Probability based volumetric "heat in-place" methods for predicting power (electricity) generation potential of liquid-dominated geothermal systems

Reserve estimation is the vital part of reservoir engineering to be able to get an idea about the accumulation and the producibility of the reserve. One should make realistic guesses about the areal extension, thickness, resource temperature, porosity, density, isobaric, volumetric specific heat capacity of reservoir rock to be able to make reserve estimations of the geothermal resource. Also recovery factor is an important input parameter having the most uncertainty in estimation of the recoverable portion of the reserve in geothermal field. It is very usual to use a volumetric probabilistic approach for the purpose of getting recoverable reserve estimations because deterministic approaches do not consider uncertainty propagation in estimation. USGS (United States Geologic Survey) and MIT (Massachusetts Institute of Technology) methods are widely used for volumetric probabilistic approaches to reserves with estimated input data distributions of the reservoir parameters (Garg and Combs, 2015), (Başel, 2010). When these methods are associated with probabilistic methods like MC (Monte Carlo) or its simple alternative Analytic Uncertainty Propagation (Onur et al. 2010), they give probabilistic results to assess the uncertainty in the estimation by using commonly used statistical markers of P10 (proved), P50 (probable) and P90 (possible). Clearly it is impossible to have an exact idea for the geothermal in-situ reserves due to the included uncertainties related to reservoir characteristics such as tectonic settlement, fracture networking, porosity, thickness, areal extension, heat flow mechanism or so. That is why researchers often use stochastic approaches instead of deterministic ones. Monte Carlo simulation or analytical based uncertainty propagation methods give results by using statistical markers P10 (proved), P50 (probable) and P90 (possible) to propagate the uncertainty. These markers represent the confidence of the estimated results belong to real geothermal reserve. "P" stands for the percentile of confidence. As an example; If one says; • The estimated P10 value is 200 MWe then that means there is at least a 90% probability (chance of occurance) that the exact, unknown reserves recovered will be equal or exceed the estimated value of 200 MWe. This is equivalent to starting that there is a 10% chance of occurance that the estimated value of 200 MWe will be equal or less than the unknown reserves recorded. • The estimated P50 value is 300 MWe then that means there is at least a 50% probability (chance of occurance) that the exact, unknown reserves recovered will be equal or exceed the estimated value of 300 MWe. This is equivalent to starting that there is a 50% chance of occurance that the estimated value of 300 MWe will be equal or less than the unknown reserves recorded. • The estimated P90 value is 400 MWe then that means there is at least a 10% probability (chance of occurance) that the exact, unknown reserves recovered will be equal or exceed the estimated value of 400 MWe. This is equivalent to starting that there is a 90% chance of occurance that the estimated value of 400 MWe will be equal or less than the unknown reserves recorded. These examples show that the P10 provides a conservative estimate of reserves, while the P90 provides an overoptimistic estimate of reserves. If the distribution of reserves is symmetrical, which is rarely the case, it could be said that P50 value has a more confidence to occur because it is closer or equal to the mean value of the real potential of reserve. However, the distribution of the reserves is usually log-normal and hence, it is usually preferable to consider the tails of the distribution (Capen 1996, 2001). For example, P10 is the value used to solicit financing from banks and these statistical markers are also required by private and government institutions or agencies to classify the potential of the oil, gas or geothermal reservoirs as well as providing finance to the projects proposed by the investors or decision makers (Capen 2001). USGS and MIT methods use arbitrarily chosen reference temperature and thermal power conversion efficiency values for their estimations without taking the second law of thermodynamic and the installed power conversion system into consideration. So this leads to overestimated results of reserve estimations, as shown by Garg and Combs (2015). Garg and Combs (2015) proposed a new method that a geothermal field must be treated with an approximation that relies on the resource temperature of the reservoir. If the reservoir temperature value is higher than 180 oC than this reservoir is proper for single or double flash power conversion system and if the reservoir temperature is between 57-180 oC then a binary power plant should be installed to be able to produce electricity from the resource efficiently. Also the input parameters used for the volumetric probabilistic reserve estimation should be chosen related and accordant to the power conversion system. Monte Carlo simulation method is the widely used method for reserve approximations in a probabilistic way and also analytical uncertainty propagation method presents an alternative analytical way to probabilistic reserve estimations giving very close resultant values in comparison to Monte Carlo simulations. In this thesis, all volumetric estimations of power generation potential based on USGS, MIT and Garg and Combs (2015) have been evaluated by MC simulation and AUP methods. The working equations of AUPM for each of the three methods have been derived in this thesis. Then, the applications of AUP method have been presented and compared with the MC simulation results applied to USGS, MIT and Garg and Combs (2015) methods. The results predicted by the new AUPM equations derived in this work showed an excellent agreement, not requiring extensive MC simulations and a commercial software use. Suggestions were made about the usage of USGS and MIT methods that give overestimated results by the usage of arbirtarily chosen reservoir input parameters. At the last section of this thesis, Garg and Combs (2015) method was applied to 25 geothermal fields of Turkey, which are amenable to power generation, by MCM and AUPM approximations. The same 25 fields were originally considered and studied by Başel in 2010 in her PhD thesis by using the MIT method. In Başel's dissertation, these geothermal fields were all treated by a reference temperature of 100 0C when estimating Turkey's power generation potential. However, today it is well known that binary power conversion plants are designed to generate electricity from resources having reservoir temperature values ranging from 57 to 180 0C. It is shown here that Garg and Combs (2015) method without any usage of arbitrary values and considering the installed power conversion system and thermodynamic properties of the produced water or the secondary fluid of the power conversion system is the best method to eliminate the subjectivity in selecting the reference (or abandonment) temperature and conversion efficiency and hence to estimate or predict the power generation of a geothermal field or country more realistically. In this thesis, it also shown that the AUPM and MC based Garg and Combs (2015) method predict the power generation of of the 25 geothermal fields of Turkey considered in Başel's PhD dissertation as P10 = 661 MWe, P50 = 1165 MWe, and P90 = 1909 MWe, based on arithmetic summation. On the other hand, for the same geothermal fields, Başel (2010) predicted the power generation potential as: P10 = 840 MWe, P50 = 1428 MWe, and P90 = 2371 MWe. As the same data for all 25 geothermal fields provided by Başel, it is clearly the MIT method used by Başel overestimates the power generation potential of Turkey based on these 25 geothermal fields.