Uncertainty Analysis in Production Forecasting-Part II: A Case Study on the Barnett and Haynesville Shale Plays

Abstract This paper evaluates the impact of decision making and uncertainty associated with production forecast for 11000+ wells completed in the Barnett & Haynesville shale plays. Existing studies show that unconventional reservoirs have complex reservoir characteristics making traditional methods for ultimate recovery estimation insufficient. Based on these limitations, uncertainty is increased during the estimation of reservoir properties, reserve quantification and, evaluation of economic viability. Thus, it is necessary to determine and recommend favorable conditions in which these reservoirs are developed. In this study, cumulative production is predicted using four different decline curve analysis (DCA) - power law exponential, stretched exponential, extended exponential and Duong models. A comparison between the predicted cumulative production from the models using a subset of historical data (0-3months) and actual production data observed over the same time period determines the accuracy of DCA's; repeating the evaluation for subsequent time intervals (0-6 months, 0-9 months,..) provides a basis to monitor the performance of each DCA with time. Moreover, the best predictive models as a combination of DCA's predictions is determined via multivariate regression. Afterwards, uncertainty due to prediction errors excluding any bias is estimated and expected disappointment (ED) is calculated using probability density function on the results obtained. In this paper, uncertainty is estimated from the plot of ED versus time for all wells considered. ED drops for wells having a longer production history as more data are used for estimation. Also, the surprise/disappointment an operator experiences when using various DCA methods is estimated for each scenario. However, it appears that in both plays power law exponential serves as the lower boundary of the forecast whilst the upper boundary – stretched exponential (SE) and Duong (DNG) method always significantly overpredicts the cumulative production in the Barnett and Haynesville respectively. The outcome of the paper will help improve the industry's take on existing uncertainty in production forecast using the concept of expected disappointment. This study suggests that effects of bias due to decision making can be much greater than what has often regarded, which can change the performance evaluation of the Barnett and Haynesville plays in terms of economic feasibility.