Stacking ensemble learning based inversion for three-dimensional distribution region of hydraulic fractures in shale


Cited 0|Views11
No score
Hydraulic fracturing is an essential technology for the efficient development of shale gas reservoir. The inversion of hydraulic fractures distribution region (HFDR) is of great significance for the fracturing evaluation and productivity analysis. In the absence of fracturing monitoring, the traditional inversion methods based on theoretical models and production data are limited by the high degree of simplification and multiple solutions. In this paper, based on the real geological, engineering and microseismic monitoring data from Weiyuan shale gas field, China, the stacking ensemble learning algorithm was applied to establish the inversion framework for three-dimensional distribution region of hydraulic fractures in every single stage. Specifically, the random forest and extreme gradient boosting models were introduced as the base estimators, and the linear regression model was introduced as the meta estimator. The random search method was adopted to optimize the main hyperparameters for best performance, and the robust inversion ability for the length, width, height, and orientation of HFDR was achieved finally, with a mean relative error of 5.1%similar to 15.1% and a coefficient of determination (R-2) of 0.792-0.836. Furthermore, the missing HFDR in Weiyuan shale gas field was inverted with the established framework. Comparing with the HFDR in northeast area, the fracture height in the southwest area is limited, and the planar extension range is larger correspondingly. This study shows the potential of ensemble learning approach in hydraulic fractures inversion and fracturing evaluation in shale reservoirs.
Translated text
Key words
Ensemble learning,Fracture inversion,Hydraulic fracturing,Shale
AI Read Science
Must-Reading Tree
Generate MRT to find the research sequence of this paper
Chat Paper
Summary is being generated by the instructions you defined