Tight gas resource has been getting more and more focus due to its huge potential all over the world. Hydraulic-fracturing treatments have become an essential technology for tight gas field development. Tight gas formations often contain natural fractures (NF). In presence of natural fractures, more complex fracture networks may form during the hydraulic treatment. The interaction between fractures may alter the way the hydraulic fracture (HF) propagates through the formation, causing a complex network of fractures, which can significantly influence the overall geometry and effectiveness of hydraulic fracture. The hydraulic fracture may cross, dilate or slip into the fracture plane upon its arrival at the natural fractures. Although there have been studies on investigation of the interaction between HF and NF, most of them cannot be applied reliably due to the inherent assumptions in the analytical formulation and numerical work. In cases of laboratory studies, the limitation of the equipment could also be a factor. With this study, three models based on fuzzy and neural network methods are presented. They are back-propagation neural network (BPNN) model, probabilistic neural network (PNN) model and multi adaptive neuro-fuzzy inference system (ANFIS) model. These models can be used to predict whether a HF will cross, dilate or slip into a NF under different conditions, i.e. under different approach angles, different differential horizontal stress, fracture overpressure, friction coefficient and so on. The advantage of these models is that they are data-based models. Therefore, as long as there are reliable data to train the models, they can work smoothly and easily, so that they can predict the interaction in real time with reliable accuracy. Hypothesis data has been generated from other people's criteria to test the prediction accuracy of these models. The results have been compared to these criteria which show good agreement. This has proven the strong mapping ability of the developed models. Insightful parametric study has also been conducted to investigate the effect of the factors on different types of interaction. Conventional HF design is based on the assumption that the rock is homogeneous and the fracture propagates symmetrically in a plane perpendicular to the minimum stress. In naturally fractured reservoirs due to interaction with NF, the fracture may propagate asymmetrically or in multiple strands or segments. The fuzzy and neural network models developed in this work give the ability to predict the possible interaction patterns under different scenarios thus to assist updating and optimizing the fracturing design
Date of Award | 2014 |
---|
Original language | American English |
---|
Supervisor | MD Rahman (Supervisor) |
---|
- Applied sciences
- Computer science
- Fuzzy systems
- Hydraulic fracturing
- Neural networks
- Petroleum engineering
- 0765:Petroleum engineering
Fuzzy and neural network models for predicting interaction between hydraulic fracture and natural fracture
Chen, P. (Author). 2014
Student thesis: Master's Thesis