Abstract:
This thesis discusses the construction and application of a suite of numerical and data-driven models to address a series of scientific problems from the economic thermal extraction of enhanced geothermal systems (EGS) to the relationship between injection and seismicity in geothermal reservoirs.
In the first part, I propose a novel ‘partially bridging multi-stage hydraulic fractures with horizontal-well EGS’ design. I tested the thermal performance of the proposed design, which suggested a higher long-term stabilised production temperature and delayed thermal breakthrough, comparing it with the commonly used fully bridging design proposed by Gringarten (1975). I built numerical models for the proposed EGS design considering Thermal-Hydraulic-Mechanical (THM) physics and contact opening fractures to test the theoretical feasibility of this design, focusing on thermoporoelastic fracture opening and associated permeability enhancement. Stress evolution indicates that this EGS design could result in a greater degree of secondary stimulation in the stimulated reservoir volume (SRV). These results indicate that the partially bridging fracture design is a promising candidate for practical EGS implementation. The separate role of thermoelastic, poroelastic and fracture opening effects in EGS reservoir stress evolution are further discussed in this study.
The second part of the thesis evaluates the techno-economic performance of the proposed partially bridging fractures EGS design for district heating applications using the FEHM simulator. The heat extraction numerical model developed in FEHM couples the reservoir and horizontal wellbores subject to seasonal heating requirements. A binary search protocol to optimize FEHM simulations of heat extraction was developed to estimate constant achievable thermal power return for a finite operational lifetime. I used this numerical framework as the basis for a Box-Behnken experimental design to develop a response surface reduced-order
model for rapid evaluation of EGS performance under different parameter combinations. The sensitivity analysis indicates that SRV permeability and initial reservoir temperature play essential roles in controlling EGS profitability. This workflow enabled me to investigate an EGS case study using Monte Carlo simulation to quantify the economic and environmental performance (emission offset) of the scheme with uncertainty.
For the third part of this thesis, I applied two classes of data-driven models (the time series feature engineering Machine Learning method, and the non-negative linear regression method) to understanding induced microseismicity in the Rotokawa geothermal field, New Zealand, and the Húsmúli reinjection area, Iceland. Towards the goal of mitigating induced seismicity hazards for better geothermal reservoir management, I tried to reveal the associations and patterns hidden by massive and chaotic field injection and seismicity events data between complex injection schedules involving multiple wells and time-varying seismicity. I showed that there is a stronger association between injection and the long-term microseismicity rate than the short-term microseismicity rate. The association between microseismicity and injection wells is evaluated with possible physical interpretation, which provides insights into reservoir connectedness. The injection rate and cumulative injection volume data stream are the most informative data for microseismicity, whereas no evidence was found that rapid injection rate changes trigger seismicity in these two geothermal reservoirs.