Improved Spectral Clustering Algorithm And Its Application In Prediction Of Oil And Gas Production

As an important research method in data mining,cluster analysis has been widely concerned by researchers.The idea is derived from the graph theory,which transforms the data clustering problem into the optimal segmentation problem in graph theory.The main purpose is to break up the data objects needed to be processed into different classes or clusters by the criterion which has the largest similarity inside class and the smallest similarity between classes.Compared with other typical cluster analysis algorithms,the spectral clustering algorithm reduces the requirement of the sample space shape and solves the local optimal problem of some algorithms effectively.On the basis of existing research,some improvements are made to the spectral clustering algorithm in this paper.In order to solve the problem of automatically determining the number of classes and selection eigenvectors in the traditional spectral clustering algorithm,a spectral clustering algorithm of automatically determining the number of classes and selection eigenvetors is proposed using eigenvalue intervals.Firstly,the Laplacian matrix of the network is constructed;then the Laplacian characteristic eigenvalues and the Laplacian characteristic eigenvectors are computed;by using the eigenvalue gaps,the clustering number is determined automatically and the useful characteristic eigenvectors are selected.Then we can get the result of community division by using K-means algorithm.In order to test the feasibility of the algorithm and the accuracy of the segmentation results,the data of common benchmark network are tested by Matlab.The test results show that this algorithm is effective and feasible.In the process of oil and gas production,the accurate prediction of oil and gas production is great significant not only to the deployment of oil exploitation but also to the further rectification plan.By modeling the cell wells in the production environment of oil and gas fields,we can get the data which can be processed by spectral clustering algorithm.Then the spectral clustering algorithm is used to classify the unit wells and we further predict oil and gas production through classification results.The yield prediction of a single well not only takes itself into account but also considers the role of other wells in the same cluster and associated with it.The results show that the spectral clustering algorithm can be effectively used in the prediction of oil and gas production.In this thesis,we get the main results as follows:(1)A spectral clustering algorithm of automatically determining the number of classes and selection eigenvetors is proposed.In order to verify the effectiveness and feasibility of the algorithm we test it with Matlab sofware.Modeling the unit well in the production environment of oil and gas field and the data model can be processed by clustering algorithm.(2)In this thesis,the spectral clustering algorithm is applied to the prediction of oil and gas production.Model the single well in the production environment of oil and gas field,The classification of single well is obtained by clustering analysis with spectral clustering algorithm.The result of classification is used to predict oil and gas yield of single well.