Innovative desalination system driven by a solar micro gas turbine for off-grid applications

Innovative Desalination System Driven by a Solar Micro Gas Turbine for Off-Grid Applications Rafael Gonz´ alez-Almenara, Lourdes Garc´ ıa-Rodr´ ıguez, Antonio Mu˜ noz, Tom´ as S´ anchez, David S´ anchez∗ Department of Energy Engineering, University of Seville, Camino de los descubrimientos s/n, 41092 Seville, Spain. Abstract Past work by the authors has suggested that Solar micro Gas Turbines (SmGTs) can be used cost-effectively to produce electric power and heat for freshwater production through desalination, mainly in off-grid locations. This is further studied in this work, providing an exhaustive modelling procedure and methodology. Additionally, a comprehensive analysis of system performance under both design and part-load conditions is provided along with the complete set of characteristics of the components of the SmGT and the desalination unit. To this end, the SmGT is assessed first, incorporating innovative solutions that enable better off-design performance: Variable Inlet Guide Vanes (VIGVs) in the compressor and Variable Nozzle Guide Vanes (VNGVs) in the turbine. The implementation of these features brings significant advantages in part-load operation, leading to a considerable increase in solar-to-electric efficiency and rangeability, particularly under certain ambient conditions; in particular, using these features yields 5% higher efficiency at the minimum load of the original engine (without variable geometry) and, from this load setting downwards, 15% lower minimum stable load is enabled. Furthermore, a sensitivity analysis to control strategies and ambient conditions has been included, exploring their impact on system performance. The water treatment system is comprised of two elements. A Reverse Osmosis (RO) desalination unit is driven by the electric power produced by the SmGT. Considering the emphasis on flexibility and part-load operation enhancement addressed in this work, a moderate recovery rate is proposed for this unit. The RO unit produces brine with high salt concentration to be partially treated further in a Zero Liquid Discharge (ZLD) unit driven by the exhaust gases of the microturbine (at about 250-300ºC), where the sensible heat of this stream is harvested by the ZLD unit to “dry” and concentrate the effluent. Finally, the potential and the operational limitations of the ZLD system are discussed, complemented with results from an experimental proof of concept where the feasibility of the concept has been verified. Keywords: SOLMIDEFF, Solar Micro Gas Turbine, Microturbine, Desalination, ZLD Nomenclature1 ∆zRotor Axial Length [m]2 ˙ C∗ iRatio of Heat Capacities [−]3 ˙mMass Flow Rate [kg/s]4 ηEfficiency [−]5 νKinematic Viscosity [m2/s]6 ΩTotal Pressure Loss Coefficient [−]7 ∗Corresponding author. Tel.: +34 954 487 241. Email addresses: [email protected] (Rafael Gonz´ alez-Almenara), [email protected] (Lourdes Garc´ ıa-Rodr´ ıguez ), [email protected] (Antonio Mu˜ noz), [email protected] (Tom´ as S´ anchez), [email protected] (David S´ anchez ) Preprint submitted to Applied Thermal Engineering July 31, 2023 ωλAbrupt Expansion Losses [−]8 ωBL Blades Losses [−]9 ωCH Aerodynamic Blockage Losses [−]10 ωcl Clearance Losses [−]11 ωcr Supercritical Mach Number Losses [−]12 ωdi f Entrance Diffusion Losses [−]13 ωHS Hub-to-Shroud Loading Losses [−]14 ωinc Incidence Losses [−]15 ωmix Mixing Losses [−]16 ωs f Skin Friction Losses [−]17 φFlow Coefficient [−]18 ΠOsmotic Pressure [bar]19 ψWork Coefficient [−]20 εEffectiveness [−]21 bin Rotor inlet passage width [m]22 CdDischarge Coefficient [−]23 dSolar Receiver Nozzle Diameter [m]24 Dcav Cavity Diameter [m]25 FFeed [−]26 IWork Input Coefficient [−]27 Ibl Blade Loading Losses [−]28 Id f Disk Friction Losses [−]29 Ilea f Leakage Losses [−]30 Irecir Recirculation Losses [−]31 jSolar Receiver Nozzle rows [−]32 kw f Thermal Conductivity of the Working Fluid [W/(m·K)]33 mbi Brine Inlet Mass Flow Rate [kg/s]34 mbo Brine Outlet Mass Flow Rate [kg/s]35 mgi Gases Inlet Mass Flow Rate [kg/s]36 NTotal Number of sub-Heat Exchangers [−]37 nTotal number of Solar Receiver nozzles [−]38 nsSpecific Speed [−]39 2 PPermeate [−]40 p0tTotal Pressure [−]41 Pjet Jet Pitch [m]42 qvVolumetric Flow Rate [m3/s]43 rVRecovery Rate [−]44 SSalinity [kg/kg]45 Sbi Brine Inlet Salinity [kg/kg]46 Sbo Brine Outlet Salinity [kg/kg]47 Tbi Brine Inlet Temperature [K]48 Tbo Brine Outlet Temperature [K]49 Tgi Gases Inlet Temperature [K]50 UGlobal Heat Exchange Coefficient [W/(m2·K)]51 vsTotal-to-Total Velocity Ratio [−]52 YBL Blade Loading Losses [−]53 Ycl Clearance Losses [−]54 YHS Hub-to-Shroud Losses [−]55 Yinc Incidence Losses [−]56 YpProfile Losses [−]57 YQMoisture Losses [−]58 Yrot Rotor Loss Coefficient [−]59 BB Billion Barrels [−]60 BCF Billion Cubic Feet [−]61 BD BlowDown [−]62 BP Booster Pump [−]63 CHP Combined Heat and Power [−]64 CSP Concentrated Solar Power [−]65 DNI Direct Normal Irradiance [W/m2]66 ERD Energy Recovery Device [−]67 GDP Gross domestic product [−]68 HDH Humidification-DeHumidification [−]69 HPP High Pressure Pump [−]70 ICE Internal Combustion Engine [−]71 3 ICH Isobaric Chamber [−]72 MD Membrane Distillation [−]73 MED Multi-Effect Distillation [−]74 mGT Micro Gas Turbine [−]75 MSF Multi-Stage Flash [−]76 ORC Organic Rankine Cycle [−]77 PCM Phase Change Material [−]78 PR Pressure Ratio, Performance Ratio [−]79 Pr Prandtl Number [−]80 PSHE Primary Surface Heat Exchanger [−]81 PTC Parabolic-Trough Collector [−]82 PV PhotoVoltaic [−]83 RO Reverse Osmosis [−]84 SmGT Solar Micro Gas Turbine [−]85 SSC Stationary Solar Collector [−]86 SWRO SeaWater Reverse Osmosis [−]87 TDS Total Dissolved Solids [ppm]88 TIT Turbine Inlet Temperature [K]89 TOT Turbine Outlet Temperature [K]90 TVC Thermal Vapour Compression [−]91 UA Conductance [W/K]92 VIGV Variable Inlet Guide Vanes [−]93 VNGV Variable Nozzle Guide Vanes [−]94 ZLD Zero Liquid Discharge [−]95 1. Introduction96 1.1. Background97 This paper presents the technical details of an innovative integrated system aimed at enabling a large reduction98 of the carbon footprint of water desalination in areas currently subjected to extreme water scarcity, in particular the99 Middle East and Northern Africa. In these regions, the characteristic boundary conditions include a very high available100 solar resource and also large regions where the grid is not available. Therefore, the conditions to integrate solar power101 generation and water desalination are met, in an attempt to tackle not only water scarcity but also a heavy reliance on102 fossil fuel for electricity supply.103 The countries in the Middle East have almost 50% of the world reserves of crude oil and they can be categorised104 in two main groups. On the one hand, certain countries in the Middle-East with very high state and individual105 4 wealth (gross GDP and GDP per capita), mostly affiliated with the Gulf Cooperation Council (Saudi Arabia, Kuwait,106 Qatar,United Arab Emirates and, to a lesser extent, Oman and Bahrain) and accounting for over 30% of the world107 reserves of crude oil; these countries have seen a remarkable economic growth, as a consequence of which the energy108 consumption per capita has rocketed not only for the increased demand but also for certain policies discouraging an109 efficient use of energy (for instance, through subsidised energy prices). On the other hand, certain countries also in110 the Middle East (Iran, Irak) and in Northern Africa (Libya, Algeria) which have not seen a similar improvement of the111 individual wealth despite their large share of oil reserves (15-20%). When natural gas and not oil is considered, the112 situation is somewhat similar even if the numbers add differently. Overall, regardless of the different characteristics of113 these two groups of countries, they share a strong dependence on fossil fuels to cover the power needs (>95% of the114 electricity comes from natural gas or coal, 100% in several cases). All this information, collected from [1] and [2], is115 shown in Table 1.116 Country Consumption Fossil Fuels Water Oil GDP Natural Gas of Electricity Energy mix Availability Reserves per capita Reserves [kWh/person·year] [%] [m3/person·year] [BB] [%] [USD2010] [BCF] [%] Algeria 1363 99 272.3 12.2 0.69 4710.6 159054 2.3 Bahrain 19597 99.7 2.679 0.09 0.01 20913 3250 0.05 Egypt 1683 90 10.25 3.3 0.19 3008.8 77200 1.1 Iran 3022 92.7 1583 155.6 8.74 5627 1201382 17.3 Iraq 1328 95.5 919.7 145.02 8.15 4837.4 111522 1.6 Kuwait 15591 100 0 101.5 5.70 32697.3 63500 0.9 Libya 1811 100 109.8 48.36 2.72 8122.2 58183 0.8 Morocco 904 84.6 811.4 0 0.00 3396.1 51 0 Oman 6446 100 302 5.37 0.30 14992.6 24910 0.4 Qatar 14792 100 21.22 25.24 1.42 62021.1 871585 12.5 Saud Arabia 9401 100 72.86 267.03 15.00 20542.2 294205 4.2 UAE 11088 100 15.96 97.8 5.50 41420.5 215098 3.1 Yemen 220 98 74.34 3 0.17 0 16900 0.2 Table 1: Energy, economic and water related indicators of selected countries in the MENA region. Oil and Natural Gas reserves are given in billion barrels [BB] and billion cubic feet [BCF] (2019 data from [1] and [2]). The aforedescribed abundance of fossil fuels is in contrast with a low availability of water resources, in most cases117 well below the water poverty or absolute water poverty lines, 1000 and 500 m3/(person·year) respectively, below118 which water becomes a primary constraint for human life. In some extreme cases, water availability is even below119 100 m3/(person·year), a limit that poses a very large and pressing threat for the sustainability of future generations.120 This is also shown in Table 1.121 In the light of the situation above, it is now agreed that this environmental emergency must be tackled simul-122 taneously for water and energy, exploiting the evident synergies between these two human needs and the abundant123 solar resource in the region. The current landscape for large-scale water technologies in this part of the world is char-124 acterised by large-capacity thermal desalination plants using technologies based on Multi-Effect Distillation (MED)125 or Multi-Stage Flash (MSF) integrated into thermal power plants, installed in the 20th century for the most part but126 already obsolete in present times as a result of the superior performance of Reverse Osmosis (RO). At the smaller-127 scale and for off-grid installations, RO powered by photovoltaic panels seems to be the preferred option despite its128 disadvantages, since these systems typically require: i) backup generators burning fossil fuels to enable operation in129 the absence of solar energy, and ii) batteries to run the system overnight if needed. Such redundancy is, of course, less130 reliable and has higher capital and O&M costs than a single, flexible system.131 This work introduces a technical solution which can possible tackle the flaws of PV-RO systems mentioned in the132 foregoing paragraph, producing freshwater with a single system able to work on solar energy and on backup fossil133 fuels (if needed) whilst also reducing the liquid effluent from the water production process (i.e., lower environmental134 impact). This system is based on the coupling of a solar dish collector and a micro gas turbine, downstream of135 which water is produced in an RO unit driven by the electric power produced by the gas turbine system. A Zero136 Liquid Discharge (ZLD) system running on the available thermal energy of the exhaust of the gas turbine evaporates137 water from the concentrate, hence ensuring lower water content in the latter and, accordingly, lower environmental138 5 impact. Although no commercial systems have been installed so far, the techno-economic viability of Solar micro139 Gas Turbines (SmGT) has been addressed in literature [3], finding better-than-expected results in all aspects [4].140 In addition, these theoretical results are supported experimentally by a demonstrator installed at ENEA Research141 Centre (Italy) within the framework of the OMSoP project [5], funded by the European Commission. This system142 demonstrated the technical feasibility of the coupling of a solar dish collector and a micro gas turbine.143 1.2. Scope of work and objectives144 Micro Gas Turbines (mGT) are compact engines for on-site power and heat generation able to achieve electrical145 efficiencies between 15% (small engines of ∼3 kWe [6]) and 40% (larger engines with unit output of ∼400 kWe [7]) at146 full load and an overall energy utilisation of more than 90% in Combined Heat and Power (CHP) configuration [6]. A147 mGT includes the following components: compressor, turbine/expander, combustor, recuperator, fuel supply system,148 bearings (most commonly with an associated lube oil system), electric generator, power conditioning and control unit,149 enclosure and balance-of-plant. Some of these components might not be present in a particular mGT, as this depends150 on the type of application. This is the case for solar micro gas turbines where a solar receiver replaces the combustor.151 Although the market for decentralised, micro energy systems is currently dominated by reciprocating Internal152 Combustion Engines (ICE), micro gas turbines are reportedly competitive against ICEs in applications requiring153 intermediate to high-grade heat and high heat-to-power ratio [8]. The micro gas turbines in these applications exhibit154 some variations with respect to large gas turbines, the utilisation of radial machinery being the most relevant feature.155 In micro gas turbines, single-stage radial compressors are standard and this limits the pressure ratio of the engine to156 values not higher than 4:1 (approximately). On the other hand, the utilisation of radial expanders prevents the use of157 blade cooling and this, in turn, limits Turbine Inlet Temperature (TIT) to values lower than 950ºC. Both pressure ratio158 and turbine inlet temperature are significantly lower than in large gas turbines, what brings about a dramatic efficiency159 drop if simple Brayton cycles are used. Accordingly, micro gas turbines typically adopt recuperative cycle layouts in160 order to increase thermal efficiency to more competitive values [9].161 Further to the aforementioned potential of micro gas turbines for decentralised power, particularly applied to162 renewable energy sources, and based on the authors’ past experience, a solar desalination system based on SmGTs163 with the techno-economic features already presented by the authors [10] seems to be a cost-effective solution with164 a promising future compared to other popular renewable desalination options that integrate PV panels [11]. This165 has already been explored for isolated units [12] but, according to recent literature on the topic, it might also apply166 to larger scales by coupling several SmGTs in parallel, or even to significantly larger scales using different power167 generation technologies as proposed in [13].168 The main features of the solar desalination arrangement proposed are:169 •The high temperature available at the turbine exhaust has the potential to drive thermal desalination processes170 based on phase change. In this case, the permeate of an upstream RO process could be blended with the distillate171 produced by distillation in order to increase the quality of the freshwater produced. Another option would be to172 direct the concentrate of the RO unit to a thermal ZLD system were the total water content of the effluent could173 be reduced, thereby minimising the environmental impact of water desalination.174 •Biofuels can be used as a back-up energy source to maintain the system operating at full capacity 24/7 and to175 compensate for cloudy/hazy sky conditions. This avoids the usual oversizing of desalination systems typically176 required in solar applications in order to achieve reasonable capacity factors. Moreover, this oversizing brings177 about higher costs that challenge the economics of the systems and stem as the main economic hurdle for the178 development of solar desalination technologies.179 •In parabolic dish systems, the integrated concentrator-receiver system can reach efficiencies as high as 80%180 and, furthermore, a short-term thermal storage based on Phase Change Material (PCM) can be considered to181 enhance the performance of the assembly [14]. Then, the complete assembly with a mGT can potentially attain182 annual solar-to-electric efficiencies of 18.3%, when turbine inlet temperature is in the order of 850ºC. [15].183 This is very close to the rated efficiency of approx 20%, or even more if a higher turbine inlet temperature is184 considered [16], which is similar to the efficiency of conventional, larger CSP plants. This ensures interesting185 market prospects of the technology when fully developed [17].186 6 In order to justify the interest of this work, Fig. 1 presents a comparison of the expected efficiency achieved187 by different solar desalination technologies in terms of estimated solar energy required [kWh] to produce 1 m3of188 freshwater. The following technologies are considered:189 •Five collector technologies: i) Stationary Solar Collectors (SSC), ii) Parabolic Trough Collectors (PTC), iii)190 heliostats, iv) Dish collectors and v) PV panels.191 •Two thermo-mechanical energy conversion technologies: Organic Rankine Cycle systems (ORC) and micro192 Gas Turbines (mGT).193 •Four desalination technologies based on phase-change: i) Membrane Distillation (MD), ii) Multi-Effect Distilla-194 tion (MED), iii) MED with Thermal Vapour Compression (MED-TVC) and iv) Humidification-DeHumidification195 (HDH) with values of Performance Ratios (PR) taken from [18] and [19], where a deep review of recent bibli-196 ography on the topic is undertaken.197 •Desalination based on membrane technology: SWRO desalination with a specific energy consumption of 2.2198 kWh/m3(as reported in recent literature on the topic [20]).199 From the comparison in Fig. 1, the interest of the proposed solution in terms of specific energy consumption200 becomes evident. Therefore, a more detailed, in depth analysis is needed to verify if the performance reported in the201 chart is actually accurate.202 Figure 1: Comparison of solar desalination based on reverse osmosis and phase change desalination. Estimates of energy efficiency of solar energy technologies are given in brackets along with that of power cycles [20]. In order to clarify the novelty of this work, a comparison with previous works in literature is presented, highlight-203 ing the progress made and the original contributions in it. Whilst the techno-economic assessment of a SmGT has204 been addressed before [4], to the author’s best knowledge, a study combining a SmGT with innovative water treat-205 ment systems like the one proposed in this work is not available in literature. Moreover, despite the existing literature206 discussing the interest of SmGT for renewable power generation [21], the innovative solutions to enhance off-design207 performance of these systems have not been assessed in this application so far: variable inlet guide vanes (VIGVs)208 at compressor inlet and variable nozzle guide vanes (VNGVs) at turbine inlet. Finally, regarding the water treatment209 system, a novel concept of thermal desalination based on ZLD technology is developed in this work, evaluating a210 proof of concept and achieving favourable results. This utilisation of waste heat from the turbine is also novel.211 The paper is structured as follows. Section 2 (Methodology) provides an overview of the system, outlining the212 general model. In Section 3 (Modelling), the authors elaborate on the modelling procedures, supported by relevant213 literature, and present the performance of each subsystem. Section 4 (Validation) describes the model validation214 process in detail. Section 5 (Results) addresses both design conditions and part-load performance, providing results215 for water and energy production. Finally, in Section 6 (Conclusions), the findings are summarised and future work is216 proposed.217 7 2. Methodology218 The system integrating the solar mGT and water treatment systems proposed, namely reverse osmosis desalina-219 tion and ZLD, has been developed conceptually in the SOLMIDEFF project. This project, entitled SOLar Micro gas220 turbine-driven Desalination for Environmental oFF-grid applications, was developed by the research team at Univer-221 sity of Seville, Spain, with funding from the National R&D programme of the Ministry of Science and Innovation.222 The reference site for the techno-economic assessment of the technology is in the Canary Islands.223 The SOLMIDEFF system is based on a 10 kWe SmGT producing electric power and heat to run a desalination unit224 in off-grid applications. The desalination unit is comprised of a SWRO unit, driven by the electric power produced by225 the engine, and the Zero Liquid Discharge making use of the available gas in the exhaust from the SmGT. A schematic226 of the SOLMIDEFF system is presented in Fig. 2, where the three main subsystems (SmGT, SWRO and ZLD) are227 easily identified. A parabolic dish collects and concentrates solar energy onto a solar receiver where a large fraction of228 this concentrated solar power is absorbed by the working fluid of an open-cycle recuperated mGT, achieving a solar-229 to-electric efficiency of almost 20%. In this micro gas turbine, atmospheric air is compressed to 3.5 bar by a radial230 compressor and then heated up to 500ºC by a recuperative heat exchanger. The energy source for this temperature231 rise is the stream of hot gases flowing out from the turbine. After pressurisation and preheating, air flows into a solar232 receiver where it achieves a turbine inlet temperature of 850ºC thanks to the concentrated solar energy absorbed. If233 this were not possible, because of a lack of solar energy supply (hazy sky or, maybe, operation at night), a combustor234 in series or parallel with the receiver would be able to burn fuel in order to ensure power and water supply. Regardless235 of the energy source used to take the hot air to the target turbine inlet temperature, the hot pressurised gas is then236 expanded across the turbine which, in turns, produces power to drive the compressor and electric generator.237 Figure 2: Schematic of the SmGT-driven desalination system (SOLMIDEFF system) The electric power produced by the SmGT is used by a Reverse Osmosis plant, whose feed stream is seawater,238 brackish water or industrial wastewater. The particular arrangement of pumps and pressure exchange devices is shown239 in Fig. 2. The feed stream is split into two flows. One of them is sent to the high-pressure pump (HPP) where pressure240 is increased to 55 bar approximately. The remainder of the feed flow is sent to an isobaric chamber (ICH) where241 pressure is also increased thanks to the high pressure available at the outlet concentrate stream of the RO unit. In242 an energy recovery device (ERD) such as an ICH, volumetric flow rates on both sides are similar and the feed outlet243 pressure (stream 14) is slightly lower than the concentrate inlet pressure (stream 18) [22]. This pressure loss comes244 about because the ERD is not ideal and there are energy losses preventing a perfect match of the energy exchange245 across the interacting streams; accordingly, a booster pump is needed downstream of the ERD (feed flow side) to246 make up for the pressure difference between the outlet from the ERD and the outlet from the HPP (inlet to the RO247 unit), as well as for the pressure drop through the membrane rack.248 8 The concentrate flows across the energy recovery device, where pressure decreases to almost atmospheric. A249 fraction of this stream is then sent to the ZLD unit, where part of the water content evaporates thanks to the thermal250 energy available in the exhaust gases of the micro gas turbine. The ZLD unit is a direct-contact heat exchanger whose251 purpose is to reduce the amount of water released for disposal [12].252 The methodology followed to model the design and performance of the SOLMIDEFF system is based on an inte-253 grated design and simulation platform which combines the design and off-design performance of each SOLMIDEFF254 component. The individual models have been implemented in Matlab environment [23] and are solved through a mod-255 ular approach which relies on the conservation of mass, energy and momentum and on set correlations to estimate256 energy losses and irreversibility of the individual components.257 The working fluid of the SmGT is air (dry real gas) whose thermodynamic properties are obtained from Coolprop258 [24]. Once the simple recuperated Brayton cycle is resolved for a given set of boundary conditions (i.e., the pressure259 ratio yielding maximum net efficiency is calculated), one dimensional (1-D) models are used to produce efficient260 and flexible turbomachinery designs (compressor and turbine) whilst non-dimensional (0-D) approaches are used to261 design the recuperator and solar receiver. The final step to completely model the SmGT is to evaluate the off-design262 performance of the engine; this is driven by compressor-turbine matching, which is implemented by a proprietary263 code making use of the methodology described by Razak [25]. In addition, several control strategies are embedded to264 ensure the safe operation of the SmGT in different load ranges and operating conditions.265 The performance of the RO system for the treatment of industrial effluents and brackish or seawater has been mod-266 elled through well-known equations from literature [26], compiled in a block model calibrated against the software267 Q+developed by LG Chem [27]. finally, still within the water treatment subsystem, the proof-of-concept ZLD unit268 has been experimentally demonstrated for this application, and the associated model has already been published by269 the authors [12].270 3. System modelling271 This section introduces the main features of the models used to characterise each main component of the SOLMID-272 EFF system. This includes the designs models as well as the models used to estimate off-design performance. The273 following subsystems are included: i) solar Micro Gas Turbine (compressor, solar subsystem, turbine/expander and274 recuperator), ii) SWRO system, and iii) ZLD system. Comments on the validation or verification of each of these275 models are also provided.276 3.1. Solar Micro Gas Turbine277 3.1.1. Compressor278 The design and performance model of the centrifugal compressor is of the one-dimensional, mean-line type and279 it is based on Aungier´s works [28]. The design process begins from the work (ψ) and flow (φ) coefficients, target280 pressure ratio and mass flow rate, along with the specific ambient conditions of the reference locations. This standard281 approach, widely used in literature, is based on a stepwise methodology where the compressor elements are designed282 independently through non-dimensional coefficients (pressure losses and work input) along the meridional coordinate:283 impeller, diffuser and volute. Since the outlet conditions from the machine are required to estimate these coefficients,284 this methodology renders an iterative process whose design flowcharts are shown in Figs. 3-4. In the latter flowchart,285 blade loading (BL) is defined by Eq. 1.286 BL =2∆W W1+W2 (1) where ∆Wrepresents the average blade velocity difference, used to estimate the number of blades of the impeller,287 and W1and W2is the relative flow velocity at impeller inlet and outlet.288 Single stage compressors are considered by default, given the low pressure ratio of this type of engines (owing289 to the recuperative cycle layout), most often with full-inducer impellers. The blades are three-dimensional, based on290 straight-line surface elements that connect hub and shroud. The flow at the inlet is assumed axial with uniform flow291 distribution, with the variable inlet guide vanes (VIGV) set in the axial direction at the design point1. The meridional292 1The utilisation of VIGVs is not mandatory. This is a case-specific design choice for enhanced flexibility. 9 Figure 10: Normalised performance map of the turbine. Figure 11: Schematic of Reverse Osmosis system, including an Energy Recovery Device (ERD). It is further assumed that the outlet flows of freshwater and brine (ppand pBD,ERDout) are at atmospheric pressure,431 pE.432 The Specific Energy Consumption (SEC) of the process shown in Fig. 11 depends on the electric power consump-433 tion of the high-pressure and booster pumps (HPP and BP) and can be calculated through the corresponding flow434 rates, qv, pressure, p, and pumps and engines efficiencies, η, as shown in Eq. (19):435 S EC =PPW,HPP +PPW,BP qV,P =qV,HPP qV,P ·pF,HPP −pF ηHPP ·ηeng,HPP +qV,ERD qV,P ·pF,HPP −pF,ERDout ηBPP ·ηeng,BP (19) Then, the energy recovery process taking place in the isobaric chamber is characterised by Eqs. (20-21), whereas436 the ratio of volumetric flow rates through each stream is defined by Eqs. (22-24):437 ηICH =qV,ERD ·pF,ERDout +qV,BD ·pBD,ERDout qV,ERD ·pF,ERDin +qV,BD ·pBD,ERDin (20) εICH =pBD,ERDout −ηICH ·pBD,ERDin ηICH ·pF,ERDin −pF,ERDout (21) qV,ERD =qV,BD ·εICH (22) qV,ERD qV,P = 1 rV −1!·εICH (23) qV,HPP qV,P =1 rV − 1 rV −1!·εICH (24) 16 In this set of equations, rvis the recovery rate of the process, defined as the ratio from permeate flow rate, qV,P, to438 feedwater flow rate, qV,F.439 Figure 12 depicts an exemplary off-design performance representative of a seawater RO system working in Atlantic440 conditions, with water temperature of 22ºC. The operating pressure is set to 60 bar throughout the whole operating441 range, what enables increasing recovery rates in partial load operation. The main specific energy consumption remains442 at around 2 kWh/m3in all cases.443 To assess the performance of the SWRO unit, the aforementioned model has been compared against the estimates444 provided by the software Q+developed by LG Chem [27], widely used by the industry, and considering a single445 pressure vessel with seven LGSW440GR membrane modules [56]. It is to note that this membrane type has been446 selected for its ability to operate with higher rejection rates, what implies that a more reliable operation within a wide447 range of power input to the SWRO unit can be achieved in all climate conditions, thus adapting to the variable electric448 power produced by the SmGT unit.449 Regarding sensitivity to boundary conditions, it is worth noting that both specific energy consumption and operat-450 ing pressure increase with the salinity of seawater whereas the operation with higher feedwater temperatures reduces451 the power input range enabling an adequate quality of the permeate.452 Figure 12: Off-design performance of the SWRO unit. 3.3. Zero Liquid Discharge453 3.3.1. Experimental proof of concept of the ZLD system454 Within the SOMIDEFF project, a proof-of-concept Zero-Liquid-Discharge system has also been developed. The455 concept is simple and relies on bubbling the hot exhaust gases from the SmGT into a well of brine through a specific456 diffuser design, as shown in Fig. 13. Operation of this system has been demonstrated, integrated into a Cussons P9001457 micro gas turbine engine (3 kWe) installed in the Laboratory of Heat Engines at University of Seville.458 A fast-acting pneumatic valve is used divert the host gases from the turbine exhaust to either the ZLD unit or the459 stack. The brine well and injection pipe are sized to minimise back pressure on the turbine, as this has a negative460 impact on the efficiency of the SmGT. The same applies to the diffuser, whose design stems from a compromise461 between the need to minimise back pressure whilst maximising the direct contact between gas and brine.462 The thermal energy carried by the gas stream is used to evaporate water in the brine, and this increases the463 concentration of salt in the well. At full load (hence peak mass flow rate of gases into the well), water drift becomes464 a concern. Hence, in order to minimise this, baffles are installed inside the well above the brine level: droplets carried465 over by the gaseous stream impinge on the baffles, returning to the well for evaporation.466 The performance of the prototype has been assessed for operation in batches and is illustrated in Fig. 14. Temper-467 ature measurements, logged with infrared imagery and thermocouples, indicate that the temperature of exhaust gases468 flowing across the ZLD drops from ∼450ºC to ∼70ºC. A picture of the ZLD in operation without the upper seal is469 also shown in Fig. 14, revealing the plume coming out from the well (high steam content), alongside a picture of the470 resulting solid waste (solid salt).471 17 Figure 13: Configuration of the low-cost ZLD prototype. a) Cross-sectional view of the ZLD unit. b) Schematic showing global arrangement. Figure 14: Graphical information of the proof-of-concept ZLD. 3.3.2. Performance modelling of the ZLD system472 As already discussed in earlier sections of the paper, Zero-Liquid-Discharge concepts have recently gained impor-473 tance, out of the growing desalination capacity worldwide and of the associated concerns about the impact that this474 highly-concentrated effluents have on the environment. The term brine is typically used to define the resulting by-475 product (waste) of desalination systems, whose salt concentration typically exceeds 55 mg/L [57]. Based on different476 technologies, ZLD systems tackle this problem by trying to increase the concentration of the brine further with the477 ideal objective to completely remove the water content of it; accordingly, the resulting waste would be in solid state478 and would therefore represent a much more manageable product, easier to store and transport and with a lower poten-479 tial to harm the environment. For these reasons, ZLD systems are increasingly being seen as mandatory requirements480 in some industrial processes [58].481 Indeed, reducing the amount of energy needed to minimise the discharge of liquid, highly-saline effluents is behind482 the growing interest in coupling membrane-based RO technology with ZLD systems, given the energy-intensive nature483 of this process [59]. In this regard, it is also to note that RO cannot be used for the further concentration of effluents484 18 given that these membranes can work with feedwaters whose salinity is not higher than 80 g/L, in order to avoid485 membrane fouling and scaling [60]. ZLD systems could include a membrane process before the brine concentration486 through a thermal treatment. Achieving a high-water recovery rate in these systems becomes challenging due to487 the early occurrence of membrane scaling and fouling. Consequently, brine pre-treatment is essential to safeguard488 the membrane and enable ZLD [61]. Otherwise, for brines with higher concentrations, a phase-change ZLD system489 becomes necessary to address the disposal of waste effectively [62].490 In the actual SOLMIDEFF system, the ZLD unit works very similarly to the proof-of-concept unit tested at the491 lab (Fig. 14): the concentrate resulting from the RO process is fed into an evaporation chamber, which is nothing but492 a direct contact heat exchanger where the thermal energy of the exhaust gases from the SmGT (at 225-325ºC) is used493 to evaporate water in the brine (therefore raising the concentration of this stream). Moreover, in order to enhance the494 performance of the ZLD unit, the brine flowing into the evaporation chamber is preheated in another chamber fed by495 the stack gases at a lower temperature of ∼60-100ºC (given that a large portion of their sensible heat has already been496 harvested in the evaporation process). This is shown schematically in Fig. 15.497 Figure 15: Conceptual representation of the Zero Liquid Discharge unit. The ZLD system is modelled through the methodology already presented and discussed by some of the authors[12].498 The ratios of mass flow mbo/mbi and salinity Sbo/Sbi are estimated for operating conditions different to the rated val-499 ues, noting that there is an operational limit of the inlet mass flow ratio (mgi/mbi). In particular, the mass flow rate500 of exhaust gases from the turbine must be, at least, 10 times higher than the mass flow rate of brine inasmuch as,501 otherwise, the stability and sustainability of the evaporation process would be compromised [63].502 Figures 16.a and 16.c, adapted from past work by the authors [12], show trends of the brine salinity ratio (Sbo 503 /Sbi), for values of inlet brine salinity of Sbi =0.07 kg/kg and Sbi =0.12 kg/kg, respectively. This ratio increases504 for higher temperatures of either the exhaust gases from the turbine (Tgi) or the inlet brine (Tbi), until saturation is505 reached, thus setting the upper limit for the concentration of brine. Pre-heating the brine to the lowest temperature (Tbi 506 =60ºC) enables reaching the saturation point if the inlet temperature of exhaust gases (Tgi) is 305ºC, and this latter507 temperature decreases to no more than 290ºC when the brine is pre-heated to 100ºC. The ratio of mass flow rates of508 the brine (mbo /mbi) is shown on the right charts, Figs. 16.b and 16.d. As either temperature increases (brine or gases),509 more water is evaporated and salinity becomes higher. Moreover, a global analysis of all charts shows that the mass510 flow ratio lines for different brine temperatures and concentrations change curvature and tend to collapse into a single511 line, corresponding with saturated brine conditions. This is an interesting result though it must be acknowledged that512 curves past saturation point are merely hypothetical.513 19 Figure 16: (a) Ratio between outlet and inlet brine salinity as a function of inlet temperatures of brine and exhaust gases from the turbine (into the evaporation chamber), for Sbi=0.07 kg/kg and mg/mb=10; (b) Ratio between outlet an inlet brine mass flow rates as a function of inlet temperatures of brine and exhaust gases from the turbine (into the evaporation chamber), for Sbi=0.07 kg/kg and mg/mb=10; (c) Ratio between outlet and inlet brine salinity as a function of inlet temperatures of brine and exhaust gases from the turbine (into the evaporation chamber), for Sbi=0.12 kg/kg and mg/mb=10; (d) Ratio between outlet an inlet brine mass flow rates as a function of inlet temperatures of brine and exhaust gases from the turbine (into the evaporation chamber), for Sbi=0.12 kg/kg and mg/mb=10. Adapted from Petit et al. [12] 4. Validation514 Given the lack of available SmGT experimental data in literature, it is not possible to validate the system presented515 in this work directly. Nevertheless, the accuracy of the model developed for each component of the system can be516 verified individually, either theoretically or experimentally, in order to ensure that the results obtained are trustful:517 •Turbomachinery: the turbomachinery modelling approach is widely known in literature and has been vali-518 dated extensively against a wide dataset of experimental results for both compressors [28] and turbines [55].519 Moreover, this approach has been employed in previous studies by the authors [16], upon which the present520 research is built.521 •Solar subsystem: the solar dish has been modelled according to the work carried out by Aichmayer et al.522 for the OMSoP project [37], whereas the methodology applied to model the impinging solar receiver follows523 the inverse design method proposed by Wang [40]. Both models were validated through experimentally at the524 Royal Institute of Technology in Stockholm (KTH) [39].525 •Recuperator: the recuperator is modelled following the well known method described by Nellis & Klein (2008)526 [49], whilst the conductance ratio method used to estimate the off-design performance of the recuperator was527 previously validated experimentally by the authors [50].528 •Reverse Osmosis: the validation of the design and off-design models of performance of the RO system is529 supported on the software Q+[27], whose accuracy is widely acknowledged within the industrial and scientific530 communities.531 20 •Zero Liquid Discharge: the reduction of the water content of the concentrate in the ZLD unit has been verified532 in the previous section, although further experimental validation for continuous operation is needed.533 5. Results534 The description provided so far in this paper depicts a tightly integrated system with a larger potential for overall535 energy utilisation than water desalination systems based on solar photovoltaic or wind power. Nevertheless, in order536 to verify and quantify this, it is mandatory to assess the performance of the system. This is presented now.537 The general boundary conditions considered in this assessment correspond to the Canary Islands, given their need538 for freshwater and high availability of solar energy; the applicability to neighbouring coastal locations in Northern539 Africa is an additional feature making this location interesting and representative of a larger area. In particular, ambient540 conditions have been generated with METEONORM v.7 [64], gathering records form the last 25 years and establishing541 95 percentile values, whereas seawater composition and temperature have been collected from the guidebook by Wilf542 and Awerbuch [65]:543 •Ambient temperature: 26.4 ºC.544 •Direct Normal Irradiance: 796 W/m2.545 •Water temperature: 22.0 ºC.546 •Total Dissolved Solids (TDS) of feedwater: 38739 ppm.547 5.1. Rated Performance548 The specifications of major equipment are set to the values listed in Table 2 for a preliminary assessment, according549 to the discussions presented in foregoing sections of the paper.550 Parameter Units Value Solar Energy Collection Calculated Dish aperture area [m2] 65.87 Solar energy input [kWt] 52.44 Collector efficiency [%] 89.87 Receiver efficiency [%] 81.87 Gas Turbine Calculated Isentropic compressor efficiency [%] 78.77 Isentropic turbine efficiency [%] 81.58 Receiver pressure drop [%] 1.37 Set Recuperator effectiveness [%] 85.0 Mechanical efficiency [%] 98.0 Generator efficiency [%] 98.0 Combustor pressure drop [%] 2.0 Recuperator hot pressure drop [%] 2.5 Recuperator cold pressure drop [%] 2.5 Air filter pressure drop [%] 2.0 Reverse Osmosis Set Isobaric chamber (ERD) efficiency [%] 97.0 High pressure pump efficiency [%] 87.0 Booster pump efficiency [%] 80.0 Table 2: Specifications of major equipment in SOLMIDEFF. The application of the design and performance models shown in earlier sections of the paper to the specifications551 in Table 2 yields the performance summarised in Table 3 for a system with the layout shown in Fig. 2. The engine552 net efficiency is around 26%, a value that is in line with the performance reported by the main Original Equipment553 21 Manufacturers of this type of engines [66]. The rated electric power is set to 10 kWe, which is mostly consumed by554 the pumps in the RO module -producing 4.68 m3/h of freshwater-; the remaining power is consumed by the auxiliary555 systems needed.556 Parameter Units Value Micro Gas Turbine Power output [kWe] 10.0 net efficiency [%] 26,04 Solar-electric efficiency [%] 19.15 Air mass flow rate [kg/s] 0.121 Pressure ratio [-] 3.5 Turbine inlet temperature [K] 1123.1 Exhaust temperature [K] 531.1 Shaft speed [rpm] 134,294 Reverse Osmosis Freshwater production [m3/h] 4.68 Recovery rate [%] 38.7 Operating pressure [bar] 59.6 Specific Energy Consumption [kWh/m3] 2.03 TDS of concentrate [ppm] 59,500 Pumping power [kWe] 9.5 Membrane area [m2] 286 Zero Liquid Discharge Mass flow rate [kg/s] 0.011 Salinity of outflow [kg/kg] 0.286 Table 3: Performance of SOLMIDEFF. Component specifications are reported in Table 2 Table 4 presents the heat and mass balance of the overall SOLMIDEFF system, supporting the results in 3. Streams557 1 to 9 correspond to the SmGT whereas the water treatment subsystem is defined by streams 10 to 21.558 The recovery rate of the RO module in SOLMIDEFF at design conditions is 39%, although this value increases559 as pump consumption decreases, sustaining the maximum pressure as described in the previous section. This recov-560 ery rate is equivalent to a Specific Energy Consumption of 2.03 kWh/m3, accounting for the RO module only (i.e.,561 excluding any water steam produced by the ZLD unit), which translates into a Solar Specific Energy Consumption of562 10.6 kWh/m3when the efficiency of the topping SmGT system is accounted for.563 Ambient temperature plays an important role in the turbomachinery design process since inlet air density is domi-564 nated by this parameer. A different ambient temperature implies that, for the same mass flow rate, a higher volumetric565 flow rate must flow through the SmGT, calling for larger scales or maynbe different design parameters (flow coef-566 ficient). Moreover, the pressure ratio that maximises net efficiency of the simple recuperated cycle also relies on567 ambient temperature, introducing a twofold influence on SmGT design. Table 5 lists some of the parameters influ-568 enced by ambient temperature. Data in the table refer to design values specific to the given ambient temperature, not569 off-design performance of a given design.570 5.2. OffDesign Performance571 This section is aimed at presenting the sizing process of the SOLMIDEFF system according to the environmental572 conditions of the site (Direct Normal Irradiance and ambient temperature) and, also, to assess how a particular system573 performs under different daily conditions in Gran Canaria (Canary Islands); to this latter end, an example of days574 with clear and hazy skies are considered. As a summary of some of these aspects, the performance map of this sizing575 process for every combination of DNI and ambient temperature is presented.576 Compressor-turbine matching in the solar micro gas turbine is adapted from Razak [25]. The original procedure is577 based on a simple, open-cycle industrial gas turbine and has been modified in this work to account for the additional578 pressure loss across the recuperator. The iterative process in the cited reference is also modified in order to have it579 controlled by the required turbine inlet temperature T IT and direct normal irradiance DNI, rather than shaft speed580 22 Micro Gas Turbine Stream p [bar] T [K] h [kJ/kg] 1 1.013 299.5 425.84 2 0.993 299.5 425.85 3 3.464 461.9 590.09 4 3.377 842.9 995.90 5 3.331 1123 1314.6 6 3.264 1123 1314.5 7 1.071 907.1 1067.3 8 1.045 531.1 661.52 9 1.013 531.1 661.52 Sea Water Reverse Osmosis Stream p [bar] ˙m[m3/h] TDS [ppm] 10 1.621 12.1 38,739 11 1.621 4.77 38,739 12 59.60 4.77 38,739 13 1.621 7.33 38,739 14 57.75 7.33 38,739 15 59.60 7.33 38,739 16 59.60 12.1 38,739 17 1.013 4.68 116.52 18 58.45 7.42 62,564 19 1.013 7.42 62,564 20 1.013 7.42 62,564 21 1.013 - 62,564 Table 4: Heat and mass balance of the reference case for the SOLMIDEFF system without ZLD operation. power output setting. Furthermore, for simplicity, inlet and exhaust losses are neglected (Pintake =Pexhaust) and the gas581 constant Rand isentropic index γare determined under the assumption of dry air. A graphical summary of changes582 from the original method is shown in Fig. 17.583 Engine operation relies on three main control strategies, aimed at preventing overheating and/or overspeeding of584 the engine under very high insolation. In particular, when direct normal irradiance DNI increase:585 •Shaft speed of turbomachinery can be increased, yielding higher throughput and pressure ratio. This yields586 lower turbine outlet temperature, therefore reducing the recuperative potential of the engine (i.e., lower temper-587 ature at receiver inlet). The cumulative effect of a higher mass flow rate and lower temperature at receiver inlet588 increases the amount of concentrated solar energy absorbed for the same turbine inlet temperature.589 •If this is not enough to manage the large solar input, turbine inlet temperature is allowed to rise to 25ºC higher590 than the design value. This strategy can nevertheless not be used for long periods of time (given the severe591 reduction of the useful life of the turbine).592 •Lastly, in case of continued operation under this undesired conditions (lower than 1% yearly as estimated by593 the METEONORM database [64]), the solar concentrator would have to be defocused. This is, unfortunately,594 not practical given that defocusing the concentrator implies increasing the spillage at the receiver window, thus595 compromising the mechanical integrity of this component. In other words, if defocusing were needed, the596 system would have to be shut down.597 On the other end of the operating range, reducing the heat input to the system implies either of the following598 alternatives: turbine inlet temperature is reduced for somewhat constant mass flow rate, or mass flow rate is reduced599 whilst keeping turbine inlet temperature at the rated value. In either case, the pressure ratio of the engine is reduced,600 hence lifting turbine exhaust temperature for constant turbine inlet temperature, and this poses the risk to overheat601 the recuperator. Therefore, a limit on turbine exhaust temperature is set to 650ºC, assuming that the recuperator602 is manufactured from Stainless Steel without specific treatments (e.g. ceramic coatings) [48]. The impact of this603 constraint on part-load efficiency is illustrated in Fig. 18.604 As discussed previously, the presence of VIGVs helps enhance the performance of the machine for operating605 conditions closer to the surge line, but the running line of a standard engine without Variable Nozzle Guide Vanes606 23 Parameter Units Value Ambient temperature [K] 293 313 Compressor &Turbine Mass Flow Rate [kg/s] 0.120 0.141 Pressure Ratio [-] 3.5 3.3 Shaft Speed [rpm] 135,594 121,722 Compressor Total-to-Total Efficiency [%] 78.75 78.34 Rotor Inlet Hub Radius [mm] 2.20 2.40 Rotor Inlet Shroud Radius [mm] 17.6 19.8 Rotor Outlet Radius [mm] 35.7 39.9 Turbine Total-to-Total Efficiency [%] 80.73 81.28 Rotor Inlet Radius [mm] 36.8 36.7 Rotor Outlet Shroud Radius [mm] 25.9 27.1 Rotor Outlet Hub Radius [mm] 3.0 3.1 Table 5: Sensitivity analysis of the sMGT turbomachinery design against ambient temperature Figure 17: Comparison of the reference compressor-turbine matching process in [25] (a) and the process followed in this work (b) (VNGVs) is far from this region of the compressor map. Indeed, at reduced loads, even if mass flow rate and/or607 turbine inlet temperature are reduced, so is pressure ratio, hence moving the operating conditions to lower regions of608 24 Figure 18: Part-load efficiency for a system with recuperator temperature limited to 650ºC. the map where the presence of VIGVs is not providing any added flexibility. Based on this, the utilisation of variable609 compressor geometry (i.e., VIGV) comes together with the adoption of variable nozzle guide vanes in the expander610 (VNGV). This combination is required to avoid derating the part-load performance of the engine through the adoption611 of a lower turbine inlet temperature once the inlet temperature to low-pressure side of the recuperator (turbine exhaust612 temperature) achieves the maximum value allowed by materials. In particular, the incorporation of Variable Nozzle613 Guide Vanes in the expander enables reducing the inlet are to the nozzles at part load, thus keeping pressure ratio at or614 close to the rated value. This higher pressure ratio than in the standard engine (without VNGVs) also yields a larger615 temperature drop across the turbine, thus enabling operation at low partial loads without the need to reduce turbine616 inlet temperature. This has a twofold beneficial effect. From a system standpoint, it means higher part-load efficiency617 due to the higher pressure ratio and turbine inlet temperature. For the compressor, it means higher pressure ratios at618 low mass flow rates, thus moving the running line closer to surge and to the higher efficiency region of the compressor619 map. Accordingly, the utilisation of VIGVs in the compressor, whose impact on compressor inlet flow was described620 in the previous section, is instrumental to enable such operation.621 The strong influence of incorporating VNGVs on system performance is better understood by observing the run-622 ning lines of compressor and turbine on their respective performance maps when VNGVs are or are not used. Figure623 19 shows this information for a representative solar micro gas turbine where DNI ranges from 250 W/m2to 900 W/m2 624 at constant ambient temperature (26.4ºC). The blue and red solid lines correspond to compressor and turbine with any625 variable geometry feature, whilst the blue and red dotted lines are those of the compressor and turbine incorporating626 VIGVs and VNGVs respectively; the low mass flow end of these latter running lines are set by the minimum TIT627 drop compliant with a surge margin of at least 10% (note that mass flow rate is reported in corrected values to account628 for the effect of inlet pressure and temperature). For the sake of clarity, turbine maps for only three different VNGVs629 angles, α2, are shown in Fig. 19: design point (fully open), 75% closed and 55% closed.630 Figure 19: Running lines of the SmGT. Blue and red lines correspond to compressor and turbine. Solid and dotted correspond to machines with fixed geometry and machines with variable geometry (VIGVs for compressor and VGNVs for turbines) respectively. The global impact of variable geometry on system performance is shown on Fig. 20, illustrating the pressure ratio631 and solar-to-electric efficiency of solar micro gas turbines with and without variable geometry across the entire load632 25