Experimental investigation of an organic Rankine cycle with liquid flooded expansion

Jera Van Nieuwenhuyse Student number: 01410277

Supervisors: Prof. dr. ir. Michel De Paepe, Dr. ir. Steven Lecompte

Master's dissertation submitted in order to obtain the academic degree of Master of Science in Electromechanical Engineering

Academic year 2018-2019

Experimental investigation of an organic Rankine cycle with liquid flooded expansion

Jera Van Nieuwenhuyse Student number: 01410277

Supervisors: Prof. dr. ir. Michel De Paepe, Dr. ir. Steven Lecompte

Master's dissertation submitted in order to obtain the academic degree of Master of Science in Electromechanical Engineering

Academic year 2018-2019 ”The author gives permission to make this master dissertation available for consultation and to copy parts of this master dissertation for personal use. In all cases of other use, the copyright terms have to be respected, in particular with regard to the obligation to state explicitly the source when quoting results from this master dissertation.”

Ghent, June 2019 Jera Van Nieuwenhuyse PREFACE v

Preface

I want to thank my supervisor, Steven Lecompte, for helping me through the process of making a thesis. Thank you for proof reading everything, for giving new insights when I got stuck and for staying optimistic. Thanks to Alihan Kaya, for reading my literature study multiple times and giving directions. I would also like to express my gratitude to the people from Kortrijk, Kenny Couvreur and Sergei Gusev, who put a lot of effort in getting the Blue ORC to run. One of the main things I will remember from this thesis, is that there is a lot more to a set-up than one would expect. Furthermore, I would also like to thank my promoter, Prof. Michel De Paepe, for giving me the opportunity to work on a subject very close to my interests.

Finally, I want to thank my friends and family for their support this last year. Special thanks to my sister for taking care of me and cooking superb dinners these last couple of months.

Experimental investigation of an organic Rankine cycle with liquid flooded expansion

Jera Van Nieuwenhuyse

Supervisors: Prof. Dr. Ir. Michel De Paepe, Dr. Ir. Steven Lecompte Counsellor: Alihan Kaya

Master’s dissertation submitted in order to obtain the academic degree of Master of Science in Electromechanical Engineering

Departement of Flow, Heat and Combustion Mechanics Chair: Prof. Dr. Ir. Michel. De Paepe Faculty of Engineering and Architecture Ghent University Academic year 2018–2019

Summary This master thesis deals with the experimental investigation of an organic Rankine cycle with liquid flooded expansion (ORCLFE). The first chapter, the introduction, elaborates on the context of low-temperature energy conversion systems. The second chapter summarizes the functioning of a basic organic Rankine cycle (ORC) and possible modifications to improve its performance. It provides a theoretical background and summarizes the research efforts on liquid flooded ORCs. The following chapter describes the experimental test set-up that was used to perform the measurements. The different parts and components of the set-up are explained in detail. It concludes with a section about why R1233zd(E) was chosen as the tested refrigerant. Thereafter, an overview of the boundary conditions and measured variables is given. The data reduction and uncertainty analysis performed on these measurements is explained. An experimental matrix and how the different settings of the ORC can be obtained is provided and the results are discussed. Chapter five focuses on theoretical ORC models and how they need to be adjusted to incorporate liquid flooding. It concludes with a summary of the performed study and the obtained results.

Keywords: organic Rankine cycle, liquid flooding, isothermal expansion, experimental, R1233zd(E) EXTENDED ABSTRACT viii

Extended abstract MASTER THESIS 2018-2019 Research group: Applied Thermodynamics and Heat Transfer Department of Flow, Heat and Combustion Mechanics - Ghent University, UGent

EXPERIMENTAL INVESTIGATION OF AN ORGANIC RANKINE CYCLE WITH LIQUID FLOODED EXPANSION

Jera Van Nieuwenhuyse, Steven Lecompte, Alihan Kaya, Michel De Paepe Department of Flow, Heat and Combustion Mechanics Ghent University Sint-Pietersnieuwstraat 41, B9000 Gent, Belgium E-mail: [email protected]

ABSTRACT NOMENCLATURE

The basic Rankine cycle, with water as the working BIVR [ ] Built-in expander volume ratio m˙ [kg/s− ] Mass flow rate fluid, converts energy of high-grade heat sources such p [P a] Pressure as heat coming from burning fossil fuels. However, Q˙ [J/s] Heat rate another thermodynamic cycle, i.e. the organic T[K] Temperature v [m3/kg] Specific volume Rankine cycle (ORC), is more efficient for the 3 conversion of low to medium-grade heat sources such V[m ] Volume W˙ [J/s] Power as geothermal, solar or waste heat. In the ORC a fluid x [ ] Mass fraction with a lower vapour pressure and boiling temperature X[−] Random variable y [−] Flooding ratio is used, such that evaporation at lower temperatures − is possible. The expander of the ORC is a critical part Greek symbols of the cycle as this is the part that extracts work from δ [ ] error − the working fluid. In this research, an experimental ηCarnot [ ] Carnot efficiency η [−] Expander isentropic efficiency investigation is conducted to check whether expander exp,is − ηexp,isoth [ ] Expander isothermal efficiency performance and efficiency can be improved through η [−] First law efficiency ORCLF E − the use of oil flooding, i.e. adding large amounts ηII,ORCLF E [ ] Second law efficiency of lubricant oil to the working fluid before it enters − the expander. An ORC test set-up has been built Subscripts aft after with a separate lubricant oil loop to be able to test bef before the influence of oil addition. Pressures, temperatures, cd condenser mass flow rates and torques are measured from cf cooling fluid which efficiencies and other performance indicators are e electric ev evaporator calculated. Measurements are conducted for different exp expander oil flooding ratios at different pressure levels, with hf heating fluid R1233zd(E) as the working fluid. Oil flooding ratios is isentropic are varied between 0 and 0.25 at pressure ratios isoth isothermal m measured ranging from 3.4 to 4.3. o lubricant oil r refrigerant

INTRODUCTION Acronyms EC Expansion chamber One of the hot topics of today is the human impact GHG Greenhouse gas on climate change and how this can and should be LFE Liquid flooded expansion mitigated. Agreements and targets were set out ORC Organic Rankine cycle ORCLFE Organic Rankine cycle with in order to achieve this mitigation. The ’20-20-20’ liquid flooded expansion climate and energy targets set out by the EU are SSE Single-screw expander one example [1]. They focus on three main targets: reducing greenhouse gas (GHG) emissions, reaching a high gross final energy use percentage of renewables been made, however to reach these goals and even and improving energy efficiency. Efforts have already more stringent ones, efforts will need to be extended. One way to have an impact on all the targets improving efficiency is by altering the working fluid, mentioned above is by altering aspects of electricity as different fluids will have different thermodynamic generation. Instead of using heat coming from burning and physical properties [6]. Selection can be done fossil fuels, waste heat or heat from renewables such as based on critical temperature and pressure, boiling solar and geothermal can be employed as heat source. temperature, specific volume, molecular weight, etc. This reduces the need for fossil fuels, which in turn [7, 8, 9]. Finally, cycle efficiency can also be improved reduces GHG emissions (as CO2 is produced while by optimizing cycle components such as the pump, burning fossil fuels) and increases the percentage of heat exchangers or expander. The focus of this works renewables used. Furthermore, by recovering waste lies on the latter. heat, energy efficiency can be improved. However, The main goal of a heat-to-power application is changing to these low-grade heat sources, which maximizing the net power output [5]. Therefore, consist of low-to-medium temperature heat (80-200°C) the expander is a key component of an ORC as [2], requires a different technology for its conversion to this component extracts the work from the working electricity. The Rankine cycle used in the conventional fluid. For smaller scale units (i.e. less than 50kW thermal power plant for this conversion is only efficient [10]), volumetric expanders are proven to be more at high temperatures [3]. A possible solution is the cost effective compared to turbo-expanders [2, 11]. ORC, which is similar to a conventional Rankine cycle, Ziviani et al. [2] investigated differences between but instead of using water as a working fluid, a fluid possible volumetric expanders and concluded that a with a lower vapour pressure and boiling temperature single-screw expander (SSE) has important advantages is used. This way, low-grade heat is still sufficient over other expanders such as high volumetric efficiency to evaporate the working fluid which subsequently and low leakages. They also stated that performance drives an expander resulting in shaft power from which is affected by internal losses (such as leakages, friction electricity can be generated. and heat loss) and the operating condition (i.e. An ORC consists of four main components: an applied pressure ratio over the expander). In addition evaporator, expander, condenser and pump, as can be to the expander, the expansion process itself can be seen in Figure 1. In the evaporator, heat from the heat altered as well. Two limiting cases are the adiabatic source is added to the working fluid, which evaporates. and the isothermal expansion, as illustrated in Figure After evaporation, the vapour subsequently drives an 2 by their p-v diagram. As can be seen in the figure, expander (or turbine). During this expansion process, the area under the curve of the isothermal process the pressure and temperature of the working fluid (where temperature is kept constant) is larger than drop. After the expander, the vapour is condensed that for the adiabatic one (a process without heat in a condenser such that it can be pressurized and transfer), meaning that more work can be extracted pumped around, closing the cycle. from the fluid using the former process. Moreover, an adiabatic process cannot be reached in practice as there is always a non-negligible heat transfer between working fluid and expander [12]. Another advantage of an isothermal expansion is that it will result in a higher expander outlet temperature. This favours the use of internal regeneration, which improves thermal efficiency. To achieve isothermal expansion, heat needs to be transferred to the working fluid while it expands [13]. This can be done by heating the expander surface or by adding a second medium with a higher heat capacity to the working fluid vapour. This fluid will act as a thermal buffer, meaning that the temperature drop due to expansion is recovered by the secondary medium. However, due to practical Figure 1: General ORC schematic [4] limitations such as finite surface area of the expander or finite heat capacity of the flooding medium, only a quasi-isothermal expansion process can be reached in The ORC is already considered a mature technology, reality. but the adoption rate in practical applications still has As friction losses and internal leakages are the potential for further growth. This can be realized, detrimental for expander performance, proper for example, by increasing efficiency and power lubrication is vital [14]. This is where liquid flooded output. Efficiency can be improved using different expansion (LFE) can have a positive influence on methods. One method is to alter the cycle layout, ORC efficiency and performance. In an ORC with e.g. by implementing internal regeneration. Internal LFE (ORCLFE), oil is mixed with the working regeneration means that the working fluid is preheated fluid vapour before entering the expander. The oil before entering the evaporator by the working fluid acts as lubrication oil, thus reducing friction and itself leaving the expander [5]. Another way of TEST FACILITY

Based on the work of Ziviani et al. [12] an ORC set-up was built. The focus of this work lies on experimentally investigating the influence of LFE and on validating the ORC cycle model that was developed. Real set-up performance will differ from the theoretical one due to assumptions which were made in the model. Some examples: pressure drops over the heat exchangers and line sets were neglected, the lubricant oil was considered to be incompressible and the ideal gas model was used for the refrigerant. The ORCLFE set-up, built using off-the-shelf Figure 2: Isothermal vs adiabatic expansion [13] components, consists out of four loops: a working fluid, heating, cooling and oil circulation loop. A hydraulic scheme of the set-up is given in Figure 3. The working fluid loop is the part of the ORC unit which contains leakage losses. In addition, the oil also acts as a the refrigerant and the basic ORC components (pump, thermal buffer thus pushing the expansion process evaporator, expander and condenser). The working towards a quasi-isothermal one, favouring the use of fluid under consideration is R1233zd(E). A bypass internal regeneration and increasing the expander is present at the regenerator in order to be able to power output. Ziviani et al. [12, 15] developed test a basic ORC (without regenerator), as well as models to investigate this on SSEs. In a first work, one with internal regeneration. As the influence of a thermodynamic cycle model was developed to oil addition is tested, an oil circulation loop with check the potential theoretical improvements on separate pump, oil cooler and heater is also present. thermodynamic performance and work output of The oil used is SAE 20W50. The oil is mixed with an ORCLFE compared to a baseline ORC with the refrigerant before entering the expander. Plate internal regeneration. The refrigerants investigated heat exchangers are used as condenser, evaporator, were R245fa and R1233zd(E). The cycle model was recuperator and oil heater. A static mixer is employed employed over a range of flooding ratios (yo) and to ensure a homogeneous mixture of refrigerant vapour built-in expander volume ratios (BIVR): and lubricant oil. Separation of the two fluids is done by a gravitational oil separator. A subcooler (also a m˙ plate heat exchanger) is present at the suction side y = o (1) o m˙ of the pump to ensure sufficient subcooling such that r there are no cavitation issues at the pump. The heat source of the evaporator and oil heater is simulated by an electric heater, having a maximum V BIVR = EC;end (2) heating capacity of 250kWe. Therminol 66 is used as VEC;start heating oil, with a maximum flow rate of 14m3/h at the maximum temperature of 340°C. Inlet heat source withm ˙ o the lubricant oil mass flow rate andm ˙ r the temperature and mass flow rate can be controlled. For refrigerant mass flow rate. In Equation 2 the BIVR is the cooling, a roof-top air cooled condenser is used, defined as the ratio of the expander chamber volume with a rated capacity of 480kW at 20°C ambient. at the end of expansion (VEC;end) to the expander The cooling liquid through the condenser is a water chamber volume at the start of expansion (VEC;start). and ethylene-glycol mixture. Cooling flow rate can be Overall, it was found that there was a significant controlled. As the inlet temperature of the cooling impact on cycle efficiency. Cycle efficiency improved liquid is directly related to the outdoor conditions, with 6.71% for R245fa and 2.90% for R1233zd(E). In a this parameter cannot be controlled. The chiller, used second work a semi-empirical model for the expander to cool the subcooler, also uses a water-etylene-glycol was incorporated to account for the presence of oil, mixture to cool down the refrigerant. friction and heat losses and internal volume ratio. The The SSE that is employed is the one elaborated refrigerants tested were R1234ze(Z), R1233zd(E) and in the work of Ziviani et al., which is already R1336mzz(Z). It was found that the optimal internal experimentally and numerically characterized [14]. A volume ratio lies between 4 and 6 for liquid flooded standard single-screw air compressor was converted conditions. The findings concerning cycle efficiency in order to operate as an expander. Some further were similar to the ones from their previous study. alterations were done to improve its performance such In addition, oil flooding of the expander resulted in as enlargement of the expander discharge port. The proper lubrication, reducing mechanical losses in the SSE has an BIVR of 5.3. Some practical limits apply expander, and led to increased net power output. to the pressure ratio that can be applied over the Figure 3: Hydraulic scheme of the test set-up

expander. For safety reasons, the upper limit is 1200 where W˙ net is the difference between the power kPa at the expander suction side and 300 kPa at the generated by the expander and the sum of the powers discharge side. An 11kWe generator is coupled to the used by the refrigerant and oil pump. Q˙ in is the sum of expander. the heat added to the refrigerant and to the lubricant Temperatures, pressures, mass flow rates and oil by the heat source. torques are measured. The ratio of the cycle efficiency and the Carnot efficiency is the Second Law efficiency:

OUTPUT VARIABLES ηORCLF E ηII,ORCLF E = (7) ηCarnot From these measured variables, efficiencies and other performance parameters of the ORCLFE can be with the Carnot efficiency the theoretical maximum derived. These are evaluated in function of the efficiency that can be reached with the cold amount of oil added to the refrigerant and the and hot source present (when assuming infinite pressure ratio over the expander. Possible variables for heat capacities), characterized by respectively the characterizing the mass flow rate of oil in the expander temperature of the cooling fluid (cf) at the condenser are the flooding ratio (yo) and the oil mass fraction (cd) inlet and the temperature of the heating fluid (hf) (xo): at the evaporator (ev) inlet: m˙ x = o (3) o Tbef,cd,cf m˙ o +m ˙ r η = 1 (8) Carnot − T The pressure ratio is defined as the pressure at the bef,ev,hf suction side of the expander (pbef,exp) divided by the Isothermal efficiency of the expander is equal to the pressure at the discharge side (paft,exp): ratio of the power generated by the actual expansion process to the power generated by an isothermal pbef,exp PR = (4) (isoth) process: paft,exp W˙ Similarly, the temperature ratio is calculated as η = exp (9) exp,isoth ˙ follows: Wexp,isoth T TR = aft,exp (5) Similary, the isentropic expander efficiency is Tbef,exp defined by the ratio of the actual power to the power The ORC cycle efficiency is defined as: generated by an isentropic (is) process:

W˙ W˙ exp net η = (10) ηORCLF E = (6) exp,is ˙ Q˙ in Wexp,is DATA REDUCTION AND UNCERTAINTY ANALYSIS

When the ORC set-up is started, it takes some time before steady-state operation is reached. Steady-state operation means that the different properties (such as pressure and temperature) don’t vary in time anymore. However, as data is taken from an experimental set-up, this condition will never be reached and a more practical approach to steady-state is required (described in detail by Lecompte et al. [16]). The standard deviations of specified properties (12 in total) within a representative steady-state zone are calculated and will serve as references. For Figure 4: w in function of y every measured sample of these 12 properties, a net forward-moving standard deviation is calculated. The latter is then compared to its reference for identifying the actual steady-state zones. Finally, the steady-state points are calculated as the average of the samples in the steady-state zones. Sensors are used to measure the different properties of the set-up. As these have a limited accuracy, there will be an uncertainty interval around the measured value (Xm): X = X δX (11) m ± with δX the uncertainty on the measurement. As the variables of interest are calculated from the measured ones, these will therefore also have a certain absolute uncertainty (U), calculated according to [17]: Figure 5: W˙ pp,o in function of y

N δK U = ( U )2 (12) K v δX · Xi ui=1 i uX t where the variable K is a function of the variables Xi; i=1..N.

RESULTS

Net specific power generated by the cycle decreases with increasing flooding ratio (Figure 4). The decrease in net power is due to the combined effect of a decrease in generated expander power and an increase in oil pumping power up to a certain level (Figure 5). Refrigerant pumping power has limited effect. At an Figure 6: ηexp,is in function of y oil pumping speed above 250 rpm, resonance occurred in the oil piping. Therefore, oil pumping speed and as a result flooding ratios that could be tested, were the potential for internal regeneration also stays limited. unaltered. Isentropic efficiency drops as well with increasing Figure 8 and 9 present the evolution of the cycle and flooding ratio (Figure 6). The addition of oil thus Second Law efficiency. Both of them decrease with increases the losses in the expander, instead of increasing flooding. The trend of ORCLFE efficiency decreasing friction and leakage losses. is similar to the one of the net specific power (Figure The margin of error on the isothermal efficiency 4). This indicates that the increase in heat added was too large to be able to deduce a trend. The to the cycle for heating the oil, is less important temperature ratio however also illustrates how much compared to the decrease in net power. the process approximates an isothermal process: Theoretically, expander and cycle efficiencies increasing the flooding has no observable impact on should increase and the expansion process should expander outlet temperature (Figure 7). Therefore, approximate an isothermal process better when liquid issue should however be solved first.

CONCLUSION

An ORC is a cycle which can be used for the conversion of low-grade heat sources such as solar, geothermal or waste heat. Modifications to the cycle layout can be done to increase efficiency and power output. Applying LFE is one possible modification. In LFE a large amount of oil is added to the refrigerant before it enters the expander. The oil acts as a thermal buffer, and pushes the expansion process towards an isothermal process. This increases the power output and increases the potential for internal regeneration. In addition, the oil reduces friction and leakage losses Figure 7: TR in function of y in the expander. A test set-up, with a separate oil circulation loop, has been built to experimentally investigate LFE. Measurements are done on a SSE and R1233zd(E) is used as refrigerant. Flooding and pressure ratios are varied and different performance parameters are calculated, as well as their uncertainty intervals. Overall, it was found that liquid flooding has a negative effect on ORC performance. Expander powers, efficiencies and cycle efficiencies decrease for increasing flooding ratio and no significant impact on expander outlet temperature can be observerd. Additional research with a different oil, higher PRs and higher flooding ratios is needed to make a final conclusion on the impact of liquid flooding on ORC performance. Figure 8: ηORCLF E in function of y REFERENCES

[1] European Environment Agency. Overall progress towards the european union’s ’20-20-20’ climate and energy targets. https://www.eea.europa.eu/themes/climate/ trends-and-projections-in-europe/trends-and- projections-in-europe-2016/1-overall-progress- towards-the, 7 November 2017. Accessed on: 29 November 2018.

[2] Davide Ziviani, Eckhard A Groll, James E Braun, and Michel De Paepe. Review and update on the geometry modeling of single-screw machines with Figure 9: ηII,ORCLF E in function of y emphasis on expanders. International Journal of Refrigeration, 92:10–26, 2018.

flooding is applied. However, based on the results [3] Ignace Vankeirsbilck, Bruno Vanslambrouck, described above, this is not necessarily the case in Sergei Gusev, and Michel De Paepe. Organic reality. In order to make a final conclusion on the rankine cycle as efficient alternative to steam impact of oil flooding, additional research is needed. cycle for small scale power generation. HEFAT, Oils which are more suited to be used in refrigeration 2011. cycles should be tested. The limits of the set-up w.r.t the pressure ratios and other performance altering [4] Oyeniyi A Oyewunmi, Aly I Taleb, Andrew J parameters should be investigated. Higher PRs might Haslam, and Christos N Markides. On the use of lead to higher efficiencies as under-expansion losses are saft-vr mie for assessing large-glide fluorocarbon reduced. Furhtermore, higher flooding ratios should working-fluid mixtures in organic rankine cycles. be tested as well. In order to do this, the resonance Applied Energy, 163:263–282, 2016. [5] Steven Lecompte, Henk Huisseune, Martijn Van Broek. Analysis of an organic rankine Den Broek, Bruno Vanslambrouck, and Michel cycle with liquid-flooded expansion and internal De Paepe. Review of organic rankine cycle (orc) regeneration (orclfe). Energy, 144:1092–1106, architectures for waste heat recovery. Renewable 2018. and Reviews, Sustainable Energy, 47:448–461, 2015. [16] Steven Lecompte, Sergei Gusev, Bruno Vanslambrouck, and Michel De Paepe. [6] Junjiang Bao and Li Zhao. A review of working Experimental results of a small-scale organic fluid and expander selections for organic rankine rankine cycle: Steady state identification and cycle. Renewable and reviews, sustainable energy, application to off-design model validation. 24:325–342, 2013. Applied Energy, 226:82 – 106, 2018. [7] Sotirios Karellas and Andreas Schuster. [17] Robert J. Moffat. Describing the uncertainties in Supercritical fluid parameters in organic rankine experimental results. Experimental Thermal and cycle applications. International Journal of Fluid Science, 1(1):3 – 17, 1988. Thermodynamics, 11(3), 2008. [8] V Maizza and A Maizza. Working fluids in non-steady flows for waste energy recovery systems. Applied Thermal Engineering, 16(7):579–590, 1996. [9] MJ Lee, DL Tien, and CT Shao. Thermophysical capability of ozone-safe working fluids for an organic rankine cycle system. Heat Recovery Systems and CHP, 13(5):409–418, 1993. [10] Muhammad Imran, Muhammad Usman, Byung-Sik Park, and Dong-Hyun Lee. Volumetric expanders for low grade heat and waste heat recovery applications. Renewable and Sustainable Energy Reviews, 57:1090 – 1109, 2016. [11] Biao Lei, Wei Wang, Yu-Ting Wu, Chong-Fang Ma, Jing-Fu Wang, Lei Zhang, Chuang Li, Ying-Kun Zhao, and Rui-Ping Zhi. Development and experimental study on a single screw expander integrated into an organic rankine cycle. Energy, 116:43–52, 2016. [12] Davide Ziviani, Sergei Gusev, Stefan Schuessler, Abdennacer Achaichia, James E Braun, Eckhard A Groll, Michel De Paepe, and Martijn van den Broek. Employing a single-screw expander in an organic rankine cycle with liquid flooded expansion and internal regeneration. Energy Procedia, 129:379–386, 2017. [13] Opubo N Igobo and Philip A Davies. Review of low-temperature vapour power cycle engines with quasi-isothermal expansion. Energy, 70:22–34, 2014. [14] Davide Ziviani, Sergei Gusev, Steven Lecompte, EA Groll, JE Braun, W Travis Horton, Martijn van den Broek, and Michel De Paepe. Characterizing the performance of a single-screw expander in a small-scale organic rankine cycle for waste heat recovery. Applied Energy, 181:155–170, 2016. [15] Davide Ziviani, Eckhard A Groll, James E Braun, Michel De Paepe, and Martijn van den

CONTENTS xvii

Contents

Preface v

Abstract vii

Extended abstract viii

Table of Contents xvii

List of Figures xx

List of Tables xxii

Nomenclature xxiii

1 Introduction 1

2 Literature Review 5 2.1 Organic Rankine cycles in general ...... 5 2.2 Improving organic Rankine cycle efficiency and performance ...... 6 2.3 The expansion process ...... 8 2.3.1 Generating shaft power with expanders ...... 8 2.3.2 Isothermal expansion ...... 10 2.3.3 Two-phase expansion ...... 11 2.4 Liquid flooded expansion ...... 13 2.4.1 Theoretical research on liquid flooded expansion ...... 13 2.4.2 Experimental efforts ...... 15 2.5 Comparing organic Rankine cycles ...... 16

3 Description of the Organic Rankine Cycle Test Facility 17 3.1 Working fluid loop ...... 17 3.2 Oil circulation loop ...... 20 3.3 Heating loop ...... 20 CONTENTS xviii

3.4 Cooling loop ...... 22 3.5 Single-screw expander ...... 22 3.6 Refrigerant selection ...... 24

4 Experimental Investigation 27 4.1 Variables ...... 27 4.1.1 Measured variables ...... 27 4.1.2 Output variables ...... 29 4.2 Data reduction and uncertainty analysis ...... 32 4.2.1 Determining steady-state ...... 32 4.2.2 Uncertainty analysis ...... 33 4.2.3 Required measuring interval ...... 38 4.3 Experimental matrix ...... 38 4.4 Altering set-points ...... 39 4.5 Results ...... 41

5 Theoretical ORC modelling 50 5.1 ORC model description ...... 50 5.2 Required adjustments for an ORCLFE model ...... 52 5.3 ORC model simulations ...... 53

6 Conclusion 60

A Performing measurements 62 A.1 Measuring equipment and data acquisition ...... 62 A.1.1 Pressure sensor ...... 62 A.1.2 Temperature sensor ...... 62 A.1.3 Flow rate sensor ...... 63 A.1.4 Torque sensor ...... 64 A.1.5 Data acquisition ...... 64 A.2 Run procedures ...... 65 A.2.1 Prior to start-up ...... 65 A.2.2 Start-up ...... 67 A.2.3 Shutdown ...... 67

B Steady-state data points and corresponding uncertainties 68

C Correlations and sensitivity analysis 73 C.1 Correlations for oil properties ...... 73 C.2 Correlations for mixture properties ...... 75 C.3 Uncertainty on correlations ...... 76 CONTENTS xix

C.4 Sensitivity analysis ...... 77

Bibliography 80 LIST OF FIGURES xx

List of Figures

1.1 Distribution of global GHG emissions in 2014 [4] ...... 2 1.2 Power plant using a Rankine cycle and fossil fuels to create the heat source [6]3

2.1 General ORC schematic and corresponding T-s diagram [20] ...... 6 2.2 T-s diagram of wet, dry and isentropic fluids [25] ...... 8 2.3 Expansion in a single-screw expander [29] ...... 9 2.4 Isothermal vs adiabatic expansion [33] ...... 11 2.5 Temperature-heat diagram [36] ...... 12 (a) Cold flow phase change ...... 12 (b) No phase change ...... 12

3.1 Test set-up ...... 18 3.2 Simple hydraulic scheme of set-up ...... 19 3.3 Chiller ...... 21 3.4 Plate heat exchanger ...... 21 3.5 Oil separator ...... 21 3.6 Static mixer ...... 21 3.7 Electric heater (left) with connecting lines to ORC (right) ...... 22 3.8 Roof-top cooling unit ...... 23 3.9 3D CAD model of single-screw expander [46] ...... 25 3.10 Expander installed on set-up ...... 25 3.11 T-s diagram of R245fa and R1233zd(E) ...... 26

4.1 Attaining settings ...... 40 4.2 T-s diagram of experimental results ...... 42

4.3 wnet in function of y ...... 44

4.4 wexp in function of y ...... 44 ˙ 4.5 Wpp,r in function of y ...... 44 LIST OF FIGURES xxi

˙ 4.6 Wpp,o in function of y ...... 44

4.7 ηexp,is in function of y ...... 45 4.8 TR in function of y ...... 46

4.9 Texp,out in function of y ...... 46

4.10 ηexp,isoth in function of y ...... 46

4.11 ηORCLF E in function of y ...... 47

4.12 ηII,ORCLF E in function of y ...... 47

4.13 ηORCLF E,exp in function of y ...... 47 ˙ 4.14 Qrecuperator in function of y ...... 48

4.15 wnet in function of y for same PR ...... 49

5.1 pout,exp in function of m˙ cf ...... 54

5.2 pin,exp in function of m˙ cf ...... 54

5.3 Tin,exp in function of m˙ hf ...... 55

5.4 Tin,exp in function of Tin,ev,hf ...... 55

5.5 pin,exp in function of Tin,ev,hf ...... 56

5.6 PR in function of Nexp ...... 56

5.7 pin,exp in function of Nexp ...... 56

5.8 m˙ r in function of Nexp ...... 56

5.9 m˙ r in function of fpp ...... 57

5.10 PR in function of fpp ...... 57

5.11 pin,exp in function of fpp ...... 57 5.12 Experimentally obtained trends ...... 59

A.1 Pressure and temperature sensor ...... 63 A.2 Mass flow rate sensor ...... 64 A.3 Torque meter ...... 64 A.4 Control cabinet ...... 65 A.5 LabVIEW scheme ...... 66 LIST OF TABLES xxii

List of Tables

3.1 Overview of components ...... 23 3.2 Geometric parameters of the single-screw expander [46] ...... 24 3.3 Properties of R245fa and R1233zd(E) [48–50] ...... 26

4.1 Overview of measured variables ...... 28 4.2 Sensor uncertainties ...... 33 4.3 Parameters ...... 39 4.4 Heat balance evaporator ...... 43 4.5 Heat losses ...... 43

5.1 Boundary conditions ...... 53

B.1 Steady-state points ...... 69 B.2 Absolute uncertainties corresponding to Table B.2 ...... 70 B.3 Steady-state points used for modeling illustration (Section 5.3) ...... 71 B.4 Absolute uncertainties corresponding to Table B.3 ...... 72

C.1 Input variables ...... 78 C.2 Uncertainty intervals in function of correlation uncertainties ...... 79 NOMENCLATURE xxiii

Nomenclature

cp specific heat capacity J/(kg.K) D diameter m f frequency Hz h specific enthalpy J/kg L length m ˙m mass flow rate kg/s MM molar mass kg/kmole N rotational speed rpm p pressure Pa Q quality - Q˙ heat rate J/s s specific entropy J/(kg.K) T temperature K u relative uncertainty - U absolute uncertainty unit of base variable v specific volume m3/kg V volume m3 V˙ volume flow rate m3/s w specific work J/kg W work J W˙ power J/s x mass fraction - NOMENCLATURE xxiv

y flooding ratio -

Greek symbols

δ error -

ηCarnot Carnot efficiency -

ηis isentropic efficiency -

ηORCLF E First law efficiency -

ηvol volumetric efficiency -

ηII,ORCF LE Second law efficiency - σ standard deviation unit of base variable τ torque J/rad

Subscripts

0 reference cd condenser cf cooling fluid corr correlation crit critical e electric eh electric heater ev evaporator exp expander g groove hf heating fluid NOMENCLATURE xxv

in inlet is isentropic k random variable L liquid phase m measured max maximum mech mechanical mix mixture o lubricant oil oa overall out outlet pp pump r refrigerant ref reference sr screw rotor sw starwheel V vapour phase vol volumetric yr year

Acronyms

APS Absolute pressure sensor CFC Chlorofluorocarbon CFM Coriolis flow meter DAQ Data acquistition EC Expansion chamber NOMENCLATURE xxvi

GHG Greenhouse Gas GWP Global warming potential HCFC Hydrochlorofluorocarbon HFC Hydrofluorocarbon HFO Hydrofluoroolefin LFE Liquid flooded expansion LFEC Liquid flooded Ericsson cycle ODP Ozone depletion potential ORC Organic Rankine cycle ORCLFE Organic Rankine cycle with liquid flooded expansion PLC Programmable logic controllers PR Pressure ratio PTFE Polytetrafluoretheen RSV Reference state value RTD Resistance temperature detector SSE Single-screw expander TC Transcritical cycle TFC Trilateral Flash cycle TLC Trilateral cycle TR Temperature ratio UFM Ultrasonic flow meter VFD Variable frequency drive INTRODUCTION 1

Chapter 1

Introduction

Nowadays, one of the big challenges is to lower the impact on climate change and to move towards a more sustainable environment. To mitigate the effects on climate change, agreements and targets were set out, a couple of which are mentioned here. The Paris Agreement set out the goal to keep the global temperature rise this century below 2°C compared to pre-industrial levels [1]. The EU set out the ’20-20-20’ climate and energy targets, which have three main goals [2]. Firstly, the greenhouse gas (GHG) emissions should be reduced with 20% by 2020 compared to the levels in 1990. The second target is to reach a gross final energy use of renewable energy of 20%. Lastly, the energy efficiency should be improved such that the primary energy use decreases with 20% compared to the 2007 Energy Baseline Scenario [3] (i.e. 13% compared to the actual 2005 level). These short-term targets are already reached or will be reached if the current trend of effort is sustained. In 2014, a reduction of 23% in GHG emissions, a gross final energy use of renewable energy of 16% and a primary energy use reduction of 12% compared to 2005 was reached.

However, to reach the two-degrees scenario and the long-term goal for 2050, i.e. reducing GHG emissions to 80-95% below 1990 levels, the efforts need to be increased. As can be seen in Figure 1.1, which presents the distribution of the key GHGs emitted by human activities, these GHG emissions consist for more than 60% out of CO2 which is released to the atmosphere by burning fossil fuels and through industrial processes. Lowering the GHG emissions to reach the targets thus means that the burning of these fossil fuels needs to be reduced. Moreover, the transport sector, heat generation and electricity production, which all use these fossil fuels, play a big part in CO2 emissions. Lowering the usage of fuels in the two latter cases can be achieved in multiple ways: lowering the overall energy INTRODUCTION 2

demand, utilizing renewable energy sources, use the energy as efficiently as possible, etc.

Figure 1.1: Distribution of global GHG emissions in 2014 [4]

As already mentioned, electricity generation can be achieved by burning fossil fuels (coal, gas or oil). This process is done in a thermal power plant as represented in Figure 1.2. Fossil fuels are burnt and the heat of the exhaust gases is used as the heat source for a Rankine cycle to produce superheated steam. This superheated steam is fed to a turbine, which is connected to a generator to produce electricity. As these fossil fuels have a high carbon content, a lot of CO2 is produced during the combustion with air, which is released into the atmosphere and contributes to the climate change. Furthermore, other pollutants can also be formed during combustion such as NOX , CO and SOX , meaning that a lot of effort and care is needed during treatment of the flue gases [5].

Instead of using fossil fuels to create the heat source, other sources can be used as well such as solar, geothermal and waste heat. These sources are termed as low-grade heat sources and consist of low-to-medium temperature heat (80-200°C) [7]. However, the Rankine cycle used in the conventional thermal power plant is only efficient at higher temperatures (i.e. for a subcritical cycle, typical steam temperature range is 535-565°C [8]), as at lower temperatures there is a constraint on maximum superheating temperature and evaporation pressure of the generated steam, resulting in limited efficiencies [9]. Therefore, another INTRODUCTION 3

Figure 1.2: Power plant using a Rankine cycle and fossil fuels to create the heat source [6] technology needs to be employed to efficiently convert the heat from these sources. Different thermodynamic cycles can be employed. These include the standard Rankine cycle or more exotic variations such as the trilateral (flash) cycle (TLC, TFC), etc [10]. The focus of this thesis lies on organic Rankine cycle (ORC) technology. An ORC is similar to a Rankine cycle, but instead of using water as a working fluid, a fluid with a lower vapor pressure and boiling temperature is used. This way the heat coming from a low temperature heat source is still sufficient to evaporate the working fluid to subsequently drive a turbine and generate electricity.

Besides being more efficient at lower temperatures, the ORC has a couple of other ad- vantages over a conventional steam Rankine cycle, such as being designed for unmanned operation with little maintenance, having favourable operating pressures and lacking the need for superheating (if isentropic or dry fluids are used; see Section 2.1) [9, 10]. Super- heating the steam in a conventional cycle is necessary, because otherwise liquid droplets formed during expansion could damage the turbine blades.

The ORC is already considered a mature technology. However, the adoption rate in prac- tical applications has the potential for further growth. This can for example be done by further increasing efficiency and power output (while keeping investment cost low) (see INTRODUCTION 4

Section 2.2). This is called thermo-economic optimization, with the aim to achieve lower cost per power output. When applying the ORC to recover waste heat, attention should also be given to the possibility of incorporating the system into an already existing plant. Not every practical application will benefit from the installation of an ORC [11]. There is also the need to cover the lack of experimental data from open literature with respect to optimization of efficiency and performance. Multiple methods can be applied for these optimizations. One way is altering the basic ORC cycle configuration, for example by implementing liquid flooded expansion (LFE). In an ORC with LFE, the expansion process is altered by mixing oil with the working fluid vapour before entering the expander. Bene- fits are that by the addition of the oil, friction losses and internal leakages of the expander are reduced. Furthermore, at the end of the expansion process, more heat is available for internal regeneration. The main drawback is that an additional oil circulation loop is required, which increases costs and total power usage of the pumps. Ziviani et al. [7,12–15] developed models to investigate LFE and did experimental research on single-screw ex- panders (SSEs). However, experimentally investigating the technology of LFE in a SSE for ORC applications still needs to be done in order to experimentally validate the model they developed and to asses the practical application of oil flooding in ORC technology.

The focus of this thesis lies on experimentally investigating the efficiency and performance improvement of an ORC set-up with a SSE through the use of LFE (see Section 2.4). LITERATURE REVIEW 5

Chapter 2

Literature Review

2.1 Organic Rankine cycles in general

In Figure 2.1 the general schematic of an ORC and its corresponding T-s diagram is presented. The overall layout and working principle is the same as for the conventional Rankine cycle, as already mentioned in the introduction. In the evaporator, heat from a heat source is added to the working fluid. Available heat resources are presented and their maturity (i.e. the extent of their development and application) is analyzed by Tchanche et al. [16]. During heat addition, the high pressure fluid evaporates and it leaves the evap- orator as saturated or superheated vapour. The vapour subsequently drives an expander (or a turbine) where a drop in pressure and temperature occurs resulting in shaft power from which electricity can be generated. After the expander, the vapour is condensed in a condenser, such that it can be pumped to a higher pressure. Finally the working fluid enters the evaporator, closing the cycle.

The working fluid can be a fluid such as hydrocarbons or refrigerants, preferably with a lower vapour pressure than water. The choice of a working fluid furthermore depends on many additional factors. Chen et al. [17] stated that besides their thermodynamic properties determining efficiency (see Section 2.2), other factors such as cost, availability, chemical stability, safety, compatibility with cycle components and environmental impact need to be considered as well . The two most important environmental factors are the global warming potential (GWP) and the ozone depletion potential (ODP). The GWP is a relative measure, compared to CO2, for the amount of heat a similar mass of the chemical traps in the atmosphere up to a specific amount of time [18]. The ODP is a relative measure, compared to trichlorofluoromethane (R-11; CFC-11), for the degradation of the 2.2 Improving organic Rankine cycle efficiency and performance 6 ozone layer caused by a similar mass of the chemical [19]. Additionally, properties such as their flammability and toxicity should be taken into account to assure safe operation of the ORC unit.

Figure 2.1: General ORC schematic and corresponding T-s diagram [20]

2.2 Improving organic Rankine cycle efficiency and performance

Efficiency of an ORC can be improved using different methods. One way is to alter the cycle design, e.g. by implementing internal regeneration, working at supercritical pressure, adding different evaporation pressure levels, etc. Internal regeneration is when the working fluid is preheated before it enters the evaporator by the working fluid itself leaving the expander [10]. Different ORC architectures for waste heat recovery and an overview of the available experimental data is given by Lecompte et al. [10].

Another way of improving efficiency is by altering the working fluid, as different working fluids will have different thermodynamic and physical properties. One way to categorize working fluids can be through the shape of their T-s diagram (see further on), and further classification follows from their thermodynamic and physical properties, which will affect cycle efficiency and performance [21]. Karellas et al. [22] state that selection of the fluid is done according to the process parameters of the cycle. The optimal fluid is the one which provides the highest thermal and cycle efficiency, according to the critical temperature 2.2 Improving organic Rankine cycle efficiency and performance 7

and pressure, as well as the boiling temperature under various pressures. Maizza and Maizza [23] mentioned low critical temperature and pressure, small specific volume, low viscosity and surface tension, high thermal conductivity, suitable thermal stability, high latent heat, low specific heat and moderate vapour pressure as desirable thermodynamic and physical properties. Lee et al. [24] stated that ORC cycle efficiency depends on the molecular weight, normal boiling point and critical pressure of the working fluid. However, as already mentioned, not only the thermodynamic and physical properties should be taken into account when selecting a working fluid. Other factors such as safety and cost are important as well. No fluid will perform optimally in all the criteria, and compromises must be made [17]. The choice of the working fluid might also influence the required ORC system architectures, e.g. the need for superheating. Depending on the slope of the vapour saturation line in a T-s diagram, a classification of fluids can be made, presented in Figure 2.2. For water, termed as a wet fluid, the slope is negative and a certain amount of superheat is needed after evaporation. If no superheating would be applied, liquid droplets would be formed during expansion in the turbine which could damage the turbine blades and lower the turbine isentropic efficiency. A dry fluid (with a positive slope) and an isentropic fluid (with a slope equal to infinity) don’t require superheat as no liquid phase is formed during expansion. Furthermore, when using a dry fluid a certain amount of superheating is achieved during expansion, which can be used for regeneration [9]. In addition, the temperature and pressure ranges where these fluids operate are important to develop efficient ORCs. As can be see in Figure 2.2, n-pentane and R245fa evaporate at much lower temperatures than water (considering the same pressure), making them more suitable for ORC applications, which are driven by low-grade heat sources.

Finally, cycle efficiency can also be improved by optimizing cycle components such as the pump, expander or heat exchangers. 2.3 The expansion process 8

Figure 2.2: T-s diagram of wet, dry and isentropic fluids [25]

2.3 The expansion process

As already mentioned, the expansion of the working fluid over an expander provides the usable shaft power which in turn can be converted into electricity. In this section possible types of expanders and expansion processes are highlighted.

2.3.1 Generating shaft power with expanders

As the main goal of a heat-to-power application is maximizing the net power output [10], a key component of an ORC system is the expander, as this extracts work from the working fluid. For large scale ORCs, turbo-expanders are normally used. However, for smaller scale units (i.e. less than 50kW [26]), positive displacement (i.e. volumetric) expanders are proven to be more cost effective [7,27]. These can run at lower rotational speeds, have

relatively high efficiency, high pressure ratio (PR) (i.e. pexp,in/pexp,out) and are tolerant to two-phase liquid/vapour operation [21]. They also have higher reliability concerning erosion durability [28], which becomes important when multi-phase expansion is considered (i.e. when liquid droplets could impact the expander).

Different volumetric expansion machines exist such as single-screw, twin-screw, scroll, ro- tary and reciprocating expanders. Ziviani et al. [7] investigated differences between possi- ble volumetric expanders, where they stated that their performance is affected by internal losses (such as leakage, friction and heat loss) and the operating condition (i.e. applied 2.3 The expansion process 9

pressure ratio over the expander). Their conclusion was that a single-screw expander (SSE) has important advantages over the other volumetric machines, such as: long working life, high volumetric efficiency, low leakages, low vibration, a simplified configuration and bal- anced loading of the main screw (since the expansion process occurs simultaneously on both sides of the rotor). In Figure 2.3 the expansion process in a SSE is presented. A SSE consists of different channels. When a channel is opened to the suction port during rotation of the screw, it is filled with vapour at suction pressure (Figure 2.3a). At a certain crank angle, the suction port gets closed off and expansion can start as the closed working channel increases in volume during further rotation (Figure 2.3b). When the end of the channel aligns with the discharge port, the vapour (which decreased in pressure due to the expansion) is discharged (Figure 2.3c).

Figure 2.3: Expansion in a single-screw expander [29]

Some important performance criteria and parameters of volumetric expanders are:

1. Volumetric efficiency: it represents how efficient the filling of the expander is (corre- lated to leakage and throttling). It is defined as the ratio of the theoretical flow rate to the actual one [30]: V˙ η = theoretical vol ˙ (2.1) Vactual The theoretical flow rate follows from the geometry of the expander. The actual flow rate is calculated based on the mass flow rate through the expander and the density of the fluid at expander inlet [31]. 2.3 The expansion process 10

2. Isentropic 1 efficiency: ratio of the actual work extracted to the theoretical isentropic work extracted: Wactual ηis = (2.2) Wisentropic

3. Built-in expander volume ratio (BIVR): the ratio of the volume of the expansion chamber (EC) at the end of expansion to the volume of the expansion chamber at the start of expansion [32]: VEC; end BIVR = (2.3) VEC; start

4. Under/over-expansion: in the case of under-expansion the pressure of the working fluid during discharge is greater than the discharge pressure. For over-expansion the opposite holds true.

2.3.2 Isothermal expansion

Two limiting cases for an expansion process are the adiabatic and isothermal expansion. During an adiabatic expansion process, energy transfer occurs without heat transfer. As there always is non-negligible heat transfer between the working fluid and the expander housing, adiabatic expansion cannot be reached [15]. In Figure 2.4 the p-v-diagrams of both the isothermal and adiabatic expansion process are presented. As can be seen in the figure, the area under the isothermal curve is bigger than the one under the adiabatic one, meaning that more work can be extracted from the fluid using an isothermal process. This isothermal operation has a second benefit: isothermal expansion will result in a higher expander outlet temperature than the adiabatic one. This favours the use of internal regeneration, which also improves thermal efficiency of the ORC cycle.

To achieve isothermal expansion, heat needs to be transferred to the working fluid while the fluid expands [33], to make up for the temperature drop during expansion. One way to achieve this isothermal expansion is by heating the surface of the expander sufficiently above the fluid temperature [33]. During expansion, heat is then transferred from the expander surface to the fluid. For this application, expanders with high surface-to-volume ratio are preferred. A second way to achieve isothermal expansion is by adding a secondary medium with a higher heat capacity to the working fluid vapour. This secondary medium will then act

1An isentropic process is a process which is both adiabatic (i.e. without heat transfer) and reversible (i.e. without irreversibilities, such as friction) 2.3 The expansion process 11

Figure 2.4: Isothermal vs adiabatic expansion [33] as a thermal buffer, meaning that the temperature drop due to expansion is recovered by the secondary fluid. For a steam engine, this secondary fluid is steam at a very high temperature which is injected into the expanding steam [34]. In an ORC, a different technology called liquid flooding can be applied. An oil or other medium is added after the evaporator, before entering the turbine/expander. As the surface area of the expander or the heat capacity of the flooding medium is finite, in reality only a quasi-isothermal expansion can be reached in both surface heating and the use of a secondary fluid, respectively.

Applying liquid flooding means two different phases are present in the expander. The vapour phase is the evaporated refrigerant and the liquid phase the secondary medium. In another expansion process applicable to ORCs, called two-phase expansion, two different phases are present as well during expansion. In this case, the two phases are the liquid and vapour phase of the refrigerant. This process is further explained in the following section.

2.3.3 Two-phase expansion

Heat coming from a waste heat source is transferred to the working fluid in the evaporator. As there is a phase change of the working fluid from liquid to vapour, a mismatch exists between the hot stream (waste heat) and cold stream (working fluid) (see Figure 2.5a). 2.3 The expansion process 12

This results in two main losses. Firstly, sensible heat (i.e. heat associated to a change in temperature) that is still present at the exit of the ORC is lost to the environment. Secondly, irreversibilities exist due to the finite temperature difference between the two streams. Reducing these irreversibilities plays a part in increasing ORC performance, which is done by minimizing the temperature difference between the two streams. In a TLC, the working fluid is heated until the point of saturated liquid. As a consequence no phase change occurs in the evaporator, which results in a better thermal match between the hot and cold stream (see Figure 2.5b). However, since no phase change occurs, higher flow rates and larger heat transfer surfaces are needed [35]. After the evaporator, the saturated liquid enters the expander, where a flash evaporation occurs [35]. Here, the working fluid expands into the wet vapour region: part of the saturated liquid evaporates due to the reduction in pressure during expansion. As a result, a two-phase fluid is present in the expander [10].

(a) Cold flow phase change (b) No phase change

Figure 2.5: Temperature-heat diagram [36]

Bianchi et al. [35] stated that in low-grade heat applications, turbines are not suited for two-phase expansion, and they (numerically) looked into a volumetric twin-screw expander as possible alternative in a TLC. They investigated a twin-screw, as it can achieve inter- mediate BIVR values and is capable of high rotating speeds without major performance drops. They looked into the influences of inlet quality of the working fluid and expander rotational speed on expander performance, creating performance maps. In a previous 2.4 Liquid flooded expansion 13

study [37] a parametric analysis was done to provide orders of magnitude and trends for BIVR requirements of positive displacement machines in TLC applications. A slight in- fluence on expander adiabatic isentropic efficiency and an increasing trend with maximum cycle temperature for BIVR was found. They also looked into the influence of different working fluids and quality conditions. Kanno et al. [28, 38] did an experimental study and a modeling study on adiabatic two- phase expansion. A cylinder with moving piston was chosen as the expander. From their experimental results they concluded that piston velocity and diameter have an impact on the generated boiling bubbles and adiabatic efficiency. Efficiency decreases with increasing diameter and velocity. Later they developed a numerical model to predict the pressure change and indicated adiabatic efficiency in two-phase adiabatic expansion.

2.4 Liquid flooded expansion

Friction losses and internal leakages are detrimental to the expander performance [14]. This is where liquid flooded expansion (LFE), i.e. mixing a large amount of oil with the working fluid vapour before it enters the expander, can have a positive influence on ORC efficiency. As lubrication is required during operation of volumetric expanders, friction and leakage losses can be reduced by adding lubrication oil. In addition, the expansion process can be moved towards a quasi-isothermal one, which results in higher power output [15] and favours internal regeneration.

2.4.1 Theoretical research on liquid flooded expansion

Bell et al. [39,40] developed models of a scroll compressor and expander for their applica- tion in a liquid flooded Ericsson cycle2 (LFEC). With a LFEC, quasi-isothermal working processes are achieved. They also acquired experimental data of a set-up to validate their mechanistic models and to tune some parameters. Overall, there was good agreement between the models and the experimental results, with however better results for the com- pressor model.

Ziviani et al. [12, 15] developed models to investigate LFE on SSEs. In a first work a thermodynamic cycle model was developed to check the potential theoretical improvements on thermodynamic performance and work output of an ORC with liquid flooded expansion

2An ideal Ericcson cycle is a thermodynamic cycle characterized by an isothermal compression, isobaric heat addition, isothermal expansion and isobaric heat rejection [40] 2.4 Liquid flooded expansion 14

(ORCLFE) compared to a baseline ORC with internal regeneration. The cycle model was employed over a range of flooding ratios and expander volume ratios in order to simulate and optimize the ORCLFE. The degrees of freedom used were the BIVR, evaporation temperature, pressure ratio over

the expander and the flooding ratio yo (i.e. the ratio of the oil mass flow rate to the refrigerant mass flow rate): m˙ o yo = (2.4) m˙ r The two refrigerants investigated were R245fa and R1233zd(E). The cycle thermodynamic performance was characterized by:

• a First law efficiency:

W˙ W˙ W˙ W˙ η = net = exp pp,r pp,o ORCLF E ˙ ˙ − ˙ − (2.5) Qin Qev,r + Qheater,o

• a Second law efficiency: ηORCLF E ηII,ORCLF E = (2.6) ηCarnot where: Tin,cd,cf ηCarnot = 1 (2.7) − Tin,ev,hf

with Tin,cd,cf the inlet temperature of the cooling liquid in the condenser and Tin,ev,hf the inlet temperature of the heating liquid in the evaporator. Overall, it was found that there was only a limited impact on actual expander specific work, but a significant impact on cycle efficiency. Cycle efficiency was improved using the flooded expansion and internal regeneration with 6.71% for R245fa and 2.90% for R1233zd(E). When achieving similar cycle efficiencies, a slightly higher power output and higher specific work was generated using R1233zd(E) for the same working conditions. The results of the thermodynamic cycle model were then used to design a test set-up (containing an independent lubricating oil loop and internal regeneration) to check the theoretical results in practice. In a second work a semi-empirical model for the expander was incorporated to account for the presence of oil, friction and heat losses and internal volume ratio. The low-GWP working fluids R1234ze(Z), R1233zd(E) and R1336mzz(Z) were used as potential replace- ments for R245fa. It was found that using LFE, the optimal internal volume ratios are generally higher than for dry-running expanders. For a large amount of oil in the working chamber (i.e. liquid flooding conditions), the optimal BIVR lies between 4 and 6. That’s 2.4 Liquid flooded expansion 15 why a SSE with an internal volume ratio of 5.3 was considered, which was already ex- perimentally and numerically characterized in a previous study [14]. The findings of this study concerning efficiency were similar to the ones from their previous one: the use of LFE led to a higher cycle efficiency. Looking at the net power output, there was however a significant difference to the previous study as reduction of the mechanical losses due to proper lubrication results in an increase of the net power output. A second important note was that the improvement strongly depends on the working fluid used: for each working fluid the temperature difference between the critical point and the heat source inlet will be different, which means that the expander volume ratio will differ for each working fluid.

2.4.2 Experimental efforts

As mentioned in the section above, a test set-up was developed for assessing the actual impact on ORC performance and efficiency, when using LFE. Real set-up performance will differ from the theoretical one due to assumptions that were made in the model [12]. For example, in the theoretical model, losses due to flashing of the flooding medium and irreversibilities due to mixing are neglected. Pressure drops over the heat exchangers and line sets were also neglected. For calculating the properties of the oil-refrigerant mixture the ideal mixture model was used. It was further assumed that the liquid and gas flows were in thermal and mechanical equilibrium and that there was a perfect separation process. Perfect separation means that after the expansion process, the vapour phase of the working fluid is entirely separated from the flooding medium by the oil separator. In reality, a fraction of the working fluid will be dissolved in the oil phase and droplets of oil will be entrained in the working fluid vapour phase. For practical applications it is necessary that lubricating oil and refrigerant are partially or wholly immiscible and insoluble to limit the oil pump work, as with increasing solubility and increased flooding, the oil pump specific work increases as well. The optimal oil flooding is therefore directly linked to the solubility fraction, which in turn is linked with the operating conditions at expander discharge. A higher discharge temperature results in a lower solubility fraction (when condenser pressure is imposed). The lubricant oil was considered incompressible and non-volatile and the real gas model3 was used to represent the refrigerant. Furthermore, it was assumed that the expander had a fixed volumetric displacement rate and a filling factor equal to unity (meaning leakages were neglected).

3The real gas model used here is an adaptation of the ideal gas law, based on the van der Waals equation of state [41] 2.5 Comparing organic Rankine cycles 16

Compared to a conventional ORC, separate lines were incorporated in the test set-up for the oil loop, which also contains an oil pump and heater. To be able to mix and separate the working fluid and the oil, a mixer and oil separator were also added. These extra components needed are a disadvantage of an ORCLFE, as they increase the cost and complexity of the installation.

2.5 Comparing organic Rankine cycles

Besides having a reliable test set-up, cycle components and measuring equipment, it is also important to know how to compare ORC settings to each other. There are multiple definitions to quantify efficiency and performance of an ORC. DiPippo [42] found exergy4 efficiency (i.e. second law efficiency) to be the most appropriate indicator to compare thermodynamic performances. Dai et al. [44] stated that thermal efficiency, mainly determined by the temperature level of the heat source and condenser conditions [9], is not a good parameter to assess an ORC as it cannot reflect the ability of converting low- grade heat into usable work. They as well think exergy efficiency is a viable way to evaluate the performance of waste heat recovery cycles. Bianchi et al. [35] state that indicated power, volumetric efficiency, adiabatic-isentropic efficiency and the mass flow rate are significant performance indicators for positive displacement machines. As the performance of an ORC is dependent on operating conditions and the working fluid [45], it is important that the different ORC settings (i.e. with or without LFE, use of different working fluids, etc.) are compared to each other with each setting under its optimal operating conditions.

Considering the efficiency and performance indicators, an assessment should be done to de- termine if adding LFE to an ORC is beneficial, and what the optimal operating conditions are.

4Exergy indicates the energy’s potential for use [43] DESCRIPTION OF THE ORGANIC RANKINE CYCLE TEST FACILITY 17

Chapter 3

Description of the Organic Rankine Cycle Test Facility

This chapter provides information on the ORCLFE set-up that was built (see Figure 3.1), using off-the-shelf components, to asses the impact of oil flooding on ORC efficiency and performance. The four loops (i.e. working fluid, heating, cooling and oil circulation loop) and their components will be discussed (see Figure 3.2). Information about the different types of sensors that were used is present in Appendix A.

3.1 Working fluid loop

The working fluid loop is the part of the ORC unit that contains the refrigerant. The work- ing fluid under consideration is R1233zd(E) (see Section 3.6 for more information about the selection of the fluid). In this subcritical ORC the working fluid is pressurized by a multi-diaphragm pump. In addition, an auxiliary filter is installed in parallel at the discharge of the pump to remove unwanted particles coming from the heat exchangers or expander. After being pressurized, the working fluid can be preheated by a regenerator before entering the evaporator. A bypass is present here in order to be able to test a basic ORC (without regenerator/recu- perator) as well as one with internal regeneration. The fluid evaporates in the evaporator before being mixed with the lubricant oil. The mixture is finally brought to a lower pres- sure by the expander, which is connected to a generator to produce electricity. After being separated from the lubricant oil, the working fluid is cooled and condensed in the con- denser to assure proper functioning of the refrigerant pump, as it is designed to pump a liquid. Initial testing of the set-up revealed that there was insufficient subcooling of the 3.1 Working fluid loop 18 Test set-up Figure 3.1: 3.1 Working fluid loop 19 Simple hydraulic scheme of set-up Figure 3.2: 3.2 Oil circulation loop 20 refrigerant before entering the pump, which resulted in cavitation issues. In order to avoid such cavitation, a subcooler is installed at the suction side of the pump. The cold source of the subcooler is a chiller (see Figure 3.3) with a mixture of water and ethylene-glycol as cooling liquid. In between the condenser and subcooler a liquid receiver is placed. This ensures a minimum liquid level at the suction side, includes an over-pressure protection and serves as a charging point for the installation. Additionally, a bypass option is present at the expander, such that multiple expanders can be tested on the same set-up.

3.2 Oil circulation loop

As the influence of oil addition is tested, an oil circulation loop is also present. SAE 20W50 is tested, which is a mineral multigrade motor oil. Similar to the working fluid, the lubricant oil, which will be mixed with the refrigerant, is pressurized by a multi-diaphragm pump and an auxiliary filter is present parallel to the discharge line of the oil pump. Before the mixing process, the lubricant oil is first heated by an oil heater. To cool down the oil, in order to not damage the pump, an oil cooler is installed which is connected to the oil pump.

Plate heat exchangers are used as condenser, evaporator, recuperator, subcooler and oil heater. An example of such a plate heat exchanger is presented in Figure 3.4. Separation of the two fluids is done by a gravitational oil separator (see Figure 3.5). A static mixer (see Figure 3.6) is employed to ensure a homogeneous mixture of refrigerant vapour and lubricant oil before entering the expander (which will be explained in more detail in Section 3.5).

3.3 Heating loop

The waste heat source and thus the heat source of the evaporator and oil heater is simu- lated by an electric heater with a maximum heating capacity of 250kWe (see Figure 3.7). m3 Therminol 66 is used as heating oil, with a maximum flow rate of 14 h at the maximum temperature of 340°C. Inlet heat source temperature can be controlled, as well as the mass flow rate. 3.3 Heating loop 21

Figure 3.3: Chiller Figure 3.4: Plate heat exchanger

Figure 3.5: Oil separator Figure 3.6: Static mixer 3.4 Cooling loop 22

Figure 3.7: Electric heater (left) with connecting lines to ORC (right)

3.4 Cooling loop

For the cooling, a roof-top air cooled condenser (see Figure 3.8) is used. It has a rated capacity of 480kW at 20°C ambient. The cooling liquid used is a water-ethylene-glycol mixture. The cooling flow rate can be controlled. As the inlet temperature of the cooling liquid is directly related to the outdoor conditions, this parameter cannot be managed.

An overview of the components is given in Table 3.1.

3.5 Single-screw expander

The SSE employed is the one elaborated in the work of Ziviani et al. [46]. A standard single- screw air compressor was converted in order to operate as an expander. More information about the general geometry can be found in Table 3.2. Figure 3.9 illustrates the SSE. Some further alterations were done to improve its performance: 3.5 Single-screw expander 23

Figure 3.8: Roof-top cooling unit

Table 3.1: Overview of components

Component Type Refrigerant pump VERDER D10XKBTHFEHH 2.2kW -1500rpm-230/400V Oil pump VERDER D10EKBTHFEHH 1.5kW -1000rpm-230/400V Expander generator IE3 160M2-2P-15KW 380V-50HZ B35 Liquid receiver FRIGOMEC 25L Oil cooler OCWI-34-3 Condenser SWEP B200THx70 Evaporator SWEP B200THx70 Recuperator SWEP B50Lx80 Subcooler SWEP B80Hx30 Oil heater SWEP B80Hx70 Chiller NESLAB ThermoFlex 10000 Static mixer VMW-040-2-05-C-INJ Oil filter ESK-Schulze F-16L Refrigerant filter Castel 084S

• Seals of the main shaft, side plates and starwheel bearing caps were altered to with- stand the higher temperatures and the refrigerant-type working fluids

• The discharge port of the expander (and thus suction port of the compressor) was enlarged in order to minimize the filling losses 3.6 Refrigerant selection 24

• Stainless steel plates (end plates and front circular cap) and flanges (suction and discharge flanges which were soldered to the piping) were employed

• Copper gaskets were used in every connection in the housing

• For the meshing pair a straight-line profile was used, which is essential for the sealing between the sides of the tooth and mating groove flanks

• A PTFE rotary shaft seal was installed to avoid leakages

Table 3.2: Geometric parameters of the single-screw expander [46]

Engaging ratio*[-] 11/6

Dsr [mm] 122

Dsw [mm] 132 3 Vg,max[cm ] 57.39 BIVR [-] 5.3

Lrotor [mm] 121 * The engaging ratio of a SSE is the ratio of the num- ber of starwheel teeth to the number of main rotor grooves.

Some practical limits apply to the pressure ratio over the expander. The expander suction side can withstand pressures up to 1400 kPa, but for safety reasons this upper limit is set to 1200 kPa. Similarly, the upper pressure limit at the discharge side is set to 300 kPa. The open-drive expander installed on the test set-up (see Figure 3.10) is coupled to an 11 kWe generator, with a torque meter in between. The electricity that is produced is injected into the grid through an inverter, which allows to vary the rotational speed of the expander.

3.6 Refrigerant selection

The different criteria which can be applied when choosing a refrigerant were already high- lighted in Section 2.1 and 2.2. One of the many criteria is the environmental impact of the refrigerant, quantified by its ODP and GWP (see Section 2.1). HFCs, such as the 3.6 Refrigerant selection 25

Figure 3.9: 3D CAD model of single-screw Figure 3.10: Expander installed on set-up expander [46] commonly used R245fa, were introduced as a replacement for ozone depleting substances such as CFCs and HCFCs. Even though they don’t impact the ozone layer, they do have a significant impact on global warming [47]. Therefore HFOs, low in both ODP and GWP, can be considered as a possible replacement for HFCs in refrigeration cycles. Additional criteria for the selection are based on the practical limits of the ORC set-up. As already mentioned in Section 3.5, the expander suction side has an upper pressure limit of 1200kPa. The chosen refrigerant must therefore have an evaporation temperature at 1200kPa within the practical limits of the temperature which can be reached by the elec- tric heater. Similarly, the saturation temperature at 300kPa must also be considered to ensure safe operation at the discharge side of the expander. Furthermore, under-pressure in the condenser must be avoided such that the condenser is protected from air infiltration. Condensation needs to occur at a temperature corresponding to higher pressure levels than the atmospheric one.

Based on the criteria described above, the refrigerant R1233zd(E) is chosen as a suitable replacement for R245fa. In Table 3.3 the properties of both R245fa and R1233zd(E) are summarized. In Figure 2.2 the T-s diagrams of both R245fa and R1233zd(E) are illustrated. From this graph, it can be deduced that both R245fa and R1233zd(E) are dry fluids (see Section 2.2). Therefore, no superheating is required to avoid condensation in the expander. 3.6 Refrigerant selection 26

Table 3.3: Properties of R245fa and R1233zd(E) [48–50]

R245fa R1233zd(E) Chemical name pentafluoropropane Trans-1-chloro-3,3,3-trifluoropropene Type hydrofluorocarbon (HFC) hydrofluoroolefin (HFO) MM [kg/kmole] 134.05 130.5

Tcrit [°C] 154 165.5

pcrit [kPa] 3651 3570

Tsat,12bar [°C] 106.6 97.7

Tsat,3bar [°C] 50.8 45.6

Tsat,1bar [°C] 18.3 15.05 ODP [-] 0 0

GWP100yr [-] 1030 7

Figure 3.11: T-s diagram of R245fa and R1233zd(E) EXPERIMENTAL INVESTIGATION 27

Chapter 4

Experimental Investigation

This chapter elaborates on the experimental investigation that was performed. The first part deals with the calculations performed on the measured variables in order to assess the performance and efficiency of the ORCLFE. In the second part the data reduction and uncertainty analysis applied to the measured data-set is explained. The next part deals with the choice of the number of set-points and settings. Finally, the experimental results are presented and interpreted.

4.1 Variables

4.1.1 Measured variables

The sensors used on the set-up are temperature, pressure, mass flow rate and torque meters. Power measurements come from the variable frequency drives (VFDs) connected to the pumps and expander (see Appendix A). An overview of the measured variables which will be used in further calculations is given in Table 4.1. The numbering corresponds to the locations illustrated in the simplified hydraulic scheme (see Figure 3.2).

Due to limited sensor input modules, not all the sensor positions indicated on the hy- draulic scheme can be used. Therefore, temperature and pressure measurements that are performed at certain locations are assigned to others, neglecting pressure and temperature drops over the connecting lines. For example, the measurement at the refrigerant inlet of the evaporator is replaced by the measurement at regenerator outlet. 4.1 Variables 28

Table 4.1: Overview of measured variables

COOLING LIQUID

T15 [°C] temperature at inlet condenser

HEATING LIQUID

T11 [°C] temperature at inlet evaporator

m˙ ev,hf [kg/s] mass flow rate through evaporator

REFRIGERANT

T5 [°C] temperature at inlet evaporator

T6 [°C] temperature at outlet evaporator

p5 [kPa] pressure at inlet evaporator

p6 [kPa] pressure at outlet evaporator

m˙ r [kg/s] mass flow rate through ORC unit

Npp,r [rpm] pump rotational speed ˙ Wpp,r [W] pump power

τpp,r [Nm] pump shaft torque

LUBRICATION OIL

m˙ o [kg/s] mass flow rate through ORC unit

T9 [°C] temperature at inlet oil heater

T10 [°C] temperature at outlet oil heater

Npp,o [rpm] pump rotational speed ˙ Wpp,o [W] pump power

MIXTURE

T7 [°C] temperature at inlet expander

T8 [°C] temperature at outlet expander

p7 [kPa] pressure at inlet expander

p8 [kPa] pressure at outlet expander

Nexp [rpm] expander rotational speed ˙ Wexp [W] expander power

τexp [Nm] expander shaft torque 4.1 Variables 29

4.1.2 Output variables

From the sensor outputs presented in Table 4.1, efficiencies and other performance param- eters of the ORCLFE can be derived.

Amount of oil

As this study focuses on the influence of a certain amount of oil added to the refrigerant, characterization of this amount of oil is important. Possible variables for characterizing the mass flow rate of oil in the expander are the oil mass fraction and the flooding ratio:

m˙ o xo = (4.1) m˙ o +m ˙ r

m˙ o yo = (2.4) m˙ r

Pressure ratio

In the experiments, the pressure ratio across the expander is also varied:

p7 PR = (4.2) p8 The pressure ratio will have an influence on the performance parameters. The expander has a certain BIVR, corresponding to a certain PR across the expander. If a higher PR is applied, over-expansion will occur. If a lower PR is applied, under-expansion will occur.

Temperature ratio

One way to quantify how close the expansion process lies to an isothermal process, is through the temperature ratio of the expander:

T8 TR = (4.3) T7

ORCLFE cycle efficiency

As already mentioned, a large amount of oil will push the expansion process towards an isothermal process. Isothermal expansion will result in a higher expander outlet temper- ature, which favours the use of internal regeneration and will result in reduced external heat addition. However, adding an increased amount of oil will also result in higher oil 4.1 Variables 30

pumping power. One possible way of comparing the benefits to the disadvantages of oil addition is through the cycle efficiency. The cycle efficiency is defined as:

W˙ η = net ORCLF E ˙ (2.5) Qin Heat added to the refrigerant by the heat source, i.e. through the evaporator and oil heater, is calculated according to:

˙ Qev,r =m ˙ r (h6,r(T6, p6) h5,r(T5, p5)) (4.4) · − ˙ Qheater,o =m ˙ o (h10,o(T10, p10) h9,o(T9, p9)) (4.5) · − ˙ ˙ ˙ Qin = Qev,r + Qheater,o (4.6)

The net amount of power generated by the ORCLFE is:

˙ ˙ ˙ ˙ Wnet = Wexp Wpp,r Wpp,o (4.7) − − with: ˙ 2π Wexp = Nexp τexp (4.8) 60 · ·

The corresponding specific powers are: ˙ Wnet wnet = (4.9) m˙ r +m ˙ o ˙ Wexp wexp = (4.10) m˙ r +m ˙ o

Second Law efficiency

Comparing the ORCLFE efficiency to the theoretical maximum efficiency that can be reached with the cold and hot source present when assuming infinite heat capacity (i.e. the Carnot efficiency), results in the Second Law efficiency:

ηORCLF E ηII,ORCLF E = (2.6) ηCarnot with: T15 ηCarnot = 1 (2.7) − T11 where both the temperatures should be expressed in Kelvin. 4.1 Variables 31

Isothermal efficiency

The isothermal efficiency of the expander expresses how close the power generated by the actual expansion process approximates the power of an isothermal process (i.e. a process with constant temperature): W˙ η = exp exp,isoth ˙ (4.11) Wexp,isoth with: W˙ =m ˙ (h (T , p ) h (T , p ) T (s (T , p ) s (T , p ))) exp,isoth r · 7,r 7 7 − 8,isoth,r 7 8 − 7 · 7,r 7 7 − 8,isoth,r 7 8 +m ˙ v (p p ) o · 7,o · 7 − 8 (4.12)

where T7 should be expressed in Kelvin and with v7,o the specific volume of the lubricant oil at expander inlet conditions. The subscript r indicates that the variable should be calculated for the refrigerant, not for the mixture of refrigerant and oil. An increase in the amount of oil added should thus result in an increase of isothermal efficiency, as the expansion process is pushed towards the isothermal process.

Isentropic efficiency

Proper lubrication results in reduction of friction and leakage losses. Increasing oil addition should thus also increase the isentropic efficiency. The isentropic efficiency expresses how close the power generated by the actual expansion process approximates the power of an isentropic process (i.e. a process with constant entropy):

W˙ η = exp exp,is ˙ (4.13) Wexp,is

with: ˙ Wexp,is = (m ˙ r +m ˙ o) (h7(T7, p7) h8,is(p8, s7)) (4.14) · −

where: s7 = f(T7, p7)

Values for the enthalpies and entropies of the refrigerant were calculated using CoolProp 6.3.0 [48]. More explanation on the calculation of the oil and the mixture properties is provided in Appendix C. 4.2 Data reduction and uncertainty analysis 32

4.2 Data reduction and uncertainty analysis

Before the above mentioned parameters can be calculated from the measurements, the measured data first needs to be filtered. Furthermore, uncertainty intervals need to be determined to express the uncertainty range of the calculated variables.

4.2.1 Determining steady-state

When the ORC is started, it takes some time until transients disappear and steady-state operation is reached. Theoretically, steady-state operation means that the different prop- erties of the set-up don’t vary in time anymore. However, as data is taken from an exper- imental set-up, this condition will never be reached. There are uncertainties associated to the sensors and measurement equipment. A more practical approach to determine steady- state operation is described in detail by Lecompte et al. [51]. This steady-state detection procedure consists of:

• Identifying a representative steady-state zone: This zone is reached after ap- proximately one hour of operation. If the identified zone does not comply with the criteria for steady-state set out by Woodland et al. [52], a subsequent zone should be used.

• Calculation of reference standard deviations: The standard deviations of cer- tain properties in this specific representative steady-state zone are calculated (based on 600 samples). These will serve as the reference standard deviations of those prop- erties and will be used further to identify the actual steady-state zones. 12 reference properties were chosen, which are: flow rate of cooling flow, flow rate of refrigerant, expander inlet and outlet temperature, expander inlet and outlet pressure, heating flow rate through evaporator, temperature of heating liquid before and after evap- orator, temperature of refrigerant after condenser, expander and refrigerant pump rotational speed.

• Calculation of forward-moving standard deviations: For every sample of these 12 properties, a forward-moving standard deviation is calculated.

• Identifying steady-state zones: A sample k is considered as steady-state, when the moving standard deviations of all the 12 reference properties are smaller than or equal to the threshold (which is 2.5 times the reference standard deviation):

σk 2.5 σref (4.15) 6 · 4.2 Data reduction and uncertainty analysis 33

A steady-state zone exists when subsequent data-samples meet this requirement.

• Calculation of steady-state points: Finally, the steady-state points are calculated as the average of the samples in the steady-state zones. All the steady-state points of the performed measurements are given in Appendix B.

4.2.2 Uncertainty analysis

The sensors which are used to measure the different properties have a limited accuracy. The actual value will lie in an interval around the measured one:

X = Xm δX (4.16) ±

δX is the uncertainty of the measurement corresponding to a certain confidence interval. The used sensors and their uncertainties are presented in Table 4.2. These uncertain- ties correspond to a 95%-confidence interval approximately, meaning that if the same measurement would be performed 100 times, the actual value, which is fixed, will lie in approximately 95 of the calculated intervals. More information on the sensors itself can be found in Appendix A. Some uncertainties are represented by their relative value. To get the absolute uncertainty, the reported values need to be multiplied with the measured quantity.

Table 4.2: Sensor uncertainties

Measured variable Type Uncertainty

m˙ r, m˙ o CFM 0.09% ± m˙ hf Pressure orifice 1% ± T1 16 RTD 0.2°C − ± p1 16 APS 1.6kP a − ± τexp,τpp,r Torque meter 0.1% ± Nexp, Npp,r, Npp,o PLC Neglected

As described in Section 4.1.2, the variables of interest are calculated from the measured ones. This however means that there is also an uncertainty on the calculated values, as the sensor uncertainty propagates throughout the calculations. The absolute uncertainty 4.2 Data reduction and uncertainty analysis 34

U of a variable K (which is a function of the variables Xi; i=1..N ) is calculated according to [53]: N δK 2 UK = ( UX ) (4.17) v δX · i u i=1 i uX t Amount of oil

For the oil mass fraction and flooding ratio, which are both a function of the refrigerant and oil mass flow rate, this results in:

δx δx o 2 o 2 Ux = ( Um˙ ) + ( Um˙ ) (4.18) o δm˙ r δm˙ o r r · o · m˙ m˙ U = ( o )2 U 2 + ( r )2 U 2 xo − 2 m˙ r 2 m˙ o (4.19) s (m ˙ o +m ˙ r) · (m ˙ o +m ˙ r) · δy δy o 2 o 2 Uy = ( Um˙ ) + ( Um˙ ) (4.20) o δm˙ r δm˙ o r r · o · m˙ 1 U = ( o )2 U 2 + ( )2 U 2 yo − 2 m˙ r m˙ o (4.21) s m˙ r · m˙ r ·

Pressure ratio

The pressure ratio has an uncertainty of:

1 p 2 2 7 2 2 UPR = ( ) Up + (− ) Up (4.22) p · p 2 · r 8 8 Temperature ratio

The uncertainty on the temperature ratio of the expander is:

1 T 2 2 8 2 2 UTR = ( ) UT + (− 2 ) UT (4.23) s T7 · T7 ·

ORCLFE cycle efficiency

The uncertainty on the ORCLFE cycle efficiency is calculated according to:

1 W˙ 2 2 net 2 2 UηORCLF E = ( ) U ˙ + (− ) U ˙ (4.24) ˙ Wnet ˙ 2 Qin s Qin · Qin · 4.2 Data reduction and uncertainty analysis 35 with: 2 2 2 U ˙ = U + U + U (4.25) Wnet W˙ exp W˙ pp,r W˙ pp,o q 2π 2π 2 2 2 2 U ˙ = ( Nexp) U + ( τexp) U (4.26) Wexp 60 τexp 60 Nexp r · · 1 W˙ U = ( )2 U 2 + ( − net )2 (U 2 + U 2 ) wnet W˙ m˙ r m˙ o (4.27) s m˙ r +m ˙ o · net m˙ r +m ˙ o ·

1 W˙ U = ( )2 U 2 + ( − exp )2 (U 2 + U 2 ) wexp W˙ m˙ r m˙ o (4.28) s m˙ r +m ˙ o · exp m˙ r +m ˙ o ·

2 2 U ˙ = U + U (4.29) Qin Q˙ ev,r Q˙ heater,o q 2 2 2 2 2 U ˙ = (h6,r h5,r) U +m ˙ (U + U ) (4.30) Qev,r − · m˙ r r · h6,r h5,r q 2 2 2 2 2 U ˙ = (h10,o h9,o) U +m ˙ (U + U ) (4.31) Qheater,o − · m˙ o o · h10,o h9,o The uncertainties of the enthalpies,q which are determined by temperature and pressure, are calculated according to:

δh δh 2 2 2 2 2 Uh = ( ) Up + ( ) UT + Uh,corr (4.32) s δp · δT ·

δh δh For the refrigerant, the derivatives δp and δT are calculated using CoolProp [48]. The additional uncertainty Uh,corr is present due to the fact that CoolProp calculates the ther- mophysical properties based on models. The values for the latter can be found in Appendix C. For the oil, the first term in Equation 4.32 equals zero as the oil properties are only in function of temperature, not pressure, see Appendix C.

Second Law efficiency

The uncertainty on the Second Law efficiency is calculated as:

1 η U = ( )2 U 2 + ( ORCLF E )2 U 2 ηII,ORCLF E ηORCLF E − 2 ηCarnot (4.33) s ηCarnot · ηCarnot · with: 1 T U = ( )2 U 2 + ( 15 )2 U 2 ηCarnot T − 2 T (4.34) s T11 · T11 · 4.2 Data reduction and uncertainty analysis 36

Isothermal efficiency

For the uncertainty on expander isothermal efficiency, it holds:

1 W˙ 2 2 exp 2 2 Uηexp,isoth = ( ) U ˙ + ( − ) U ˙ (4.35) ˙ Wexp ˙ 2 Wexp,isoth s Wexp,isoth · Wexp,isoth ·

˙ Wexp,isoth is a function of m˙ r, m˙ o, T7, p7 and p8. The errors on the correlations for enthalpies, entropies and specific volume should also be taken into account, resulting in: ˙ ˙ ˙ δWexp,isoth 2 2 δWexp,isoth 2 2 δWexp,isoth 2 2 ( ) UT + ( ) Up + ( ) Up v δT7 · δp7 · δp8 · u u δW˙ δW˙ δW˙ u + ( exp,isoth )2 U 2 + ( exp,isoth )2 U 2 + ( exp,isoth )2 u m˙ r m˙ o u δm˙ r · δm˙ o · δh7,r · U ˙ = (4.36) Wexp,isoth u ˙ ˙ u δWexp,isoth δWexp,isoth uU 2 + ( )2 U 2 + ( )2 u h7,r,corr δh h8,isoth,r,corr δs u 8,isoth,r · 7,r · u ˙ ˙ u 2 δWexp,isoth 2 2 δWexp,isoth 2 2 uUs + ( ) Us + ( ) Uv u 7,r,corr δs · 8,isoth,r,corr δv · 7,o,corr u 8,isoth,r 7,o t The expressions for the partial derivatives are: ˙ δWexp,isoth δh7,r δh8,isoth,r δs7,r δs8,isoth,r =m ˙ r (( ) (s7,r s8,isoth,r) T7 ( )) δT7 · δT7 − δT7 − − − · δT7 − δT7 δv7,o +m ˙ o (p7 p8) · δT7 · − (4.37) ˙ δWexp,isoth δh7,r δs7,r δv7,o =m ˙ r ( T7 ) +m ˙ o ( (p7 p8) + v7,o) (4.38) δp7 · δp7 − · δp7 · δp7 · − ˙ δWexp,isoth δh8,isoth,r δs8,isoth,r =m ˙ r ( + T7 ) m˙ o v7,o (4.39) δp8 · − δp8 · δp8 − · ˙ δWexp,isoth = h7,r h8,isoth,r T7 (s7,r s8,isoth,r) (4.40) δm˙ r − − · − ˙ δWexp,isoth = v7,o (p7 p8) (4.41) δm˙ o · − ˙ δWexp,isoth =m ˙ r (4.42) δh7,r ˙ δWexp,isoth = m˙ r (4.43) δh8,isoth,r − 4.2 Data reduction and uncertainty analysis 37

˙ δWexp,isoth = m˙ r T7 (4.44) δs7,r − · ˙ δWexp,isoth =m ˙ r T7 (4.45) δs8,isoth,r · ˙ δWexp,isoth =m ˙ o (p7 p8) (4.46) δv7,o · −

Isentropic efficiency

The uncertainty on the expander isentropic efficiency: ˙ ˙ ˙ δWexp,is 2 2 δWexp,is 2 2 δWexp,is 2 2 ( ) UT + ( ) Up + ( ) Up v δT7 · δp7 · δp8 · u ˙ ˙ ˙ u δWexp,is 2 2 δWexp,is 2 2 δWexp,is 2 U ˙ = u + ( ) U + ( ) U + ( ) (4.47) Wexp,is u m˙ r m˙ o u δm˙ r · δm˙ o · δh7 · u ˙ ˙ u 2 δWexp,is 2 2 δWexp,is 2 2 uUh7,corr + ( ) Uh + ( ) Us7,corr u δh · 8,is,corr δs · u 8,is 7 t with the partial derivatives: ˙ δWexp,is δh7 δh8,is δs7 = (m ˙ r +m ˙ o) ( ) (4.48) δT7 · δT7 − δs7 · δT7 ˙ δWexp,is δh7 δh8,is δs7 = (m ˙ r +m ˙ o) ( ) (4.49) δp7 · δp7 − δs7 · δp7 ˙ δWexp,is δh8,is = (m ˙ r +m ˙ o) (4.50) δp8 − · δp8 ˙ δWexp,is δh7 δxo δh8,is δxo = h7 h8,is + (m ˙ r +m ˙ o) ( ) (4.51) δm˙ r − · δxo · δmr − δxo · δmr ˙ δWexp,is δh7 δxo δh8,is δxo = h7 h8,is + (m ˙ r +m ˙ o) ( ) (4.52) δm˙ o − · δxo · δmo − δxo · δmo ˙ δWexp,is =m ˙ r +m ˙ o (4.53) δh7 ˙ δWexp,is = (m ˙ r +m ˙ o) (4.54) δh8,is − ˙ ˙ δWexp,is δWexp,is δh8,is = (4.55) δs7 δh8,is · δs7

The uncertainties corresponding to the steady-state points are presented in Appendix B. 4.3 Experimental matrix 38

4.2.3 Required measuring interval

The trend of the efficiencies and performance indicators in function of the flooding ratio and pressure level is investigated. However, in order to correctly assess the influence of oil flooding or pressure ratio over the expander, the measuring interval should be large enough such that the measurements are representative. This means that the difference between subsequent set-points should be larger than the uncertainty on those set-points.

Equation 4.21 can be simplified, knowing that Um˙ = 0.0009 m˙ o and Um˙ = 0.0009 m˙ r: o · r · m˙ o Uyo = 0.00127 (4.56) · m˙ r The highest absolute uncertainty thus occurs for maximal oil mass flow rate and mini- mal refrigerant mass flow rate. As this ratio will never be larger than one, choosing oil flooding ratio set-points with an interval larger than 0.00127 will guarantee representative measurements.

Simplification of Equation 4.2, with Up=1600Pa, gives: 1600 p 7 2 UPR = 1 + ( ) (4.57) p · p 8 r 8 The highest absolute uncertainty will occur for highest expander inlet pressure and lowest expander outlet pressure. As already mentioned, there are some restrictions to the pressures which can be applied over the expander. The maximum allowed value for p7 is 1200kPa and by taking an absolute minimum of 100kPa for p8, the maximum uncertainty on PR is 0.193.

The above calculations of the measuring intervals only take into account the uncertainties of the measuring equipment. However, the accuracy of the settings themselves needs to be incorporated as well. The flooding ratio and pressure ratio cannot be controlled independently, but are a function of different directly controllable variables such as pump and expander rotational speed. The precision with which these variables can be tuned also effects the measuring interval that can be achieved.

4.3 Experimental matrix

Measurements are performed in function of flooding and pressure ratio. To determine the number and range of set-points, not only the measuring interval but also the time required 4.4 Altering set-points 39 and available to perform measurements needs to be taken into account. In addition, also the possible working ranges of the set-up and working fluid must be respected when choosing the set-points. Based on these considerations, an experimental matrix is outlined. Pressure ratio and oil flooding ratio will be varied to investigate their influence on the performance of the expander (see Table 4.3). When the flooding ratio is varied, some variables are held constant: Nexp, Tin,exp, pin,exp and Npp,r.

Table 4.3: Parameters yo [-] 0 0.15 0.3 0.5 PR[-] 3.5 5 6.5

In reality, it is however difficult to reach the exact settings mentioned in the table. There- fore, the experimental matrix should be considered more as a guideline for the settings of the ORC.

4.4 Altering set-points

With the exception of expander and refrigerant pump rotational speed, the different set- tings such as Tin,exp and pin,exp and the ones presented in Table 4.3, cannot be controlled independently and thus need to be obtained by altering other parameters.

Pressure ratio

The pressure ratio across the expander is a direct function of the pressure before and after the expander. The pressure ratio can thus be altered using two different methods, i.e. changing the inlet pressure of the expander or changing its outlet pressure. The latter is determined by the condensing pressure. Therefore, adjusting the condenser settings will result in adjusting the PR. The only controllable parameter at the condenser is the mass flow rate of cooling medium. An increase in cooling mass flow rate will result in a decrease in condensing pressure, and vice versa, as the heat balance (see Equation 4.58; neglecting heat losses to the environment) between refrigerant and cooling liquid is always met.

m˙ cf cp,cf (Tout,cd,cf Tin,cd,cf ) =m ˙ r (hin,cd,r(pcd,r) hout,cd,r(pcd,r)) (4.58) · · − · − 4.4 Altering set-points 40

Flooding ratio

The flooding ratio is directly determined by the mass flow rate of oil and mass flow rate of refrigerant. The latter is set by the refrigerant pump and expander rotational speed. As both these variables are fixed, the flooding ratio will need to be altered by altering the mass flow rate of oil, which is done by altering the oil pump rotational speed.

Expander inlet temperature

Expander inlet temperature can be controlled by controlling the heat source. By adjusting the mass flow rate or inlet temperature of the heating liquid, a certain expander inlet temperature can be set. This follows from a heat balance across the evaporator.

Expander inlet pressure

Analogously to the temperature, the expander inlet pressure can also be set by adjusting the heat source conditions. Altering the pump rotational speed or the expander speed will also have an influence on the expander inlet pressure.

On the used set-up (see Section 3), the expander outlet pressure can not be altered signifi- cantly by changing the condenser settings. Therefore, a new PR is set by altering the pump and expander speed, and the heat source conditions. Figure 4.1 provides an overview of the order at which the different ORC parameters need to be altered to obtain the different set-points. As some are influenced by others, iteration will need to be applied to acquire the correct settings. Changing the flooding ratio will for example alter the pressure and temperature at the inlet of the expander. The dependence of the different settings on each other will be illustrated through simulations (see Section 5.3).

Figure 4.1: Attaining settings 4.5 Results 41

4.5 Results

Set-up related restrictions

Measurements were performed in function of different pressure and flooding ratios. The steady-state points and their uncertainties are tabulated in Appendix B. The range of flooding ratios that could be achieved was however limited. When the rotational speed of the oil pump was increased above 250 rpm, resonance occurred in the oil piping. Due to time limitations, this issue could not be resolved and experiments were therefore performed with a maximum oil pump speed of 250 rpm. In total, 14 steady-state points were acquired with flooding ratios between 0 and 0.24 and pressure ratios varying between 3.36 and 4.29. The steady-state points and their uncertainties are tabulated in Appendix B.

The logging of the expander and pump torques and powers did not occur correctly. There- fore, when a steady-state operation was reached, the powers had to be read directly from the drives, where they were represented as a percentage of the nominal powers.

T-s diagram

Figure 4.2 is a T-s diagram based on live measurements (refrigerant: orange, hot source: red, cold source: blue). The cycle begins with the expansion of the refrigerant (1-2). After the expander, the refrigerant flows to the recuperator where it exchanges some of its heat (2-3). The refrigerant enters the condenser and subcooler subsequently and is cooled (3-4). Afterwards, it is pressurized by the pump (4-5) and preheated by the recuperator (5-6). After the recuperator the refrigerant enters the evaporator, where the fluid is evaporated and superheated (6-1). The superheat after the evaporator of this example is limited. The subcooling however is significant. The refrigerant line crosses the cold source line, which is never possible when only a condenser is used. The further subcooling of the refrigerant is realized by the chiller. This significant amount of subcooling is not necessary. The chiller should be set such that an amount of subcooling is reached which is just sufficient for the pump to be able to pressurize the refrigerant.

Heat balance check

One way to check whether some heat inputs, temperatures and pressures are logged cor- rectly is to evaluate the heat balance over the evaporator. The heat added to the refrigerant should be equal to the heat input of the electric heater (when neglecting losses) in the evap- orator. The heat added to the refrigerant is calculated according to Equation 4.4. The 4.5 Results 42

Figure 4.2: T-s diagram of experimental results head added by the electric heater is calculated according to:

˙ Qev,hf =m ˙ hf cp,hf (Tev,hf,in Tev,hf,out) (4.59) · · −

From Table 4.4, which represents the results for the 14 acquired steady-state points, it can be deduced that something went wrong with the logging of the heat input to the refrigerant. Firstly, the heat taken up by the refrigerant is larger than the heat added to it, which is not possible. The reverse is possible, due to heat losses occurring in the evaporator. Secondly, the difference between the two heat rates is too large to be a consequence of measuring errors. Therefore, the logging of the data should be checked to see whether and where a mistake is made. A similar evaluation should be performed for the oil heater and condenser. However, due to the limited number of sensor input modules, not all the temperatures and pressures necessary for this evaluation were present.

The set-up is not isolated, and therefore heat can escape to the environment. An estimation of the overall heat losses can be calculated by evaluating the overall heat balance of the cycle on the refrigerant and oil side:

˙ ˙ ˙ ˙ ˙ ˙ ˙ ˙ Qlosses = Wpp,r + Wpp,o + Qev,r + Qheater,o Wexp Qcooling,r Qcooling,o (4.60) − − − 4.5 Results 43

Table 4.4: Heat balance evaporator

˙ ˙ ˙ Point Qev,hf [kW] Qev,r [kW] ∆Q [kW] 1 60.8 78.2 -17.4 2 62.2 77 -14.8 3 63.6 77.6 -14 4 63.1 77.1 -13.9 5 58.9 69.1 -10.2 6 59 68.1 -9.1 7 59.1 67.9 -8.8 8 58.2 67.2 -9 9 58 66.9 -8.9 10 57.6 66.5 -8.9 11 56.8 65.7 -8.9 12 74.6 79.5 -4.9 13 73.7 80.1 -2.9 14 73.3 80.8 -3.1

The cooling heat rate of the refrigerant is determined by combining the effect of the chiller and condenser. The oil cooling should also be taken into account, but this cooling can not be measured on the set-up. Therefore, it is approximated by the oil heating rate. The results are tabulated in Table 4.5.

Table 4.5: Heat losses

˙ ˙ Point Qlosses [kW] Point Qlosses [kW] 1 8.3 8 4.9 2 7.3 9 4.9 3 7.5 10 4.8 4 7.4 11 4.5 5 6.7 12 3.3 6 5.8 13 1.1 7 5.3 14 1 4.5 Results 44

Influence on expander and pumping powers

The net specific power generated by the cycle decreases when the flooding ratio is increased (Figure 4.3). For higher flooding ratios, the decrease in net power is more moderate. The decrease in net power is due to the combined effect of a decrease in generated expander power and an increase in oil pumping power up to a certain flooding ratio (Figure 4.4-4.6). The decrease in expander power is dominant as the oil pumping power even decreases from a certain flooding ratio. The rotational speed of the oil pump is limited to 250 rpm, which is far below its nominal rotational speed. Therefore, investigating higher flooding ratios, with more favourable oil pump rotational speeds, can be interesting. The impact of flooding on the refrigerant pump is limited (Figure 4.5).

Figure 4.3: wnet in function of y Figure 4.4: wexp in function of y

˙ Figure 4.5: W˙ pp,r in function of y Figure 4.6: Wpp,o in function of y 4.5 Results 45

Influence on isentropic efficiency

Isentropic efficiency drops with increasing flooding ratio (Figure 4.7). In theory, oil ad- dition should increase isentropic efficiency as leakage and friction are reduced by proper lubrication. In reality, this is however not the case and the addition of oil increases the losses in the expander.

Figure 4.7: ηexp,is in function of y

Influence on isothermal expansion

Figure 4.8 and 4.9 illustrate the effect of flooding ratio on the temperature ratio over the expander and expander outlet temperature, for different pressure ratios. Increasing flooding ratio has no observable impact on the expander outlet temperature. The slight increase that can be observed is due to the fact that during the measurements, it was difficult to keep the temperature at expander inlet at a constant level, for different flooding ratios. At some flooding ratios, the expander inlet is slightly higher, resulting in a more favourable temperature ratio for the same temperature difference.

Figure 4.10 illustrates the evolution of expander isothermal efficiency. The margin of error on these calculations is too large to deduce a trend. A further explanation on this is given in Appendix C. 4.5 Results 46

Figure 4.8: TR in function of y Figure 4.9: Texp,out in function of y

Figure 4.10: ηexp,isoth in function of y

Influence on cycle efficiency

The trend of the cycle and Second Law efficiency (Figure 4.11 and 4.12) is similar to the trend of the net specific power (Figure 4.3). This indicates that the increase in heat added to the cycle for heating the oil, is less important compared to the decrease in net power. The cycle efficiencies are low, even without flooding, due to the bad efficiencies of the pumps. When cycle efficiency is calculated based only on the expander power, it increases with up to 2% (Figure 4.13). 4.5 Results 47

Figure 4.11: ηORCLF E in function of y Figure 4.12: ηII,ORCLF E in function of y

Figure 4.13: ηORCLF E,exp in function of y

The heat exchanged in the recuperator is independant of the oil flooding ratio (Figure 4.14). As already mentioned, there is no significant impact on expander outlet temperature when the flooding ratio is increased. Therefore the potential for internal regeneration is unaltered. 4.5 Results 48

Figure 4.14: Q˙ recuperator in function of y

Additional required research

Theoretically, expander and cycle efficiencies should increase and the expansion process should approximate an isothermal process better when liquid flooding is applied. How- ever, based on the results described above, this is not necessarily the case in reality. In order to make a final conclusion on the effect of liquid flooding on ORC performance, ad- ditional research is required. The oil which is used for the experiments is a standard motor oil. An oil (such as ACD100FY) which is more suited to be used in refrigeration cycles should be tested. The type of oil could have an influence on ORCLFE performance, as dif- ferent oils have different viscosities, different solubility effects with the refrigerant, etc. It is also important to investigate the limits of the set-up w.r.t. the pressure ratios and other performance altering parameters. The highest PR that was tested, is around 4.3. Higher PRs might lead to higher efficiencies as the under-expansion losses are reduced. Higher temperature levels and expander speeds also increase power output and performance. Fig- ure 4.15 illustrates this effect. The net specific power is drawn for measurement sets with the same PR (around 4.1) but different expander speeds and temperature levels. Further- more, higher flooding ratios should be tested as well. In order to do this, the resonance issue should however be solved first. 4.5 Results 49

Figure 4.15: wnet in function of y for same PR THEORETICAL ORC MODELLING 50

Chapter 5

Theoretical ORC modelling

Multiple theoretical ORC models have been developed at Ghent University. This chapter focuses on their general structure, how they reach a solution and what adjustments need to be made to incorporate liquid flooding. The last part of this chapter is an illustration on how the models can be used to predict certain trends of an ORC or ORCLFE.

5.1 ORC model description

ORC boundary conditions

In order to run the simulations, a number of input variables must be defined. These are either boundary conditions that cannot be controlled (e.g. ambient temperature) or setpoint values that can be controlled. The only variables that can be controlled directly are the temperature of hot oil, the mass flow rate of hot oil, the mass flow rate of cooling medium and the pump and expander rotational speed.

Components initialization

For every component of the ORC a detailed simulation model is necessary. Dependant on the type of pump or expander present in the set-up, different correlations are used to calculate their output variables. These correlations can be found in literature or can be derived from experimental results. The heat exchangers (i.e. the evaporator, condenser and, in the case of internal regeneration, the recuperator) need to be initialized with their specific characteristics such as plate thickness, plate conductivity, number of plates and inclination angle for plate heat exchangers. These specifications are needed to calculate the heat transfer. The liquid receiver is characterized by an internal diameter and height. The 5.1 ORC model description 51 copper piping connecting the different components is characterized by its length, internal and external diameter, its thermal conductivity and the thermal conductivity and thickness of the insulation material surrounding it. In the models, pressure and temperature drops over these lines are neglected.

Property and output calculations

When all the inputs are present, outputs of components and properties of fluids can be calculated. All the fluid properties are calculated using CoolProp [48]. Dependent on the type of heat transfer phenomenon that occurs (e.g. phase change or not), different correlations are used to determine the convective heat transfer coefficient and pressure drops of the heat exchangers. The -NTU method is employed to calculate the different outlet temperatures and heat transfer rates.

Initial conditions

Initial ORC conditions are required to run the simulations. Guess values for the evapo- ration and condenser outlet pressures and enthalpies are chosen based on heating liquid and cooling liquid inlet conditions, respectively. A certain initial amount of subcooling is assumed at the exit of the condenser. Based on these assumptions, inlet and outlet pressures and outlet enthalpies can be calculated for the pump and expander, resulting in initial values for pump and expander power and mass flow rate. The outputs then serve as inputs for other components such as the evaporator and condenser.

Mass and heat balance

After one iteration, all the ORC cycle variables are calculated. Based on these, a mass and heat balance can be calculated which are updated after every performed iteration. The mass balance compares the mass flow rate between the pump and expander. The heat balance compares the pump power input and evaporator heat input to the expander power output and condenser heat output.

Solution method

After every iteration, the input variables (e.g. evaporating and condensing temperature or pressure) are updated. A solution to the simulation is reached when the residuals on the mass and heat balance reach their threshold value or when the relative difference between two subsequent iterations is smaller than a certain tolerance value. 5.2 Required adjustments for an ORCLFE model 52

5.2 Required adjustments for an ORCLFE model

The model described above can be applied to basic ORCs. To incorporate liquid flooding, some adjustments need to be made to this basic model.

Additional components

As an ORCLFE contains a separate oil circulation loop, additional components such as an oil pump, oil heater, mixer, separator and piping in between need to be simulated as well. The simulation of the pump, heater and piping is similar to the ones in the refrigerant loop. To reduce complexity, the mixing and separation can be assumed ideal, at constant temperature and pressure. If a more detailed model is desired, solubility effects of refrigerant in oil can be incorporated if these correlations are known. Simulating ideal mixing means that the oil enters the mixer at the same temperature and pressure as the refrigerant. The same applies to the exit of the separator. If the set-up described in Chapter 3 is simulated, a subcooler and its chiller should also be incorporated.

Additional boundary conditions

The addition of these components also requires extra boundary conditions that need to be specified. The oil pump rotational speed needs to be defined, as well as the cooling rate of the oil cooler. Based on this cooling rate, the oil pump mass flow rate and inlet temperature, the oil pump outlet temperature can be calculated, which is an input of the oil heater. The mass flow rate of heating liquid through the oil heater should be a boundary condition as well.

Fluid properties

Not every lubricant oil is incorporated in CoolProp [48]. Therefore, separate correlations need to added to calculate the thermodynamic properties of the oil. Correlations for ACD100FY are presented in Appendix C. Moreover, not all the properties such as thermal conductivity are present in CoolProp for certain refrigerants. If, for example, R1233zd(E) is used as refrigerant, correlations need to be included for this refrigerant or a different library needs to be used. Correlations for calculating the oil-refrigerant mixture properties can be found in Appendix C. 5.3 ORC model simulations 53

Alteration of balances

The mass and heat balance, which determine when a solution is reached, need to be adjusted. The mass balance should also include the injection of oil. The heat balance should incorporate the oil cooling and heating and the oil pumping power.

5.3 ORC model simulations

In Section 4.4 trends in ORC behaviour are based on theoretical reasoning. This section illustrates how simulations can validate this reasoning and estimate different behaviours. The reference boundary conditions that are used, which are needed to run the simulations, are tabulated in Table 5.1. When one variable is changed, the other ones are set to these reference values. The simulations are run for an ORC with internal regeneration and a fixed subcooling of 1°C, with R245fa as refrigerant, Therminol 66 as heating liquid and water as cooling liquid. Under these conditions, the simulations were more stable, resulting in a wider possible range of the altered variables.

Table 5.1: Boundary conditions

fpp,r [Hz] 30

Nexp [rpm] 3000

Tin,cd,cf [°C] 28.59

pin,cd,cf [kPa] 200

m˙ cf [kg/s] 4.166

Tin,ev,hf [°C] 125

pin,ev,hf [kPa] 140

m˙ hf [kg/s] 3.4

Influence of cooling liquid flow rate

As already mentioned in Section 4.4, changing the flow rate of the cooling liquid should theoretically lower the condensing pressure and thus the pressure at the outlet of the expander. Figure 5.1 illustrates this influence. There is a decreasing trend of the pressure with increasing flow rate, however the effect is minimal. Moreover, the pressure at expander inlet also decreases with increasing flow rate (see Figure 5.2). This indicates that solely 5.3 ORC model simulations 54 increasing the cooling flow rate, will not result in an increase of PR. Countermeasures to the increase in expander inlet pressure need to be taken as well.

Figure 5.1: pout,exp in function of m˙ cf

Figure 5.2: pin,exp in function of m˙ cf

Influence of heating liquid flow rate

In Figure 5.3, the influence of the heating liquid flow rate on the expander inlet tem- perature is illustrated. Increasing or decreasing the flow rate does not have a significant impact on the temperature. The simulated evaporator is oversized and is able to increase the temperature of the refrigerant up to its theoretical maximum (i.e. 125°C; the inlet temperature of the heating liquid), even at low flow rates. Therefore, increasing the flow rate even further has no impact. 5.3 ORC model simulations 55

Figure 5.3: Tin,exp in function of m˙ hf

Influence of heating liquid inlet temperature

An increase in heating liquid inlet temperature results in a significant increase in expander inlet temperature and a minimal increase in expander inlet pressure (see Figure 5.4 and 5.5). Increasing the heating inlet temperature increases the working fluid temperature at the outlet of the evaporator. This thus results in an increase in expander inlet temperature.

Figure 5.4: Tin,exp in function of Tin,ev,hf

Influence of expander rotational speed

Higher expander rotational speeds lead to lower pressure ratios (see Figure 5.6). The reduction in PR is a consequence of the reduction in expander inlet pressure (Figure 5.7). The mass flow rate through the expander increases with expander rotational speed (Figure 5.8). 5.3 ORC model simulations 56

Figure 5.5: pin,exp in function of Tin,ev,hf

Figure 5.6: PR in function of Nexp Figure 5.7: pin,exp in function of Nexp

Figure 5.8: m˙ r in function of Nexp

Influence of pump rotational speed

An increase in pump rotational speed will increase the refrigerant mass flow rate (see Figure 5.9) as the mass flow rate through the pump is directly determined by the rotational speed. Increasing the pump rotational speed will also increase the pressure ratio over the expander (see Figure 5.10). This is due to the large increase in pressure at expander inlet side (Figure 5.3 ORC model simulations 57

5.11). The pressure after the expander also increases with rotational speed, but this effect is only limited.

Figure 5.9: m˙ r in function of fpp

Figure 5.10: PR in function of fpp Figure 5.11: pin,exp in function of fpp

In the models, the mass flow rate through the expander and pump are not only determined by the expander and pump rotational speed, respectively. In addition, the pressure at the inlet and the pressure ratio over the expander or pump have an influence as well. A solution to the simulation is reached when the mass flow rates approximate each other. Therefore, an increase or decrease in one of the mass flow rates (by changing the rotational speed of the corresponding component) will result in changes to the pressures, in order to reach a solution. 5.3 ORC model simulations 58

Experimentally obtained trends

As an illustration, some trends which were obtained experimentally on the described set-up (see Section 3) are presented here. The steady-state data points and their uncertainties are tabulated in Appendix B. It should be noted that the ORC which was modeled differs from the real ORC.

Changing the pump rotational speed of the real set-up has a different effect compared to the modeling: pressure ratio decreases with increasing pump rotational speed. This is due to a significant increase in expander outlet pressure. The condenser compensates for any increase or decrease in heat input or net power input as a result of the increased mass flow rate, by increasing or decreasing its condensing pressure. The effect on refrigerant mass flow rate is however similar to the modeling. An increase in rotational speed results in an increase in mass flow rate through the ORC. Changing the rotational speed of the expander has no real impact on the mass flow rate.

The influence of the heating liquid inlet temperature corresponds to the one obtained through modeling: an increase results in an increase in both expander inlet pressure and temperature. When the mass flow rate of heating liquid is changed, the inlet temperature is altered as well. The evaporator used in the real set-up is not oversized. Therefore, increasing heating liquid flow rate still has an impact on the heat transferred to the refrig- erant.

Simulations can thus be used to estimate some trends of real set-ups. However, it should be kept in mind that simulations are not 100% correct, and real set-up behaviour will be dif- ferent from results obtained through modeling. This emphasizes the need for experimental validation. 5.3 ORC model simulations 59

Figure 5.12: Experimentally obtained trends CONCLUSION 60

Chapter 6

Conclusion

The goal of this master thesis was to perform an experimental investigation on an ORCLFE.

The relevance of using ORCs and other low-temperature energy conversion systems in power generation is explained. A literature study was performed to understand the basic ORC functioning and how its efficiency and performance can be improved, more specif- ically by the application of liquid flooded expansion. Therefore, the expansion process was discussed in more detail and a theoretical background on liquid flooded ORCs was provided. LFE theoretically results in an increase in isentropic efficiency, as leakage and friction losses are reduced due to proper lubrication. In addition, more heat is available at the expander outlet, favouring the use of internal regeneration which is beneficial for thermal efficiency. It concludes with a summary of the research efforts on ORCLFEs.

Next, a thorough description of the test facility, used to perform the measurements, was given. In addition to a basic ORC, a separate oil circulation loop (with oil heater, pump and cooler) is included. A mixer and separator are added as well, to mix and separate the oil-refrigerant mixture, respectively. Information about the used SSE is provided, as well as the reasoning for choosing R1233zd(E) as refrigerant. R1233zd(E) has a low ODP and GWP, and meets the requirements set out by the practical limits of the set-up.

An overview of the boundary conditions and the measured variables was given. Based on these measurements, efficiencies and other performance parameters were derived. An explanation about the data reduction and uncertainty calculations is added. An exper- imental matrix and how the different settings of the ORC can be obtained is provided as well. The performance parameters were evaluated in function of flooding and pressure ratios, while keeping refrigerant pump rotational speed, expander rotational speed and ex- CONCLUSION 61 pander inlet temperature and pressure constant at a certain pressure ratio. Flooding ratios varied between 0 and 0.25, and pressure ratios between 3.4 and 4.3. The obtained results were presented and discussed. Overall, it was found that liquid flooding has a negative effect on ORC performance. Expander powers, efficiencies and cycle efficiencies decrease for increasing flooding ratio and there is no significant impact on expander temperature. Additional research with a different oil, higher PRs, higher flooding ratios and higher ex- pander speeds an temperature levels is needed to make a final conclusion on the impact of liquid flooding on ORC performance.

The general structure of theoretical ORC models and how they reach a solution was ex- plained. In order to incorporate oil flooding, additional components and boundary condi- tions need to be specified. In addition, the balances used to reach a solution need to be updated. The ORC models were then used to simulate different ORC behaviours. These results illustrate how settings of a real ORC can be tuned. PERFORMING MEASUREMENTS 62

Appendix A

Performing measurements

This appendix provides information on the different sensors that are used and how their output is processed. It also contains a description on how the start-up and shutdown procedure of the set-up work.

A.1 Measuring equipment and data acquisition

A.1.1 Pressure sensor

The different pressures are measured by absolute pressure sensors (APS). The pressure transducer converts the pressure into an electrical output. The transducer considered here is a resistive one, meaning that the electrical signal comes from a deflection of a diaphragm which results in physical deformation of strain gages bounded to it, producing an electrical resistance dependant on the applied pressure [54]. The sensors used are the series A-10 or S-20 from Wika. They can measure pressures ranging from 0 to 1600kPa.

A.1.2 Temperature sensor

Pt-100 Resistance temperature detectors (RTD) are used to measure the temperatures. Their principle is based on temperature dependant resistivity. As the temperature in- creases, the resistance of the metal (Platinum in this case) increases as well [55]. The type 514-942 from manufacturer TC-Direct is used, which has a range from -50 to 200°C.

Figure A.1 illustrates the pressure and temperature sensors. A.1 Measuring equipment and data acquisition 63

Figure A.1: Pressure and temperature sensor

A.1.3 Flow rate sensor

The mass flow rates of the working fluid and the lubricant oil are measured by Coriolis flow meters (CFM). Their functioning is based on the Coriolis effect. The fluid flows through a tube which is put into an oscillating motion. Sensors are present at inlet and outlet of the tube, which measure a sine wave. For the case without flow, both the sine waves are in phase. When there is flow through the tube, the tube will twist more due to the inertia of the fluid. This results in a phase shift between the two sine waves, which is proportional to the mass flow rate through the tube [56]. Sensor type OPTIMASS 1000-S 25 (for refrigerant) and 1000-S 15 (for oil) from Khrone are used, with a range from 0 to kg 1.8 s (see Figure A.2).

The volumetric flow rate of the cooling fluid is measured by an ultrasonic flow meter (UFM), which is based on the Doppler effect. Sound waves are emitted. If there is no flow, the reflected wave has the same frequency as the emitted one. In the presence of flow, the reflected wave frequency will be different [57]. The used sensor is manufactured by Siemens, the model is Sitrans FUS 380.

A pressure orifice is used to measure the hot source flow rate, which is based on the Bernoulli principle. The pressure drop over an orifice plate is measured, which is an indicator for the flow rate passing through the orifice plate. Model 3051 from Rosemount is installed. A.1 Measuring equipment and data acquisition 64

A.1.4 Torque sensor

Both on the refrigerant pump shaft and expander shaft, a torque meter is present, using shaft-to-shaft in-line placement (see Figure A.3). The sensor for the expander additionally has an encoder, such that besides torque also speed and angle can be measured. TQ513- 500-EN1024-MB for the expander (range: 0-53Nm) and TQ513-200-MB for the refrigerant pump (range: 0-22Nm) are used, both from manufacturer Omega.

Figure A.2: Mass flow rate sensor Figure A.3: Torque meter

A.1.5 Data acquisition

The electrical sensor outputs are connected to NI input modules, which in turn are con- nected to the LabVIEW software on the PC (through an ethernet connection). Dependent on the type of NI module, the conversion of the electrical signal into the physical quantity either takes place in the module itself or in the software.

The entire installation is controlled by a programmable logic controller (PLC), which is also connected to LabVIEW (through an ethernet connection). Each pump has its own VFD, connected to the PLC using the MODBUS protocol. This enables to monitor the power, torque and rotational speed of each drive. The expander generator is controlled by a 15kWe regenerative drive, also connected to the PLC. In the LabVIEW software the different rotational speeds of the pumps and expander need to be specified. These values are then sent to the PLC, where they are converted and used for the control of the system.

The heating and cooling loop are controlled by a separate PLC, and the hot source tem- perature is maintained with a PI-controller. A.2 Run procedures 65

The control cabinet, containing the NI modules, drives and PLC, is presented in Figure A.4. The LabVIEW interface used to implement the different settings and to keep track of the different measured variables is given in Figure A.5.

Figure A.4: Control cabinet

A.2 Run procedures

A.2.1 Prior to start-up

Air removal

When every component of the test set-up is installed, the refrigerant needs to be added to the working fluid loop. Before the system can be charged, it should however be drawn into a vacuum first. This is required for two reasons. Firstly, it serves as a test to check whether every component is installed correctly and no leaks are present in the system. Secondly, all the air needs to be removed to ensure that only refrigerant is present in the ORC, instead of an air-refrigerant mixture. Air in the system would impact performance A.2 Run procedures 66

Figure A.5: LabVIEW scheme of the ORC, and would make calculations based on pure refrigerant properties not reliable. The vacuum is realized by a vacuum pump.

ORC charging

When the ORC is drawn in vacuum, the refrigerant can be added. This is done by con- necting the refrigerant container to the liquid receiver. Initially, the under-pressure in the system realizes the flow of refrigerant. However, this under-pressure ceases quickly and the refrigerant needs to be heated to ensure further flow. Sight glasses are installed on the liquid receiver to keep track of the refrigerant level. A.2 Run procedures 67

A.2.2 Start-up

Separator preheating

In the separator, the oil and refrigerant are split. As it is a gravitational separator, the fluid with the highest density will lie at the bottom, the lighter one will float on top. At start- up, both the refrigerant and lubricant oil are at room temperature. At this temperature, the lubricant oil has a smaller density than the liquid refrigerant. If no measures would be taken, this would result in a large amount of oil in the working fluid loop and a large amount of refrigerant in the oil loop. Therefore, heating wire is wrapped around the separator which is turned on at start-up. This way, only vapour refrigerant is present and the separation of the mixture can occur correctly. When, after some time, it is assured that only vapour is present, the heating wire can be turned off.

Hot source charging

The electric heater should be turned on to heat up the Therminol 66. Either the desired temperature or the heating power can be defined.

Expander start-up

While the heating oil is warmed up and circulated, the refrigerant, oil and cooling liquid pump are turned on as well. The expander is bypassed until the desired temperatures and amount of superheat at the expander inlet are reached.

A.2.3 Shutdown

First the expander is switched off and the expander bypass is opened. Then, the heating power is turned off. All the pumps (i.e. the refrigerant, oil, cooling liquid and heating liquid pump) should run until the hot source is cooled down, and finally they can be shut down as well. STEADY-STATE DATA POINTS AND CORRESPONDING UNCERTAINTIES 68

Appendix B

Steady-state data points and corresponding uncertainties STEADY-STATE DATA POINTS AND CORRESPONDING UNCERTAINTIES 69 net w II,ORCLF E η ORCLF E η exp,is η exp,isoth η TR Steady-state points C] [-] [-] [-] [-] [-] [J/kg] exp,in ° T PR Table B.1: in,exp p y pp,o N pp,r N exp N [rpm] [rpm] [rpm] [-] [bar] [-] [ Point 12 17703 838 17704 838 17705 0 838 17706 107 838 15007 0 160 0.1 742 15008 227 0.142 9.66 742 9.51 15009 9.68 0 0.175 742 3.74 240010 9.65 100 3.91 3.82 100.3 99.4 742 2400 100.811 0 229 3.75 0.104 0.56 2400 742 0.56 100.912 10.19 0.55 0 0.27 0.22 742 0.27 2400 0.25 0.57 4.22 10.2813 100 9.96 742 0.26 2400 103.5 212 4.2814 0 0.2 0.106 0.18 102.4 919 0.58 0.27 4.09 2400 250 0.209 8.35 0.28 103.3 0.16 0.57 0.017 0.017 919 2400 8.2 0.024 0 8.56 0.236 0.28 0.58 3.45 0.015 919 8.26 100 0.29 95.5 0.21 3.52 0.085 3.37 0.086 0.118 0 226 95.8 0.31 3.36 95.7 0.077 0.019 0.62 0.079 96.4 12 0.16 0.24 0.085 0.029 0.64 0.62 12.14 3395 3403 12 0.24 5001 0.018 0.64 0.24 0.09 4.14 4.14 0.23 3051 0.18 0.143 124.5 124.5 4.04 0.094 0.26 0.14 0.01 0.77 0.75 126 0.12 3754 0.02 0.33 0.33 0.01 6129 0.009 0.78 0.054 3597 0.32 0.27 0.37 0.105 0.055 0.05 0.032 0.043 0.26 2007 4091 0.033 1891 0.146 0.187 1696 0.147 6094 8439 6118 STEADY-STATE DATA POINTS AND CORRESPONDING UNCERTAINTIES 70 net w II,ORCLF E η ORCLF E η exp,is η exp,isoth η TR C] [-] [-] [-] [-] [-] [J/kg] exp,in ° T PR Absolute uncertainties corresponding to Table B.2 in,exp p y Table B.2: pp,o N pp,r N exp [rpm] [rpm] [rpm] [-] [bar] [-] [ N Point 12 03 04 0 05 0 06 0 0 07 0 0 08 0 0 0 09 0.0002 0 0 010 0.016 0.0001 0 0 0 0.01611 0.024 0.016 0 0.0002 0 0.027 0 0.212 0.024 0.016 0.2 0 0 0 0 0.213 0.025 0 0.002 0 0.0001 0 0.214 0.002 0.134 0.016 0 0.002 0 0.0003 0.128 0 0.016 0.14 0.029 0.016 0 0 0.002 0 0 0.025 0.029 0.2 0.14 0.08 0.028 0 0.2 0.0004 0 0.0001 0.2 0.042 0.0003 0.0007 0.016 0 0 0.002 0.0004 0.002 0.016 0.016 0.031 0.002 0.134 0.024 0.0003 0 0.002 0.136 0.003 0.024 0.0004 0.023 0.2 0.016 0.141 0.002 0 0 0.2 0.2 0.04 0.023 0.002 0.09 7.1 0 0.002 0.2 0.021 0.0005 0.002 8.9 0.002 0.133 0.0001 0.0008 0.0005 0.016 7.9 0.133 0.137 0.016 0.002 0.002 0.023 0.016 7.7 0.04 0.004 0.13 0.002 0.022 0.2 0.09 0.02 0.022 0.2 0.0003 0.2 0.0006 0.0003 8.9 0.018 0.002 10.8 0.002 0.002 8.1 0.166 0.0002 0.002 0.003 0.162 0.001 0.167 0.001 0.105 0.052 6.5 0.001 0.057 7.8 0.001 5.8 0.001 5.4 0.006 0.005 0.005 13.5 11.1 11.4 STEADY-STATE DATA POINTS AND CORRESPONDING UNCERTAINTIES 71 hf m C] [kg/s] in,ev,hf ° T C] [ exp,in ° T PR in,exp p r m pp,r N Steady-state points used for modeling illustration (Section 5.3) exp [rpm] [rpm] [kg/s] [bar] [-] [ N Point 12 24003 858 22004 853 20005 0.375 853 9.74 24006 0.375 852 9.32 24007 0.375 3.6 852 9.52 24008 0.377 3.67 104.4 100 851 9.09 24009 127.9 0.375 3.84 100.6 852 8.23 240010 0.375 3.63 123.1 1.57 123.2 99.4 851 7.27 150011 0.376 3.35 1500 1.56 93.5 919 7.55 122.6 1.57 0.375 3.11 838 1500 87.3 8.69 116.2 0.402 3.21 1.56 742 0.369 89.3 10.59 108.4 3.53 1.56 10.59 0.328 4.19 96.8 107 1.56 4.32 104.2 10.56 111.2 105 126.4 4.53 1.96 106.8 2.61 126.6 1.57 126.6 1.56 1.56 Table B.3: STEADY-STATE DATA POINTS AND CORRESPONDING UNCERTAINTIES 72 hf m U in,ev,hf T C] [kg/s] ° U exp,in T C] [ ° U PR U in,exp p U r m U Absolute uncertainties corresponding to Table B.3 pp,r N U exp N Table B.4: U [rpm] [rpm] [kg/s] [bar] [-] [ Point 12 03 04 0 05 0 06 0 0.00003 07 0.016 0 0.00003 08 0.016 0.022 0 0.00003 09 0.2 0.016 0.024 0 0.00003 010 0.2 0.016 0.026 0 0.00003 011 0.2 0.2 0.016 0 0.024 0 0.00003 0.2 0.2 0.016 0 0.023 0 0.00003 0.2 0.2 0.0157 0 0.016 0.022 0.00003 0.2 0.2 0.0156 0 0.016 0.023 0.00004 0.2 0.2 0.0157 0.00003 0.016 0.024 0.2 0.016 0.2 0.0156 0.00003 0.027 0.2 0.016 0.029 0.2 0.0156 0.2 0.2 0.032 0.0156 0.2 0.2 0.0196 0.2 0.0261 0.2 0.0156 0.0156 0.0156 CORRELATIONS AND SENSITIVITY ANALYSIS 73

Appendix C

Correlations and sensitivity analysis

This appendix provides more information on how the different properties of the oil and the oil-refrigerant mixture are calculated. In addition, the values of the uncertainties of these correlations is presented and a sensitivity analysis is performed to look at their influence on the calculations.

C.1 Correlations for oil properties

ACD100FY

The oil which would originally be tested was ACD100FY, which is a polyolester. Correla- tions for the different properties are obtained from Zhelezny et al. [58].

The heat capacity of the oil, in the temperature range of 273K

dho = cp,o dT (C.2) · T 3 2 ho = (1304 + 1.035 T + 2.801 10− T )dT (C.3) · · · ZT0 3 1.035 2 2 2.801 10− 3 3 ho = 1304 (T T0) + (T T0 ) + · (T T0 ) + hRSV (C.4) · − 2 · − 3 · − with hRSV =200kJ/kg. Entropy is calculated according to: dT dso = cp,o (C.5) · T C.1 Correlations for oil properties 74

T 3 2 dT so = (1304 + 1.035 T + 2.801 10− T ) (C.6) · · · T ZT0 3 T 2.801 10− 2 2 so = 1304 ln( ) + 1.035 (T T0) + · (T T0 ) + sRSV (C.7) · T0 · − 2 · − with sRSV =1kJ/(kgK) The density follows from:

3 ρo = 1.18631056 10 0.731369048 T (C.8) · − · which results in the specific volume: 1 vo = (C.9) 1.18631056 103 0.731369048 T · − · SAE 20W50

Due to circumstances, another oil (i.e. SAE 20W50) was purchased and tested. As no correlations for the different properties were found in literature, assumptions had to be made.

Ugochukwu et al. [59] determined the specific heat capacity of five samples of SAE 20W50. In the calculations described in Section 4, it is assumed that the specific heat capacity of the lubricant oil is constant and equal to the average of the reported specific heat capacities of these five samples.

cp,o = 2625.276 [J/(kgK)] (C.10)

The density is assumed to be constant as well and equal to the value reported on the datasheet of the purchased oil [60]:

3 ρo = 855 [kg/m ] (C.11)

This yields in:

ho = 2625.276 (T T0) (C.12) · − T so = 2625.276 ln( ) (C.13) · T0

1 3 vo = [m /kg] (C.14) 855 C.2 Correlations for mixture properties 75

C.2 Correlations for mixture properties

The knowledge of the mixture properties is only required at the inlet and outlet of the expander. The oil-refrigerant mixture before and after the expander is assumed to be ideal, such that the properties can be calculated according to the ideal mixture rule. For the enthalpy, this results in [61]:

hmix = xL,r hL,r + xV,r hV,r + xo ho (C.15) · · ·

i.e. the weighted average of the enthalpies of the different phases present. hL,r and hV,r

are the saturation enthalpy of refrigerant liquid and vapour, respectively. xL,r and xV,r are the weight fractions of those phases and are calculated according to:

m˙ L,r xL,r = (C.16) m˙ L,r +m ˙ V,r +m ˙ o

m˙ V,r xV,r = (C.17) m˙ L,r +m ˙ V,r +m ˙ o

with m˙ L,r +m ˙ V,r =m ˙ r.

The mass flow rates of vapour phase and liquid phase written in function of the quality Q of the refrigerant: m˙ V,r m˙ V,r Q = = (C.18) m˙ L,r +m ˙ V,r m˙ r

m˙ L,r = (1 Q) m˙ r (C.19) − ·

m˙ V,r = Q m˙ r (C.20) · Equation C.15 to Equation C.20 result in:

hmix = xr hr + xo ho (C.21) · · with: m˙ r xr = (C.22) m˙ o +m ˙ r The equation for the entropy calculation is in correspondence with Equation C.21.

In these calculations the oil solubility, i.e. the dissolution of refrigerant in the oil, is neglected as only values for ’ideal’ situations such as isothermal or isentropic expansion are calculated based on these correlations. C.3 Uncertainty on correlations 76

C.3 Uncertainty on correlations

Uncertainty of oil correlations

Zhelezny et al. [58] stated that the uncertainties on temperature and heat capacity did not exceed 0.2K and 1.5%, correspondingly. Based on Equation 4.17 and Equations C.4, C.7 and C.9, this results in: δh o 2 Uh,o,corr = ( UT ) (C.23) r δT · 3 2 Uh,o,corr = (1304 + 1.035 T + 2.801 10− T ) UT (C.24) · · · · δs o 2 Us,o,corr = ( UT ) (C.25) r δT · 1 3 Us,o,corr = (1304 + 1.035 + 2.801 10− T ) UT (C.26) · T · · · δv o 2 Uv,o,corr = ( UT ) (C.27) r δT · 0.731369048 Uv,o,corr = UT (C.28) (1.18631056 0.731369048 T )2 · − · with UT =0.2K.

The correlations used for SAE 20W50:

Uh,o,corr = 2625.276 UT (C.29) · 2625.276 Us,o,corr = UT (C.30) T ·

Uv,o,corr = 0 (C.31) For the specific volume and heat capacity, no uncertainty is taken into account as no correlations were found expressing these properties in function of temperature. These assumptions can be justified based on the fact that the oil addition is limited, and the effect of the uncertainty on the lubricant oil properties will thus be limited in the overall uncertainty calculations.

Uncertainty of refrigerant correlations

As the different thermodynamic properties of the refrigerant calculated using CoolProp [48] are based on models, there is a certain uncertainty associated to the these values. Neglect- ing the uncertainties related to the correlations could lead to large errors in the overall C.4 Sensitivity analysis 77 uncertainty calculation. The uncertainties corresponding to the modelling of refrigerant R1233zd(E) are not tabulated. Lemmon and Span [62] did however tabulate relative un- certainties for other fluids, such as R245fa. For R245fa, the maximum values were:

uh,r,corr = 1.18%(gaseous) (C.32)

us,r,corr = 1.18%(gaseous) (C.33)

uT,r,corr = 1.18%(gaseous) (C.34)

uh,r,corr = 0.1%(liquid) (C.35)

us,r,corr = 0.1%(liquid) (C.36)

As an approximation, these maximum values are chosen for the uncertainty calculation of R1233zd(E). As this is only approximate, the next section focuses on how sensitive the total uncertainty calculation is in function of these assumptions.

Uncertainty of mixture correlations

Combining the correlation uncertainties of the oil and of the refrigerant, results in expres- sions for the oil-refrigerant mixture:

2 2 Uh,corr = (xr Uh,r,corr) + (xo Uh,o,corr) (C.37) · · q 2 2 Us,corr = (xr Us,r,corr) + (xo Us,o,corr) (C.38) · · q C.4 Sensitivity analysis

To investigate the influence of the previous assumptions, the uncertainty intervals of the output variables are calculated for different values of the correlation uncertainties for the refrigerant (i.e. Equation C.32-Equation C.36). The input variables used for this analysis can be found in Table C.1. These values are chosen arbitrarily and serve solely for an illustrative purpose. In Table C.2 the different uncertainty intervals are presented for varying relative correlation uncertainties.

Changing the uncertainty on the correlation for the liquid phase of the refrigerant has limited impact on the calculation of the intervals. This correlation is only used for calcu- lating the evaporator heat input and its influence on the First and Second Law efficiency is C.4 Sensitivity analysis 78 small. However, the correlations for the vapour phase of the refrigerant are used frequently. Therefore, its impact on the uncertainty calculations is significant. The uncertainty on isen- tropic expander efficiency doubles when the relative uncertainty is increased from 0.5 to 1%. However, relative to the value of the variable itself, the uncertainty interval still remains small enough if the correlation uncertainty is not too high. For the isothermal expander efficiency, this is not the case. The uncertainty interval amounts to approximately 25% for a relative correlation uncertainty of 1.18%, rendering the calculation of the efficiency unreliable. This is due to the fact that for calculating the isothermal expander power, four values need to be calculated using CoolProp [48]: the enthalpy and entropy at the inlet of the expander and the isothermal enthalpy and entropy at the outlet. The value for the power is too small to compensate for the high uncertainty on it (second term of Equa- tion 4.35), resulting in a high overall uncertainty on isothermal efficiency. Therefore, the temperature ratio of the expander is a better variable to quantify how well the expansion process approximates an isothermal one, as the uncertainty on it is only determined by the uncertainty on temperature measurements.

Table C.1: Input variables

Temperatures Tin,cd,cf 20 °C Tin,ev,hf 125 °C

Tin,exp 100 °C Tout,exp 80 °C

Tin,ev,r 60 °C Tout,ev,r 100 °C

Tin,heater,o 30 °C Tout,heater,o 100 °C

Pressures pin,exp 10 bar pout,exp 2 bar

pin,ev,r 10 bar pout,ev,r 10 bar ˙ ˙ Powers Wexp 4000 W Wpp,r 500 W ˙ Wpp,o 100 W

Mass flow rates m˙ r 0.2 kg/s m˙ o 0.05 kg/s C.4 Sensitivity analysis 79 [%] exp,isoth η [W] exp,isoth ˙ W [%] exp,is η [W] exp,is ˙ W [%] II,ORCLF E η [%] Uncertainty intervals in function of correlation uncertainties ORCLF E η Table C.2: [W] in ˙ Q 87 0.02 0.09 71 0.22 38 0.32 * x/y%: x is the relative uncertainty for the vapour phase, y is the relative uncertainty for the liquid phase Variable 0/0% 0/0.1%0/0.5% 1060/1% 3170.5/0.1% 0.021/0.1% 513 617 0.051.18/0.1% 1010 1190 0.081.5/0.1% 0.1 0.12/0.1% 1510 0.16 0.19 0.25/0.1% 2011 0.24 0.32 0.34 5022 0.61 0.32 0.72 71 0.8 71 0.91 589 71 1.21 0.22 1171 1381 3.03 0.22 1754 1.82 38 0.22 3.62 4.27 2338 38 1266 5.42 5843 38 2532 2987 7.23 0.32 3797 18.06 0.32 10.56 5063 21.11 24.91 0.32 12657 31.66 42.21 105.53 Value 46261 7.35 27.87 11377 35.16 6926 57.75 * BIBLIOGRAPHY 80

Bibliography

[1] UNFCC. The paris agreement. https://unfccc.int/process-and-meetings/ the-paris-agreement/the-paris-agreement, 22 October 2018. Accessed on: 29 November 2018.

[2] European Environment Agency. Overall progress towards the euro- pean union’s ’20-20-20’ climate and energy targets. https://www. eea.europa.eu/themes/climate/trends-and-projections-in-europe/ trends-and-projections-in-europe-2016/1-overall-progress-towards-the, 7 November 2017. Accessed on: 29 November 2018.

[3] Directorate-General for Energy, Transport (European Commission), N Tasios, V Pa- pandreou, L Mantzos, and Capros P. Europen energy and tranport - trends to 2030: update 2007, 2008.

[4] EPA. Global greenhouse gas emissions data. https://www.epa.gov/ghgemissions/ global-greenhouse-gas-emissions-data, 2014. Accessed on: 11 November 2018.

[5] World Nuclear Association. Electricity generation - what are the options? http://www.world-nuclear.org/nuclear-basics/ electricity-generation-what-are-the-options.aspx. Accessed on: 11 Novem- ber 2018.

[6] http://www.leam.faa.illinois.edu/up466/lab-images/copy_of_lab-image/ Rankine%20Cycle.jpg/image_view_fullscreen. Accessed on: 11 November 2018.

[7] Davide Ziviani, Eckhard A Groll, James E Braun, and Michel De Paepe. Review and update on the geometry modeling of single-screw machines with emphasis on expanders. International Journal of Refrigeration, 92:10–26, 2018. BIBLIOGRAPHY 81

[8] Bruce G. Miller. 7 - clean coal technologies for advanced power generation. In Bruce G. Miller, editor, Clean Coal Engineering Technology, pages 251 – 300. Butterworth- Heinemann, Boston, 2011.

[9] Ignace Vankeirsbilck, Bruno Vanslambrouck, Sergei Gusev, and Michel De Paepe. Organic rankine cycle as efficient alternative to steam cycle for small scale power generation. HEFAT, 2011.

[10] Steven Lecompte, Henk Huisseune, Martijn Van Den Broek, Bruno Vanslambrouck, and Michel De Paepe. Review of organic rankine cycle (orc) architectures for waste heat recovery. Renewable and Reviews, Sustainable Energy, 47:448–461, 2015.

[11] Rebecca McGinty, Giuseppe Bianchi, Obadah Zaher, Steven Woolass, David Oliver, Christopher Williams, and Jeremy Miller. Techno-economic survey and design of a pilot test rig for a trilateral flash cycle system in a steel production plant. Energy Procedia, 123:281–288, 2017.

[12] Davide Ziviani, Eckhard A Groll, James E Braun, Michel De Paepe, and Martijn van den Broek. Analysis of an organic rankine cycle with liquid-flooded expansion and internal regeneration (orclfe). Energy, 144:1092–1106, 2018.

[13] Davide Ziviani, Sergei Gusev, Steven Lecompte, EA Groll, JE Braun, WT Horton, Martijn van den Broek, and Michel De Paepe. Optimizing the performance of small- scale organic rankine cycle that utilizes a single-screw expander. Applied energy, 189:416–432, 2017.

[14] Davide Ziviani, Sergei Gusev, Steven Lecompte, EA Groll, JE Braun, W Travis Hor- ton, Martijn van den Broek, and Michel De Paepe. Characterizing the performance of a single-screw expander in a small-scale organic rankine cycle for waste heat recovery. Applied Energy, 181:155–170, 2016.

[15] Davide Ziviani, Sergei Gusev, Stefan Schuessler, Abdennacer Achaichia, James E Braun, Eckhard A Groll, Michel De Paepe, and Martijn van den Broek. Employ- ing a single-screw expander in an organic rankine cycle with liquid flooded expansion and internal regeneration. Energy Procedia, 129:379–386, 2017.

[16] Bertrand F Tchanche, Gr Lambrinos, Antonios Frangoudakis, and George Papadakis. Low-grade heat conversion into power using organic rankine cycles - A review of various applications. Renewable and Reviews, Sustainable Energy, 15(8):3963–3979, 2011. BIBLIOGRAPHY 82

[17] Huijuan Chen, D Yogi Goswami, and Elias K Stefanakos. A review of thermodynamic cycles and working fluids for the conversion of low-grade heat. Renewable and reviews, sustainable energy, 14(9):3059–3067, 2010.

[18] Global warming potential. https://en.wikipedia.org/wiki/Global_warming_ potential, 8 November 2018. Accessed on: 29 November 2019.

[19] Ozone depletion potential. https://en.wikipedia.org/wiki/Ozone_depletion_ potential, 24 November 2018. Accessed on: 29 November 2018.

[20] Oyeniyi A Oyewunmi, Aly I Taleb, Andrew J Haslam, and Christos N Markides. On the use of saft-vr mie for assessing large-glide fluorocarbon working-fluid mixtures in organic rankine cycles. Applied Energy, 163:263–282, 2016.

[21] Junjiang Bao and Li Zhao. A review of working fluid and expander selections for organic rankine cycle. Renewable and reviews, sustainable energy, 24:325–342, 2013.

[22] Sotirios Karellas and Andreas Schuster. Supercritical fluid parameters in organic rankine cycle applications. International Journal of Thermodynamics, 11(3), 2008.

[23] V Maizza and A Maizza. Working fluids in non-steady flows for waste energy recovery systems. Applied Thermal Engineering, 16(7):579–590, 1996.

[24] MJ Lee, DL Tien, and CT Shao. Thermophysical capability of ozone-safe working fluids for an organic rankine cycle system. Heat Recovery Systems and CHP, 13(5):409– 418, 1993.

[25] Mohammad Bahrami, Ali A Hamidi, and Soheil Porkhial. Investigation of the effect of organic working fluids on thermodynamic performance of combined cycle stirling-orc. International Journal of Energy and Engineering, Environmental, 4(1):12, 2013.

[26] Muhammad Imran, Muhammad Usman, Byung-Sik Park, and Dong-Hyun Lee. Volu- metric expanders for low grade heat and waste heat recovery applications. Renewable and Sustainable Energy Reviews, 57:1090 – 1109, 2016.

[27] Biao Lei, Wei Wang, Yu-Ting Wu, Chong-Fang Ma, Jing-Fu Wang, Lei Zhang, Chuang Li, Ying-Kun Zhao, and Rui-Ping Zhi. Development and experimental study on a single screw expander integrated into an organic rankine cycle. Energy, 116:43–52, 2016. BIBLIOGRAPHY 83

[28] Hiroshi Kanno and Naoki Shikazono. Experimental and modeling study on adiabatic two-phase expansion in a cylinder. International Journal of Heat and Transfer, Mass, 86:755–763, 2015.

[29] Guo-Dong Xia, Ye-Qiang Zhang, Yu-Ting Wu, Chong-Fang Ma, Wei-Ning Ji, Shan- Wei Liu, and Hang Guo. Experimental study on the performance of single-screw expander with different inlet vapor dryness. Applied Thermal Engineering, 87:34–40, 2015.

[30] Bernard Aoun and Denis F Clodic. Theoretical and experimental study of an oil-free scroll vapor expander. 2008.

[31] José Galindo, Vicente Dolz, Lucía Royo-Pascual, Regine Haller, and Julien Melis. Modeling and experimental validation of a volumetric expander suitable for waste heat recovery from an automotive internal combustion engine using an organic rankine cycle with ethanol. Energies, 9(4):279, 2016.

[32] E Oralli, Md Ali Tarique, C Zamfirescu, and I Dincer. A study on scroll compressor conversion into expander for rankine cycles. International Journal of Low-Carbon Technologies, 6(3):200–206, 2011.

[33] Opubo N Igobo and Philip A Davies. Review of low-temperature vapour power cycle engines with quasi-isothermal expansion. Energy, 70:22–34, 2014.

[34] SP Cicconardi, E Jannelli, A Perna, and G Spazzafumo. A steam cycle with an isothermal expansion: the effect of flowvariation. International journal of hydrogen energy, 24(1):53–57, 1999.

[35] Giuseppe Bianchi, Stuart Kennedy, Obadah Zaher, Savvas A Tassou, Jeremy Miller, and Hussam Jouhara. Numerical modeling of a two-phase twin-screw expander for trilateral flash cycle applications. International Journal of Refrigeration, 88:248–259, 2018.

[36] Sadik Kakaç, Hongtan Liu, and Anchasa Pramuanjaroenkij. Heat exchangers: selec- tion, rating, and thermal design. CRC press, third edition, 2002.

[37] Giuseppe Bianchi, Rebecca McGinty, David Oliver, Derek Brightman, Obadah Zaher, Savvas A Tassou, Jeremy Miller, and Hussam Jouhara. Development and analysis of a packaged trilateral flash cycle system for low grade heat to power conversion applications. Thermal Science and Progress, Engineering, 4:113–121, 2017. BIBLIOGRAPHY 84

[38] Hiroshi Kanno and Naoki Shikazono. Modeling study on two-phase adiabatic expan- sion in a reciprocating expander. International Journal of Heat and Transfer, Mass, 104:142–148, 2017.

[39] Ian H Bell, Vincent Lemort, Eckhard A Groll, James E Braun, Galen B King, and W Travis Horton. Liquid-flooded compression and expansion in scroll machines–part i: Model development. International Journal of Refrigeration, 35(7):1878–1889, 2012.

[40] Ian H Bell, Vincent Lemort, Eckhard A Groll, James E Braun, Galen B King, and W Travis Horton. Liquid flooded compression and expansion in scroll machines–part ii: Experimental testing and model validation. International Journal of Refrigeration, 35(7):1890–1900, 2012.

[41] Nicola Casari, Alessio Suman, Davide Ziviani, Martijn Van Den Broek, Michel De Paepe, and Michele Pinelli. Computational models for the analysis of positive displacement machines: Real gas and dynamic mesh. Energy Procedia, 129:411–418, 2017.

[42] Ronald DiPippo. Second law assessment of binary plants generating power from low- temperature geothermal fluids. Geothermics, 33(5):565–586, 2004.

[43] Michael J Moran, Howard N Shapiro, Daisie D Boettner, and Margaret B Bailey. Fundamentals of engineering thermodynamics: SI version. John Wiley & Sons, eighth edition, 2015.

[44] Yiping Dai, Jiangfeng Wang, and Lin Gao. Parametric optimization and compara- tive study of organic rankine cycle (orc) for low grade waste heat recovery. Energy Conversion and Management, 50(3):576–582, 2009.

[45] FI Abam, EB Ekwe, SO Effiom, and MC Ndukwu. A comparative performance anal- ysis and thermo-sustainability indicators of modified low-heat organic rankine cycles (orcs): An exergy-based procedure. Energy Reports, 4:110–118, 2018.

[46] Davide Ziviani, Ian Bell, Martijn van den Broek, and Michel De Paepe. Comprehensive model of a single-screw expander for orc-systems. 2014.

[47] CCA Coalition. Hydrofluorocarbons (hfc). http://www.ccacoalition.org/fr/ slcps/hydrofluorocarbons-hfc. Accessed on: 22 May 2019. BIBLIOGRAPHY 85

[48] Ian H. Bell, Jorrit Wronski, Sylvain Quoilin, and Vincent Lemort. Pure and pseudo- pure fluid thermophysical property evaluation and the open-source thermophysical property library coolprop. Industrial & Engineering Chemistry Research, 53(6):2498– 2508, 2014.

[49] Honeywell. Genetron ®245fa (r-245fa). https://www.honeywell-refrigerants. com/europe/product/genetron-245fa/. Accessed on: 4 April 2019.

[50] Honeywell. Solstice ®zd (r-1233zd). https://www.honeywell-refrigerants.com/ europe/product/solstice-zd/. Accessed on: 10 May 2019.

[51] Steven Lecompte, Sergei Gusev, Bruno Vanslambrouck, and Michel De Paepe. Exper- imental results of a small-scale organic rankine cycle: Steady state identification and application to off-design model validation. Applied Energy, 226:82 – 106, 2018.

[52] Brandon J Woodland, James E Braun, Eckhard A Groll, and W Travis Horton. Ex- perimental testing of an organic rankine cycle with scroll-type expander. 2012.

[53] Robert J. Moffat. Describing the uncertainties in experimental results. Experimental Thermal and Fluid Science, 1(1):3 – 17, 1988.

[54] Omega™. Pressure transducers and transmitters. https://www.omega.co.uk/ prodinfo/pressure-transducers.html. Accessed on: 21 May 2019.

[55] Pt100. https://nl.wikipedia.org/wiki/Pt100, 24 March 2019. Accessed on: 10 April 2019.

[56] Endress+Hauser. Coriolis massaflowmeters. https://www.be.endress.com/ nl/flowmeting-niveaumeting-drukmeting-temperatuurmeting-analysemeting/ Flowmeting-vloeistoffen-gassen-stoom/Flowmeting-met-coriolis-massaflowmeter. Accessed on: 10 April 2019.

[57] Universal Flow Monitors. Ultrasonic flowmeter technology. http://www.flowmeters. com/ultrasonic-technology. Accessed on: 10 April 2019.

[58] VP Zhelezny, Yu V Semenyuk, SN Ancherbak, AJ Grebenkov, and OV Beliayeva. An experimental investigation and modelling of the solubility, density and surface tension of 1, 1, 1, 3, 3-pentafluoropropane (r-245fa)/synthetic polyolester compressor oil solutions. Journal of Fluorine Chemistry, 128(9):1029–1038, 2007. BIBLIOGRAPHY 86

[59] Joseph Ugochukwu, Chigbo Nwamaka, Anyadiegwu Chukwunweike Faustinus, and Olinya Christopher. Determination of the specific heat capacity of different sae 20w- 50 engine oil samples using the electrical method.

[60] raxol® motoroils. Eco flow 20w50. https://www.raxol.eu/en/oilproducts/ trucks-2/motor-oil-1/mineral-motor-oil-22/eco-flow-20w50-1027/?p=1027. Accessed on: 12 June 2019.

[61] M Youbi-Idrissi, J Bonjour, M-F Terrier, C Marvillet, and F Meunier. Oil presence in an evaporator: experimental validation of a refrigerant/oil mixture enthalpy calcula- tion model. International Journal of refrigeration, 27(3):215–224, 2004.

[62] Eric W Lemmon and Roland Span. Short fundamental equations of state for 20 industrial fluids. Journal of Chemical & Engineering Data, 51(3):785–850, 2006.