Probabilistic assessment of equivalent fracture aperture constrained on quasi-real-time drilling mud loss data

Abstract

We provide a rigorous workflow to quantify the effects of key sources of uncertainty associated with equivalent fracture aperture estimates w constrained through mud loss information acquired while drilling a well in a reservoir. A stochastic inverse modeling framework is employed to estimate the probability distribution of w. This choice is consistent with the quantity and quality of available data. The approach allows assessing the probability that values of w inferred from mud loss events exceed a given threshold. We rely on a streamlined analytical solution to model mud losses while drilling. We explicitly consider uncertainties associated with model parameters and forcing terms, including drilling fluid rheological properties and flow rates, pore fluid pressure, and dynamic drilling fluid pressure. A synthetic scenario is considered to provide a transparent reference setting against which our stochastic inverse modeling workflow can be appraised. The approach is then applied to a real-case scenario. The latter is associated with data monitored on a rig site. A direct comparison of the impact of data collected through two common techniques (respectively, relying on flow meter sensors or pump strokes) on the ensuing probability of w is provided. A detailed analysis of the uncertainty related to the level of data corruption is also performed, considering various levels of measurement errors. Results associated with the field setting suggest that the proposed workflow yields probability distribution of w that are compatible with interpretations relying on traditional analyses of image logs. Results stemming from direct and indirect flow data display similar shapes. This suggests the viability of the probabilistic inversion methodology to assist quasi-real-time identification of equivalent fracture apertures on the basis of routinely acquired information during drilling.