A critical assessment of the reliability of predicting subsidence phenomena induced by hydrocarbon production

Abstract

Predictive analysis of the spatial and temporal evolution of the subsidence induced by hydrocarbon exploitation could be affected by a high uncertainty also related to the a priori identification of the most realistic deformation parameter values. In fact, deformation behaviour of the involved formations is not constant but evolves according to the increasing strain caused by fluid pressure drop. Especially shallow medium depth clastic formations (which contain most of the reserves in the largest known reservoirs) could exhibit an important non-linear influence of strain on formation stiffness.The scope of this research is to provide insight into how reliable the prediction of the subsidence could be as a function of adopted pseudo-elastic values related to the non-linear poro-elastic behaviour of clastic formations. To this end, a series of multi-variable sensitivity analyses were developed to quantify the discrepancy in subsidence forecast deriving from the assumption of constant (static or dynamic) pseudo-elastic parameters values along all reservoir production life instead of inputting into the model a decay curve, which describes the progressive degradation of the pseudo-elastic parameters. The case study, even if synthetic, is a compound of standard features representative of gas-production from a clastic reservoir so as to analyse the response of a macro category.The results of all the analysed scenarios (which differ in terms of reservoir depth, Gas Originally in Place and shape factor) allow the identification of the reference confidential intervals of static and dynamic assumptions vs the decay curve hypothesis. Furthermore, they show that even if the mechanical response of the reservoirs evolves towards the static modulus during production, the relative induced subsidence is approximated with higher accuracy by the constant dynamic modulus assumption. The results of the sensitivity analysis on synthetic models were validated via the analysis of a real gas reservoir.