Optimal Matching of Multi-stage Fracturing Parameters with Geological Characteristics in Shale Gas Well Based on Deep Learning Algorithm

Abstract Multi-stage hydraulic fracturing in horizontal well is a pivotal technology for the efficient development of shale gas reservoirs. Given the pronounced heterogeneity along the horizontal section, precise matching of fracturing parameters with geological characteristics in each fracturing stage is imperative for enhancing well productivity. In this paper, a data-driven productivity prediction model at the stage level was established based on field data and the Gate Recurrent Unit (GRU) algorithm. Fracturing and geological features in each stage was integrated as the model input matrix, and an additional mask layer was employed to accommodate varying stage numbers among different wells. To achieve optimal matching between fracturing parameters and geological characteristics, a novel variable-dimension optimization algorithm was proposed. Aiming at maximizing well productivity, this algorithm can determine the optimal number and positions of fracturing stages according to the geological profile and generate corresponding fracturing parameters. Trained with real data from over 120 wells in Sichuan basin, China, the productivity prediction model achieved a robust performance with a mean relative error (MRE) of 10.1%. This MRE represents a 34.3%~67.7% improvement over traditional multilayer perceptron and random forest models, demonstrating the significance of extracting features in each stage. Optimization results for a representative well case indicate that to maximize productivity, the optimal stage number is 28, with stage length ranging from 51m to 91m, and fracturing parameters tailored to the geological characteristics in each stage. Furthermore, post-fracturing optimization, micorseimic monitoring reveals a 32.3%~44.5% increase in stimulated reservoir volume compared to adjacent wells, resulting in a productivity boost of 21.2%~35.7%. Unlike manual stage division and separate fracturing parameters optimization in each stage, the approach proposed in this paper can identify the optimal stage number, stage positions and fracturing parameters in each stage perfectly matching the geological characteristics. This study provides a novel framework for the fracturing parameters optimization in shale gas wells.