Distributed static pressure measurements are routinely used in exploration settings to delineate reservoirs, determine reserves, and help make development plans and in development settings to monitor reservoir performance. However, the detailed analysis of such data is typically cumbersome and left to specialists in this field. Frequently, this leads to a non-optimal integration of valuable and expensive datasets into reservoir models. A new method is discussed in this paper that automatically analyzes distributed pressure datasets acquired in single or multiple wells. The results are compared to frequently employed manual interpretation techniques.
The new solutions consist of determining hydraulic units, fluid types, and free-fluid levels in an automated and optimal manner. First, in the time domain, the test type—valid or invalid—is identified; for instance, from the pressure time response, dry or lost-seal tests are identified and discarded. Secondly, a measure of the confidence in the pressure measurements for each valid test is quantified using a set of six criteria. These criteria are combined to form a single score for each test. The scores are subsequently used as weights in the statistical tests employed. Thirdly, in the depth domain, an algorithm using statistical measures on all the data is run sequentially to determine pressure gradient groups that define unique fluid types and hydraulic units. The procedure then calculates free-fluid levels in an optimized manner using the reservoir architecture model results computed from the previous step. Finally, once reservoirs in a multiwell context have been delineated and their associated fluid types identified, uncertainties in the free-fluid levels can be estimated using either Monte Carlo simulation with given pressure and depth error distributions or by computing directly the standard error around the free-fluid level from the standard errors and gradient values of the two fluids in equilibrium.
A database of over 500 tests was used to validate the criteria used to establish the uncertainty quantifications made around each time domain measurement. The new workflow was used on formation tester datasets and the resulting optimized solutions were compared to the results from the traditional manual pressure gradient evaluation technique. The study showed that the new, automated method is capable of predicting the architecture of reservoirs through the identification of hydraulic units and free-fluid levels for each reservoir. The results also showed the positive impact of weighting the gradient data by the quality control scores.
The significance of this work and its novelty is the development of a simple analysis technique that may be consistently applied in the interpretation of distributed pressure data or that may be used as a starting point for subsequent interpretations.