble algorithm, to reduce the data dimension and select the key features of power big data. The case studies are based on the load data of residential areas in multiple states of the United States. The simulation results show that the proposed algorithm can quickly determine clustering center and effectively control the number of clusters. Moreover, the improved algorithm can largely reduce the time complexity, and does not produce any intermediate variables. It is useful for precise features selection and less space occupation.