This article, written by JPT Technology Editor Chris Carpenter, contains highlights of paper SPE 200466, “Proxy-Based Assisted History Matching and Well-Spacing Optimization in Shale Gas Development of a Real Field Case,” by Chuxi Liu, The University of Texas at Austin; Wei Yu, Sim Tech and The University of Texas at Austin; and Cheng Chang, PetroChina, et al., prepared for the 2020 SPE Improved Oil Recovery Conference, originally scheduled to be held in Tulsa, 18–22 April. The paper has not been peer reviewed.
A robust, reliable work flow for well-spacing optimization in a shale reservoir development incorporating various types of uncertainties and detailed economics analysis is necessary for achieving sustainable unconventional production. In the complete paper, the authors describe a novel well-spacing-optimization work flow based on the results of assisted history matching and apply it to a real shale gas well, incorporating uncertainty parameters such as matrix permeability, matrix porosity, fracture half-length, fracture height, fracture width, fracture conductivity, and fracture water saturation.
The work of well-spacing optimization is significant because it will subsequently dominate the planning of the drilling job and completion job and ultimately will affect recovery efficiency. The purpose of well-spacing optimization serves to maximize either capital revenue or ultimate recovery. The greatest challenge for well-spacing optimization is how to interpret the uncertainties associated with unconventional reservoirs. Stimulated reservoir volume and external reservoir volume, effective fracture half-length vs. propped half-length, matrix permeability, and complex structural geology are examples of such challenges. Therefore, developing an efficient and trustworthy work flow for optimizing well spacing in any shale reservoir is critical.
Previous work on unconventional shale well-spacing optimization includes operator data analysis and numerical and analytical simulation. However, almost all previous studies ignored the effects of uncertainties. In addition, most studies require input information regarding the reservoir of interest. One method to obtain such information is to history match the production data, and a few history-matching methods have been explored and analyzed. Nevertheless, traditional history-matching methods could not overcome the problem of high-dimensional uncertainty space, as is commonly seen in unconventional development. Because of this, more- stochastic approaches have been developed and applied. These methods use the concept of proxy to minimize simulation runs and are also able to obtain as many, or more, history-matching realizations. Furthermore, Markov-chain Monte Carlo (MCMC) algorithms usually are coupled with the proxy model in assisted history matching. This method could be helpful in finding the complex posterior distributions of multiple uncertainty variables with ease.