Research on Forecasting Method of Oil Well Liquid Production
Title
Research on Forecasting Method of Oil Well Liquid Production
Subject
Linear regression
Neural networks
Oils
Liquids
Production
Training
Prediction methods
long short-term memory
error back propagation
normalized treatment
prediction of well fluid production
Description
In the middle or late stage of oilfield exploitation in China, the storage speed of oil well does not match pumping speed, resulting in pumping unit in the state of “non-full pumping” or “empty pumping”, which wastes lots of electricity. This paper used overdetermined equation and neural networks to predict liquid production after eliminating dimensional influence among oil well feature data with normalization method. Comparing the results with the regression equation, the effect of neural network is better than the other two methods in the case of large amounts of data. Even under the same training set, the results of the two neural networks are quite different. According to the prediction results of regular data, the effect of neural networks are significantly affected by the characteristics and quantity of the sample data. Therefore, appropriate prediction method in the prediction of liquid production can provide reference for oilfield management.
304-307
Publisher
2021 6th International Conference on Intelligent Computing and Signal Processing (ICSP)
Date
2021
Contributor
T. Liu
X. Yang
T. Li
Y. Huang
H. Peng
Type
conferencePaper
Identifier
10.1109/ICSP51882.2021.9408715
Collection
Citation
“Research on Forecasting Method of Oil Well Liquid Production,” Lamar University Midstream Center Research, accessed May 18, 2024, https://lumc.omeka.net/items/show/23147.