A Holistic Methodology to Quantify Product Competitiveness and Define Innovation Requirements for Micro Gas Turbine Systems in Hydrogen-Based Energy Storage Applications

A HOLISTIC METHODOLOGY TO QUANTIFY PRODUCT COMPETITIVENESS AND DEFINE INNOVATION REQUIREMENTS FOR MICRO GAS TURBINE SYSTEMS IN HYDROGEN-BASED ENERGY STORAGE APPLICATIONS Giuseppe Tilocca1,∗ , D. S´ anchez1, M. Torres Garc´ıa1, A. Escamilla Perej´ on1, S. Minett2 1Department of Energy Engineering, University of Seville, Camino de los descubrimientos s/n, 41092 Seville, Spain 2Challoch Energy Rue Eug` ene Castaigne 29, 1310 La Hulpe, Belgium ABSTRACT Micro gas turbines are an on-site power and heat generation technology with a small footprint, low gaseous (NOx) and acoustic emissions, low maintenance and high-grade heat. They entered the market at the dawn of the twentieth century; nevertheless, they achieved minimal success and a marginal role in the microgeneration market. Reciprocating internal combustion engines raised considerable barriers hindering their market deployment, and Fuel Cells are also set to compete in this segment. In this scenario, this work presents an analysis of competitiveness grounded in the Theory of Constraints. To this end, a specific Key Performance Indicator has been produced, which combines technical, economic, and operational factors according to the end-user requirement. This indicator is a function of several penalty factors representing technology and market barriers, which aims to yield a unique insight into the most competitive technology for a given application, accounting for the uncertainty deriving from technical and economic elements. This novel methodology is applied to a new potential niche market: Power-to-Hydrogen-to-Power for remote applications. The methodology is applied to an independent rural community in South Wales, for which a backup power system is assessed. Four technologies are considered in the analysis: reciprocating engines, fuel cells and two different microturbines layouts. Finally, this work provides an overview of the possible R&D&I paths necessary to increase the competitiveness of micro gas turbines in certain markets. Nomenclature −R,−ICR Recuperated, Intercooled-Recuperated ηElectric Efficiency CCost COP Coefficient of Performance DALY loss of Disability Adjusted Life Year FC Fuel Cell HC Heat Consumption ICE Internal Combustion Engine k,KPreference range Index, Preference ranges count KPI Key Performance Indicator LB,lb Lower Bound LCoE Levelised Cost of Electricity LoLE Loss of Load Expectation ∗Address all correspondence to this author. Email: [email protected] 1 GTP-23-1412 - Tilocca Accepted Manuscript Not Copyedited Journal of Engineering for Gas Turbines and Power. Received August 07, 2023; Accepted manuscript posted November 18, 2023. doi:10.1115/1.4064061 Copyright (c) 2023 by ASME Downloaded from http://asmedigitalcollection.asme.org/gasturbinespower/article-pdf/doi/10.1115/1.4064061/7060164/gtp-23-1412.pdf by Universidad De Sevilla user on 06 December 2023 MGT Micro Gas Turbine MPMT Mean Preventive Maintenance Time MT BF Mean Time Between Failure MT TR Mean Time to Repair n,NRequirement Index, Total count of requirements O&MOperation and Maintenance OEM Original Equipment Manufacturer P2PPower to Power PC Power Consumption PEM(FC)Polymeric Exchange Membrane (Fuel Cell) R&D(&I)Research and Development (and Innovation) RAMS Reliability Availability Maintainability & Safety SD Standard Deviation SI Service Interval TIT Turbine Inlet Temperature TOC Theory of Constraints UA Unavailability UB,ub Upper Bound UQ Uncertainty Quantification INTRODUCTION Microturbines or Micro Gas Turbines are small on-site power and heat generators. Their small scale and modularity make them eligible as a technology to help the transition from centralised to distributed energy systems, thereby contributing to a low-carbon economy [1]. Their contribution might be relevant to achieve the ambitious objectives of the European Union, whose Green Deal plans to make Europe the first climate-neutral region in the world by 2050 [2]. MGTs have been in the market for over two decades. Nonetheless, their deployment could not match the expectations of the market (end-users) and the industry (technology providers) for this promising technology inasmuch as the long-lasting presence of reciprocating internal combustion engines (ICE), the established technology in the energy and transportation sector, raised substantial barriers against MGT commercialisation. This paper is part of a corrective actions scheme [3] that aims to identify the root causes of this experience and propose some remedial actions to enable the successful market deployment of microturbines. In particular, rather than relying on a speculative methodology only, as it is usually the case for this type of action, it aims to create a numerical approach that can mimic the Root Cause Analysis process. To this end, it is mandatory to devise a methodology that can model the competitiveness of MGTs compared to other commercial technologies. In literature, Weighted Decision Matrices are a standard process to select the best candidate amongst a set of different options [4, 5, 6, 7]. In Weighted Decision Matrices, multiple parameters are weighted depending on their importance. Nevertheless, they are specific to their application and only provide information about that individual decision; for instance, Sanchez et al. created a compound indicator to identify potential markets for small-scale Concentrating Solar Power systems based on hybrid microturbine and parabolic dish arrangements [8]. Then, Tilocca and al. combined the fitness indicator of Sanchez et al. and a traditional Weighted Decision Matrices [9]. The resulting methodology assessed the competitiveness of an extensive range of commercial prime movers for several cogeneration applications. This work builds on Tilocca et al.’s approach by introducing a penalty factor based on the Theory of Constraints. The authors seek to improve the previous work by creating a flexible tool for technology innovation that removes arbitrariness and helps drive engineering according to market demand. The tool identifies and quantifies factors necessary for competitiveness in a specific market segment and provides guidance on areas where innovation is needed. Lastly, it serves as a multi-purpose means to advise industry and academia on how to target market needs. The paper’s first section presents the Theory of Constraint problem-solving technique. The next part introduces the three competing technologies under analysis: Micro Gas Turbines, reciprocating Internal Combustion Engines and Fuel Cells. Once the context of the paper is set, the following section presents the methodology that defines the Key Performance Indicators. After that, the models used to simulate the performance of Power-to-Power systems accounting for boundary conditions like weather and microgrid for the case study are presented. Then, the following section introduces the approach to uncertainty quantification adopted, including the definition of the methodology and the range of uncertainty of the parameters used. The last two sections discuss the results and provide an overview of the necessary technological improvements and innovation. The conclusive remarks also mention the prospects for future related studies. 2 GTP-23-1412 - Tilocca Accepted Manuscript Not Copyedited Journal of Engineering for Gas Turbines and Power. Received August 07, 2023; Accepted manuscript posted November 18, 2023. doi:10.1115/1.4064061 Copyright (c) 2023 by ASME Downloaded from http://asmedigitalcollection.asme.org/gasturbinespower/article-pdf/doi/10.1115/1.4064061/7060164/gtp-23-1412.pdf by Universidad De Sevilla user on 06 December 2023 METHODOLOGY Theory of Constraints The Theory of Constraints (TOC) was initially drafted by Goldratt [10] and then grew into an effective problem-solving process holistically integrating intuitive and analytical thinking [11]. The approach followed in this work is mainly inspired by the work by Dettmer [12]. The Theory of Constraints aims to improve systems: a set of processes or entities co-acting to achieve a goal. Therefore, defying a problem only makes sense when comparing the system’s performance to its expectations. The principle at the core of the TOC is that ”a chain is no stronger than its weakest link”; i.e., each system is affected by its constraint, suggesting that the weakest component does negatively affect the entire system [11]. According to Dettmer [12], the most undesirable effects within a system result from one or a few root causes. These high-level causes are hidden at the end of the cause-effect chain and are usually hard to detect. They result in several undesirable effects connected by an intricate network of cause-effect relationships. Removing them individually (i.e. the symptoms or the low-level causes) delivers a short-term solution that cannot resolve the issue. On the contrary, targeting the root causes eliminates the undesired effects along their cause-effect chain. A penalty-based approach to quantify competitiveness Tilocca et al.’s work defined a KPI to monitor competitiveness under the name of Global Penalty Factor [9]. The indicator is the weighted sum of single linear penalty functions and enables a model of the competitiveness of different technologies for specific applications. The cited work evaluated the effectiveness of this KPI to compare an extensive range of commercial engines. Their definition of weights models the importance of each feature monitored by the linear penalty functions, yet, it can introduce some arbitrariness. Moreover, ranking the objectives and determining the weights is a simplified representation of more complex problems and system dynamics. From this observation and experience, the methodology in this paper translates the principle of the TOC into a mathematical formula, making the utilisation of arbitrary weights and parameter ranking unnecessary. The principle behind the TOC is that each system is affected by its constraints, the chain’s weakest link, so it does not matter how strong all the other links are. Following this principle and the overarching objective of this paper, as put forward in the Introduction, the following requirements that the new numerical method should follow are set: •The KPI monitoring user/market perception should be a maximising function taking values between [0,1]. The KPI should be easy to read and compare. •The contributions to the KPI are penalty factors. Each penalty simulates the addition of a cause to the effect. Following the TOC problem-definition scheme, they are computed by comparing the system performance against the user’s requirements. •The KPI must lie in a range which is not better than the worse single penalty, according to the TOC principle. •The KPI and the penalties should be continuous functions to create a valuable numerical tool suitable for optimisations, uncertainty and sensitivity studies. The mathematical function that satisfies the aforementioned requirements for a minimisation problem is the exponential function defined as: penaltyn=        1 if x<xideal Nk+   x−lbk ubk−lbk  if xideal <x<xun f easible NKif x>xun f easible (1) where Kis the number of preference intervals, lbkand ubkare the lower and upper bound for each sub-range krespectively, and Nis the number of user requirements (i.e. penalties) of the problem considered. The global penalty and the competitiveness KPI are then defined as: GlobalPenalty = N ∑ n=1 penaltyn(2) KPI =1− logN(GlobalPenalty)−1 K(3) 3 GTP-23-1412 - Tilocca Accepted Manuscript Not Copyedited Journal of Engineering for Gas Turbines and Power. Received August 07, 2023; Accepted manuscript posted November 18, 2023. doi:10.1115/1.4064061 Copyright (c) 2023 by ASME Downloaded from http://asmedigitalcollection.asme.org/gasturbinespower/article-pdf/doi/10.1115/1.4064061/7060164/gtp-23-1412.pdf by Universidad De Sevilla user on 06 December 2023 (a) (b) (c) FIGURE 1: This case involves assigning penalties and calculating the Key Performance Indicator (KPI) for K=3 user preference intervals and N=3 user requirements. Five examples are given on the left-hand side of subfigure a) to show the score (x) against the user requirements. The right-hand side of a) displays the conversion to a single penalty factor (Eq. 1). The penalties are then added in subfigure c) using Eq. 2. Subfigure d) exemplifies the calculation of the KPI using Eq. 3. The five examples demonstrate different scenarios, with Example 5 showing that the highest penalty for Requirement A brings the solution to an unfeasible region (Eq. 4) according to the TOC principle, regardless of the x score against Requirements B and C. User requirements The present work considers three different user requirements: N=3. They are relative to economic performance, energy system availability and reliability, and the impact of emissions on human health. For each requirement, three preference intervals are defined (K=3). In this case, the final KPI could be interpreted as:      1.00 <KPI <0.67 =>best range 0.67 <KPI <0.33 =>f easible range 0.33 <KPI <0.00 =>un f easible range (4) Where KPI =1 represents the ideal solution. Figure 1 demonstrates the framework for a case having N=3 and K=3. The figure of merit for the system’s economic performance is the Levelised Cost of Electricity (LCoE), according to the definition by the International Energy Agency [13]. The wholesale price of electricity predicted for 2030 sets the ideal range threshold [14] whilst the preferred, acceptable and feasibility thresholds correspond to the 2030 projection of LCoE for Secondary Response, Peaker, and Seasonal Storage [15]. The authors selected the Loss of Load Expectation (LoLE) to represent the reliability of the independent microgrid. LoLE is a commonly adopted approach in generation planning studies. It assesses the need for generation reserves as a function of load magnitude, factor and shape, generation availability, and the size and number of generating units [16]. LoLE is computed following the scheme provided by Vijayamohanan Pilla [17]: LOLE =∑ i Piti(5) 4 GTP-23-1412 - Tilocca Accepted Manuscript Not Copyedited Journal of Engineering for Gas Turbines and Power. Received August 07, 2023; Accepted manuscript posted November 18, 2023. doi:10.1115/1.4064061 Copyright (c) 2023 by ASME Downloaded from http://asmedigitalcollection.asme.org/gasturbinespower/article-pdf/doi/10.1115/1.4064061/7060164/gtp-23-1412.pdf by Universidad De Sevilla user on 06 December 2023 The probability of an event taking place (event occurrence probability, Pi) derives from the achieved availability, including forced (MTBF and MTTR) and planned outages (SI, MPMT). In contrast, the event duration (ti) is calculated based on the yearly load profile. It represents the number of load hours a specific outage would potentially affect in a year; in other words, given a yearly load profile and a capacity outage scenario, the number of hours with a load demand greater than the leftover system capacity. LoLE’s preference ranges are taken from techno-economic, and reliability studies on microturbines by Davis [16]. The ideal LoLE is 0 (i.e., the system operates without interruption). The preferred value is the usual requirement for electric grids, equivalent to 0.1 days per year (2.4 h). The acceptable range is marked by a loss of load expectation of 1 day (24 h), and more than 3 days is deemed unacceptable. The impact of the calculated emissions of the prime movers follows the damage assessment scheme adopted by Macor and Bennato [18]. This procedure aims to estimate the damage caused by each emitter by evaluating it in DALY units1. This work considers NOx emissions only, given that the stacks run on pure hydrogen, so the cited emission assessment is qualified for the Respiratory Inorganics damage category. Nonetheless, this parameter can holistically estimate the impact of contaminants for conventional and unconventional fuels other than H2. Effectively, it can integrate the impact of different contaminants based on the effects on human health for several damage categories. This technique does not need further weighting factors. The characterisation is computed as follows: DALY =sd∑ i C fixi(6) where xis the mass of each contaminant, Cfis its characterisation factor, and sdis the specific damage in kg−1 eq for each impact category. The values are taken from [18]. Since there is no aggregate emission regulatory requirement based on health assessment, the DALY preference ranges are taken from the natural gas and biogas measurements provided by Macor and Bennato. TABLE 1: Preference ranges for the considered user requirements. n Requirement k=1 k=2 k=3 Reference LB UB LB UB LB UB 1 LCoE 0.074 0.2 0.8 2.0 [14, 15] 2 LoLE 0.0 2.4 24 72 [16] 3 DALY 0.0 6.80E-5 2.34E-4 4.68E-4 [18] Description Ideal Preferred Acceptable Unfeasible Rural Community Power-to-Power model This work takes over a feasibility study by Challoch Energy, aimed at investigating the potential supply of green hydrogen to a rural energy community under the ’Hydrogen in Rural Energy Systems (HyRES)’ project. The reference is a local community in South Cornelly, Bridgend County Borough (South Wales). In the same village, a Whole System Business Research Innovation for Decarbonisation (WBRID) demonstration project also investigates how to organise low-carbon communities forming local energy markets (LEM), which trade in locally produced and consumed renewable energy. The concept combines locally produced hydrogen with a portfolio of renewables, including wind, photovoltaic and batteries. The analysis presented herein applies to the backup power sub-system only. The village is comprised of 220 family houses and a nursing house. The residential power demand is estimated from the readings taken by the authors from nine houses between February and May 2022. In this period, consumption profiles proved uniform. Natural gas boilers provide heating to all the sampled houses; hence, the demand for electric power registered is only due to lighting and appliances. These considerations are assumed to be representative of the whole village. Therefore, the final consumption profile is the average measured profile for the nine houses multiplied by the total number of places. In the current scenario, the residential houses rely on natural gas boilers, and the nursing house has a heat pump installed for heating. For this study, the authors assume the whole village is electrified, and both residential and nursing houses have installed heat pumps. The 1DALY: disability-adjusted life year. One DALY represents the loss of the equivalent of one year of full health. 5 GTP-23-1412 - Tilocca Accepted Manuscript Not Copyedited Journal of Engineering for Gas Turbines and Power. Received August 07, 2023; Accepted manuscript posted November 18, 2023. doi:10.1115/1.4064061 Copyright (c) 2023 by ASME Downloaded from http://asmedigitalcollection.asme.org/gasturbinespower/article-pdf/doi/10.1115/1.4064061/7060164/gtp-23-1412.pdf by Universidad De Sevilla user on 06 December 2023 residential heat demand model considers three different heat profiles based on the house layout. These layouts are assigned following an on-site assessment: •Terraced houses 10000kWh/year: 66 houses •Semi-detached house 13000kWh/year: 121 houses •Detached house 16000kWh/year: 33 houses The 2.77 GWh/year residential heat demand closely agrees with the gas provider readings. The care home hourly heat consumption is computed from literature by combining a monthly consumption [19] and the hourly normalised profile [20]. The nursing house’s electricity profile is the sum of the heat pump power demand and the electricity consumption for lighting and appliances PPcarehome. The latter is derived from the elderly care wards’ power profile per m2reported in [21]. PPcarehome is assumed constant throughout the year following the same assumption as for the residential houses. The overall power consumption is then computed as follows: PCtot =PChouses +PCcarehome +(HCcarehome +HChouses)/COP (7) The effective monthly COP as a function of the ambient temperature is assumed according to the values reported by Staffell et al. in their review of domestic heat pumps [22]. The Power-to-Power (P2P) model, which ultimately calculates the levelised cost of hydrogen, is an in-house software developed by Escamilla et al. at the University of Seville [23]. Weather data and models for renewable energy technologies are taken from the System Advisor Model (SAM) of The National Renewable Energy Laboratory (NREL) [24], using irradiation data from the Photovoltaic Geographical Information System (PVGIS) of the European Commission’s Joint Research Centre [25] and wind data from NREL’s National Solar Radiation Database [26], all for a geographic location near South Cornelly in Bridgend County Borough (South Wales). Costs relative to the P2P sub-system are gauged from vendors’ data, as listed in Table 2. The model accounts for the hourly RES generation and power demand, after which the hydrogen production and the residual power demand to the prime mover are computed. TABLE 2: Economic assumptions of the P2P model. Item Capital Cost O&M Cost Wind Turbine 1695 e/kW 51 e/kW year PV 1000 e/kW 9e/kW year Electrolyser 1500 e/kW 20 e/kW year Water - - 4.9 e/m3 HP Vessel 70 e/Nm3- - Compressor 4000 e/kW 240 e/kW year The model simulates each hour of the year for the total lifespan of the equipment. If renewable power production exceeds the electric demand, surplus energy drives the electrolyser, producing hydrogen stored in high-pressure tanks. Contrariwise, if the demand exceeds the available renewable power, the prime mover (MGT, ICE or FC) runs on stored H2to fill the energy gap. This work considers an arrangement comprised of two parallel 400 kWeprime movers to satisfy the LoLE feasibility requirement inasmuch as an arrangement with one engine only would not provide sufficient availability, as observed in an early study by Davis [16]. Figure 2 displays the installed capacity and power demand profiles for the characteristic year considered. With this configuration, the systems operate with a 12% reserve margin. Technical specifications under Uncertainty In many engineering problems, the uncertainty of input data and parameters is a significant and critical concern. The analysis of complex energy systems commonly assumes a series of deterministic input parameters. Nevertheless, many of these variables are subject to uncertainty since they may not be measured, determined or predicted with enough precision, and they can strongly affect the final output. The evaluation of competitiveness based on a techno-economic analysis is no exception, and, therefore, the assessment of market competitiveness through the KPI presented in this work incorporates Uncertainty Quantification. Given the relatively quick KPI 6 GTP-23-1412 - Tilocca Accepted Manuscript Not Copyedited Journal of Engineering for Gas Turbines and Power. Received August 07, 2023; Accepted manuscript posted November 18, 2023. doi:10.1115/1.4064061 Copyright (c) 2023 by ASME Downloaded from http://asmedigitalcollection.asme.org/gasturbinespower/article-pdf/doi/10.1115/1.4064061/7060164/gtp-23-1412.pdf by Universidad De Sevilla user on 06 December 2023 FIGURE 2: Annual power demand profile and reserve margin. The power demand for the prime mover is the total demand minus the available renewable power. evaluation (around 0.25s in an average i7 laptop) and the level of accuracy needed, the authors opted for a direct application of the Monte Carlo Method. Table 3 summarises the information about Uncertainty Quantification. The installation cost is assumed to be a normal distribution with a mean value of 830e/kW and 20% standard deviation (SD) for all technologies. The microgrid failure rate of 0.2 failures per year and the Mean Time to Repair (MTTR) of 12 h is taken from Kumar et al. [27]. Micro Gas Turbines Micro Gas Turbines are miniature gas turbines with rated electric output between 1 and 500 kWeapproximately [28]. Due to their reduced volumetric flow, MGTs primarily use single-stage radial turbomachinery for compression and expansion. Since radial turbine cooling is not commercially available, microturbines feature low TITs of around 950ºC. Also, given the adoption of singlestage turbomachinery, pressure ratios do not exceed 4-5:1). These are different specifications to larger gas turbines, which feature TITs higher than 1500◦C[29] and pressure ratios higher than 20:1. The low TITs and pressure ratios of micro gas turbines limit the maximum thermal efficiency of a simple Brayton cycle, whose low performance is further exacerbated by the small scale of turbomachinery [30, 31, 32]. Accordingly, the most widespread solution is the recuperated MGT (MGT-R) which uses an internal heat exchanger to enhance its thermal performance substantially. However, this is at the expense of a costly and bulky component. Another configuration that could enhance the performance of the microturbine further is the intercooled and recuperated MGT (MGT-ICR). This combines intercooling between two compressor stages and internal heat recovery, further increasing electrical efficiency [33]. The performances of the micro gas turbines are taken from representative commercial products; Capstone’s C200 [34] and Ansaldo Green Tech’s AE100 [35] for MGT-R and Aurelia Turbine’s A400 [36] for MGT-ICR. The R&D community is exploring the possibility of burning hydrogen on MGTs, which is currently not commercial. In a recent study, Banihabib and Assadi [37] declared to have operated a T100 (100 kWe) MGT on pure H2and NG-H2blends, making use of a FLOX®combustor designed by the German Aerospace Centre (DLR). The measurements reveal that NOx emissions did not grow significantly, certainly not exceeding the regulated levels, nor was any significant shift in performance experienced when running on pure H2. Based on this experience, the authors of this paper assume that both (C200 and T100) MGTs can operate on hydrogen without significant changes to emissions because the working cycle (TIT, pressure ratio, layout) of both engines are similar. The A400 engine operates with slightly higher TIT (around 1070ºC [33, 38]) and, according to a conversation between representatives of Aurelia Turbines and the authors of this paper, the engine is also being tested with a similar FLOX®combustor developed by DLR. Accordingly, the authors assume that the specifications of the engine running on hydrogen are similar to those of the factory engine running on NG, though with a higher degree of uncertainty due to the different cycles and TIT to the T100 engine. Following these considerations, the efficiency curve for the recuperated MGT is taken from the specifications of the C200 engine [34]. A normal probability density distribution is applied to the peak efficiency value for UQ, with the efficiency corrections developed 7 GTP-23-1412 - Tilocca Accepted Manuscript Not Copyedited Journal of Engineering for Gas Turbines and Power. Received August 07, 2023; Accepted manuscript posted November 18, 2023. doi:10.1115/1.4064061 Copyright (c) 2023 by ASME Downloaded from http://asmedigitalcollection.asme.org/gasturbinespower/article-pdf/doi/10.1115/1.4064061/7060164/gtp-23-1412.pdf by Universidad De Sevilla user on 06 December 2023 by Galanti and Massardo to upscale from 200kWeto 400 kWe[30]. The efficiency curve for the intercooled-recuperated MGT is taken from the specifications of the A400 engine [36], with a standard deviation of SD =0.01 applied to peak efficiency. For both MGTs, the standard deviation is a cubic polynomial with an equivalent SD =0 at 100% power and SD =0.1 at 10% power. Planned maintenance data are taken from Capstone’s and Aurelia’s product maintenance schedules. UQ is applied to the Mean Preventive Maintenance Time (MPMT) only, through a log-normal distribution such that the MPMT is higher than 0.8·MPMTaverage and 80% lower than 1.2·MPMTaverage 95% of the time. The logarithmic failure rate distribution for both MGTs is taken from Robinson et al.’s assessment of the reliability of distributed energy resources [39]. Given the very low failure rate, the results turn out to be weakly affected by MTTR. Equipment and O&M costs are taken from the MGT cost functions presented previously by Tilocca et al. for a 400kW engine [9]. The Gaussian standard deviation distribution is equal to 10% of the mean value. For the MGT-ICR, equipment cost is the sum of the OEM ex-works price plus the fuel compression cost function for the NG engine, according to Tilocca et al.’s study. O&M cost is assumed to be the same as for MGT-R. Reciprocating Internal Combustion Engines ICEs are the most widespread technology for distributed (thermal) power generation, cogeneration and biofuel applications [40]. Their superiority relies on their high versatility, flexibility, availability, costeffectiveness and efficiency. The mean value and standard deviation of the performance of ICEs working at rated conditions derive from a normal regression of the Natural Gas engines used in the study by Tilocca et al. [9], for units with rated electric output between 250 kWeand 550 kWe. According to recent assessments, the performance of ICEs running on hydrogen should not experience relevant deviations from that of conventional fuels [41]. The normalised part-load efficiency curve is taken from literature, both for the mean value and standard deviation [42, 43], which is deemed representative of a broad spectrum of ICEs. ICEs already utilise after-treatment of the exhaust gases to meet the stringent NOx limits set by regulations. As a result, their mean NOx emissions are equivalent to the threshold set by the European Commission’s Stage V. In this case, the uncertainty applies to the current post-exhaust treatment costs [44]. The service interval of reciprocating engines is set to 2000 h, following the maintenance schedules of most OEMs. The upper and lower bounds of MPMT are estimated at 47.8 h and 17.8 h, respectively. The author’s assessments lead to a log-normal distribution such that the cumulative probability of MPMT values being lower than 17.8 h is 5%, and that of being greater than 47.8 h is 85%. The uncertainties applicable to failure rate and MT T R are quantified according to NREL’s measurements for well-maintained engines [45]. Equipment and O&M costs are from the cost function of ICEs developed by Tilocca et al. for a 400kW engine [9], with a 10% standard deviation. Fuel Cells FCs generate electric power through electrochemical reactions; thus, the Second Law of Thermodynamics does not limit their performance [46]. Fuel cells consume fuel -typically hydrogen- and an oxidiser, with water being the end product. FCs are a clean technology and do not rely on combustion [47]. A study from Napoli et al. [48] demonstrated that PEMFC is the most suitable FC technology for load-following operations, thanks to their quick-response capability and steep ramp rates. For this reason, this work considers PEMFC only. For FCs, the total cost of ownership is -likely- the most decisive system constraint. Yet, it presents a very high uncertainty due to the low availability of price data for stationary applications of PEMs and to the erratic price trends owing to a fast-evolving market and supply chain. The capital cost is taken from the European Scenario for stationary FCs of Hydrogen Europe for large PEMFC [49]. The reference values are those provided for State-of-Art (1900e/kWe) and the 2030 prediction (900e/kW), values that do not include power conversion systems and margins. The cost of power electronics is set according to the FC manufacturing Cost Analysis of the Department of Energy [50], whose maximum and minimum values based on production volumes are taken as worse and best-case scenarios, respectively. The gross sales margin is set to 30%. The logarithmic distribution used to quantify the high uncertainty is defined such that the cumulative probability of the equipment cost being lower than the minimum value (1569e/kW) is 5%, that of being lower than the maximum value (3259e/kW) is 85%. Uncertainty Quantification for the O&M cost2, electrical efficiency3and degradation profile are defined similarly, taking the State of the Art and 2030 values from Hydrogen Europe’s report [49]. The standard deviation function for part-load performance is a polynomial regression of the standard deviation from the normalised part-load efficiency reported 2O&M costs include the potential stack replacement. 3DC/AC Conversion efficiency is assumed to be 93%. 8 GTP-23-1412 - Tilocca Accepted Manuscript Not Copyedited Journal of Engineering for Gas Turbines and Power. Received August 07, 2023; Accepted manuscript posted November 18, 2023. doi:10.1115/1.4064061 Copyright (c) 2023 by ASME Downloaded from http://asmedigitalcollection.asme.org/gasturbinespower/article-pdf/doi/10.1115/1.4064061/7060164/gtp-23-1412.pdf by Universidad De Sevilla user on 06 December 2023 TABLE 3: Uncertainty distribution of input parameters. The Probability Density Functions (PDF) are marked as D(deterministic), N(normal) and LOG (log-normal). The PDFs that use variable mean and Standard Deviation (SD) functions are marked with V. For the LOG category, the reported mean and standard deviation are those of the logarithm of the normal distribution; the corresponding mean is computed as: MEAN =exp(MEANLOG +SD2). Product Parameter Units PDF Mean SD Ref. MGT ICR ηel(Rated)[-] N 0.4056 0.005 [36] ηel(PartLoad)[-] N-V - - [36] NOx(Rated)g/kWh N 0.99 0.0382 [37] MTBF h LOG 9.267 0.324 [39] MTTR h LOG 2.47 1.59 [-] SI h D 1904 - [36] MPMT h LOG -0.130 0.171 [36] CEquipment e/kW N 1148 115 [36] CO&Me/MWh N 17.3 1.73 [9] MGT R ηel(Rated)[-] N 0.3336 0.005 [34] ηel(PartLoad)[-] N-V - - [34] NOx(Rated)g/kWh N 0.99 0.0191 [37] MTBF h LOG 9.267 0.324 [39] MTTR h LOG 2.47 1.59 [-] SI h D 4000 - [34] MPMT h LOG 0.513 0.162 [34] CEquipment e/kW N 1448 145 [9] CO&Me/MWh N 17.3 1.73 [9] ICE ηel(Rated)[-] N 0.4027 0.022 [9] ηel(PartLoad)[-] N-V - - [42, 43] NOx g/kWh N 0.607 0.051 [-] MTBF h LOG 7.389 0.312 [45] MTTR h LOG 2.47 1.59 [45] SI h D 2000 - [9] MPMT h LOG 3.53 0.396 [-] CEquipment e/kW N 680 68 [9] CExhaust e/kW LOG 3.29 0.274 [44] CO&Me/MWh N 26.0 2.60 [9] FC ηel(Rated)[-] LOG -0.708 0.0548 [49] ηel(PartLoad)[-] N-V - - [51] dηel/dt [%/1000h] LOG -5.79 0.259 [49] UA [-] LOG -4.10 0.285 [51] CEquipment e/kW LOG 7.87 0.316 [49, 50] CO&Me/kWh LOG -3.35 0.342 [49] General CInstallation e/kW N 830 166 [9] by NREL’s statistical measurements for stationary FCs above 100 kWe[51]. The uncertainty of Reliability, Availability, Maintainability and Safety (RAMS) is quantifiable from the same NREL’s statistical assessment measuring the operational availability of stationary FCs above 100 kWe. RESULTS Discussion of results The convergence study highlighted in Fig. 4 shows that 1000 evaluations are enough to achieve an accuracy of 10−3for MGT-R, MGT-ICR and ICEs, whereas FCs need 10000 evaluations to reach a 10−3precision. Within the scope of the analysis in this work, the 1000 seed’s accuracy is deemed sufficient. 9 GTP-23-1412 - Tilocca Accepted Manuscript Not Copyedited Journal of Engineering for Gas Turbines and Power. Received August 07, 2023; Accepted manuscript posted November 18, 2023. doi:10.1115/1.4064061 Copyright (c) 2023 by ASME Downloaded from http://asmedigitalcollection.asme.org/gasturbinespower/article-pdf/doi/10.1115/1.4064061/7060164/gtp-23-1412.pdf by Universidad De Sevilla user on 06 December 2023