Development of applied data-driven modeling approaches for SAGD operations design

蒸汽辅助重力泄油（SAGD）已被广泛应用，并被认为是一种可行的和有效的方法，用于商业化的油砂储量的现场生产。然而，这种提取方法需要通过天然气的燃烧产生过量的蒸汽。与蒸汽产生相关的水消耗、能源使用和碳排放成本高昂，并对环境产生不利影响，威胁到加拿大油砂开发项目的可持续性。为了提高加拿大非常规石油和天然气工业的竞争力，在最佳条件下操作该技术以实现对宝贵资源的清洁开采至关重要。目前的工作流程，需要静态储层建模和数值流动模拟通常只提供近似的解决方案，恢复响应，因为许多简化和假设必须调用。建模过程本身也相当繁琐和耗时，限制了其在实时优化中的应用。拟议的研究旨在通过统计分析和人工智能技术开发改进和应用的数据驱动建模方法，为SAGD操作的性能预测，设计和实时决策提供实用的工具。这些工具将用于油藏特征描述、采收性能预测、不确定性评估、SAGD操作的定量排名和评价以及最佳开发或生产策略的设计。这项研究的成果将提高我们的能力，了解和预测SAGD油藏动态。新的建模程序，详细的工作流程指南，涉及现场数据的案例研究示例，以及通过本研究探索的数值算法将立即应用于提高现场规模井规划和工厂运营的效率，同时减少当前和未来油砂项目的碳足迹。

Steam assisted gravity drainage (SAGD) has been employed extensively and considered as a viable and effective method for commercial in-situ production of oil sands reserves. This extraction method, however, requires generation of excessive amount of steam by combustion of natural gas. The water consumption, energy use, and carbon emissions associated with steam generation are costly and have adverse environmental impacts, threatening the sustainability of oil sands development projects in Canada. In order to enhance the competitiveness of Canada's unconventional oil and gas industries, it is extremely critical that the technology be operated at optimal conditions to achieve a cleaner exploitation of the valuable resources. Present workflow that entails static reservoir modeling and numerical flow simulations usually provide only approximate solutions to recovery responses, as numerous simplifications and assumptions must be invoked. The modeling process itself is also quite cumbersome and time-consuming, limiting its application in real-time optimization. The proposed research aims to develop improved and applied data-driven modeling approaches by means of statistical analysis and artificial intelligence techniques that would provide practical tools suitable for performance prediction, design, and real-time decision-making for SAGD operations. These tools will be used for the purposes of reservoir characterization, recovery performance prediction, uncertainty assessment, quantitative ranking and evaluation of SAGD operations, and design of optimal development or production strategies. The outcomes of this research would improve our ability to understand and predict SAGD reservoir performance. New modeling procedures, detailed workflow guidelines, case study examples involving field data, and numerical algorithms explored through this study would find immediate application in enhancing efficiencies in field-scale well planning and plant operations, while reducing carbon footprint of current and future oil sands projects.