Integrated Geo-Modeling and Ensemble History Matching of Complex Fractured Carbonate and Deep Offshore Turbidite Fields, Generation of Several Geologically Coherent Solutions Using Ensemble Methods

Abstract

Abstract History matching reservoir models has always been tedious as it involves many uncertain parameters and requires many trial and error iterations. Frequently the modifications introduced seem artificial and may destroy geological concepts, only one matched model is obtained and the forecast of such a model may quickly be invalidated by new data. Eventually imperfect models lead to imperfect decisions. Assisted History matching with ensemble methods has received a lot of attention in the past decade. In this methodology with an ensemble of models the correlation between all uncertain model parameters and the selected production data is assessed and using this correlation the ensemble of the models are modified to reduce the difference between simulated and real historical data in an iterative manner. Ensemble methods are recognized to perfectly perform on the continuous Gaussian parameters, but their application on discrete geological parameters like facies and rock types has been a challenge for several years. Different solutions proposed to tackle this issue showed the importance of integrated workflows and the implementation of an assisted history matching loop in close relationship with the geo-modeling tools. After several years of research, an assisted history matching tool based on ensemble method has been developed in Total via the integrated platform of geo-modelling Sismage-CIG. This tool has been industrialized early 2016 with the first operational study performed on a giant gas field in the Middle-East. Ensemble methods are known to be relatively insensitive to the size of the model, number of uncertain parameters to be handled, number of wells and the length of historical data, but the industrialization of the tool to operate on huge complex fields with very large number of the wells needed several rounds of code optimization using state of art algorithmic approaches. This tool showed outstanding performance on several types of models such as turbiditic deep-offshore and complex fractured carbonate fields. The latest history matching study performed with this method on a Middle-East field modeled with a grid containing 20 million cells, around 200 wells and more than 25 years of production history, involved more than 130 million uncertain parameters in each realization. The use of assisted history matching with ensemble methods allows not only to take into account cell by cell heterogeneities as uncertainty in a coherent manner but it also delivers an ensemble of hundred matched models which creates a huge opportunity for forecast and decision making process. Moreover all models fully respect the geological a priori knowledge and the duration of history matching study has been drastically reduced (weeks instead of months if not years) using much less manpower.