Reservoir simulation using multi-variate machine learning techniques

Novel use of statistical and machine learning techniques in this study show that improved reservoir simulation results, history matched production and reserve estimates can be improved by integrating all available reservoir simulation, infill log and 4D seismic data. This can be set-up for real-time analysis without waiting for many iterations of the inputs (e.g. rigorous re-processing and inversion of the 4D seismic) with manual reservoir simulations. The key is these multi-variate techniques can integrate all data simultaneously in 3D to mimic the many steps of manual calibration.