| In this work, a general automatic history matching procedure for naturally fractured reservoirs is developed. The procedure can be used to generate maximum a posteriori (MAP) estimate of reservoir parameters by conditioning the static geostatistical model to reservoir production data. The development is based on a fully implicit, three-dimensional, oil/water two-phase, dual porosity single permeability model. The reservoir parameters that can be estimated include fracture gridblock permeability and porosity, matrix gridblock permeability and porosity, and fracture/matrix coupling factor. Several types of model parameters can be estimated simultaneously. The production data can be bottom-hole flowing pressure, water-oil ratio or the combination of the two kinds of data. The objective function is derived from Bayesian inversion frame. Levenberg-Marquardt optimization algorithm is used to minimize the objective function. Gradient simulator and adjoint methods of computing sensitivity coefficients are implemented for the dual porosity reservoirs. The computational efficiency comparison of adjoint and gradient simulator methods is performed. Highly efficient sparse LU decomposition is applied to solve the matrix equations to improve the computational efficiency of sensitivity coefficients.; A two-step algorithm of automatic history matching is proposed for the reservoirs where the priori knowledge of the model parameters is not available.; A novel approach to integrate the production data into the description of fracture intensity is presented and validated by a case of Middle East naturally fractured field. |
