Research On Data Mining Of Power System Based On Cloud Computing

Electric power system is closely related to people's life and industrial production.The stable operation of power system is related to the normal production and the normal life of the national industry. With the arrival of information and digital era, the input of various intelligent terminal equipment in power system, the construction of smart grid and smart energy system, make the power system emerged a lot of power big data. These large amounts of power data are hidden behind a lot of power system operation of valuable information. How to dig out these valuable information has become an urgent problem to be solved in the power system.Data mining is the use of mathematical methods to dig out the value of the hidden behind the information, to complete the transformation from knowledge to value.Therefore, using data mining method to analyze the power data and it has always been a reference for power decision-making and power data analysis. But the traditional mining method is based on the single node serial mining and it can not meet the needs of mass data mining. Cloud computing relies on distributed storage and parallel computing. It can combine many inexpensive computers into clusters to deal with massive data.In this paper, we first study the open source Hadoop cloud computing platform. The K-means and Canopy clustering algorithms are studied, and the process of two kinds of traditional algorithms are introduced. The two kinds of clustering algorithms are used to analyze the rules of power users. At the same time, the identification algorithm based on the K-means clustering algorithm is the basis of the gap statistics method and the elbow criterion. By using the characteristics of the two algorithms, a combination of the two algorithms is proposed to achieve the purpose of power system identification. Parallel to all algorithms based on Map Reduce model, make sure it has the ability of data processing.Through experiments, all the parallel algorithms are applied to the example and verify the proposed parallel algorithm is efficient and feasible.