An Efficient Approach for Automatic Complex Fractured Networks Parameter Inversion Based on Surrogate Model and Deep Reinforcement Learning

Abstract

Complex fracture networks in subsurface are of great significances to the management of ground water, carbon sequestration, and petroleum resources. The first and primary task is to understand their transport characteristics and properties. Like pumping test interpretation, well test interpretation provides a convenient method to identify reservoir flow regime and estimate reservoir parameter, which purpose is to obtain unknown parameters by matching theoretical and measured pressure curves by adjusting theoretical model parameters (Bourdet, 2002;Z. Chen et al., 2018). However, due to the influence of human factors, the interpretation process is prone to produce non-unique solutions (Khadivi & Hassanzadeh, 2021;Xiao et al., 2021). In addition, with the formation of complex fractures during the process of horizontal well fracturing, which hinders the accurate estimation of reservoir properties, it is more difficult to inverse parameters (Sun et al., 2016). As a result, automatic interpretation methods have received increasing attention.At present, a variety of deep learning algorithms are being integrated with energy development, which has led to a significant increase in computational efficiency and a prominent reduction in solution design cycles (Y. Li et al., 2019;Sun & Zhang, 2020). At the same time, many automatic interpretation methods have been proposed, including gradient-based optimization algorithms (