Optimization of Proppant Placements for Larger Propped Areas

Proppant placement determines the long-term well productivity. In this work, a genetic algorithm (GA) optimization module is integrated with a simplified three-dimensional displacement discontinuity method (S3D DDM) hydraulic fracture model. The proposed workflow is used to algorithmically optimize completion parameters to maximize the propped area. The S3D DDM hydraulic fracture model simulates 3D geometries of multiple simultaneously propagating hydraulic fractures. The associated proppant transport model uses an Eulerian-Lagrangian regime to solve slurry flow and proppant transport. Using numerical case studies, we analyze the effects of slurry properties and pumping schedules on proppant placement within a single hydraulic fracture. Additionally, we examine the impact of cluster placement on proppant distribution in multi-cluster stages. Results show that optimizing one pumping parameter improves proppant placement, and optimizing these parameters altogether supersedes optimizing an individual one. In the case of multi-cluster stage, the optimized cluster placements increase the effective propped areas in all scenarios with different numbers of clusters. Interestingly, although not all fractures in the optimized stages have propped areas, adding a cluster alters the fracture geometries and may increase the total effective propped area of the stage. 1. INTRODUCTION Propped hydraulic fractures serve as pathways for hydrocarbon in unconventional reservoirs, and propped area has been a metric to quantify hydraulic fracturing effectiveness (Novotny, 1977). Many studies have been done to simulate proppant transport in multi-cluster stages. Mao et al. (2021) used an efficient three-dimensional (3D) multiphase particle-in-cell (MP-PIC) method to simulate proppant transport among multiple fractures at the field scale. Sinkov et al. (2021) applied a numerical model to simulate the propagation of multiple hydraulic fractures with proppant transport. Yang et al. (2023) established a displacement discontinuity method (DDM) hydraulic fracture model with an Eulerian-Eulerian approach proppant model to investigate proppant transport in multi-cluster and multi-well completions. These numerical models have also supported the optimization of completion designs, while most of the studies on completion design optimization used manual-tuning of completion parameters. For example, Pei and Xiong (2022) optimized well performance and operator profit by manually setting up various completion designs and summarizing the beneficial effects. Modern complex unconventional reservoir development has gained complexity such as the involvement of parent-child well and multi-stage completion (Pei et al. 2023). Therefore, the number of input parameters of a numerical model also increases to accommodate increased amount of simulated physics, which promotes the needs of algorithmically optimization approach. Yi et al. (2020) used a genetic algorithm (GA) to optimize perforation and pumping schedules to create an even distribution among all perforation clusters. Wang et al. (2021) built a proppant transport experimental device and a back propagation neural network. They trained the neural network model with the experimental results and used the trained model to locate the optimal pumping schedule giving the best proppant filling ratio.