Ragnar Hauge

Daesoo Lee, Oscar Ovanger, Jo Eidsvik et al.

Creating accurate and geologically realistic reservoir facies based on limited measurements is crucial for field development and reservoir management, especially in the oil and gas sector. Traditional two-point geostatistics, while foundational, often struggle to capture complex geological patterns. Multi-point statistics offers more flexibility, but comes with its own challenges related to pattern configurations and storage limits. With the rise of Generative Adversarial Networks (GANs) and their success in various fields, there has been a shift towards using them for facies generation. However, recent advances in the computer vision domain have shown the superiority of diffusion models over GANs. Motivated by this, a novel Latent Diffusion Model is proposed, which is specifically designed for conditional generation of reservoir facies. The proposed model produces high-fidelity facies realizations that rigorously preserve conditioning data. It significantly outperforms a GAN-based alternative. Our implementation on GitHub: \url{https://github.com/ML4ITS/Latent-Diffusion-Model-for-Conditional-Reservoir-Facies-Generation}.

Oscar Ovanger, Ragnar Hauge, Jacob Skauvold et al.

This paper presents PointSSIM, a novel low-dimensional image-to-image comparison metric that is resolution invariant. Drawing inspiration from the structural similarity index measure and mathematical morphology, PointSSIM enables robust comparison across binary images of varying resolutions by transforming them into marked point pattern representations. The key features of the image, referred to as anchor points, are extracted from binary images by identifying locally adaptive maxima from the minimal distance transform. Image comparisons are then performed using a summary vector, capturing intensity, connectivity, complexity, and structural attributes. Results show that this approach provides an efficient and reliable method for image comparison, particularly suited to applications requiring structural analysis across different resolutions.

Gabriele Martinelli, Jo Eidsvik, Ragnar Hauge

This paper has been withdrawn by the authors. We present a framework for sequential decision making in problems described by graphical models. The setting is given by dependent discrete random variables with associated costs or revenues. In our examples, the dependent variables are the potential outcomes (oil, gas or dry) when drilling a petroleum well. The goal is to develop an optimal selection strategy that incorporates a chosen utility function within an approximated dynamic programming scheme. We propose and compare different approximations, from simple heuristics to more complex iterative schemes, and we discuss their computational properties. We apply our strategies to oil exploration over multiple prospects modeled by a directed acyclic graph, and to a reservoir drilling decision problem modeled by a Markov random field. The results show that the suggested strategies clearly improve the simpler intuitive constructions, and this is useful when selecting exploration policies.