Most reservoir development or management decisions require a three-dimensional numerical representation of the reservoir. Intuitively, that digital representation consists of a single model to which all subsurface disciplines contribute to the best of their knowledge and technology.
However, direct measurements of the reservoir are limited to a few boreholes, while indirect measurements, typically seismic data, have low resolution and can be very noisy. Due to the sparseness and ambiguity of the reservoir data, choices need to be made during the data interpretation and modeling phases of the digital representation construction.
An extremely perilous practice
Interpretation and modeling uncertainties make it impossible to expect a best technical case model to be a fully truthful representation of the reservoir, especially for green fields. The proof is that new production data rarely match the model predictions. It often requires updating or rebuilding the model from scratch. Making multi-million-dollar decisions based on a single reservoir model, even built by the best subject matter experts using the best technologies, is an extremely perilous practice.
Asset managers need to account for the uncertainties associated with their projects to optimise development decisions and mitigate potential risks. This means that subsurface uncertainties need not only be identified and quantified for their impact on volume and production, but also communicated.
Generating an ensemble of alternative models capturing subsurface uncertainties and providing probabilistic forecasts is the key to sound reservoir development and management decisions, helping to deliver results on-target and on-time.
To provide reliable probabilistic volumetric and production predictions, ensemble modeling requires four fundamental ingredients:
- Seamless workflow orchestration: flexible enough to offer a variety of modeling options and parameters, but with sufficient user guidance to avoid the complexity and error-proneness of conventional workflow management solutions based on computer programming principles.
- Robust subsurface uncertainty quantification: using machine learning-based solutions for interpretation automation and uncertainty assessment, removing personal bias, coupled with the integration of contextual information such as regional depositional trends through multiple modeling scenarios.
- Uncertainty-driven data assimilation: offering fast and efficient algorithms for production and 4D seismic data integration that minimise the number of computationally intensive flow simulations, while accounting for all subsurface uncertainties. A gradual modeling approach is essential, starting with simple coarse resolution models integrating first-order structural and stratigraphic uncertainties, and then refining concepts and adding geological details as needed.
- Insightful post-processing: providing clear ensemble analytics for decision support and allowing the quick testing of alternative development scenarios to explore all opportunities.
In conclusion, ensemble modeling helps energy companies to move to more geologically challenging areas, while decreasing project turnover time from years or months to weeks or days. In other words, if you are not using ensemble modeling, you are most likely leaving money on the table.
