Modeling Viscosity Response of Fracturing Fluids from Flowback and Produced Water using Advanced Cooperative Optimization Algorithms

摘要

As water management becomes a more prominent aspect of completion and production strategies within the oil and gas industry, some operators use flowback and produced water in place of fresh water during fracturing operations. These strategies help reduce the volume of fresh water used, thus lowering the environmental impact during completion and production operations. Because the highly optimized legacy formulations that provide optimal proppant transport have been formulated with relatively pure water sources, using flowback and produced water (which can contain dissolved minerals from the formation or byproducts of spent fracturing fluid) can prove challenging when trying to obtain predictable fracturing fluid properties. Reusing flowback or produced water typically involves reformulating the fracturing fluid through optimization for the specific source water, which can be a time-intensive process with high uncertainty associated with limited design experiments. This paper presents a method for determining the chemical formulation to achieve a specific viscosity and time profile for a fracturing fluid based on the chemical constituents of the source water. The process uses neural network as a basis for modeling fracturing fluid viscosity over time after mixing. Once the viscosity is calculated from the empirical formulation, the chemical components are re-estimated through inverse neural network models to validate the previous selection. The optimization of the fluid formulation is implemented with a multiobjective genetic algorithm to determine the best selection of chemical components necessary for producing a specific viscosity profile. The results from the fluid simulations and actual testing are also discussed to demonstrate the different applications.