History Matching in a Reservoir Model Using an Automatic Approach

Abstract

History matching may be seen as an optimization problem based on minimizing an objective function that measures the mismatch between reservoir history and simulated data. Manual methods of history matching are largely cumbersome and grossly ineffective especially when the optimization parameters are large. Recourse to the manual method leads to development of models which cannot accurately predict the reservoir behaviour and thus are not suitable for predicting future behaviour of the reservoir. The only way out is the use of automatic methods especially those backed by artificial intelligence. The study aims at applying an automatic method to perform history matching in a reservoir model. The objectives will be to; Perform automatic history matching using ABC, match permeability distribution in the reservoir using oil production and bottom hole flowing pressure data and compare the effectiveness and convergence speed of the algorithm. History matching aims at fine-tuning the parameters used in building a reservoir model to closely match that of the real field. In this study, a very promising novel optimization algorithm has been employed to history a well-known reservoir model namely the PUNQ-S3 model. The model used for this study is the popular PUNQ-S3 reservoir model. PUNQ (Production forecasting with Uncertainty Quantification) is a joint industrial-academic project with the aim of developing efficient history matching and uncertainty quantification methods. Results obtained proves the algorithm used to be a very efficient optimization tool as the data used as the history of the study is nearly equaled by the optimization tool. We therefore conclude that the ABC algorithm be employed in performing tasks that demand high degree of accuracy.