The Improved Algorithm Of K-means Method And Its Applications In Data Analysis Of Oil Recovery

Petroleum enterprises are the backbone of a country, among which domestic petroleum production has play an important role. Most of china petroleum enterprises are confronted with medium and late production phase and the producing capacity decreased year by year. In order to improve oil production further and provide scientific basis for decision making of development, meticulous reservoir description of wells is need. We should divide the data that is acquisited from the oil fields into different kinds of classes or clusters to analysis.Data Mining, which extracts useful information from large data sets or databases, is a rising cross contents. Clustering is an important field for research in Data Mining, and also an important method in data partition or data grouping. In the research, the algorithm of K-means, which is a widely used partition method in clustering, is used to classify the data of oil recovery. The algorithm is suitable for the huge amount of data, however, it has its own inherent deficiency. This research has made a consequential improvement of it.The aim of the paper is to improve the traditional algorithm of K-means and apply it to data analysis of oil recovery. The paper testes and verifies effectiveness of the research model by the history data from Oil Production Plant in HuaBei Petroleum Company. The main comtents are as follows:First, the conception and theory of data mining and the basic principles of clustering are introduced and the basic data types in clustering are described in general.Second, all kinds of the clustering algorithms are described in brief. The limitations and the existing treatment of the algorithm of K-means are described and analyzed and the improved scheme is provided. The basic principle and the usage of the prim algorithm in minimum cost spanning tree, which is used in the improved scheme, is discussed clearly.Last, the paper verifies the feasibility of the algorithm which has been improved and proves that the improved algorithm has more interpretable and accuracy of classification than traditional algorithm.With the development of the technology of data mining, the method of clustering analysis in oil recovery will perfect gradually.