Gradient techniques are used predominantly in History Matching and Optimization. In this paper gradient technique was used in estimation of multiple packer test data (permeability distribution of very low permeable formations). A high pressure gas chamber has been released into the formation and pressure changes in this chamber and two nearby observation chambers have been measured.
A simulation model has been set up for this experiment and Eclipse black-oil simulator was used for simulation. By defining region multipliers close to packer tools and validating the model, the Gradient Optimization technique of SIMOPT-History Matching software has been used to find permeability multiplier in these regions.
Akaike's information criterion (AIC) concept which is the tradeoff between minimum of objective function and the necessary complexity of the reservoir description could be used plausibly to assess the uniqueness of the selected model. Although the Objective Function is reduced with increasing the number of parameters, AIC could estimate the most reliable unique solution avoiding an over-parameterization.
The proposed technique offers a way to estimate reservoir parameters at In-Situ reservoir conditions, accounting for heterogeneities and could prove reasonably the uniqueness of the selected model among different sets of identified solutions.