Flow Based Pattern Recognition for Data Exploration − Identification of Key Parameters Affecting Reservoir Performance

The identification of key parameters governing reservoir performance is a significant step towards realistic production forecasting. To identify such key parameters, one classical method consists in performing dynamic simulations covering a wide range of reservoir models. In stratigraphically complex reservoirs such as deepwater turbidites, conducting such dynamic simulations and the analysis of their results is an extremely computationally expensive and time consuming effort. Moreover, the interpretation of the results is complicated when some of the parameters are categorical (i.e. parameters that cannot be quantified such as different geological scenarios or well-placement scenarios). The study documented in this report develops a preliminary workflow capable of automatically analyzing a huge number of dynamic simulations in order to understand and estimate how the input parameters, including categorical ones, affect their results. This quantitative analysis can then be used to quickly identify key parameters.