Automatic Placement of Infill Wells During Simulation Run-Time

Abstract

Well planning is the determination of the number, types and locations of wells required to efficiently extract a reservoir's hydrocarbons. It is a manual, time consuming process that is influenced by the engineer's experience. The process involves a search for model locations that would provide best instantaneous oil production rate while minimizing interference with neighboring wells. Traditionally, this is done by loading the model into a 3D visualization package to identify target locations. Cross-sections are then created to identify the reservoir layers to be targeted and in what direction the well's lateral should be oriented. The well is then designed and its connections are exported into a simulator. In the current study, our goal is to incorporate all these processes into a flow simulator to be performed during run-time. The current work presents the development of a novel automatic well placement logic (AWPL) that can detect potential reservoir targets and design wells in the course of a simulation runtime. AWPL allowed rapid sensitivity scenarios to be conducted on factors such as, the minimum perforation thickness, maximum initial water saturation, minimum permeability of target reservoir intervals, and well length. These scenarios resulted in different well locations and well counts, and consequently, different recovery volumes that could be used to decide on the optimum field development scenario. In large reservoirs with significant permeability heterogeneities that require hundreds of development wells, conducting such scenario evaluations could be time-consuming using traditional approaches, resulting in the partial evaluation of scenarios whose results may not allow for optimum decision-making. The objective function used by AWPL is the sweet-spot defined as porosity*log10(permeability)*thickness*(1-sw-sor). Depending on the number of sweet-spot zones identified at a location, a single or multilateral well is proposed. In an example application, it was found that AWPL autonomously selected the reservoir targets that would have been targeted by an expert. More so, AWPL was found to be faster and capable of doing more sensitivities within a shorter period as compared to a human. Time savings resulting from this methodology reached 98% in relation to traditional methods. AWPL also created a well location risk map, which is a measure of how persistent a given well location sweet-spot is regardless of the geo-model realization considered. Well's locations that results in good performance regardless of geo-model scenario are given higher drilling priority, all other things being equal. All prior art in the domain of automatic well placement have focused on using 2D maps to identify well locations, while the landing depths of the wells are based on a predetermined user input. The present work is the first to go further, and automatically determine the optimum landing depth of the well, both single and multi-lateral wells are supported. Additionally, prior studies focused on the placement of vertical wells, our approach is capable of placing vertical, horizontal and multi-lateral wells.