Abstract

Utilization of direct fracture mapping technologies has greatly increased over the last few years, from the first real-time measurement of fracture dimensions in 1997 to the routine mapping of fracture dimensions on more than 500 fracture treatments per year in 2004. The development of commercial technologies1-5 to routinely measure fracture growth has greatly improved our fracture modeling capabilities, enabling us to distill the essential fracture growth behavior in many environments into calibrated fracture models. 

Calibrated fracture models combine complimentary strengths and weaknesses of fracture mapping and modeling. Fracture models provide the ability to predict how changes to a fracture treatment should alter fracture geometry6, but suffer from a tenuous and generally unknown relationship with reality.  Fracture mapping provides a direct measurement of fracture geometry from a given treatment, but cannot be used to predict what might happen under a different set of conditions. By combining direct measurements with models, we can create calibrated frac models with superior predictive capabilities.

Calibrated models have been developed for various regions and formations, and the improvements in predictive modeling capabilities have lead to a proliferation of calibrated fracture models throughout the industry. This has provided improved insight into fracture growth behavior in a diverse set of environments including the North Texas Bossier and Barnett shale, the East Texas Cotton Valley sands, and various formations in the Rockies, the San Juan basin and California.

This paper discusses how fracture growth models can be improved using direct fracture geometry measurements and what changes in models have been necessary to accomplish this. We will also discuss minimum data requirements for calibrated models, discuss their main limitations and capabilities, and the strategies that are used to define calibrated models. Finally, we will present several case histories – comprising the results of over a hundred mapped treatments – to illustrate how these findings have been used to improve fracture treatment design, execution and economics in various formations and regions across the United States.

Why Bad Models Happen to Good Engineers

Predictive modeling of fracture dimensions and the associated production stimulation has often been more an art than a science.  Many good engineers have turned their back on modeling efforts to find better strategies to stimulate a specific target zone, and have reverted back to the safe-haven of the “cookie-cutter” design that “worked just fine” in many instances.

The first problem of fracture models is that the inputs for the model are typically not very well defined. The most critical input parameters for fracture models are generally the Young’s modulus of the rock, the permeability and the fracture closure stress, all to be determined along the depth interval of interest. All of these parameters greatly affect fracture height growth, but we typically do not measure these critical parameters.