QUANTITATIVE INVERSION OF MICROSEISMIC DATA: BAYESIAN MODEL SELECTION USING FAST PROXIES FOR FRACTURING AND WAVE PROPAGATION
Open Access
- Author:
- Singh, Manik
- Graduate Program:
- Energy and Mineral Engineering
- Degree:
- Doctor of Philosophy
- Document Type:
- Dissertation
- Date of Defense:
- March 13, 2019
- Committee Members:
- Sanjay Srinivasan, Dissertation Advisor/Co-Advisor
Sanjay Srinivasan, Committee Chair/Co-Chair
Shimin Liu, Committee Member
Hamid Emami-Meybodi, Committee Member
Charles James Ammon, Outside Member - Keywords:
- Microseismic
Hydraulic Fracturing
Bayesian Model Selection
Quantitative Inversion
Tight Oil/Gas
Petroleum Engineering
Reservoir Engineering - Abstract:
- Hydraulic fracturing has brought a revolution in the oil and gas industry. It has had a tremendous impact especially on oil and gas production of the USA. The potential of hydraulic fracturing in increasing production from existing reservoirs and make tight shale plays producible has been firmly established. There is a 25 fold increase in the recovery estimates of gas-in-place by adapting technologies like hydraulic fracturing. Hydraulic fracturing has attracted a lot of research in the past decades and is still very attractive area of research. With the increasing use of hydraulic fracturing, use of microseismic monitoring has also increased. A propagating hydraulic fracture on interaction with a natural fracture gives rise to seismic waves which are monitored at locations near the fracturing zone. These monitoring tools are used along with models for the fracturing processes to monitor the growth of a hydraulic fracture in the reservoir. In current workflows, microseismic interpretation is used for qualitative description of the reservoir. It gives the information about the presence of natural fractures locations by inverting for the location of the microseismic event. Microseismic waves carry information about the reservoir it travels in. Which is highly heterogeneous and scarcely sampled. The interpretation of this information has a lot of uncertainties in terms of location of the natural fractures. The uncertainties stem from various reasons like the distribution of reservoir properties, discrete fracture network (DFN) and direction of propagation of the hydraulic fracture. A deterministic inversion does not give an idea about the uncertainties associated with the inversion. To correctly infer the uncertainties in such reservoirs there is a need to run multiple computer simulations for every possible DFN configuration and compare the synthetic and observed seismogram. This poses serious challenges as the state of the art numerical forward models are computationally intensive. Hence, we need models, proxies, which are sufficiently able to capture the physics of the problem while being computationally fast. We use analytical models for fracture propagation and seismic wave propagation. These models while having more assumptions than a comprehensive numerical 3D model are considerably faster. However, there is no single analytical or semi-analytical model which would cover the full range of processes involved in the problem. Hence, several standalone models need to be coupled to each other in order to have proxy model, that can yield synthetic seismograms generated corresponding to the propagation of a hydraulic fracture in a DFN. Model selection and assessment of uncertainties requires running the proxy model on a suite of reservoir models with different DFN configurations. Computed seismograms when compared to the observed seismograms yields the posterior set of models which give responses similar to observed microseismic. The Bayesian approach gives the uncertainties associated with the interpretation. Selected DFN configurations can be simulated later for validation using a full physics simulator. The aim of the research is to quantitatively interpret the microseismic data for the description of DFN and quantify the uncertainty associated with the description. Knowledge of DFN and associated uncertainty with interpretation is particularly important in planning drilling, stimulation or fracture jobs in the future.