Staging optimization of multi-stage perforation fracturing based on unsupervised machine learning 基于无监督机器学习的多段射孔压裂的分段优化

In the development of unconventional shale oil and gas resources, it is necessary to conductmulti-stage perforated fracturing in low permeability reservoirs and form fracturing net sweep regions or fracture clusters through artificial fractures in order to establish effective paths for oil and gas flow from the reservoir to the production wellbore. In this study, an unsupervised k-means clustering algorithm based on the Euclidean distance was used to predict the reservoir fluid flow and geo-mechanical parameters in order to identify the regions which can be fractured to form fracture clusters, so as to ensure the effectiveness of the perforation and improve the efficiency of the perforated fracturing. The location and area of the fracturing clusters and fracture pattern can be obtained by the numerical simulation of the perforated fracturing process using the proposed k-means model. The simulation results show that the fracture clusters divided and designed by the k-means algorithm can be used to identify the suitable fracturing regions, and the trained k-means can be applied to predict the perforated fracturing stages for similar wells in the same block. The method proposed in this paper can optimize the division of fracturing stages and the selection of perforated fracturing locations to improve the fracturing efficiency.