Intelligent model management and Visualization for smart oilfields

Simulation models are commonly used as an aid to decision making for oilfield development and operations. Such models represent information uncertainty and alternate operational strategies and form a design space, which is used by engineers to explore different “what-if” scenarios. As the composition of the engineering team changes over the lifetime of the oilfield and new modeling requirements emerge, it becomes important for engineers to be able to quickly “mine” from and understand the distribution of the models in the design space, and also to know if a particular scenario was already modeled in the past or if a new model needs to be created. In this paper, we describe a technique to analyze arbitrarily large set of simulation models, identify similarities and differences between model parameters, and automatically cluster the models based on similarity in an n-dimensional design space. The major contribution of this work is a vector space based approach for automatic model clustering without human intervention. Building on this contribution, our system provides a smart browse, search and visualization capability over a legacy model catalog in a non-proprietary manner. The system also performs automatic analysis of available models in order to discover their underlying basic structure and, if possible, represents the models as variations of this basic structure. We demonstrate the application of our algorithm to a set of Integrated Production Modeling (IPM) models. However, our use of a standard, non-proprietary network model abstraction as an intermediate representation means that our analysis technique can be applied to models created using a variety of modeling and simulation tools. The broader significance of this work is in the context of knowledge management for smart oilfields, specifically focused on extracting meaningful information from legacy simulation models, and making this information available and useful to the domain experts.