An integrative approach for conditioning, robust design and control in the subsurface
When predicting processes in the subsurface, the need to quantify prediction uncertainty and to per-form probabilistic risk assessment is evident. Subsequent challenges include calibration, robust de-sign, monitoring and control. These follow-up tasks have to account for the uncertainty inherent in the system description. Numerical Monte-Carlo simulation is one of the most popular options for stochastic modelling. It is simple and almost universally applicable. Unfortunately, it has vast computational costs that may become strictly prohibitive when joining it with follow-up tasks. Polynomial chaos expansion (PCE) is a promising new approach and receives a quickly increasing attention, because it is computationally much more efficient. However, only little work has been done to make PCE available to the above follow-up tasks.The proposed work will advance and extend PCE, producing a single, integrative and efficient frame-work for stochastic modelling and its advanced follow-up tasks. In the new framework, all involved quantities will be treated via an integrative response surface that approximates the system's behaviour under all probable states and within the entire feasible range of design or control variables. The computational costs of stochastics and optimization or inversion will no longer multiply. The drastic gain in computational efficiency will finally allow performing advanced follow-up tasks at their full level of complexity to full-scale real-world problems. This will be demonstrated by application to CO2 injection into the deep subsurface. In this example, site characterization, site selection, design and control of injection strategies, as well as optimal monitoring of CO2 leakage to the surface will be performed within the new framework, leading to better quantification and management of the involved risks.
- Project manager
- Research assistants
German Research Foundation (DFG NO805/3-1)
- Cooperation partners
Prof. H. G. Matthies and Dr. A. Litnivenko, Institute fopr Scientific Computing, TU Braunschweig, Germany
Prof. D. Tartakovsky, Department of Mechanical and Aerospace Engineering, UC San Diego