Abstract

This study aims to develop a rapid workflow that combines a multi-resolution reduced physics model and multi-objective evolutionary algorithm (MOEA) for application to a multi-well unconventional reservoir. Hydraulic fractures are modeled using a dual porosity dual permeability (DPDP) system generated by an embedded discrete fracture model (EDFM), and a novel fast marching simulation method was used to reduce the computational cost and substantially speed up history matching. A genetic Algorithm was used to calibrate reservoir properties and fracture parameters using field production data. A substantial reduction in the uncertainties in the reservoir and hydraulic fracture parameters was observed after history matching. A broad range of integrated monitoring technologies was installed to characterize the fracture network. The hydraulic fracture locations were interpreted from the warm-back analysis of the distributed temperature sensing (DTS) data and incorporated into the simulation model as embedded discrete fractures. A fast-marching-based multi-resolution model was set up by partitioning the reservoir into local and shared domains based on the concept of "Diffusive-Time-of-Flight" (DTOF). The local domain is discretized with a multi-resolution scheme. The original 3D grids are preserved near the wells, and the rest of the domain is transformed into 1D grids to accelerate the simulation. Before history matching, the most influential parameters were first identified by a detailed sensitivity analysis. Finally, the model is calibrated with production data using a multi-objective evolutionary algorithm and used for performance forecasting. The fast-marching-based multi-resolution simulation is feasible for multi-well models and highly efficient, reducing the computational time by more than an order of magnitude with minimal loss of accuracy. The reduction in computation time significantly speeded up the history-matching process, where hundreds of simulations were required. The most sensitive history-matching parameters were found to be fracture geometry and conductivity, fluid saturations, and rock compressibility in the SRV regions. A significant uncertainty reduction of tuning parameters was observed after history matching. Finally, the history-matched model is used for drainage volume visualization and performance forecasting.

This content is only available via PDF.
You can access this article if you purchase or spend a download.