Parameterization Using Sensitivity Methods for Global History Matching Techniques
B. Al-Shamma, O. Gosselin and P. King
Event name: 76th EAGE Conference and Exhibition 2014
Session: Optimisation Approaches Applied to Geomodel and Recovery
Publication date: 16 June 2014
Info: Extended abstract, PDF ( 1.47Mb )
Price: € 20
For any history-matching method, an efficient optimisation method is required, but more importantly an effective selection of parameters. The parameterisation assists in reducing a large number of possible parameters, in the absence of available data measurements, lowering also the number of altered parameters. This paper describes the implementation of flexible integrated parameterization and optimization methods, tested on the PUNQS3 synthetic model, an iterative series of parameterisation, as a pragmatic strategy, and a comparison between various parameterization methods: layer-based, gradient-based, median-based, and distribution-based. The chosen parameters are regions or zones where permeabilities and porosity are adjusted using a common multipliers. The selected parameters are then utilized as search parameters to minimize an objective function, which quantifies the mismatch between the observed and simulated production data, using a so-called global minimisation algorithm. Successive parameterizations can be used, as part of an iterative process, where the history match is improved by further parameterisation, based on the previous “best match”. The optimisation techniques cannot perform well without a suitable and effective parameterisation method. This study shows a pragmatic combination of a global technique and various parameterisation methods. It emphasizes, that a low objective function can be far from the true models, and not predictive.