In several of ADCO's fields, potential gains from an improved recovery mechanism are large. Several such projects are therefore being tested to optimise the next phase of field development. Gas injection pilots have been designed to verify the feasibility and estimate the efficiency of gas injection for enhanced recovery of oil left behind by natural depletion and water flooding. Conclusive evaluation of these pilots requires a good understanding of the various sources of error that affect monitoring data.
Initially, the main purpose is to detect gas breakthrough at the observers. This is generally straightforward, as long as the wells have been correctly designed, and a suitable set of logs is acquired on a regular basis.
As the gas flood progresses, the hydrocarbon properties in the reservoir change; oil becomes heavier as some intermediate hydrocarbons are carried over to the gas phase. When gas, oil and water saturations have stabilized; the question arises of how accurate the log-derived saturations really are. This is important since the residual oil saturation is one criterion by which the efficiency, and thus commercial viability, of the project can be assessed.
A lot of effort goes into reducing log uncertainty by acquiring numerous passes and using systematically the same logging tool over the monitoring period. Quantitative log interpretation requires careful environmental corrections and using the right properties for various formation and fluid components.
This paper presents a method to help reduce uncertainty in residual oil saturation to a level where the results can be used in reservoir simulation. The uncertainty in this number is estimated by analytical and numerical means.
For the past five years, Neutron and Sigma (pulsed neutron) measurements have been regularly recorded in several observers to monitor the progress of these gas injection pilots. Qualitatively, the method has been very successful at detecting gas breakthrough, showing variations in gas saturation with time and defining the gas sweep.
However, as the project matures and better knowledge of the changing fluid properties becomes available, accurate estimation of the residual oil saturation becomes important to evaluate the efficiency and commercial viability of the project. Sources of uncertainty and efforts to minimize their effects will be discussed before an example of application is described in detail, from which conclusions will be drawn.
Sources of uncertainty
The sources of uncertainty can be split into three types:
Log uncertainty. Every log sensor has an intrinsic accuracy and precision. Accuracy indicates how close to the true value the measurement is. Downhole, the true value is difficult to define, since there is no reference to compare with. Good accuracy is ensured by calibrating sensor response to known standards followed by regular master calibrations. Accuracy can sometimes be checked downhole in known lithology layers like anhydrite.
Precision is a measure of log repeatability. Generally linked to the type of sensor, precision also depends on the properties of the formation. For example, a neutron sensor has a much better precision in low porosity than in high porosity; because count rates are much higher when the Hydrogen index is low. This is actually a favorable effect in our case, since gas has a low Hydrogen index. Nuclear measurements being statistical in nature have relatively poor precision compared to resistivity logs for example. However, with slow logging speed and averaging of several passes, precision can be improved to become a very small contributor to overall uncertainty.
Tool response. Even if a measurement is perfect, environmental corrections need to be applied. Other measurements are provided without any correction and need to be fully and explicitly corrected with dedicated software. Neutron logs have a large number of individual corrections.