Fracture Network Mapping Using Integrated Micro-Seismic Events Inverse with Rate-Transient Analysis

In the development of shale oil and gas reservoir, hydraulic fracture treatments may induce complex network configuration, which is very challenging to characterize. The existing fracture properties interpretation methods mostly rely on simplifying assumptions and are typically empirical in nature. The aim of this work is therefore to introduce an integrated framework involving fractal theory, inverse analysis of micro-seismic events (MSE), and rate-transient analysis to map the heterogeneity and distribution of fracture properties. In this work, a general framework is proposed to characterize both the geometry configuration and the owing properties of the complex fracture network (CFN). The CFN characterization framework is naturally divided into two stages: characterize the fracture geometry network by microseismic data and characterize the fracture dynamic properties by production data. In the fracture configuration characterization stage, a stochastic fractal fracture model based on an L-system fractal geometry is applied to describe the CFN geometry. Moreover, the genetic algorithm (GA) as a mixed integer programming (MIP) algorithm are applied to find the most probable fracture configuration based on the microseismic data. As to the owing properties characterization stage, we introduced embedded discrete fracture model (EDFM) for the computational concern and a Bayesian framework is used to quantify these fracture dynamical properties e.g., conductivity, porosity and pressure dependent multiplier by assimilating the production data. In addition, rate-transient analysis is also applied to calibrate the total fracture length and estimate effective stimulated-reservoir volume (ESRV). In order to validate this framework, a synthetic numerical case is developed. The result indicates that our integrated framework is able to characterize both CFN configuration and properties by assimilating microseismic and production data sequentially. The proposed workflow shows that the characterized CFN model would yield reasonable probability predictions in unconventional production rate.