Data-driven multi-objective optimization design method for shale gas fracturing parameters

Accurate design of fracturing parameters in shale gas reservoirs does not only increase the productivity and profitability of wells but also minimize the damage fracturing fluids cause to fluid-sensitive formations. However, obtaining a highly accurate and optimized fracturing parameters with rapid as well cost-effective simulation run is difficult to achieve. This study proposed a framework for optimizing shale gas fracturing parameters based on least squares support vector regression (LSSVR) prediction model and non-dominated sorting genetic algorithm II (NSGA-II). This framework was named the multi-objective optimization prediction model (MOO-PM). In the proposed framework, the flowback ratio of fracturing fluid (FBR) and first month gas production (PROD) were considered as the objective functions while the fracturing parameters including horizontal length, number of fractured sections, fractured length, fracturing fluid injection rate, fracturing fluid viscosity, volume of fracturing fluid and proppant amount were chosen as the optimization variables. Firstly, based on field data the LSSVR prediction model was established. Secondly, the accuracy of the model was verified by the remaining field data. Subsequently, the Latin Hypercube method was used to construct a sample set of fracturing parameters within a predetermined range of variables and the corresponding objective functions were calculated by the LSSVR method. Finally, Pareto front was used to generate flowback ratio and first-month production by the NSGA-II algorithm. To the best of our knowledge, this is the first time that LSSVR prediction model and multi objective method are applied simultaneously to optimize shale gas fracturing parameters. Furthermore, the novel multi-objective optimization framework was applied to data from the Weiyuan as well as the Changning shale gas reservoirs in Sichuan basin, southwest China. It was concluded that this framework could optimize fracturing parameters of shale gas reservoirs with higher accuracy and efficiency.

Automated summary

The study proposed a framework for optimizing shale gas fracturing parameters based on LSSVR prediction model and NSGA-II algorithm, named MOO-PM. It is the first time that LSSVR prediction model and multi-objective method are applied simultaneously to optimize shale gas fracturing parameters, providing higher accuracy and efficiency. The novel multi-objective optimization framework was successfully applied to data from shale gas reservoirs in Sichuan basin, southwest China.