Integrated modelling framework for enhancement history matching in fluvial channel sandstone reservoirs

Modelling lithofacies and petrophysical properties are the challenging processes at the beginning of exploration and production of hydrocarbon reservoirs. However, the limited amount of well data and core data are the main issues facing conventional modelling processes. In this study, Artificial Neural Network (ANN), Sequential Gaussian Simulation (SGS), Co-kriging and object-based modelling (OBM) were integrated as the enhancement framework for lithofacies and petrophysical properties modelling in the fluvial channel sandstone reservoir. In the OBM, multiple fluvial channels were generated in the lithofacies model. The result of this model represented all the characteristic of the fluvial channel reservoir. The model was then distributed with channels, crevasse, and leeves depositional facies with background shale. Multiple geological realizations were made and cross-validation to select the most suitable lithofacies distribution. This model was cross-validated by modelling the porosity and permeability properties using Sequential Gaussian Simulation. Thereafter, the modelling process continued with Artificial Neural Network. Petrophysical properties (mainly porosity and permeability) were predicted by training various seismic attributes and well log data using the ANN. Applying the co-kriging algorithm, the predicted ANN model was integrated with OBM simulated lithofacies model to preserve the fluvial features of the geological system. To achieve full field history matching, the final geological model was upscaled to serve as input data in dynamic history matching. An excellent and nearly perfect history matching with a least mismatch was obtained between the measurement and simulated bottom hole pressure from well test and production history. The results indicated that an efficient integrated workflow of ANN and other geostatistical approaches are imperative to attaining an excellent history matching.