Modeling and optimizing for operation of <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML" display="inline" id="d1e3368" altimg="si10.svg"><mml:mrow><mml:mi>C</mml:mi><mml:msub><mml:mrow><mml:mi>O</mml:mi></mml:mrow><mml:mrow><mml:mn>2</mml:mn></mml:mrow></mml:msub></mml:mrow></mml:math>-EOR project based on machine learning methods and greedy algorithm
Rongquan He, Weizhong Ma, Xinyu Ma, Yuchen Liu
Abstract
The carbon dioxide-enhanced oil recovery (CO2-EOR) project operation optimization methods that can improve the efficiency and profitability have recently become a prominent demand. Global warming is a puzzle and threat to the world today. The primary reason for global warming is CO2 emissions, produced during the combustion of fossil fuels. CO2-EOR project has been valued, conducted, and developed by many countries in recent years due to its dual advantages of reducing the emission of the greenhouse gas CO2 and enhancing of the oil recovery efficiency. We propose a CO2-EOR project operation optimization workflow based on machine learning methods and heuristic optimization algorithms. Profitability of the CO2-EOR project was significantly improved by this workflow. This workflow includes a power consumption prediction exponential-gaussian process regression (GPR) model, an oil production prediction nonlinear autoregressive neural network with exogenous inputs (NARX) model, and an operation optimization model. The accuracy, effectiveness efficiency, and reliability of this workflow were confirmed in a practical project in northern China. 16-day (from April 15th to 30th, 2020) operation optimized results show that the optimized CO2 injection and storage total volume is 4668 tons, which is 70.3% more than the unoptimized; the optimized total oil production is 925.5 tons, which is 14.1% more than the unoptimized; and the optimized profit enhancement is 40,007.9 U.S. dollar (USD), which is 22.1% more than the unoptimized, and the optimized profit is 103,917.1 USD.