Petroleum exploration activities demand the creation of lots of models to represent the uncertainties of a given reservoir, aiming to guide decision-making regarding exploration strategies. As part of these activities, a data assimilation process may refine the model set and discard models that do not honor the history data of the reservoir, given an error tolerance level. Reservoir engineers may use visualizations to analyze the partial results of iterative data assimilation processes and make decisions such as accepting the models selected in an iteration or refining the model set. Among a set of available visualizations for this process, we did not find in the literature the use of interactive heatmaps based on reorderable matrices. In this paper, we propose to add this kind of visualization to the analysis toolset for data assimilation, aiming to help engineers to better overview data assimilation datasets and how their models honor the history data of a reservoir. The proposed visual mappings use a set of heatmaps that enable the engineers to analyze a misfit measure (NQDS) for a given iteration, regarding models, wells, and their dynamic attributes. Users may reorder and filter these heatmaps according to their needs. We implemented our proposal in a web-based prototype that was submitted to a preliminary evaluation with real users in a synthetic scenario. This evaluation revealed positive opinions about the potential of the tool, which may be used to complement other approaches, and provided some opportunities for improvement.
Tag: Reorderable Matrix
Using reorderable matrices to compare risk curves of representative models in oil reservoir development and management activities
Methodologies of oil reservoir development and management demand the creation of a set of reservoir models that represents the uncertainties of a reservoir. This set of uncertainties is often simplified by the use of Representative Models (RMs), i.e., models that represent the full set. Comparison of risk curves (a.k.a. complementary cumulative distribution functions) of reservoir variables is an approach used for helping oil engineers to select a set of RMs. A typical comparison chart of a given variable of interest superposes two risk curves: one of the entire model set, and another from the set of RMs. The level of similarity between the curves in this chart indicates the representativeness of the set of RMs regarding the entire model set. A visualization with some of these charts may help to compare the representativeness of the set of RMs regarding more variables (reservoir properties, production data etc.) or distinct sets of RMs. However, this kind of chart is not enough to provide an overview of these comparisons if the number of variables or sets of RMs increase. This paper shows how the use of reorderable matrices, depicted as heatmaps, can provide an overview of this dataset that can be helpful to engineers to make decisions. We propose to represent the dissimilarity of pairs of risk curves instead of the curves themselves. This solution enables our visualization to increase the number of sets of RMs and the number of variables to represent. We show the usefulness of our proposal in three case studies of oil reservoir benchmarks, and discuss the pattern we found in these cases.