Identification of Working Condition From Sucker-Rod Pumping Wells Based On Multi-View Co-Training and Hessian Regularization of SVM

Abstract

In view of the oil extraction production system of sucker-rod pumping wells in big data and IOT of oil and gas production, to solve the limitations in working condition recognition and to further enhance the accuracy and practicality, a novel method based on multi-view co-training and Hessian regularization of SVM to identify the working condition of sucker-rod pumping wells is proposed. Firstly, the features of the measured dynamometer cards and electrical power data are extracted based on mechanism analysis, prior information and expert knowledge. Then a working condition recognition model with multi-view co-training algorithm based on Hessian regularization of SVM is established. The proposed method was applied to the recognition of eleven kinds of typical working conditions from sixty sucker-rod pumping wells in a certain block in Shengli Oilfield and compared with traditional recognition methods based on measured dynamometer cards, electrical power data and multi-sources of feature connection respectively. The experimental results show that the recognition rates are increased by 3.2%, 4.3% and 7.4% respectively. The performance is even much better for the cases of only small amount of marked training samples, obviously in the case of 10%.