@article{LIU2026, 
author = {Huajie LIU and Zhaopeng LI and Sergey CHERNYSHOV and Liming ZHANG and Wenxiang LIN and Fuquan DING and Theis Ivan SOLLING},
title = {A machine learning-based optimization design method for fracturing fluid flow rate targeting equilibrium proppant bank height},
year = {2026},
journal = {Experimental Technology and Management},
volume = {43},
number = {4},
pages = {22-29},
keywords = {machine learning, displacement rate optimization, plate experiment, Belt and Road Initiative, green and intelligent oilfields},
url = {https://www.sciopen.com/article/10.16791/j.cnki.sjg.2026.04.003},
doi = {10.16791/j.cnki.sjg.2026.04.003},
abstract = {ObjectiveThe Middle East is a key region for oil and gas within the Belt and Road Initiative, but it faces challenges because most oil fields lack optimized fracturing designs. To achieve the desired stimulation results, high-displacement fracturing fluid operations are often used, which can lead to significant groundwater contamination from broken gel fluid. Visual plate experiments are commonly employed to optimize fracturing parameters; however, they are time-consuming and require substantial material, which hinders the development of green and intelligent oilfields. Therefore, this study aimed to develop an efficient machine learning model to predict the equilibrium proppant bank height within primary fractures and to use the output to inversely optimize the fracturing fluid displacement rate.MethodsAfter a comprehensive review of existing literature, a dataset was established, and data standardization was carried out. By evaluating predictive performance using five-fold cross-validation and a test set, four machine learning algorithms were compared, and the best predictive model was selected. A method leveraging this optimal model to improve both the plate experiment process and the fracturing fluid displacement rate was then proposed.ResultsThe random forest model showed the best predictive performance, achieving a coefficient of determination of 0.947, a root mean square error of 3.62, and a mean absolute error of 2.27 on the test set. For the validation set with 21 samples, the absolute error was within 1.50 cm, with a fitted curve slope of 1.07 and an intercept of −0.81. The inversely designed, optimized fracturing fluid displacement rate was significantly lower than the empirical rate, saving at least 106.2 m3 of fracturing fluid per hour while still ensuring effective fracture filling.ConclusionsThis study not only addresses current limitations of plate experiments but also offers guidance for designing fracturing fluid displacement rates in field operations within the Middle East.}
}