Reservoir-History-Matching

This code is a collection of MATLAB scripts used for inverse modelling of petroleum reservoirs generally termed as “history matching” in the reservoir engineering literature. The code is applied to 4 synthetic reservoir models. MASTER0,SPE10,PUNG and NORNE. It couples 19 different methods generated during the course of my PhD,ranging from Level set methods for image reconstruction,supervised and unsupervised learning for parameter sub space parametrisation,the JPEG lossy image technique discrete cosine transform parametrisation technique.The code is easy to use and requires the availability of the Eclipse reservoir simulator and MATLAB research licence. The code begins with defining the size of the reservoir model,ensemble construction and size (designed a priori from SGeMS),the true well,seismic or electromagnetic data,over complete learned dictionaries(unsupervised learning) and initialization of the ensemble. The forward problem is run in parallel for computational time reduction,utilising all cores of the local machine being utilised for running the Eclipse simulator.The inverse modelling,data processing with production curves and petro physical maps are generated automatically after the end of the history matching loop.