Author: Morgan F. Ames

Publisher:

Published: 2016

Total Pages:

ISBN-13:

One of the most significant open problems in geothermal reservoir engineering is the development of a reliable and accurate method to predict thermal breakthrough. Such a method would enable more informed decisions to be made regarding reservoir management. Methods developed at present include analytical models and solute tracers, both of which have limitations. The use of particles as temperature-sensitive tracers is a promising approach due to the high degree of control of the physical and chemical properties of nanomaterials and micromaterials. Additionally, particles experience less matrix diffusion than solute tracers and tend to stay in high velocity fluid streamlines, which results in earlier particle breakthrough in the absence of significant particle deposition. These properties could potentially be exploited to infer temperature and measurement location, which could in turn provide useful information about thermal breakthrough. In order to assess whether particle tracers can provide more useful information about future thermal behavior of reservoirs than existing solute tracers, models were developed for both solute tracers and particle tracers. Three existing solute tracer types were modeled: conservative solute tracers (CSTs), reactive solute tracers with temperature-dependent reaction kinetics (RSTs), and sorbing solute tracers that sorb reversibly to fracture walls (SSTs). Additionally, three particle tracers which have not been developed in practice were modeled: dye-releasing tracers (DRTs) that release a solute dye at a specified temperature threshold, threshold nanoreactor tracers (TNRTs) with an encapsulated reaction that does not begin until a specified temperature threshold is reached, and temperature-time tracers (TTTs) capable of recording detailed temperature-time histories of each particle. In this study, TTTs represent the most informative tracer with respect to thermal breakthrough. These models were used in the context of an inverse problem in which synthetic tracer data were calculated for several "true" discrete fracture networks. Next, computational optimization was used to match these data by adjusting fracture location, length, and orientation for a variable number of fractures. Finally, the thermal behaviors of the fracture networks with the best fits to the data were compared to those of the true fracture networks, and the tracers were ranked according to their forecasting ability. Overall, thermal breakthrough forecast error was found to increase with fracture network complexity. However, in all cases, all tracers forecasted thermal breakthrough with unrealistic accuracy. This is partly due to neglecting thermal interference between closely spaced fractures in the thermal model. In all three cases, CSTs were found to be the least informative tracer type because they are insensitive to temperature. SSTs were also modeled as insensitive to temperature in this work, but they performed better than CSTs because sorption is sensitive to surface area, which is also closely related to a reservoir's thermal performance. In order to fully understand the relative informativity of these solute and particle tracers, a second study was performed using a uniform parallel fracture reservoir model that accounts for interference between fractures in both thermal and tracer transport. In this study, a seventh type of tracer test was also considered in which all three solute tracer types (CSTs, RSTs, and SSTs) were used simultaneously to gain the benefits of all three tracer types. This tracer type was designated ALLSOL, which is short for "all solutes." As with the discrete fracture network modeling study, synthetic data were generated and matched using optimization, after which thermal breakthrough forecasts were calculated. The decision variables used in optimization were the number of fractures and fracture length, width, aperture, and spacing. Two inverse problem scenarios with different fracture spacings were examined: 15 meter spacing and 5 meter spacing. In both scenarios, all individual solute tracers had significant error, particle tracers and ALLSOL forecasted thermal breakthrough more accurately than individual solute tracers, and ALLSOL had slightly more accurate forecasts than particle tracers. In the 15 meter spacing scenario, both RST and TNRT had very inaccurate forecasts because the temperature distribution is somewhat insensitive to fracture spacing at early time when fracture spacing is sufficiently large. This resulted in good matches and small objective function values for inaccurate estimates of fracture spacing. In order to determine if other tracers besides RST and TNRT are insensitive to spacing at early time when spacing is sufficiently large, the objective function values of all tracer types were evaluated using the optimal solution for TNRT in the 15 meter spacing scenario. Low objective function values and good fits to the data were observed for every tracer type except for TTT, indicating that TTT is the only tracer type considered that is capable of detecting differences in spacing at early time when the true fracture spacing is large. This is because the temperature is measured directly by the TTT rather than inferring the temperature from the return curve, as is the case for all other tracer types. In the 5 meter spacing case, the RST had a very inaccurate thermal breakthrough forecast because its return curve has a nonunique relationship with the temperature distribution (i.e. the RST return curve was matched by a reservoir with a significantly different temperature distribution from the true reservoir, which happened to result in the same amount of reaction). Forecast error was generally larger in the uniform parallel fracture modeling scenarios than in the discrete fracture network modeling scenarios. This demonstrates the importance of accounting for thermal interference in temperature-sensitive tracer modeling.