Over the past decade, technological advancements in horizontal drilling and multistage fracturing enable natural gas to be economically produced from tight shale formations. However, because of limited availability of the production data as well as the complex gas-transport mechanisms and fracture geometries, there still exist great uncertainties in production forecasting and reserves estimation for shale gas reservoirs. The rapid pace of shale gas development makes it important to develop a new and efficient probabilistic-based methodology for history matching, production forecasting, reserves estimates, and uncertainty quantification that are critical for the decision-making processes.
In this study, we present a new probabilistic approach with the Bayesian methodology combined with Markov-chain Monte Carlo (MCMC) sampling and a fractional decline-curve (FDC) model to improve the efficiency and reliability of the uncertainty quantification in well-performance forecasting for shale gas reservoirs. The FDC model not only can effectively capture the long-tail phenomenon of shale gas-production decline curves but also can obtain a narrower range of production prediction than the classical Arps model. To predict the posterior distributions of the decline-curve model parameters, we use a more-efficient adaptive Metropolis (AM) algorithm in place of the standard Metropolis-Hasting (MH) algorithm. The AM algorithm can form the Markov chain of decline-curve model parameters efficiently by incorporating the correlation between the model parameters. With the predicted posterior distributions of the FDC model parameters generated by the AM algorithm, the uncertainty in production forecasts and estimated-ultimate-recovery (EUR) prediction can then be quantified. This work provides an efficient and robust tool that is based on a new probabilistic approach for production forecasting, reserves estimations, and uncertainty quantification for shale gas reservoirs.