Study on Working Diagnosis of Rod Pumping Wells Based on Machine Learning
Volume 8, Issue 1, February 2020, Pages: 7-11
Received: Apr. 8, 2020;
Published: Apr. 17, 2020
Views 67 Downloads 23
Peng Gao, Institute of Oil and Gas Recovery, Engineering Technology Research Institute of Huabei Oilfield Company, Renqiu, China
Xuefeng Jin, Institute of Oil and Gas Recovery, Engineering Technology Research Institute of Huabei Oilfield Company, Renqiu, China
Jieyu Du, Institute of Engineering Technology, The Fourth Exploit Factory of Huabei Oilfield Company, Langfang, China
Yu Han, Institute of Engineering Technology, The Fourth Exploit Factory of Huabei Oilfield Company, Langfang, China
Yanhui Zhu, Institute of Oil and Gas Recovery, Engineering Technology Research Institute of Huabei Oilfield Company, Renqiu, China
Follow on us
The working condition diagnosis of pumping wells is an important basis for judging it's running condition. For a long time, indicator diagram is used to solve the wave equation by mathematical method to get the downhole pump indicator diagram which is obtained to judge and analyze the working condition of oil pumping equipment. In recent years, with the development of AI (artificial intelligence) technology, how to combine artificial intelligence technology with rod pumping wells lifting technology to provide a new way of thinking for pumping wells lifting and improve the intelligence level to achieve real-time control of the well has become a difficult problem in front of technicians. After several years of research, technicians in huabei oilfield innovatively proposed to establish working diagnosis model of pumping wells through machine learning neural network algorithm. The model can realize working diagnosis of oil well by using a large number of labeled indicator diagram and working diagnostic data as training samples. As long as the labeled data provided is of high accuracy and large amount, the accuracy of the model can reach more than 90%. The model is simple in structure, fast in operation speed, and has high application value in the field. Several rod pumping wells were randomly selected from a working area of huabei oilfield, and the above model was used to diagnose the working conditions. The results showed that the average error between the diagnosis results and the real ones was only 7%, less than 10% of the common requirements, which met the needs of engineering application.
Machine Learning, Rod Pumped Well, Neural Network, Working Diagnostic
To cite this article
Study on Working Diagnosis of Rod Pumping Wells Based on Machine Learning, Science Discovery.
Vol. 8, No. 1,
2020, pp. 7-11.
严长亮，彭勇.泵示功图单井自动量油技术研究[J].西安石油大学学报(自然科学版) ，2006，21(6)： 92-95。
S. G. Gibbs. Computer Diagnosis of Down-hole Conditions in Sucker-Rod Pumping Wells [J]. Journal of Petroleum Technology, 1966, 18 (1): 27-29。
S.G. Gibbs. A Method of Determining Sucker Rod Pump Performance [J]. U. S. Patent. 1967, 3 (3): 18-20。
Dory, D. R and Schmidt Z. An Improved Model for Sucker Rod Pumping[ J]. SPE Engineers Journal, 1983, 2 (2): 33-40。
R. R. Dickinson. The Use of Pattern Recognition Techniques in Analyzing Down Hole Dynamometer Cards [J]. A Thesis for the Degree of Master of science, 1987, 3 (6): 187-192。