Integrated uncertainty assessment - from seismic and well logs to flow simulation

It is commonly acknowledged today that uncertainty exists in all aspects of the subsurface and most disciplines are doing a good job at assessing, understanding, and capturing the uncertainty within their own sphere of influence. However, propagating that uncertainty across the E&P spectrum is still a challenge. It is often expected that upstream disciplines will provide their best results and that downstream disciplines in turn also need the best possible input either because they may not know how to deal with multiple possibilities or it is just what one thinks they expect. Petrophysicists may have looked at uncertainty in well log interpretation for a long time, but more often than not will only pass on one set of logs to their geologists. Geophysicist have problems transferring velocity uncertainty to the interpreters, interpreters can “see” the uncertainty associated with positioning a fault on a seismic volume but rarely capture it or pick alternative positions for the same fault. Reservoir modelers have always claimed to do a great job at assessing uncertainty by generating multiple equi-probable models, yet in general they rely on one set of input logs and one reservoir structure. Engineers have used varied experimental design techniques to quantify the uncertainty in production forecasts, but rarely consider more than one geological model. A 3D model is the center-piece of any integrated uncertainty study as it numerically combines data from petrophysical analysis, seismic and geologic interpretation with geological concepts. To be complete, it must account for the uncertainty associated with all of these and therefore there cannot be just one or a handful of models, there must be many alternative representations. Furthermore, a 3D model is built hierarchically, capturing the geological and physical dependencies of the elements that constitute it. It will therefore highlight the compounded effect of the various sources of uncertainty without artificially inflating the overall uncertainty. This presentation will discuss two often overlooked fundamentally uncertain elements of the construction of a 3D model and their impact on both hydrocarbon in place estimates and production forecasts. (i) We will first look at uncertainty in fault position due to velocity and interpretation uncertainty and its consequences. Structural uncertainty assessment has generally been limited to horizons as it is relatively straight forward to deform a reservoir model accordingly. Varying the position of multiple faults simultaneously while preserving the integrity of the fault network has until now been technically tedious and required the manual reconstruction of the reservoir grid. Thus multiple realizations of a fault network are never considered, at best only a handful of alternative scenarios. A new approach based on the UVT Transform enables the stochastic simulation of a complete structural model including multiple horizons and faults. Structure being a first order element, this approach enables to quantify the impact of structure on reservoir volumes and connectivity. (ii) We will then discuss uncertainty in petrophysical analysis and their impact. A typical reservoir model relies on well data to infer all the parameters it needs: quantity of sands, dimension of geological bodies, spread and spatial correlation of porosity and permeability values. By considering only a single set of well logs, the reservoir model is essentially locked in. Yet there is a large uncertainty associated with well log processing, interpretation and modeling. We will illustrate how well log uncertainty affects in-place volumes and flow simulation results.