In this work the Ensemble Kalman Filter (EnKF) is used to history match the simulation model and to assess the remaining uncertainty on production forecast for a deep-water under saturated oil reservoir. The reservoir was originally matched by the company asset team using a traditional but effective trial and error process driven by engineering and geological best practices. The asset team history matched model is used to benchmark EnKF results.
EnKF is implemented assuming porosity and permeability as the dominant model uncertainty. Prior porosity uncertainty is simulated using a variogram-based statistics. To preserve the facies model, Gaussian simulated permeability is properly truncated to associate connate water saturation and relative permeability. Then, the ensemble of stochastically generated permeability and porosity grids is sequentially updated by means of EnKF to get a posterior ensemble which reflects production data. This work provides further evidence that the EnKF is an effective and cost-efficient alternative to the usual time-consuming conventional history match process. Moreover, both the traditional history matching achieved by the asset team and the calibrated ensemble of models converge towards a common understanding of the reservoir behaviour, while the benchmark between the two methodologies shows an over-all superior quality of EnKF in terms of objective function values. This work is also a real-case example of the possibility to quantify uncertainty on production forecast and assess the value of the information acquired along field life by means of the EnKF.
Reservoir management is based on 3D simulation models which are used to evaluate production performance at different stages of the field development. Before production, reservoir models are built using all the available data and development strategies are chosen according to models output. In the past one model was deemed enough for practical purposes, but nowadays reservoir professionals in the different companies agree that the subsurface uncertainty is such that many, different, models are needed to estimate the uncertainty in the forecast due to the lack of knowledge of the reservoir. These models reflect our prior knowledge, that is to say what is known about the reservoir before any production activities. Missing key uncertainties may lead to cost underestimation which compromises the economy of the development project.
Depletion activities represent a clear source of information about reservoir geology. Well rates, shut-in pressure values, Modular Formation Dynamic (MDT) test, Repeat Formation Test (RFT) and down-hole metering data provide implicit information on the reservoir. Reservoir management is then challenged by two closely related issues: first to integrate production data in the geological model as a part of history matching and then to quantify the decrease of uncertainty due to the indirect information on the subsurface properties gained with production data. The common understanding is that the acquisition of production data may decrease the technical risk of the field.
In real life, translating these concepts into practice is not easy. History matching represents one of the most challenging, time intensive phases in reservoir management and often the final results of the calibration consists of one or few history matched models. Sometimes history matched models are often a result of a trial and error process where modifications are patched to the simulation data-set without achieving an acceptable geological consistency. Even though this approach may lead to well calibrated reservoir models, the reliability is often questionable.