Improved Estimation of Net Pay and Gas to Oil Ratio Using Intuition AI with Limited PVT Data

Abstract

Summary Efforts have been ongoing for decades to extract valuable reservoir insights by interpreting the mud-gas logging data in combination with drilling and other petrophysical parameters. Unlike dedicated wireline operations, mudlogging data are readily available during drilling, and if reliable, they can provide an alternate method for assessing fluid properties such as Gas-Oil Ratio (GOR) and reservoir quality indicators such as net pay thus reducing the need for costly wireline operations especially in situations resulting from high pressure, high temperature, wellbore instability, tar, low pay or high angle wells, where wireline operations can be difficult or too expensive to justify. Conventional mud gas interpretation techniques are mostly based on empirical cut-offs and formulae, and unless tuned very locally, these will generally produce unreliable and inaccurate results compared to the pressure-volume-temperature (PVT) measurements. There have been attempts to use machine learning (ML) models to improve accuracy, but these have also had limited success due to the limited size and range of data that is generally available for model training. Such models have inherent difficulty catering to the broad range of fluid property distributions that exist in nature. Intuition artificial intelligence (AI) is a novel causation AI methodology that iterates experts’ hypotheses and creates a scientific fabric to model system behavior. It has the ability to interpret complex changes in the state of a system by converging multiple views and eliminating hard-to-detect situational fluctuations. Intuition AI was applied to enhance the mudlogging-based reservoir characterization, with highly encouraging results. Importantly, it is able to do so with limited data sets. In our case, based on the data processed from 5 wells containing 15 pay clusters, the estimation of GOR and net pay using intuition AI fell within the 10% deviation range when compared with the ground truth, the PVT results.