Automatic history matching of reservoirs using the streamline-based EnKF method

The present paper combined ensenble Kalman filter(EnkF) with a streamline method as an optimization algorithm for the dynamic estimation of reservoir parameters. In the realization of this algorithm, a streamline simulator with higher accuracy and faster computing speed was chosen to predict the performance, and a reservoir model was automatically decoupled into a series of one-dimensional models along streamlines, which could minimize the numerical dispersion and the effect of grid generation while maintaining a sharp displacement front. A group of geologic models were generated by using the sequential Gaussian co-simulation method, and by means of assimilating the real-time observation data(output, water cut, etc.), both of the static parameters(porosity, permeability, etc.) and the dynamic parameters(pressure, saturation, etc.) of a reservoir model were updated dynamically and continuously, in the meantime, the streamline distribution was also updated, thus, moving traces of reservoir fluids between injection and production wells could be visually reflected. The mathematical principle of the streamline-based EnKF method was elaborated by the Bayes theory, and the validity of this method was verified by means of fitting calculation of a 2D water-drive reservoir model.