Summary
Numerical modeling of the in-situ conversion process (ICP) is a challenging endeavor involving thermal multiphase flow, compositional pressure/volume/temperature (PVT) behavior, and chemical reactions that convert solid kerogen into light hydrocarbons and are tightly coupled to temperature propagation. Our investigations of grid-resolution effects on the accuracy and performance of ICP simulations demonstrated that ICP-simulation outcomes (e.g., oil/gas production rates and cumulative volumes) may exhibit relatively large errors on coarse grids, where “coarse” means a gridblock size of more than 3 to 5 m. On the other hand, coarse-scale models are attractive because they deliver favorable computational performance, especially for optimization and uncertainty quantification workflows that demand a large number of simulations. Furthermore, field-scale models become unmanageably large if gridblock sizes of 3 to 5 m or less have to be used. Therefore, there is a clear business need to accelerate the ICP simulations with minimal compromise of accuracy.
We developed a novel multiscale-modeling method for ICP that reduces numerical-modeling errors and approximates the fine-scale simulation results on relatively coarse grids. The method uses a two-scale adaptive local-global solution technique. One global coarse-scale and multiple local fine-scale near-heater models are timestepped in a sequentially coupled fashion. At a given global timestep, the global-model solution provides accurate boundary conditions to the local near-heater models. These boundary conditions account for the global characteristics of the thermal-reactive flow and transport phenomena. In turn, fine-scale information about heater responses is upscaled from the local models, and used in the global coarse-scale model. These flow-based effective properties correct the thermal-reactive flow and transport in the global model either explicitly, by updating relevant coarse-grid properties for the next timestep, or implicitly, by repeatedly updating the properties through a convergent iterative scheme. Upon convergence, global coarse-scale and local fine-scale solutions are compatible with each other.
We demonstrate the much-improved accuracy and efficiency delivered by the multiscale method by use of a 2D cross-section pattern-scale ICP simulation problem. The following conclusions are reached through numerical testing: (1) The multiscale method significantly improves the accuracy of the simulation results over conventionally upscaled models. The method is particularly effective in correcting the global coarse-scale model through the use of the fine-scale information about heater temperatures to regulate the heat-injection rate into the formation more accurately. The effective coarse-grid properties computed by the multiscale method at every timestep also enhance the accuracy of the ICP simulations, as demonstrated in a dedicated test case, in which a constant heat-injection rate is enforced across models of all investigated resolutions. (2) Multiscale ICP models result in accelerated simulations with a speed-up of four to 16 times with respect to fine-scale models “out of the box” without any special optimization effort. (3) Our multiscale method delivers high-resolution solutions in the vicinity of the heaters at a reduced computational cost. These fine-scale solutions can be used to better understand the evolution of the fluids and solids (e.g., kerogen conversion and coke deposition) in the vicinity of the heaters (several-feet-long spatial scale). Simultaneously, with the fine-scale near-heater solutions, the local-global coupled multiscale model provides key commercial ICP performance indicators at the pattern scale (several-hundred-feet-long spatial scale) such as production functions.
Modeling the past, present, and future distributions of endangered white abalone (Haliotis sorenseni) to inform recovery efforts in California
