Geosteering based on resistivity data and evolutionary optimization algorithm

Abstract

Currently, the oil and gas industry faces numerous challenges in addressing geosteering issues in horizontal drilling. To optimize the extraction of hydrocarbon resources and to avoid penetration in aquifers, industry experts frequently modify the drilling trajectory using real-time measurements. This approach involves quantifying subsurface uncertainties in real-time, enhancing operational decision-making with more informed insights but also adding to its complexity. This paper demonstrates an approach to decision making for trajectory correction based on real-time formation evaluation data and the differential evolution algorithm. The approach uses volumetric resistivity log data and data from reservoir models, such as porosity. The provided methodology suggests corrections for planned well trajectories by maximization of the objective function. The objective function operates with a calculated hydrocarbon saturation environment as the decision-making system in a virtual sequential drilling process. To demonstrate the accuracy and reliability of our approach, we compared the simulations of the corrected trajectory with the preliminary trajectory drilled in the same area. In addition, we conducted several experiments to tune the hyper-parameters of the differential evolution algorithm to select the optimal parameter set for our case study and compared proposed differential evolution algorithm with particle swarm optimization and pattern search algorithms. The results of our experiments showed that the real-time formation evaluation data combined with the differential evolution algorithm outperformed a trajectory provided by the drilling engineers. Differential evolution algorithm demonstrated strong performance compared to others optimization algorithms. We have implemented a complete pipeline from generating resistivity and porosity cubes, using the Archie equation to estimate oil saturation, and consequently generating a corrected trajectory in this cube based on near-well data, angle constraints and predefined hyper-parameters set prior to well trajectory planning. The methods developed were validated on synthetic and real datasets. Our decision-making system shows better cumulative oil saturation values than the preliminary provided horizontal well.

Keywords

• Geosteering
• Well trajectory optimization
• Artificial intelligence
• Evolutionary algorithms
• Deep resistivity logging