Integrated Uncertainty Modelling Workflow – A Phenomenological Approach

Abstract The success of hydrocarbon maturation projects hinges on three key factors: making informed decisions, managing uncertainty, and addressing risks. Effective management of uncertainty is crucial because of the complex risks posed by geological, technical, economic, and environmental factors. To manage these uncertainties well, it is important to identify, evaluate, and mitigate risks throughout the entire lifecycle of the project. A crucial component of managing uncertainty is integrated subsurface uncertainty modeling. This approach involves several key elements: (a) Geological Data: Understanding the structure, types of rock (facies), and distribution of resources; (b) Geophysical Data: Analyzing seismic data and well logs to understand subsurface conditions.; (c) Petrophysical Data: Assessing rock properties and characteristics; (d) Dynamic Aspects: Using simulations and historical data matching to predict reservoir behavior. To manage uncertainties effectively, these elements need to be combined through coordinated strategies. This involves integrating data from various sources, using probabilistic and stochastic modeling, and continually refining models through iterative processes. By doing so, companies can create more accurate and reliable subsurface models, reduce uncertainties, and make better decisions, leading to improved project outcomes. However, traditional methods often involve using different tools for each aspect of subsurface data analysis and modeling. This usually requires transferring data between tools, which can take between 4 to 15 weeks of man-hours. The industry has recognized this issue, and some have moved towards integrated workflows that use single applications for these tasks. But this solution may not be ideal for every situation or company. The current study introduces a new approach to uncertainty modeling. This method uses available applications to create an automated and dynamic workflow that integrates different elements of uncertainty modeling. It allows for a more thorough exploration and quantification of uncertainties while reducing the need for manual work and speeding up data processing. This approach not only saves time but also lowers the chance of human error, resulting in more reliable subsurface models and faster project development. The approach is built on principles of surrogate modeling, utilizes high fidelity reservoir simulation model in the core ensuring accurate modeling of fluid flow physics and the workflow is augmented with application of designs of experiment and neural networks to achieve envelope of uncertainty in reasonable time. To enhance flexibility in the scale of development, the workflow encompasses a method for deconvoluting the uncertainty envelope specific to subsets of reservoir characterization. The integration is achieved through pre simulation scripting to enable automatic updates to the geomodel within the dynamic simulation model. A neural network proxy model performs statistical variance analysis on simulation outputs to identify parameters most affecting the development goal, such as optimizing oil recovery. A given field "A" with limited data and numerous uncertainty parameters affecting subsurface outcomes is used in this study. Key critical uncertainties include the lateral continuity of the reservoir, log interpretation, saturation height function, and well productivity.