Abstract
Multistage fracturing is a common stimulation technique for unconventional development in the oil and gas industry. Understanding this complex process to extract hydrocarbons optimally throughout field development is a primary goal. Companies create these multistage fractures by pumping hydraulic fracturing fluid at high pressure to fracture the formation and create flow pathways into the wellbore. This process induces small scale movements or microseismic events, and fracturing operators place geophones in an offset well to detect these microseismic events. Companies often collect these data in real time to assess the effectiveness of the stimulation job and adjust the fracturing program to enhance capital efficiency. Moreover, reservoir engineers need to know fracture properties, such as fracture size and orientation, to build well-calibrated reservoir-simulation models. However, predicting fracture properties is challenging because fractures can have complex shapes and orientations. In addition, hydraulic-fracture size and orientation can change when they intersect existing natural fractures in the reservoir. To reduce model complexity, engineers often assume uniform properties and that could reduce the accuracy of the simulation result.
Here we develop a novel workflow that uses microseismic data to better estimate the uncertainty in fracture properties. The approach uses a variety of techniques: density-based spatial clustering of applications with noise (DBSCAN), surface fitting, embedded discrete fracture modeling (EDFM), and proxy-based assisted history matching (AHM). We first fit a fracture model to the microseismic data and change the size of the fractures using scaling factors. Then, we use the scaling factors as history-matching parameters to address the uncertainty in the fracture model. The workflow can predict fractures properties such as dip angle, direction, fracture half length, and fracture height.
This approach leads to building a realistic fracture model and a well-calibrated reservoir model. We also found that the initial fracture fit over-predicts the fracture size suggested by the AHM workflow. We expect that simulation models built using the proposed workflow will be more accurate compared to the ones built using uniform fracture properties.
One important outcome of this work is the establishment of a direct method to build a discrete fracture model using microseismic events. Achieving this outcome would lead to several field applications. For example, we can compare fracture models for several wells in the same field and rank them based on the total fracture surface area. Then, by comparing the fracture procedure in each well, we can determine which fracturing procedure resulted in the most successful fracture model. Establishing an optimum fracturing procedure is extremely valuable for future wells. In addition, using the fracture orientation from our model, we can estimate the orientation of the minimum principle stress, which is an important parameter in geomechanics and fracture treatment design.
