2020/01/11 by Hardikkumar Zalavadia, Zalavadia, Hardikkumar, Eduardo Gildin +1
Computer Science · Engineering · Mathematics · #Algorithm #Channelized #Computational Engineering #Computer science #FOS: Computer and information sciences #Finance #Hydraulic Fracturing and Reservoir Analysis #Mathematical optimization #Mathematics #Oil and Gas Production Techniques #Parametric statistics #Point of delivery #Proper orthogonal decomposition #Reduction (mathematics) #Reservoir Engineering and Simulation Methods #Simulation #Statistics #Workflow #and Science (cs.CE) #cs.CE
paper · pdf · doi:10.48550/arxiv.2001.05061
published in arXiv (Cornell University) (Cornell University)
openalex publication_date 2020/01/11 · arxiv created 2020/01/12 · arxiv updated 2020/01/16 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28
The objective of this paper is to develop a global non-intrusive Parametric\nModel Order Reduction (PMOR) methodology for the problem of changing well\nlocations in an oil field, that can eventually be used for well placement\noptimization to gain significant computational savings. In this work, we\npropose a proper orthogonal decomposition (POD) based PMOR strategy that is\nnon-intrusive to the simulator source code and hence extends its applicability\nto any commercial simulator. The non-intrusiveness of the proposed technique\nstems from formulating a novel Machine Learning (ML) based framework used with\nPOD. The features of ML model are designed such that they take into\nconsideration the temporal evolution of the state solutions and thereby\navoiding simulator access for time dependency of the solutions. We represent\nwell location changes as a parameter by introducing geometry-based features and\nflow diagnostics inspired physics-based features. An error correction model\nbased on reduced model solutions is formulated later to correct for\ndiscrepancies in the state solutions at well gridblocks. It was observed that\nthe global PMOR could predict the overall trend in pressure and saturation\nsolutions at the well blocks but some bias was observed that resulted in\ndiscrepancies in prediction of quantities of interest (QoI). Thus, the error\ncorrection model that considers the physics based reduced model solutions as\nfeatures, proved to reduce the error in QoI significantly. This workflow is\napplied to a heterogeneous channelized reservoir that showed good solution\naccuracies and speed-ups of 50x-100x were observed for different cases\nconsidered. The method is formulated such that all the simulation time steps\nare independent and hence can make use of parallel resources very efficiently\nand also avoid stability issues that can result from error accumulation over\ntimesteps.\n