Prediction of well performance in SACROC field using stacked Long Short-Term Memory (LSTM) network
作者：
Highlights：
• Long short-term memory (LSTM) is trained to predict rates from the SACROC formation.
• Production and operational data are divided into training (83%) and test (17%) sets.
• Performances are evaluated based on MSE and mean absolute percentage error.
• Uncertainties in the training of model are quantified using Box-Whisker plots.
• LSTM is efficient in multi-step ahead forecasting of complex and oscillating rates.
摘要
•Long short-term memory (LSTM) is trained to predict rates from the SACROC formation.•Production and operational data are divided into training (83%) and test (17%) sets.•Performances are evaluated based on MSE and mean absolute percentage error.•Uncertainties in the training of model are quantified using Box-Whisker plots.•LSTM is efficient in multi-step ahead forecasting of complex and oscillating rates.
论文关键词：Enhanced Oil Recovery (EOR),Long Short-Term Memory (LSTM),SACROC unit,Water Alternating CO2 Injection,Mean Absolute Percentage Error (MAPE),Production Performance
论文评审过程：Received 24 December 2021, Revised 18 May 2022, Accepted 27 May 2022, Available online 3 June 2022, Version of Record 10 June 2022.
论文官网地址：https://doi.org/10.1016/j.eswa.2022.117670
