Predicting production performance of Coalbed Methane reservoirs with Long Short-Term Memory Networks

International Conference on Manufacturing Technology, Materials and Chemical Engineering

机械制造;材料科学;化学工业

Chen, Dong^1 ; Zhang, Shu^2 ; Guo, Dali^2 ; Zeng, Xiaohui^3

China United Coalbed Methane National Engineering Research Center, Beijing

100095, China^1

School of Science, Southwest Petroleum University, Chengdu, Sichuan

610500, China^2

State Key Laboratory of Oil and Gas Reservoir Geology and Exploitation, Southwest Petroleum University, Chengdu, Sichuan

610500, China^3

关键词: Coalbed methane reservoir;    Complex geological condition;    Fracturing treatments;    Long-term dependence;    Long-term prediction;    Production performance;    Short term prediction;    Time series prediction;   

Others  :  https://iopscience.iop.org/article/10.1088/1757-899X/392/6/062090/pdf
DOI  :  10.1088/1757-899X/392/6/062090

学科分类：材料科学（综合）

来源: IOP

【 摘 要 】

Coalbed Methane (CBM) is a clean and highly efficient energy source. The prediction of its production performance is difficult because of the complex geological conditions and exploitation processes. The CBM well deliquification always has a long duration. Therefore, long-term prediction is more instructive than short-term prediction for fracturing treatment parameters. But there is a long-term dependence problem in the time series prediction. Therefore, in this paper long short-term memory networks (LSTM) are proposed to overcome the problem. The experimental results show that in the case of similar accuracy, the LSTM predict longer than artificial neural networks. And the LSTM are more accurate than the artificial neural networks in the same output time period.