Advanced Methods for Determining the Value of Information of Surveillance Data Acquisition Using Predictive Analytics

Abstract

In stacked reservoirs with commingled production, achieving an understanding of relative contributions of the flow units is fundamental to reservoir management, most notably for conformance management of reservoirs under water flood or enhanced oil recovery (EOR) scheme. To that effect, the desired surveillance data usually includes: reservoir layer pressures, phase distribution profiles (through PLTs) in flow units, and monthly well test data (water cut, gas oil ratio, oil rate etc.). These measurements will form the basis of well by well flow unit production allocation; all necessary information for classical engineering analysis and reservoir simulation. The enduring challenge of value-effective reservoir management is to determine the ‘adequate’ frequency and selection of well and flow units data acquisition. Industry practice shows clearly that there is no consistent answer to this challenge. In the authors' opinion, this is due to the unavailability of a methodology and tools to rigorously define the Value of Information (VOI) associated with surveillance data acquisition. VOI is defined as the net present value (NPV) difference between the total production & costs outcomes with the benefit of information, and the total production & costs outcomes without this information. In some cases, the VOI can also indirectly translate to critical understanding of subsurface integrity such as unintentional communication of deeper, higher pressure gas reservoir with shallower reservoir units having a much lower fracture gradient that if left unattended could subsequently lead to subsurface blowout scenario. In this paper, we set out to define surveillance data acquisition decisions as an optimisation problem: where is the optimum cost versus reward for a field, given allocated well production and the usual (partial) understanding of reservoir layer absolute and relative permeability at the well, from logs and core. We present how novel predictive analytic algorithms, coupled with multi-phase deliverability models, material balance analysis, and global optimisation search methods are integrated to assess the resulting uncertainty in layer-phase allocation, in presence of different surveillance datasets. We use a representative synthetic field simulation model to formulate reservoir outcomes. As a precursor to a full VOI, we define an allocation uncertainty versus the data acquisition frequency, and provide general recommendations in terms of data frequency and type that can be generically used as findings.