Spatial Clustering and Reservoir Analysis: An Expert-Guided Synergy Dynamic Time Warping (DTW) Machine Learning Technique on Volve and Norne Fields

Jakub Marek Cebula, Mohamed Hassan Abdalla Idris, Shamsul Masum, Jebraeel Gholinezhad, Edward Smart

This study introduces an expert-guided application for clustering production wells using Machine Learning (ML), focusing on the Volve and Norne Field datasets to optimise reservoir analysis and decision-making. The Dynamic Time Warping (DTW) algorithm was employed for clustering and further enhanced by spatial visualisation through Voronoi polygons on topographic maps. The study presents a workflow, integrable before reservoir characterization, significantly reducing time for spatial property distribution analysis. The findings indicate that DTW, especially integrated with domain expertise, emerges as an efficient method, offering flexibility and quick results within this workflow. The approach was validated using the Silhouette Score as a reliable clustering metric guide. Alternative methods like K-Shape Clustering and Frequency Domain Aggregated Clustering (Wavelet and FFT) were also examined. Although offering distinct insights, DTW was preferred for its flexibility in capturing temporal shape similarity directly and its better integration into the proposed expert-guided synergistic workflow, despite known computational considerations. The study highlights the potential application of these techniques in different reservoirs of various geological settings.