Applied Energy 138 (2015) 505–516

Contents lists available at ScienceDirect

Applied Energy

journal homepage: www.elsevier .com/locate /apenergy

Experimental validation of a thermodynamic boiler model under steadystate and dynamic conditions

http://dx.doi.org/10.1016/j.apenergy.2014.10.0310306-2619/� 2014 Elsevier Ltd. All rights reserved.

⇑ Corresponding author at: Bioenergy 2020+, Gewerbepark Haag 3, 3250Wieselburg Land, Austria. Tel.: +43 7416 5223859.

E-mail address: elisa.carlon@bioenergy2020.eu (E. Carlon).

Elisa Carlon a,b,⇑, Vijay Kumar Verma a, Markus Schwarz a, Laszlo Golicza a, Alessandro Prada b,Marco Baratieri b, Walter Haslinger a, Christoph Schmidl a

a Bioenergy 2020+, Gewerbepark Haag 3, 3250 Wieselburg Land, Austriab Free University of Bozen-Bolzano, Universitätsplatz – Piazza Università 5, 39100 Bozen-Bolzano, Italy

h i g h l i g h t s

� Laboratory tests on two commercially available pellet boilers.� Steady state and a dynamic load cycle tests.� Pellet boiler model calibration based on data registered in stationary operation.� Boiler model validation with reference to both stationary and dynamic operation.� Validated model suitable for coupled simulation of building and heating system.

a r t i c l e i n f o

Article history:Received 2 July 2014Received in revised form 6 October 2014Accepted 7 October 2014Available online 19 November 2014

Keywords:Pellet boilerDynamic test methodBuilding energy simulationHVAC boiler modelModel validation

a b s t r a c t

Nowadays dynamic building simulation is an essential tool for the design of heating systems for residen-tial buildings. The simulation of buildings heated by biomass systems, first of all needs detailed boilermodels, capable of simulating the boiler both as a stand-alone appliance and as a system component. Thispaper presents the calibration and validation of a boiler model by means of laboratory tests. The chosenmodel, i.e. TRNSYS ‘‘Type 869’’, has been validated for two commercially available pellet boilers of 6 and12 kW nominal capacities. Two test methods have been applied: the first is a steady state test at nominalload and the second is a load cycle test including stationary operation at different loads as well as tran-sient operation. The load cycle test is representative of the boiler operation in the field and characterisesthe boiler’s stationary and dynamic behaviour. The model had been calibrated based on laboratory dataregistered during stationary operation at different loads and afterwards it was validated by simulatingboth the stationary and the dynamic tests. Selected parameters for the validation were the heat transferrates to water and the water temperature profiles inside the boiler and at the boiler outlet. Modellingresults showed better agreement with experimental data during stationary operation rather than duringdynamic operation. Heat transfer rates to water were predicted with a maximum deviation of 10% duringthe stationary operation, and a maximum deviation of 30% during the dynamic load cycle. However, forboth operational regimes the fuel consumption was predicted within a 10% deviation from theexperimental values.

� 2014 Elsevier Ltd. All rights reserved.

1. Introduction

The recent EU policies encourage energy efficiency in heatingsystems and promote the use of energy from renewable sources[1,2]. In this framework small scale biomass combustion devices(boilers and stoves) are currently considered a promising

technology to supply heating and domestic hot water for the resi-dential sector in Europe [3,4]. Because of the ongoing renovationsof the residential building stock and because of the increasing pop-ularity of low-energy and passive houses, the demand is graduallyshifting to small scale heating devices. Boilers are the core ofhydronic central heating and domestic hot water supply systems.Hot water leaving the boiler is delivered to one or more space heat-ing circuits and to the hot water storage tank. The technology ofbiomass boilers is already well-established, with efficiencies reach-ing 90% and low emission factors [5]. Heat production can also be

Nomenclature

cp,fg specific heat capacity of the dry flue gases (kJ kg�1 K�1)cp,vap specific heat capacity of water vapour (kJ kg�1 K�1)cp,w specific heat capacity of water (kJ kg�1 K�1)Ctherm thermal capacitance of the boiler (kJ K�1)dTnom, dThx,wat parameters for the determination of the flue gas

temperature (K)facA,lam, facB,lam parameters for the determination of the excess

air ratio (–)HHV higher heating value of the fuel (kJ kg�1)_mfg mass flow rate of dry flue gases (kg h�1)_mfuel;dry mass flow rate of dry fuel (kg h�1)_mCO;fg mass flow rate of CO in the flue gases (kg h�1)_mw water mass flow rate (kg h�1)_mw;fg mass flow rate of water vapour in the flue gases (kg h�1)_mw;nom water mass flow rate at nominal load (kg h�1)

Pel electricity absorbed by the boiler (W)Pel,MAX electricity absorbed by the boiler, at nominal load (W)Pel,MIN electricity absorbed by the boiler, at minimum load (W)Pel,OFF electricity absorbed by the boiler, in stand-by mode (W)Pmin minimum firing thermal output of the burner (W)Pmax maximum firing thermal output of the burner (W)Pnom nominal thermal output of the boiler (W)Pstart firing output of the burner during start-up phase (W)_Qair heat transferred from the boiler’s body to the forced

draught air (W)_Qamb heat transferred from the boiler’s body to the environ-

ment (W)_Qash heat loss to the unburnt constituents in the residues

(W)_Qbc;camb heat transferred from the combustion chamber to the

environment (W)

_Qenvelope heat transferred from the boiler’s envelope to theenvironment (W)

_Qfg;chem chemical heat loss in the flue gases (W)_Qfg;lat latent heat loss in the flue gases (W)_Qfg;sens sensible heat loss in the flue gases (W)_Qfuel energy input of the fuel, based on its higher heating

value (W)_QGCV

fuel;nom energy input of the fuel at nominal load (W)_Qhx heat transferred from flue gases to water (W)_QMtherm heat stored in the boiler’s body (W)_Qwout heat transferred to the water (W)

Tboiler boiler’s temperature (�C)Tfg outlet temperature of the flue gases (�C)Troom room temperature (�C)Tw,in water return temperature (�C)Tw,out water outlet temperature (�C)t time (h)UAamb heat transfer coefficient to the environment (W K�1)UAamb,ON heat transfer coefficient to the environment, during

burner operation (W K�1)UAamb,OFF heat transfer coefficient to the environment, during

standby (W K�1)Volwat volume of water inside the boiler (m3)Wel electricity consumption during start-up phase (Wh)

Greek lettersk excess air ratio (–)s duration of test sequence (h)DHvap,w latent heat of vaporisation of water (kJ kg�1)DHCO,comb standard enthalpy of combustion of carbon monoxide

(kJ kg�1)

506 E. Carlon et al. / Applied Energy 138 (2015) 505–516

supported by solar collectors or heat pumps, thus forming a hybridsystem [6,7].

Dynamic building simulation is an essential tool for the reliabledesign of appropriate system solutions that suit the demands ofspecific buildings, since it is capable to describe the dynamic inter-actions among building, energy systems, occupants and outdoorenvironment. The analysis of the building, coupled with its heatingsystem, provides key information for choosing, sizing and control-ling the system’s components in order to ensure comfort condi-tions in the house for the whole heating season. Previous studies[7–9] investigated the installation of combined solar and pelletheating system in single-family houses. The authors modelledand compared different system configurations and found improvedcontrol strategies to increase system efficiencies. The simulation ofa heating system equipped with a biomass boiler needs a detailedboiler model, capable of describing the boiler both as a stand-aloneappliance and as a component of the house’s heating system.Nonetheless, detailed models require a large number of parameterswhose imprecise definition can undermine the reliability of simu-lation results. Hence, the model must be calibrated to represent thespecific boiler, both under steady state and dynamic conditions.

Currently, in many countries, regulations and related technicalstandards characterise the behaviour of biomass boilers only insteady state conditions for few load factors, thus making it difficultto calibrate dynamic simulation models. For example, efficienciesand emission factors of biomass boilers commercially available inthe EU are currently assessed by means of standard laboratorytests [10–13]. Test methods comprise steady state operation at fulland partial load (i.e. 30% of the nominal load) but no dynamic testshave been standardised yet. Dynamic tests in cycling operation

have been performed on various types of boilers and stoves, inorder to characterise the start-up and stop sequences as well asthe burner’s cycling operation [12–15]. In a recent work of Glembinet al. [16], gas and oil boilers were tested under cycling operationand load transitions. Another type of dynamic test is representedby dynamic load cycles, which reproduce the time variable heatingand domestic hot water demand of a house, so that, even if the testis carried out in the laboratory, it is representative of the actualboiler operation in the field [17–20]. In the last decade, newdynamic test methods have been developed in order to test thecombination of system components (i.e. boilers, heat pumps, solarcollectors and hot water storage tanks) in different configurations,thus allowing to reproduce the complete heating and hot watersupply system at the test bench [21–25].

Only few studies have formally dealt with the validation of sim-ulation models under unsteady state conditions. For instance,Persson et al. [26] presented the validation of a thermodynamicmodel, suitable for pellet stoves and boilers, under steady stateconditions and during dynamic tests in which the boilers under-went ‘‘on/off’’ cycles or heating up and cooling down curves [27].

To our knowledge, no data about the validation of a boilermodel under dynamic tests, which reproduce in detail the fieldoperation of the boiler in single family houses, are available inthe literature yet.

This paper presents the experimental validation of a boilermodel, suitable for dynamic building simulations, for two commer-cially available pellet boilers, whose nominal capacities are 6 and12 kW. The boilers have been simulated in the TRNSYS simulationsuite. TRNSYS is a software used for the dynamic simulation of sys-tems, in particular for buildings and energy systems [28]. The

E. Carlon et al. / Applied Energy 138 (2015) 505–516 507

boilers can be simulated in TRNSYS first of all as a stand-aloneappliance and later as component of the building’s heating system.To our knowledge, at the moment there are three TRNSYS compo-nents suitable to model biomass boilers and stoves: Type 370 [29],Type 210 [26] and Type 869 [30]. As the latter is the most extensivemodel, in the present study the two pellet boilers have been sim-ulated by means of Type 869 (version 502), a non-standard TRNSYSType distributed by SPF Institut fur Solartechnik [31].

For the experimental validation, two different test methodshave been adopted. The first test method consists of a steady statetest at nominal load. The second method is a dynamic load cycleincluding stationary operation at different loads, as well as tran-sient operation. The cycle reproduces the variable heating anddomestic hot water demand of a single family house [23]. Theparameters required to calibrate the model were calculated basedon laboratory data registered in stationary conditions at differentloads. Successively, additional laboratory tests carried out both instationary and dynamic regimes provided reference data for themodel validation.

This work has been performed in the frame of the BioMaxEffproject [32], a European FP7 project aiming at the demonstrationof biomass boilers under real life operating conditions. The pelletboilers analysed in this study are operated in the laboratory andmonitored in several houses where they are installed.

The aim of this work is first of all to evaluate if the Type 869 boi-ler model is suitable for the simulation of the concerned boilers, instationary as well as in dynamic operational regime. For this pur-pose, the accuracy of the simulation results needs to be assessedin comparison to the experimental data registered both in the sta-tionary test and in the dynamic load cycle test.

A second aim of this study is to verify if the adopted tests meth-ods, described in detail by Heckmann et al. [23], provide a set oflaboratory data which is adequate to completely calibrate the Type869 boiler model. After being calibrated, the boiler model could beimplemented in the TRNSYS simulation of the monitored houses, inorder to simulate the building-system behaviour and thus assess-ing the performance of the boiler in the field.

2. Methods

2.1. Boiler model

In this study the pellet boilers have been simulated by means ofthe TRNSYS Type 869 boiler model, which is described in detail in[14,30]. The model calculates the mass and energy balance of theboiler under time variable inputs, thus describing the boiler oper-ation in dynamic conditions. Under the assumption of completemass conversion, the pellet chemical composition and the excessair ratio determine the flue gases chemical composition. Theenergy balance of the boiler is set up in two steps: the first repre-sents the combustion chamber and the second the heat exchangebetween flue gases, water and boiler’s structure. In the combustionchamber the energy released from the pellet combustion is sepa-rated into several fractions, according to Eq. (1).

_Q fuel ¼ _Q ash þ _Q fg;lat þ _Q fg;sens þ _Q fg;chem þ _Qbc;amb þ _Q hx ð1Þ

A fraction of the energy input ( _Q fuel) is contained in the unburntconstituents in the residues ( _Q ash) and a second fraction is trans-ferred from the combustion chamber to the surrounding environ-ment ( _Q bc;camb). The heat lost in the flue gases is the sum of theirlatent heat ( _Qfg;lat), of their sensible heat ( _Qfg;sens) and of chemicallosses due to unburned carbon content ( _Q fg;chem). After having sub-tracted all the energy losses in the combustion chamber, the

remaining energy fraction ( _Qhx) is transferred to the boiler’s heatexchanger, where the energy balance reported in Eq. (2) isimplemented.

_Qhx ¼ _Q w;out þ _Q amb þ _Q Mtherm ð2Þ

The hot flue gases leaving the combustion chamber exchangeheat with the water flowing through the boiler ( _Qwout) and withthe boiler’s body ( _QMtherm). The term _QMtherm comprises the heattransferred both to the boiler’s structure and to the water insidethe boiler. The term _Qamb represents the heat loss from the boilerenvelope to the ambient.

2.2. Experimental setup

Two commercially available pellet boilers of 6 and 12 kWnominal capacities, manufactured by the company WindhagerZentralheizung Technik GmbH [33], were investigated. The boilersconsist of a steel body with a fully insulated cladding. The boilerscan modulate the power output in the range between 30% and100% of the nominal power. Pellets were loaded manually into ahopper and fed through a screw auger into the top feed burner.The burner pot is made of high-temperature-resistant stainlesssteel and equipped with automatic ignition and automatic ashremoval. A speed controlled vacuum fan regulates the primaryand secondary air supply. Hot flue gases generated during combus-tion pass through a vertical heat exchanger, which is automaticallycleaned by a spiral mechanism, and deliver heat to the circulatingwater. The control concept of the combustion process is based onthe flue gas temperature, which is measured directly at the exitof the combustion chamber (i.e. thermo-control combustion con-trol concept).

The tests have been carried out at the laboratory of Bioenergy2020 + GmbH, located in Wieselburg an der Erlauf, Lower Austria(48.117�N 15.136� E, 270 m above mean sea level). A schematiclayout of the setup designed for the experiments is presented inFig. 1. This experimental setup was designed to conform as muchas possible to the requirements of EN 303-4 [34] and EN 303-5[35]. Deviating from the procedures under EN 303-5 [35], transientoperating conditions, start and stop phases and the consumption ofauxiliary energy of the boiler have been measured and evaluated aswell.

During the experiments, boiler inlet and outlet temperatureswere set respectively at 55 �C and 70 �C according to the test meth-odology. Class B Pt100 resistance thermometers in a 4 wires con-figuration were used to measure the flue gas temperature and amicro manometer was used for the measurement of chimney draft(stabilised at �12 Pa). The water flow through the boiler was mea-sured by an inductive flow meter and temperatures of water in for-ward and return directions were measured using Pt100temperature sensors. The boiler was mounted on a balance witha precision of 0.02 kg in order to determine the fuel consumptionrate. The flue gas velocity was determined using ‘‘Höntzsch’’ vortexflow meter, having an accuracy of 0.15 m/s. During the tests, mea-sured values of flue gas velocity ranged from 0.75 m/s to 3 m/s,depending on different test sequences. Gaseous emissions werecontinuously measured using a ‘‘M&A’’ Gas Analyser (ModelServomex 4900 for CO, CO2, O2, NOx and SOx) and a Flame Ioniza-tion Detector (Model Thermo-FID PT63FH/LT for CxHy). A heatedsampling line (JPES portable gas sampling probe by JCTAnalysentechnik GmbH) was employed. The measurement princi-ples of the gas analysers were non-dispersive infra-red (CO, CO2,NOx, SOx) and paramagnetism for O2. Each gas analyser was cali-brated with appropriate gas on zero and span points, before andafter the measurements. The measured concentrations of O2 andCO2 in the flue gases have an accuracy of 0.2% and 1% respectively.

Fig. 1. Experimental setup.

Table 1Properties of wood pellets used during the experimental tests.

Property Unit Value Testing standard

1 Gross calorific value MJ kg�1a 20.07 EN 14918 [38]2 Carbon content w-%a 50.31 EN 15104 [39]3 Hydrogen content w-%a 6.05 EN 15104 [39]4 Oxygen content w-%a 43.10 By difference5 Nitrogen content w-%a 0.16 EN 15104 [39]6 Sulphur content w-%a <0.01 EN 15289 [40]7 Ash content w-%a 0.38 EN 14775 [41]8 Moisture content w-%b 5.94 EN 14774-1 [42]

a Dry basis.b As received.

508 E. Carlon et al. / Applied Energy 138 (2015) 505–516

All the experiments were conducted using ENplus [36] certifiedwood pellets manufactured by the Austrian company RZ PelletsGmbH [37]. Wood pellets were delivered in plastic bags of 15 kgcapacity during July 2012. The chemical composition and thermalproperties of the pellets were determined by means of laboratorytests conducted according to the testing standards reported inTable 1. The efficiency of the boiler was determined during station-ary operation as well as through the reference load cycle within atolerance threshold of ±3% (as also specified in EN 303-5 [35]).

Fig. 2. Load profile in the cycle test method.

2.3. Test methods

In this study, two different test methods have been adopted. Inthe first test series the boiler is operated in stationary conditions atnominal load for a total time of 6 h [35]. The second test method isa dynamic load cycle, during which the boiler is tested in station-ary operation at different loads and in transient regime of opera-tion. The load cycle was developed as a laboratory test based ona typical daily profile of heating and domestic hot water demandof a single family house (Fig. 2) [23]. The cycle consists of five timeintervals of steady state operation, whose power level was chosenaccording to DIN 4702-8 [43], supplemented with a 100% peak loadat the beginning of the test. Each load change is obtained throughprogrammed load gradients suitable to test the dynamic behaviourof the boiler, as reported in Table 2. The duration of the load cycle,without start-up and switch-off phases, is 8 h.

The methodology adopted for the laboratory tests is shown inFig. 3. At the beginning of the test the water circulation was startedand, with the assistance of external heating, the boiler was taken toa stable initial state with both supply and return temperatures at45 �C (time = t0). At t0 the boiler was switched on and, when theincreasing rate in the flue gas temperature was lower than 0.5 Kper minute (time = t1), the actual measuring cycle (stationary mea-surement at nominal load or load cycle) started. During the boileroperation, the inlet water was maintained at 55 �C while the outletset temperature was 70 �C. By varying the water mass flow,different loads were imposed on the boiler, which modulated

Table 2Duration of the test sequences in the load cycle test.

Sequence type Percentage of nominal load (%) Duration (hh:mm:ss)

Peak 100 00:04:52Gradient 100–48 00:26:00Stationary 48 00:50:15Gradient 48–39 00:09:00Stationary 39 01:13:44Gradient 39–63 00:12:00Stationary 63 00:24:38Gradient 63–30 00:22:00Stationary 30 00:53:14Gradient 30–13 00:34:00Stationary 13 02:50:17

Fig. 3. Temperature and load profiles during the laboratory tests.

E. Carlon et al. / Applied Energy 138 (2015) 505–516 509

continuously the power output in order to maintain the outletwater temperature at 70 �C. The boiler operation continued withconstant flow and return temperatures until the end of the mea-suring cycle (time = t2). At t2 the boiler was switched off and thefirst cooling sequence started: heat extraction was carried out at20% of nominal heat output until the flow temperature loweredto 55 �C (time = t3). At this time (t3) the water circulation wasstopped, thus starting the second cooling sequence. The end of thissequence was set eight hours after the moment at which the boilerturned off (t4 = t2 + 8 h).

The load cycle was developed to reproduce the boiler operationfor heating single family house with hydronic heating systems(floor heating and radiators). In these system a three-way valvemixes hot water flow coming from the boiler with the cold waterflow returning from the radiators or the floor heating system.The boiler control system is set to maintain a constant outlet watertemperature, while the inlet water flow is continuously varied bythe three-way valve. This control algorithm is reproduced in thelaboratory by the load cycle test: with a constant set temperaturevalue at the boiler outlet, the load modulation and consequentchanges of fuel supply rate are both a function of the variablewater mass flow.

2.4. Data elaboration

The data registered during the tests were analysed in order tocalculate the energy balance of the boiler and to compare theexperimental values with the results of the model. By postprocess-ing the laboratory data, we calculated the energy input to the boi-ler, the heat losses to the flue gases and the heat transferred to thewater at every timestep of data acquisition (1 s). Mass measure-ments through a calibrated balance provided the mass flow of fueldelivered to the combustion chamber ( _mfuel) and consequently, theenergetic input to the boiler ( _Qfuel):

_Q fuel ¼_mfuelHHV

3:6 � 103 ð3Þ

The heat losses to the flue gases were calculated in accordancewith the European Standard EN 14785 [45]. The sensible heat lossis a function of the flue gas volume flow and chemical composition(Eq. (4)).

_Qfg;sens ¼ð _mfgcp;fg þ _mw;fgcp;vapÞ � ðTfg � TroomÞ

3:6 � 103 ð4Þ

In which _mfg and cp;fg are the mass flow rate and specific heat ofthe dry flue gases, _mw;fg and cp;vap are the mass flow rate and spe-cific heat of the water vapour, Tfg and Troom are respectively the fluegas temperature and the room temperature.

The chemical heat loss to the flue gases is due to their content ofunburned CO ( _mCO;fg) and was calculated with Eq. (5):

_Qfg;chem ¼_mCO;fgDHCO;comb

3:6 � 103 ð5Þ

The water vapour content of the flue gases ( _mw;fg) causes also aloss of latent heat ( _Qfg;lat) as reported in Eq. (6) [44]:

_Qfg;lat ¼_mw;fg � DHvap;w

3:6 � 103 ð6Þ

The measurements of the water temperature at the boiler inlet(Tw,in) and outlet (Tw,out) lead to the determination of the heattransferred to the water, as specified in EN 14785 (Eq. (7)) [45].

_Qwout ¼_mwcpw

ðTw;out � Tw;inÞ3:6 � 103 ð7Þ

In which _mw is the water mass flow rate and cpwits specific heat

capacity.

3. Results and discussion

3.1. Model calibration

The data required by TRNSYS components are either parame-ters, whose value is kept constant throughout the simulation, orinputs, whose values change in time. The more detailed the model,the higher the number of parameters and inputs required. TheType 869 boiler model requires 18 inputs and 51 parameters. Someparameters can be found in the boiler’s technical documentationprovided by the manufacturer [46], others must be determinedeither experimentally or default values must be assumed [47,48].By means of the laboratory tests, we determined the parametersthat are needed to calculate:

Fig. 4. Correlation between flue gas temperature and water mass flow rate.

Fig. 5. Correlation between fuel mass flow rate and excess air ratio (k).

510 E. Carlon et al. / Applied Energy 138 (2015) 505–516

– Outlet temperature of the flue gases (Tfg).– Excess air ratio (k) at different loads.– Heat losses from the boiler envelope to the environment ( _Qamb).– Electricity consumption (Pel).

The Type 869 boiler model could also be used for calculating theCO emissions of the boiler. However, as the main focus of thisstudy is the thermal performance of the boiler, the calculation ofthe emission factors was not included in our analysis. The resultsof the model calibration for the two boilers under study arereported in Table 3.

3.1.1. Flue gas temperature and excess air ratioThe parameters necessary to calculate the flue gas temperature

and the excess air ratio have been calibrated using the data mea-sured during the steady-state operation of the boiler. These datacorrespond to the test at 100% load and to the steady state timeintervals in the load cycle test.

The flue gas temperature (Tfg,out) has been calculated with the‘‘empirical delta T approach’’ [9,29]. According to this approach,the flue gas temperature is calculated as a function of the returnwater temperature (Tw,in), of the water mass flow rate ( _mwat) andof the fuel’s mass flow rate. As previously explained in Section2.3, in our laboratory tests the load modulation has been imposedonly by varying the water mass flow, with the boiler control sys-tems adjusting the fuel supply rate depending on the current load.In these conditions the fuel supply rate is a function of the watermass flow, therefore the expression of the empirical delta Tapproach is reduced to:

Tfg;out ¼ Tw;in þ dTnom þ dThx;wat 1�_mwat

_mwat;nom

� �ð8Þ

The parameter dTnom represents the difference between flue gastemperature and return water temperature at nominal load andwas obtained from the data registered during the stationary testat nominal load. To determine the parameter dThx,wat, the flue gastemperatures measured during the load cycle test in stationaryconditions were correlated to the corresponding water mass flowrates and dThx,wat was calculated with a linear regression (Fig. 4).

The excess air ratio (k) was assumed to be a linear function ofthe fuel mass flow ( _mfuel;dry):

k ¼ facA;lam þ facB;lam � _mfuel;dry ð9Þ

The two coefficients facA,lam and facB,lam have been determinedby correlating the fuel consumption during stationary operationwith the corresponding excess air ratios (Fig. 5). The uncertainties

Table 3Parameters resulting from the calibration of the boiler model.

Parameter name 6 kW boiler 12 kW boiler

Pmax 6.9 13.2Pmin 2.3 4.4Pstart 6.9 13.2Ctherm 80 80Volwat 0.03 0.03UAamb,OFF 5.09 2.90UAamb,ON 3.47 1.27Pnom 6 12dTnom 42.15 97dThx,fg 0 0dThx,wat �0.50 �1.07facA,lam 5.53 3.86facB,lam �2.18 �0.97Pel,MIN 21 27Pel,MAX 36 50Pel,OFF 7 7Wel 122 126

of the experimental temperature differences and fuel mass flowrates shown in Figs. 4 and 5 have been calculated by propagatingthe uncertainties of the measurement devices used in the labora-tory [49].

3.1.2. Heat losses from the boiler envelope to the environmentIn the Type 869 boiler model, the heat losses from the boiler

envelope to the environment are represented by the term _Q amb

(Section 2.1). Under the assumption that the boiler is a fully mixed

Unit Source

kW Boiler technical documentation [46]kW Boiler technical documentation [46]kW Laboratory datakJ K�1 Boiler technical documentation [46]m3 Boiler technical documentation [46]W K�1 Laboratory dataW K�1 Laboratory datakW Boiler technical documentation [46]K Laboratory dataK Laboratory dataK Laboratory data– Laboratory datah kg�1 Laboratory dataW Boiler technical documentation [46]W Boiler technical documentation [46]W Boiler technical documentation [46]Wh Laboratory data

E. Carlon et al. / Applied Energy 138 (2015) 505–516 511

water body, _Qamb is calculated as a linear function of the differencebetween boiler temperature and room temperature:

_Q amb ¼ UAamb � ðTboiler � TroomÞ ð10Þ

The coefficient UAamb assumes two different values, dependingon the burner operation. If the burner is on, UAamb is set to UAamb,ON,otherwise, the value changes to UAamb,OFF. The change of this coef-ficient, and in particular the increase from UAamb,ON to UAamb,OFF

allows to account for additional cooling effects occurring afterthe boiler is switched off, such as the forced air flow through theboiler induced by the chimney drought. In application to our tests,UAamb,OFF has been calculated from the data registered in thesecond cooling sequence (between t3 and t4), during which thereis no water flow through the boiler and no combustion takes place.In these conditions, the energy balance of the boiler reported in Eq.(2) can be reduced to Eq. (11):

0 ¼ _Q amb þ _QMtherm ð11Þ

with:

_Q amb ¼ UAamb;OFF � ðTboiler � TroomÞ ð12Þ

The coefficient UAamb,OFF has been calculated by means of Eq.(13), which can be obtained by integrating Eq. (12) along time, inaccordance with [14]. In Eq. (13), Ctherm is the boiler’s thermalcapacitance, s2 is the total duration of the second cooling sequence(from t3 to t4), Tboiler,start and Tboiler,end are the temperatures of thewater in the boiler respectively at the beginning (t3) and at theend (t4) of the cooling sequence, Tamb,start and Tamb,end are the corre-sponding values of the room temperature.

UAamb;OFF ¼ �Ctherm

s2� ln Tboiler;end � Tamb;end

Tboiler;start � Tamb;start

� �ð13Þ

During the second cooling sequence, the term _Q amb is the sum ofthe heat transferred from the boiler’s body to the forced air flowinduced by the fan ( _Q air) and to the external environment( _Q envelope).

_Q amb ¼ UAamb;OFF � ðTboiler � TroomÞ ¼ _Q envelope þ _Q air ð14Þ

As the term _Qenvelope represents the heat loss from the boiler’senvelope to the room air, it can be written as:

_Q envelope ¼ UAamb;ON � ðTboiler � TroomÞ ð15Þ

At every time step, the experimental value of the sensible andlatent heat loss to the cooling air ( _Qair) have been calculated fromthe laboratory data, according to Eqs. (4) and (6), therefore thecoefficient UAamb,ON could be estimated by solving Eq. (16)

ðUAamb;OFF � UAamb;ONÞ � ðTboiler � TroomÞ ¼ _Qair ð16Þ

3.1.3. Electricity consumptionThe operation of pellet boilers requires auxiliary electric energy

for fuel supply, ignition, forced draught fan and cleaning installa-tions [23]. In the Type 869 boiler model, the auxiliary electricpower is calculated as a linear function of the boiler’s load. Theelectricity supplied at maximum and minimum load define twopoints, (respectively Pel,MAX and Pel,MIN) whose fitting line allowsto calculate the auxiliary electric power in the range of powermodulation. In our analysis, the points were defined at 100% and30% load, which are the limits of the boiler’s power modulation.

During the start-up phase, the model includes the auxiliaryelectric power needed for the pellet ignition (Wel) into the energybalance of the boiler. After the start-up phase, it is assumed thatmost of the heat generated by the electric current (Joule effect) isinstantaneously lost to the ambient, and therefore does not affectthe energy balance of the boiler. For the concerned 6 and 12 kW

boilers we measured a maximum electricity consumption of36 W and 43 W respectively, which can be reasonably considereda negligible fraction of the energy balance during the whole boileroperation. Finally, when the boilers are in stand-by mode, theirelectricity consumption (Pel,OFF) decreases down to 7 W.

3.2. Model validation

In order to validate the model, both the stationary test at nom-inal load and the load cycle test have been performed several timesin the laboratory, under different ambient conditions of room tem-perature (varying between 18 �C and 30 �C) and air pressure (vary-ing between 96,900 mbar and 100,400 mbar). The same tests havebeen simulated by means of the Type 869 boiler model, to assessthe agreement between the predictions of the model and theresults of the experiments. Measured values of water mass flow,water return temperature, room air temperature, pressure and rel-ative humidity were used as input for the model, whereas theparameters identified with the calibration were kept constant.Both the stationary and the dynamic test were simulated underthe assumption that, during the boiler operation, the water tem-perature at the boiler outlet is equal to the water temperatureinside the boiler. Moreover, all the losses from the boiler to theenvironment have been grouped in the term _Q amb which representsthe heat transfer from the boiler envelope to the room air, thuseliminating from Eq. (1) the term _Qbc;camb (heat losses from thecombustion chamber). The losses to the unburnt constituents inthe residues ( _Qash) have been set to 0.02% points of efficiencyaccording to EN 14785 [45]. The technical documentation of theboiler informed that, if the water temperature inside the boilerbody exceeds 85 �C, the boiler turns off automatically.

As already mentioned in Section 2.3, during the tests the settemperature value at the boiler outlet was 70 �C. The boiler’s con-trol system adjusted the fuel supply rate and modulated the poweroutput to maintain the water outlet temperature at 70 �C. Thesame control strategy was also adopted for the simulations: theboiler model adjusted continuously the fuel supply ( _mfuel) to main-tain a constant water outlet temperature. The model was validatedby comparing the results predicted by the simulations with thevalues measured in the laboratory tests. A first evaluation regardedthe heat fluxes characterising the energy balance of the boiler, inparticular the heat transferred to water and the heat loss to the fluegases. Furthermore, the measured profiles of the outlet water tem-perature and of the water inside the boiler were compared with theprofiles resulting from the simulations. Finally, the overall fuelconsumption calculated by the model was compared with theactual fuel consumption measured by the balance. As the balancedid not register a significant mass change at every time step of dataacquisition, the comparison was done on a cumulative basis.

3.2.1. Model validation for the stationary testDuring the stationary tests, the boilers operated in steady con-

ditions at nominal load for six hours. Laboratory data registeredduring a test on the 12 kW boiler are reported in Fig. 6 togetherwith the results obtained from the simulation of the same test withthe Type 869 boiler model.

At the beginning of the test the boiler started operation with itsmaximum firing output, which was maintained for approximately15 min (Start-up phase). Simulation results show that most of theheat released by the pellet combustion was stored in the boiler’sstructure and only a minor part was transferred to the water. Asthe system approached the stationary state, the energy stored inthe boiler’s structure decreased, whereas the fraction transferredto water increased. In stationary conditions the heat flowsmaintained a constant profile with small deviations around theaverage values: approximately 85% of the energy input from the

Fig. 6. Boiler energy balance and temperature profiles during the test at nominal load.

512 E. Carlon et al. / Applied Energy 138 (2015) 505–516

fuel ( _Qfuel) was transferred to the water and 14% was lost in the fluegases. No heat was stored in the boiler’s structure, which main-tained a constant temperature. During stationary operation, thesimulated profile of the heat transferred to the water shows a rootmean square deviation (RMSD) 0.37 kW from the experimentaldata, corresponding to 3.5% of the average experimental value,whereas the heat loss to the flue gases has a RMSD 0.15 kW(7.6% of the average experimental value). The boiler turned off attime = t2, while the water circulation continued until the end ofthe first cooling sequence (until time = t3). The boiler body, stillhot, released heat to the water, therefore the boiler’s temperaturedecreased from 70 �C to 55 �C in approximately one hour. This heattransfer is shown in Fig. 6: the negative flow characterising the boi-ler body, which loses heat, corresponds to the positive heat flow tothe water.

The simulations have been performed under the assumptionthat the water outlet temperature is equal to the water tempera-ture inside the boiler body. Laboratory data show that this assump-tion is acceptable as long as there is water circulation through theboiler (until time = t3). After t3 the measured outlet water temper-ature is lower than the water temperature in the boiler: this occursbecause the measurement point of the water outlet temperature isinside the exit pipe at approximately 30 cm from the boiler. Afterthe water circulation stops, the water in the exit pipe cools downmore rapidly than the water inside the boiler, therefore, the tem-perature profile predicted by the model can be compared only withthe water temperature inside the boiler. During the second coolingsequence (between t3 and t4), the boiler body transferred heat tothe forced draught air and to the surrounding environment. Asalready explained in Section 3.1.2, the boiler model combines thesetwo mechanisms in the term ( _Qamb). An accurate calibration of theUAamb,OFF coefficient resulted in a good agreement between theexperimental temperature profile and the profile predicted by

the model, with a maximum deviation of 1.5 K (Fig. 6). As in thesecond cooling sequence there was no water circulation, the cool-ing of the boiler proceeded very slowly, with a temperature gradi-ent of 10 K in 8 h. For the stationary test, the fuel overallconsumption measured by the balance was 16.3 kg, whereas themodel calculated 15.0 kg, resulting in 90% accuracy.

3.2.2. Model validation for the load cycle testFor the load cycle test, a comparison of the laboratory data and

of the simulation results is reported in Fig. 7. In the load cycle test,a 100% peak load is imposed at the beginning of the test, followedby a load gradient leading to a stationary state at 48% load. Fig. 6shows that the boiler would require a longer time to reach themaximum heat transfer rate to the water, because at the beginningof the test most of the heat released by the pellet combustion isstored in the boiler’s structure. As the test proceeds, the boilerpasses through the programmed load gradients and stationarystates. The highest loads correspond to the highest heat transferrates to water. However, with increasing firing output, the fluegas temperature increases as well, resulting in higher heat loss tothe flue gases.

During the stationary states, the simulation results and theexperimental values show a RMSD of 0.59 kW for the heat transferrate to water and of 0.12 kW for the heat loss to the flue gases.During the time intervals of dynamic operation, the RMSD arerespectively 0.70 kW and 0.48 kW, therefore the model can repro-duce more accurately the stationary states.

In accordance with the test method, approximately after 5.5 hfrom the beginning of the load cycle, the load imposed on the boi-ler decreased to 13%, which is lower than minimum modulationrange of the boiler (30%). Because the experimental set-up didnot include a buffer storage tank, the temperature of thewater started to increase in all the circuit. As soon as the water

Fig. 7. Boiler energy balance and temperature profiles during the load cycle test.

E. Carlon et al. / Applied Energy 138 (2015) 505–516 513

temperature inside the boiler reached 85 �C, the boiler turned offautomatically and the boiler body started to release heat to the cir-culating water as shown by the negative heat flow in Fig. 7. Due tothis heat exchange, the water temperature decreased quickly andreached 55 �C after 1.5 h. At this time the boiler turned on againand maintained the maximum power output during the start-upphase. After a short time interval of load modulation, the watertemperature inside the boiler reached again 85 �C, and conse-quently the boiler turned off a second time. After the second stop,the boiler body cooled down until the end of the 8-h cycle(time = t2), when the water temperature inside the boiler was57 �C. According to the test method, the first cooling sequencelasted from time = t2 until the outlet water temperature reached55 �C. The 2 K decrease of the outlet water temperature tookapproximately five minutes. The profile of the heat flow to the boi-ler body shows that, during the start-up sequence, the majority ofthe heat released by the pellet combustion is stored in the boileritself. During stand-by mode, the heat transferred from the boilerbody to the water resulted in a fast temperature decrease whichaccelerated the cooling of the boiler. However, as soon as the watercirculation stopped (time = t3), the decrease of the boiler’s temper-ature was much slower. The experimental profile of the water tem-perature inside the boiler and the simulated profile arecharacterised by a low RMSD, as reported in Table 4. Moreover,the model predicts correctly the boiler’s starts and stops, basedon the input data of water mass flow and water inlet temperature.For the load cycle test, the overall fuel consumption estimated bythe model is 7.6 kg, which overestimates of 8.5% the experimentalvalue of 7.0 kg.

3.2.3. Comparison of stationary test and load cycle testIn order to compare the performance of the model when simu-

lating the boiler’s stationary and dynamic operation, laboratory

data have been correlated with simulation results. Referenceparameters were the heat transfer rate to water and the profilesof the water temperature inside the boiler. Each test sequencewas analysed separately, to characterise the model’s performancefor different operational regimes:

– The start-up phase, from t0 to t1.– The test sequence (stationary operation at nominal load or load

cycle test, from t1 to t2).– The stand-by phase, after the boiler is turned off (from t2 to t4).

Fig. 8 shows that for both the stationary test and the load cycletest, the model overestimates the heat transfer rate to water duringthe start-up phase, especially in the load cycle test. During thestart-up phase, the major contribution in the boiler’s energy bal-ance is the heat stored in the boiler’s structure, whose temperatureincreases from 45 �C to 70 �C, as reported in Figs. 6 and 7. As theboiler operation begins (at time = t0), the heat released by the pelletcombustion starts to diffuse from the combustion chamber into theboiler body, thus resulting in a non-uniform temperature in theboiler. On the contrary, the Type 869 model assumes that all theboiler body has the same temperature, equal to the outlet watertemperature. For the start phase, the initially overestimated boilerbody temperature leads to the overestimation of the heat transferrate to the water and to underestimation of the heat stored in theboiler body. The difference between simulation results and exper-imental data decreases as the boiler approaches the stable state atend of the start-up phase (time = t1). At this point, the model over-estimates the heat transfer rate to the water of 10% for the station-ary test and of 30% for the load cycle test.

For the stationary test, the experimental values of the heattransfer rates to water during operation at nominal load are inthe range from 9 to 12 kW, whereas the rates estimated by the

Table 4Results of model validation for the two boilers.

12 kW boiler 6 kW boiler

Stationary test Load cycle test Stationary test Load cycle test

Water temperature inside boilerMax. deviation during start up phase (K) 1.75 6.09 7.15 2.96Max. deviation during test sequence (t1–t2) (K) 2.62 4.39 4.99 6.59Max. deviation in stand-by mode (t2–t4) (K) 5.18 2.82 3.79 2.67Correl. Coefficient during start up phase (–) 0.996 0.970 0.954 0.965Correl. Coefficient during test sequence (t1–t2) (–) 0.712 0.836 0.603 0.832Correl. Coefficient in stand-by mode (t2–t4) (–) 0.983 0.994 0.988 0.996RMSD during start up phase (K) 0.71 3.86 4.87 2.12RMSD during test sequence (t1–t2) (K) 1.36 2.63 1.45 2.21RMSD in stand-by mode (t2–t4) (K) 1.50 3.46 1.05 0.53

Heat transfer rate to waterMax. deviation during start-up phase (kW) 1.64 5.03 1.78 1.02Max. deviation during test sequence (t1–t2) (kW) 1.31 2.46 0.60 1.18Max. deviation in stand-by mode (t2–t4) (kW) 0.83 1.68 0.41 0.36Correl. Coefficient during startup phase (–) 0.991 0.871 0.929 0.877Correl. Coefficient during test sequence (t1–t2) (–) 0.290 0.865 0.011 0.867Correl. Coefficient in stand-by mode (t2–t4) (–) 0.976 0.977 0.872 0.890RMSD during start up phase (kW) 0.23 3.22 1.13 0.74RMSD during test sequence (t1–t2) (kW) 0.37 0.67 0.30 0.31RMSD in stand-by mode (t2–t4) (kW) 0.44 0.14 0.19 0.14

0

2

4

6

8

10

12

14

0 2 4 6 8 10 12 14

Start Phase

Load Cycle

Standbyy = x

0

2

4

6

8

10

12

14

0 2 4 6 8 10 12 14

Start Phase

Stationary

Standby

y = x

STATIONARY TEST LOAD CYCLE TEST

Heat to water – Experimental kW Heat to water – Experimental kW

Hea

t to

wat

er –

Mod

el k

W

Hea

t to

wat

er –

Mod

el k

W

Fig. 8. Correlation between experimental values and simulation results – heat transferred to water.

514 E. Carlon et al. / Applied Energy 138 (2015) 505–516

model are between 10.5 and 12 kW (Fig. 8). All the points referredto the operation at nominal load are included in a margin of ±10%from the identity line, having the equation ‘‘y = x’’. For the loadcycle test, the overall correlation coefficient (R2) between labora-tory data and simulation results, calculated throughout the 8-hcycle, is 0.865. The maximum deviation from the identity line is30%, showing that the model reproduces stationary operation bet-ter than dynamic operation.

The cooling sequences (from t2 to t4), are characterised by highcorrelation coefficients both in the stationary test (R2 = 0.976) andin the load cycle test (R2 = 0.977), showing that the heat exchangebetween boiler body and surrounding environment, which occursafter the boiler has turned off, was accurately modelled.

The profiles of the water temperature inside the boiler show astrong correlation for both tests (Fig. 9) and during all the testsequences. The maximum deviation between simulation resultsand laboratory data is 6 K and corresponds to the beginning ofthe start phase. As explained above, this is due to the assumptionof a uniform temperature of the boiler body, which leads to a lessaccurate energy balance during the start-up phase.

3.2.4. Comparison of 12 kW and 6 kW boilerIn this study, the Type 869 boiler model was validated for two

commercially available pellet boilers, having a nominal capacity

of 6 kW and 12 kW. A comparison of the validation results, for testsperformed on different boilers, is reported in Table 4.

As demonstrated by the high correlation coefficients and lowRMSD, the simulated water temperature profiles generally showa good agreement with the experimental data for both boilers:the maximum deviation (7 K) was registered for the 6 kW boiler.The heat transfer rates to water have low deviations from theexperimental data in both the test sequence (stationary operationor load cycle) and in the cooling sequences. During stationary oper-ation, the model calculated a constant heat transfer rate, whereasthe experimental data showed small fluctuations around theaverage value. These fluctuations are not correlated with thesimulation results as evidenced by the low correlation coefficients(0.29 and 0.01 for the 12 kW and the 6 kW boiler respectively).Moreover, the low RMSD (0.37–0.30 kW) show that the constantheat transfer rates estimated by the model are very representativeof the average experimental values. As already mentioned inSection 3.2.3, during the start-up phase, the model overestimatesthe heat transfer rates to water, especially for the 12 kW boiler.

4. Conclusions and future work

In this work the TRNSYS boiler model Type 869, suitable fordynamic building simulations, has been calibrated and validated

0

10

20

30

40

50

60

70

80

90

0 10 20 30 40 50 60 70 80 90Boiler Temperature - Experimental °C

Boi

ler

Tem

pera

ture

-M

odel

°C

Start PhaseLoad CycleStandbyy = x

0

10

20

30

40

50

60

70

80

90

0 10 20 30 40 50 60 70 80 90Boiler Temperature - Experimental °C

Boi

ler

Tem

pera

ture

-M

odel

°C

Start PhaseStationary Standbyy = x

STATIONARY TEST LOAD CYCLE TEST

Fig. 9. Correlation between measured and simulated water temperature profiles inside the boiler.

E. Carlon et al. / Applied Energy 138 (2015) 505–516 515

for two commercially available pellet boilers, with reference to lab-oratory data. Results showed that the model described with a highaccuracy the stationary operation of the boiler as well as the heatexchange between the boiler body and the surrounding environ-ment occurring after the boiler had turned off. The simulation ofthe boiler’s dynamic operation was characterised by a lower accu-racy, especially in the start-up sequence. The heat transfer rates towater were predicted with a maximum deviation of 10% duringstationary operation, and a maximum deviation of 30% duringthe dynamic load cycle. For both tests the fuel consumption waspredicted within a 10% deviation from the experimental values.

Concerning model calibration, the test methods provided suffi-cient data for the complete calibration of the model. The flue gastemperature was calculated with a reduced expression of the‘‘empirical delta T approach’’, as a function of the return watertemperature and of the water mass flow. The function was cali-brated with experimental data registered during the dynamic loadcycle test. This procedure can be adopted only when boiler’s con-trol algorithm is based on the water’s set temperature at the outlet.This type of control is typical for boilers integrated in hydronicheating systems. For control algorithms imposing a variable com-bustion power at constant water mass flow rate, the reducedexpression of the ‘‘empirical delta T approach’’ would introducean uncertainty in the calculation of the flue gas temperature in loadmodulation regime.

This work showed that the Type 869 boiler model is suitable todescribe the stationary and dynamic operation of the analysed pel-let boilers as well as the heat transfer from the boiler body to thesurrounding environment during stand-by mode. Moreover, thisstudy developed a procedure to calibrate the Type 869 boilermodel by means of laboratory tests. The adopted test methods(in particular the load cycle test) could be used as standard testfor the calibration and validation of boiler models under dynamicconditions.

The next step of this research will be the dynamic simulation ofsingle family houses, heated by the same boilers pellet as thoseconsidered in this study. The simulation shall include the buildingenvelope and the heating and domestic hot water supply system,equipped with a pellet boiler. It is expected that simulation resultswill give a significant contribution to assess the field performanceof the boilers under investigation.

Acknowledgment

This study has been performed in the frame of the BioMaxEffproject [32]. The research leading to these results has received

funding from the European Union Seventh Framework Programme(FP7/2007-2013) under Grant Agreement n� 268217.

References

[1] Directive 2009/28/EU of the European Parliament and of the Council of 23 April2009 on the promotion of the use of energy from renewable sources andamending and subsequently repealing Directives 2001/77/EC and 2003/30/EC.

[2] Directive 2010/31/EU of the European Parliament and of the Council of 19 May2010 on the energy performance of buildings (recast).

[3] Míguez JL, Morán JC, Granada E, Porteiro J. Review of technology in small-scalebiomass combustion systems in the European market. Renew Sustain EnergyRev 2012;16(6):3867–75.

[4] Fiedler F. The state of the art of small-scale pellet-based heating systems andrelevant regulations in Sweden, Austria and Germany. Renew Sustain EnergyRev 2004;8(3):201–21.

[5] Verma VK, Bram S, De Ruyck J. Small scale biomass heating systems: standards,quality labelling and market driving factors – an EU outlook. BiomassBioenergy 2009;33(10):1393–402.

[6] Raffenel Y, Fabrizio E, Virgone J, Blanco E, Filippi M. Integrated solar heatingsystems: from initial sizing procedure to dynamic simulation. Sol Energy2009;83:657–63.

[7] Fiedler F, Nordlander S, Persson T, Bales C. Thermal performance of combinedsolar and pellet heating systems. Renewable Energy 2006;31(1):73–88.

[8] Persson T, Nordlander S, Rönnelid M. Electrical savings by use of pellet stovesand solar heating systems in electrically heated single-family houses. Energyand Buildings 2005;37(9):920–9.

[9] Haller MY. Combined solar and pellet heating systems – improvement ofenergy efficiency by advanced heat storage techniques, hydraulics, andcontrol. PhD Thesis, Graz University of Technology; 2010.

[10] Rabaçal M, Fernandes U, Costa M. Combustion and emission characteristics ofa domestic boiler fired with pellets of pine, industrial wood wastes and peachstones. Renewable Energy 2013;51:220–6.

[11] Carvalho L, Wopienka E, Pointner C, Lundgren J, Verma VK, Haslinger W, et al.Performance of a pellet boiler fired with agricultural fuels. Appl Energy2013;104:286–96.

[12] Haberl R, Haller MY, Konersmann L. Solar & pellet heating: specifications forhigh efficiency system design 2011 ISES Solar Word Congress, <http://www.solarenergy.ch/fileadmin/daten/publ/Haberl_Solar_Pellet_Heating_SWC_2011.pdf> [accessed 10.06.14]

[13] Verma VK, Bram S, Gauthier G, De Ruyck J. Evaluation of the performance of amulti-fuel domestic boiler with respect to the existing European standard andquality labels: Part-1. Biomass Bioenergy 2011;35:80–9.

[14] Haller MY, Paavilainen J, Konersmann L, Haberl R, Dröscher A, Frank E, et al. Aunified model for the simulation of oil, gas and biomass space heating boilersfor energy estimating purposes. Part II: parameterization and comparison withmeasurements. J Build Perform Simulat 2011;4(1):19–36.

[15] Persson T, Fiedler F, Nordlander S. Methodology for identifying parameters forthe TRNSYS model Type 210 – wood pellet stoves and boilers. Version17.04.2012. <http://www.diva-portal.org/smash/get/diva2:522996/FULLTEXT01.pdf> [accessed 10.06.14].

[16] Glembin J, Rockendorf G. Abschlussbericht: Integration von Heizkessel inWärmeverbundsysteme mit großen Solaranlagen – Teil 2: Kesselmodell undParametrierung, <http://www.delta-q.de/export/sites/default/de/downloads/Endbericht_T2_Kesselmodell.pdf>; 2012 [accessed 10.06.14].

[17] Verma VK, Bram S, Delattin F, De Ruyck J. Real life performance of domesticpellet boiler technologies as a function of operational loads: a case study ofBelgium. Appl Energy 2013;101:357–62.

516 E. Carlon et al. / Applied Energy 138 (2015) 505–516

[18] Verma VK, Bram S, Delattin F, Laha P, Vandendael I, Hubin A, et al. Agro-pelletsfor domestic heating boilers: standard laboratory and real life performance.Appl Energy 2012;90(1):17–23.

[19] Win KM, Persson T, Bales C. Particles and gaseous emissions from realisticoperation of residential wood pellet heating systems. Atmos Environ2012;59:320–7.

[20] Ozgen S, Caserini S, Galante S, Giugliano M, Angelino E, Marongiu A, et al.Emission factors from small scale appliances burning wood and pellets. AtmosEnviron 2014;94:144–53.

[21] Haller MY, Haberl R, Persson T, Bales C, Kovacs P, Chèze D, et al. Dynamicwhole system testing of combined renewable heating systems – the currentstate of the art. Energy and Buildings 2013;66:667–77.

[22] Panaras G, Mathioulakis E, Belessiotis V. A method for the dynamic testing andevaluation of the performance of combined solar thermal heat pump hot watersystems. Appl Energy 2014;114:124–34.

[23] Heckmann M, Friedl G, Schwarz M, Rossmann P, Hartmann H, Baumgartner Het al. Bestimmung von Jahresnutzungsgrad und Emissionsfaktoren vonBiomasse-Kleinfeuerungen am Prüfstand, <http://www.bioenergy2020.eu/files/publications/pdf/815650_Jahresnutzungsgrad_Endbericht.pdf>; 2010[accessed 10.06.14].

[24] Bales C. Combitest – a new test method for thermal stores used in solarcombisystems PhD thesis, Department of building technology. Göteborg,Sweden: Chalmers University of Technology; 2004.

[25] Haberl R, Frank E, Vogelsanger P. Holistic system testing-10 years of concisecycle testing. In: Proc. of the solar world congress 2009, Johannesburg,South Africa, <http://www.solarenergy.ch/fileadmin/daten/publ/202_Haberl_Frank_Holistic_System_Testing_FullPaper.pdf>; 2009 [accessed28.08.14].

[26] Persson T, Fiedler F, Nordlander S, Bales C, Paavilanen J. Validation of adynamic model for wood pellet boilers and stoves. Appl Energy 2009;86(5):645–56.

[27] Persson T. Combined solar and pellet heating systems for single-family houses– how to achieve decreased electricity usage, increased system efficiency andincreased solar gains, PhD thesis, Department of Energy and EnvironmentalTechnology KTH – Royal Institute of Technology Stockholm, Sweden, <http://www.diva-portal.org/smash/get/diva2:523356/FULLTEXT01.pdf>; 2006[accessed 28.08.14].

[28] Klein SA et al. TRNSYS 17, a TRaNsient SYstem Simulation program – volume 1getting started. Madison, USA: Solar Energy Laboratory, University ofWisconsin-Madison, <http://www.transsolar.com/__software/download/en/trnsys_shortinfo_en.pdf>; 2009 [accessed 10.06.14].

[29] Fiedler T, Knirsch A. Add-on to hitherto existing gas fuelled boiler-modelfor a wood-fired boiler with calorific value usage, <http://www.transsolar.com/__software/download/en/ts_type_370a_en.pdf>; 1996 [accessed 10.06.14].

[30] Haller MY, Paavilainen J, Konersmann L, Haberl R, Dröscher A, Frank E, et al. Aunified model for the simulation of oil, gas and biomass space heating boilersfor energy estimating purposes. Part I: model development. J Build PerformSimulat 2011;4(1):1–18.

[31] SPF Institut für Solartechnik, Oberseestrasse 10, 8640 Rapperswil, Switzerland<www.solarenergy.ch> [accessed 16.08.14].

[32] BioMaxEff project website <www.biomaxeff.eu> [accessed 30.06.14].[33] Windhager Zentralheizung Technik GmbH, Anton-Windhager-Straße 20, 5201

Seekirchen, Österreich <http://www.windhager.com/at/kontakt/> [accessed30.06.14].

[34] EN 303-4 Heating boilers – Part 4: heating boilers with forced draughtburners; special requirements for boilers with forced draught oil burners withoutputs up to 70 kW and a maximum operating pressure of 3 bar;terminology, special requirements, testing and marking; 1999.

[35] EN 303-05 Heating boilers – Part 5: heating boilers for solid fuels, manuallyand automatically stoked, nominal heat output of up to 500 kW – terminology,requirements, testing and marking; 2012.

[36] European pellet council handbook for the certification of wood pellets forheating purposes; 2013. <http://www.enplus-pellets.eu/wp-content/uploads/2012/01/ENplus-Handbook-2.0.pdf>; 2013 [accessed 10.06.14].

[37] RZ Pellets GmbH, Bahnhofstraße 32, A-3370 Ybbs, Lower Austria www.rz-pellets.at [accessed 11.06.14].

[38] EN 14918 Solid biofuels – determination of calorific value; 2010.[39] EN 15104 Solid biofuels – determination of total content of carbon, hydrogen

and nitrogen – instrumental methods; 2011.[40] EN 15289 Solid biofuels – determination of total content of sulfur and

chlorine; 2011.[41] EN 14775 Solid biofuels – determination of ash content; 2009.[42] EN 14774-1 Determination of moisture content – oven dry method Part 1:

total moisture – reference method; 2010.[43] DIN 4702-8: Central heating boiler; determination of the standard efficiency

and the standard emissivity Deutsches Institut Fur Normung; 1990.[44] Neuenschwander P, Good J, Nussbaumer T. Combustion efficiency in furnaces

with flue gas condensation. In: Proc. of the 10th European conference andtechnology exhibition – biomass for energy and industry, Würzburg, <http://www.bfe.admin.ch/php/modules/enet/streamfile.php?file=000000000374.pdf&name=000000190339>; 1998 [accessed 10.06.14].

[45] EN 14785 Residential space heating appliances fired by wood pellets –requirements and test methods; 2006.

[46] Operating Manual of Windhager VarioWIN Pellet central heating boilerpneumatic + direct <http://www.windhager.com/fileadmin/user_upload/VarioWIN_OP_Manual0810.pdf> [accessed 30.06.14].

[47] Huber-Fauland H, Wertz D, Ponweiser K, Haider M. Simulatonsvergleich vonPelletskesselmodellen Pages 222-227. In: Proc. of BauSIM-IBPSA, Wien,<http://info.tuwien.ac.at/bausim/conftool/huber-fauland-2010-simulationsv-ergleich_von_pelletskesselmodellen-130.pdf>; 2010 [accessed 29.08.14].

[48] Stift F, Hartl M, Ferhatbegovic T, Aigenbauer S, Simetzberger A. Model basedoptimization of a combined biomass-solar thermal system. Energy Procedia2011;48:681–8.

[49] JCGM 100:2008 GUM 1995 with minor corrections evaluation of measurementdata – guide to the expression of uncertainty in measurement, <http://www.bipm.org/utils/common/documents/jcgm/JCGM_100_2008_E.pdf>; 2008[accessed 17.06.14].