An Integrated Approach for History Matching of Complex Fracture Distributions for Shale Oil Reservoirs Based on Improved Adaptive Particle Filter

Abstract

SummaryThe application of horizontal well drilling technology and volume fracturing technique makes the economic development of shale oil reservoirs feasible. The unknown fracture networks lead to severe nonlinearity and high uncertainty during fracture characterization. Moreover, the reservoir parameters usually exhibit a highly non-Gaussianity. Therefore, the key challenges for history matching in fractured shale oil reservoirs are effectively representing the fracture network and coping with the non-Gaussian distribution of reservoir-model parameters. In this work, a new characterization method for complex fracture networks is established, in which the distribution of connected fractures of the reservoir domain is represented by some statistical parameters such as fracture dip angle, fracture azimuth, and fracture half-length and some deterministic parameters such as the coordinates of fracture center points. In the uncertainty quantification and history-matching process, an integrated approach that combines the particle filter and an improved kernel density estimation (KDE) based on its Shannon entropy (SE) for estimating fracture distributions and physical parameters is presented. An adaptive mechanism based on Kullback-Leibler divergence (KLD) is introduced in the proposed history matching workflow, which automatically adjusts the number of particles to reduce the computational burden. Two examples of 3D shale oil production were constructed to validate the efficiency and accuracy of the proposed method. Results showed that the method was capable of capturing the main features of the fracture distributions in the reference cases. The proposed method has the potential to be applied in more complex cases such as multiple wells and multiphase flow.