AI-Assisted Automation of Well Type Selection and Pad Optimization During Well Placement in Yan'an Tight Gas Reservoir

Abstract

Abstract The Yan'an tight gas field, situated in the Ordos basin in central China, is emblematic of a typical tight gas reservoir. The field's development planning has been confronted with numerous challenges arising from conditions characterized by high geological heterogeneity, low porosity, multiple production formations, low lateral sand continuity, and a complex topography. Among the pivotal tasks in field development that necessitate addressing these challenges are the screening of well types and the optimization of pad locations. The judicious selection of well types, including horizontal, deviated, or vertical wells, in conjunction with the optimization of pad locations, not only facilitates the control of reserves on a larger scale but also significantly diminishes well construction costs. Conventionally, the manual determination of well types and the optimization of well pad locations rely on the analysis of 2D maps depicting sand distribution and other geological properties. This conventional approach heavily depends on individual experience and meticulousness, requiring weeks or even months to reach a final decision, with results that may lack consistency. To address the aforementioned challenges, two algorithms have been developed, harnessing both domain expertise and a state-of-the-art machine learning (ML) algorithm. The first algorithm concentrates on well type selection based on considerations such as well targets, geological conditions, and drilling constraints. A heuristic algorithm, in conjunction with the nearest neighbors method, has been devised to identify optimal combinations of various well types with the least number of wells. The second algorithm focuses on optimizing well pad locations, taking into account surface, drilling, and cost constraints, using a modified K-means clustering method tailored to address real problems under diverse conditions. Through the implementation of these developed algorithms, the planning of hundreds of infill wells in the Yan'an tight gas field has been achieved with minimal human involvement in as little as half a day, in contrast to the traditional timeline of months. The developed algorithms serve to significantly enhance well placement efficiency, reduce operational expenditures, and consistently yield reliable results.