A Novel Algorithmic Design and Implementation for Predicting Crude-Oil PVT Properties

Abstract

Abstract With the cusp in the advances in Artificial Intelligence and Machine Learning, there have been several development of AI-based PVT models. Although they have offered some advantages in one way or the other, their easy implementation and flexibility have been limited due to computational complexities, and the relatively vast amount of memory allocation required. Also, some of the classical PVT models are wrought with certain setbacks. Some of these models are only accurate within specific conditions of crude-oil type and properties, and hence fail when there are small variations in crude oil properties or when there are different genetic sources or prevalence for the crude oil type. Here in this paper, we have developed a novel algorithmic design and implementation based on a combination of mathematical combinatorics and linear algebra. Using the superposition principle in an ingenious way, we built new, improved and better models from existing PVT correlations and models. Our resulting algorithm known as IntelliPVT offers more robust, more accurate and faster PVT model predictions. IntelliPVT has been written both in MATLAB and Python. Using real field data published in literatures, journals and books, and compared with other realistic conventional PVT correlations, IntelliPVT outperformed these correlations and predicted Crude-Oil PVT properties of Formation Volume Factors, Solution Gas-Oil Ratios, Bubble Point Pressures, Viscosities, and Crude oil compressibilities to the least Root Mean Square Error, R-Squared values between 0.91 and 0.99, and the lowest minimum mean square error with very high computational speed. The novel algorithm offers more accurate results with a faster implementation and applicability to all crude-oil types including heavy oils. Its implementation will ensure a better subsurface characterization, improved reservoir management as well as surface design along the whole spectrum of the oilfield lifecycle.