A practical and efficient iterative history matching workflow for shale gas well coupling multiple objective functions, multiple proxy-based MCMC and EDFM

Abstract

A manual history matching is a time-consuming process and too risky for an investment decision because it cannot capture the uncertainty of subsurface information and production forecast. Therefore, Assisted History Matching (AHM) was adopted to obtain multiple reservoir realizations for the probabilistic production forecast. Markov Chain Monte Carlo (MCMC) is one of the AHM in the Bayesian statistics that have benefits of quantifying uncertainty without bias, not being trapped in any local minima. However, the AHM in shale reservoirs has not been widely studied due to the unconventional shale's complexities such as fracture modeling. The continuum approach cannot accurately model the connection between hydraulic fractures and matrix blocks in shale reservoirs. While, Discrete Fracture Modeling (DFM) by using local grid refinement can overcome the issue of the continuum approach, there is still a limitation of computational efficiency. Therefore, Embedded Discrete Fracture Modeling (EDFM) has been developed to combine the benefits of dual-continuum model and DFM. While DFM is an intrusive technique to model fractures, EDFM can account the mass transfer between matrix cells and fractures without being intrusive to the model. Besides, EDFM provides more efficient computational time than DFM. In this study, we proposed a practical and efficient iterative AHM workflow by integrating the benefits of using multiple objective functions, multiple proxy-based MCMC algorithm, EDFM and commercial reservoir simulator. Then, we demonstrated an application of the proposed workflow to a real shale gas well by matching both gas and water rates. The field application shows that the proposed AHM workflow efficiently finds the history matching solutions of 64 out of 300 simulation runs (21%) with a total computational time of one day. Furthermore, the workflow captures the uncertainty of reservoir and fracture parameters in terms of posterior distribution. Also, we performed the probabilistic production forecast and obtained P10-50-90 estimated ultimate recovery of gas production.