Quantification of production recovery using probabilistic approach and semi-analytical model for unconventional oil reservoirs
MetadataShow full item record
Decline curve analysis is widely applied for production forecasting in oil & gas industry. However, many models do not work for super-tight, unconventional wells with dominant fracture flows. Some novel decline models have been introduced for unconventional plays, but the transition time between the transient and pseudo-steady flow period is difficult to model with such pure empirical relations. Consequently, the decline projections are often inaccurate and furthermore, they are difficult to quantify the uncertainty associated with the predictions. To address these issues, a combined probabilistic approach is proposed that uses a dual-porosity semi-analytical decline model within an extended bootstrap framework in order to provide estimates for the P10, P50 and P90 production profiles. The probabilistic method employed in this research is a data-generative approach that employs modified bootstrap method to generate multiple decline model projections. The semi-analytical model is an approximate decline model that optimizes parameters describing flow in matrix-fracture systems using the observed production profile. In the proposed method, probabilistic approach and semi-analytical decline model are combined. The modified approach is compared to the performances developed with Arps’ hyperbolic model. Both models are fitted by optimizing respective parameters and 50 synthetic data sets are used to draw confidence interval projections. The probabilistic approach is extended by proposing alternate blocking techniques – variance of the mean and analysis of the variance (ANOVA), in place of a scheme based on the autocorrelation exhibited by the decline data, originally implemented by other researchers. The cumulative production and forecast period production errors are calculated for these alternative schemes. For all proposed applications, two unconventional, horizontal oil wells are used to test the results. Both these wells exhibit sharp decline in production rate in the first few months that is related to fracture flow regimes. The results show that the proposed application of semi-analytical model with probabilistic approach significantly improved the projections. The implementation of alternate blocking techniques also show improvement in confidence interval projections, The resultant uncertainty distributions are more accurate and precise than those obtained using the autocorrelation based schemes. The combined results show that ANOVA blocking technique outperformed the other two techniques.