This article, written by JPT Technology Editor Judy Feder, contains highlights of paper SPE 202460, “History Matching of Production Performance for Highly Faulted, Multilayered, Clastic Oil Reservoirs Using Artificial Intelligence and Data Analytics: A Novel Approach,” by Nis Ilyani Mohmad and Dipak Mandal, SPE, Petronas, and Hadi Amat, Petroliam Nasional, et al., prepared for the 2020 SPE Asia Pacific Oil and Gas Conference and Exhibition, originally scheduled to be held in Perth, Australia, 20-22 October. The paper has not been peer reviewed.
History matching is a critical step for dynamic reservoir modeling to establish a reliable, predictive model. Numerous approaches have emerged over decades to accomplish a robust history-matched reservoir model. As geological and completion complexity of oil and gas fields increase, building a fully representative predictive reservoir model can be arduous to almost impossible. The complete paper outlines an approach to history matching that uses artificial intelligence (AI) with an artificial neural network (ANN) and data-driven analytics. The new approach has been used to mitigate history-matching challenges in a mature, highly geologically complex field offshore Malaysia.
The complete paper describes a step-by-step methodology for building a reservoir model and a history-matching process using ANN. The paper discusses data preparation and quality assurance and control (QA/QC), spatiotemporal database formulation, reservoir model design, ANN architecture design, model training, and history-matching strategy. A case study of implementation in an offshore Malaysian field is presented, wherein good-to-fair history-matching quality was obtained for all production parameters.
Field A is a highly geologically complex, 25×75-km oil sandstone reservoir with more than 200 major and minor faults and more than 30 reservoir layers. It has been producing for more than 25 years.
The challenges of history matching this field lie not only in its geologically complex structure and corresponding subsurface uncertainties, but also in a production strategy that has involved commingled dual-string production with several integrity issues that exacerbate the field’s complexities. To date, Field A has no fieldwide history-matched reservoir model because the complexity of history matching has precluded using conventional numerical simulation methods. This challenge has been mitigated by implementing an AI-based reservoir model and data analytics. The new approach is estimated to be more time- and cost-efficient than the conventional method.
The complete paper compares the AI-based and conventional numerical reservoir modeling approaches and highlights the advantages, limitations, and areas of improvement of the new methodology. The authors present their case for AI-based reservoir modeling as a complement or alternative to conventional numerical modeling to create time-efficient reservoir models while reducing risk in field-development plans.