Attributes selection and reservoir prediction based on support vector machine
Zhang Changkai1, Jiang Xiudi2, Zhu Zhenyu2, Yin Haiyan2, Lu Wenkai1
1. Department of Automation,State Key Laboratory of Intelligent Technology and Systems,Tsinghua National Laboratory for Information Science and Technology,Tsinghua University,Beijing 100084,China;
2. CNOOC Research Institute,Beijing 100027,China
Abstract:This paper applies feature selection algorithm based on SVM to select seismic attributes.According to the oil and gas yielding of oil wells,samples of seismic attributes are divided into two kinds:high-yielding well and low-yielding well.After these samples are trained by SVM,the attributes sensitive to oil and gas will be selected by screening the weight corresponding to each attribute,and then be taken advantage to predict reservoirs.The detailed process can be described as:①Extract certain seismic attributes;②Obtain samples according to the given information of some oil wells and train them by using SVM;③Calculate the weight of every attribute;④Choose out the attributes whose weight absolute value are respectively large;⑤Apply the support vector regression(SVR) to the chosen out attributes and predict reservoir.Our application of this algorithm on real seismic data shows that the algorithm is able to choose out valid seismic attributes and effectively predict reservoirs at the same time.