stefan wischhusen, bruno lüdemann, gerhard schmitz … · 2010. 10. 23. · s. wischhusen, b....

Proceedings

of the 3rd International Modelica Conference, Linköping, November 3-4, 2003,

Peter Fritzson (editor)

Paper presented at the 3rd International Modelica Conference, November 3-4, 2003, Linköpings Universitet, Linköping, Sweden, organized by The Modelica Association and Institutionen för datavetenskap, Linköpings universitet All papers of this conference can be downloaded from http://www.Modelica.org/Conference2003/papers.shtml Program Committee Peter Fritzson, PELAB, Department of Computer and Information Science,

Linköping University, Sweden (Chairman of the committee). Bernhard Bachmann, Fachhochschule Bielefeld, Bielefeld, Germany. Hilding Elmqvist, Dynasim AB, Sweden. Martin Otter, Institute of Robotics and Mechatronics at DLR Research Center,

Oberpfaffenhofen, Germany. Michael Tiller, Ford Motor Company, Dearborn, USA. Hubertus Tummescheit, UTRC, Hartford, USA, and PELAB, Department of

Computer and Information Science, Linköping University, Sweden. Local Organization: Vadim Engelson (Chairman of local organization), Bodil Mattsson-Kihlström, Peter Fritzson.

Stefan Wischhusen, Bruno Lüdemann, Gerhard Schmitz Department of Technical Thermodynamics, TU Hamburg-Harburg; Imtech Deutschland GmbH, Germany: Economical Analysis of Complex Heating and Cooling Systems with the Simulation Tool HKSim pp. 259-268

Economical Analysis of Complex Heating and Cooling Systemswith the Simulation Tool HKSim

Dipl.-Ing. St. Wischhusen1�

Dr.-Ing. B. Lüdemann2† Prof. Dr.-Ing. G. Schmitz1

1 Technical University Hamburg–Harburg, Department of Technical Thermodynamics,Denickestr. 17, D–21073 Hamburg.

2 Imtech Deutschland GmbH & Co. KG, Zentrale Ingenieurtechnik,Tilsiter Str. 162, D–22047 Hamburg.

Abstract

Dynamic simulations of energy systems are essentialwhen it comes to transient analysis and design of com-plex plants. Besides the choice of efficient subcom-ponents, like boilers, pumps or chillers, the controlstrategies have a large impact on the running costs ofa cooling, heating or combined heating and coolingplant. This paper describes an applied simulation toolfor heating and cooling systems. The economical ben-efits are discussed by means of a typical application:the simulation and optimisation of a complex indus-trial energy system.

1 Introduction

In cooperation with Imtech Deutschland GmbH & Co.KG (formerly known as Rudolf Otto Meyer GmbH &Co. KG and Rheinelektra Technik) a research projectwas conducted. The aim of the project was to developa simulation tool, called HKSim [1, 2], for heating(Fig. 1) and cooling systems in building applications.This tool enables configuration studies and dynamicsystem simulations with time scales from a few sec-onds up to one year. For this purpose the simulationenvironment of Dymola [3], containing the object-oriented modelling language Modelica, is used tomodel complex heterogeneous systems. The graphicaluser interface, including the integration of Dymolaand a data base for project management, was createdby the department “Zentrale Ingenieurtechnik” ofImtech Deutschland while the model libraries [4] weredeveloped at the Department of Technical Thermody-namics at the Technical University Hamburg–Harburg.

�

[email protected], tel.: +49–40–42878–3267†[email protected], tel.: +49–40–6949–2546

Heating centre

Consumers

Electricmeters

Gasmeter

Heatingrequirement(ext. files)

Pipework

Figure 1: System schematic of a heating centre withdistributed consumers and an earth–laid pipework

The component models are focused on the simulationof entire years. Therefore, the model equations haveto be formulated as efficient as possible. The modeldesign philosophy, which results from this importantrequirement, will be discussed in detail with respect totypical system components. The components can bemostly parameterised using manufacturer informationor values resulting from own measurements. Thehandling of the models is primarily focused on userswho want to use the models as they are provided orwith alternative parameter settings. Expert–users areable to exchange model equations (e.g. models for

S. Wischhusen, B. Lüdemann, G. Schmitz …Analysis of Heating and Cooling Systems with HKSim

The Modelica Association Modelica 2003, November 3-4, 2003

medium properties, pressure losses and heat transfer)by means of replaceable models and develop theirown components from the existing base classes. Theboundary conditions of the system simulation can besupplied by data files from a building simulation oreven measurement data.

Due to the separation of the building from the systemsimulation an efficient calculation for performancestudies is realised when simulating complex plants.

The simulation tool HKSim was used successfully inseveral projects for the economical analysis of en-ergy systems. In this article a typical project willbe described, beginning with the selection of compo-nent models, followed by the consideration of individ-ual control elements and determination of necessaryboundary conditions. The last item usually consists oflocal weather data and calculated or measured heatingand / or cooling requirement.

2 Current Library Content

So far, the most important components of cooling andheating systems have been supplied by the model li-braries. All components are compatible by using iden-tically defined hydraulic interfaces. Some elementswhich have been modelled and integrated into the li-braries are [4]:

� normal, low-temperature and condensing boilers,

� cogeneration plants,

� consumers for heating, cooling and domestic wa-ter,

� pipes and storage tanks,

� splits and joints, mixing valves,

� controlled and uncontrolled pumps,

� heat exchangers [1],

� mechanical driven chillers,

� absorption chillers,

� cooling towers and dry coolers,

� special controllers (beyond: utilisation of Model-ica’s standard libraries [5]),

� electric and gas meter, oil supply.

Nevertheless, the development of new parts is contin-uing. In future, a model for different types of fuel cellswill also be offered.

Figure 2: Screenshot of selected packages of typicalheating and cooling components

In practice, it is important to have models with differ-ent levels of control and efficiency descriptions. Usu-ally, a new project starts with the modelling of a sys-tem base layout. This first approach takes a lot of timebecause a correct understanding of the system’s ther-mal and hydraulic behaviour is required. Generally,every single component which is part of the whole sys-tem should be modelled as accurate as possible. Un-fortunately, this requirement contradicts most compo-nent manufacturers’ information policies. Addition-ally, information of existing plants is often incompleteor not available. For this reason, models with a differ-ent depth of physical description are supplied by the li-braries. For example, one can start with a boiler modelwhich has a constant efficiency with the opportunity toreplace it with a refined model later on.For the user it is important, that the models use pa-rameters, which are easily available or which can bedetermined without model knowledge.

3 Applied Thermal and HydraulicModel Concept

The models are designed to enable a quick synthesis ofplant models. This requirement is already consideredin the modelling process in such a way that emphasisis placed on the calculation of the thermal behaviour.The hydraulic behaviour of a plant is not neglected but



simplified in that direction, that the mass flow can bedirectly influenced at split and joint elements. The dis-play of a transient pressure is not possible in favour ofa fast and stable simulation. In combination with theconcept of a load dependent simulation, designed tosatisfy thermal loads, the calculation of the mass flowrate ṁ is implemented in the model of the thermal con-sumers. Applying the first law of thermodynamics, theheat (or cooling) requirement Q̇build of the building (orconsumer, resp.) equals:

Q̇build � ṁ � c � � ϑ f � ϑr �� (1)If the feed temperature ϑ f of the liquid and its heatcapacity c is given two unknowns remain in this alge-braic equation. By a second equation the dependencyof the mass flow and the return temperature ϑr can bemodelled applying heat transfer laws or measurementdata. A simple but efficient approach is a proportionalgain k of that mass flow rate, which is theoreticallynecessary to deliver the needed heat under the assump-tion of a perfect heat transfer (ϑr � ϑbuild )

ṁ � k � Q̇buildc�ϑ f � ϑbuild � � (2)

The constant gain factor k may vary between 1 . . . 3.Another problem is that the mass flow rate shouldnever be higher than the rated capacity of the installedpumps. In order to take this important limitation intoaccount the rated mass flux of the pump ṁmax is addedto the hydraulic interfaces (see Fig. 3). Furthermore,this value is divided at splitting elements with regardto the actual load q � Q̇build of parallel consumers.

Figure 3: Input and output signals of the consumermodel

In addition to this, the pumps head H must overcomethe pressure losses of the plant. This is checked duringan initial calculation assuming worst case conditions.The pressure check functionality is implemented in themodel of an expansion vessel, which is also used as asink for the algebraic signals (e.g. mass flow) in closedloops.

Expansion Vessel

Boiler

Gas meter Electric meter

Pump Feed pipe

Consumer

Return pipe

Heat demand

Temperature

Data sink Data source

Figure 4: Simple heating plant

The base configuration for a simple heating plant isshown in Fig. 4. A minimum composition for a de-termined equation system must at least consist of thefollowing three components:

� a consumer model for the calculation of a massflow signal,

� a pump model for limiting the maximum massflow rate and

� an expansion vessel which checks for a suffi-cient pump head (under worst case conditions)and avoids algebraic signal loops (e.g., mass flowrate).

Since the hydraulic simplifications lead to a het-erogeneous data flow the user has to follow a fewrules during the plant model generation. Those rulesare visually supported by coloured interfaces whichidentify sources (triangle with black background),sinks (grey background) and neutral components(no background). Following the rules even complexplant models can be composed by the user withoutcausing over- or underdetermined system models.Furthermore, some models must allow a flexible dataflow due to numerical reasons. For example, the pumpmodel can be switched into a mass flow source, whenit is used in independent circuits without consumers.This is a common situation in complex applicationswhere hydraulic bypasses are needed to realise a failsafe control strategy. By introducing a structuralparameter it is possible to switch the model equationson demand. With regard to this feature on one handand the connection rules on the other hand it would



be very convenient for the model developer and themodel user to have graphical annotations which canbe controlled by structural parameters. Additionally,an update of graphical icons, when replaceable objectslike interfaces are exchanged, would be very helpful.

To verify the used hydraulic concept, it was validatedwith measurement data from a complex coolingsystem displayed in Fig. 5. The chillers are shownon the left hand side of the picture including oneconventional water chiller and two heat exchangerswhich are part of a chiller / ice storage circuit. Thetotal capacity is 1,600 kW at a system temperatureof 7 � C/ 15 � C. The cold water pump of chiller #3operates with a constant speed in contrast to all otherpumps, which are controlled. Therefore, a hydraulicbypass between the feed and return duct is necessary(see Fig. 5).

Chiller 3

HX of chiller 2

HX of chiller 1

Bypass

7°C

7°C

15°C

Consumers

dp-const pumps

J

Qdot

mdot

mdot

mdot

J

J

J

Figure 5: Schematic of the original plant and locationof measuring points

As an input to the simulation the output temperaturesof the chiller and heat exchangers were provided aswell as the mass flow rates of the controlled pumps.Furthermore, the cold water mass flow through thebuildings and the cooling requirement is known.Since the returning water’s temperature is an im-portant system variable it can be compared with themeasurement to determine the quality of the consumermodel description.

The corresponding system model to the descriptionabove is shown in Fig. 6. In this model the cold watertemperature and flow is already merged to a singleflux in the source model on the left hand side. Apotential overshoot of cold water can be passed tothe return side through the bypass. Moreover, thefour feed pumps and five buildings are modelled as asingle feed model with the same capacity since a local

deviation cannot be resolved by the measurementdata. The consumer model refers to the equations 1and 2. The model can be adjusted rather easily bytwo parameters: the constant gain, which was setto k � ϑr � ϑ fϑbuild � ϑ f � 1 � 75 and the building referencetemperature ϑbuild � 21 � C. The simulation wascarried out with measurement data of one week duringNovember with a peak load of approximately 530 kWand base load of 300 kW at the weekend (16th and17th of Nov.).

Pump Feed pipe

Consumer

Return pipe

Electric meter

Cooling demand

Average roomtemperature

Bypass

Source

Sink

Figure 6: Modelica model of the cooling plant

Comparing the temperature of the returning waterafter the bypass in Fig. 7, a very good agreement withthe measured values can be found. Some very fewexceptions are due to the lack of information aboutthe exact switch off times of chiller #3. Here onlythe output temperature was available. Instead, it wasassumed that a cold water temperature of more than7 � C indicates a turned off chiller. Knowing that thewater temperature during operation usually variesbetween 6.0 and 6.8 � C and the outlet temperature dueto a good insulation increases from 6 � C to 20 � C inmore than 2 days, the temperature deviation shown inFig. 7 can be explained.

The validation of the mass flow rate shown in Fig.8 reveals that this value depends strongly on thecooling load. Obviously, there are periods where theagreement of measurement and simulation is verygood but on some days there is a higher deviation. Ina sensitivity analysis it was found that the buildingreference temperature of ϑbuild � 21 � C may have alarge influence on the mass flow rate. Especially, Nov.



17.11 18.11 19.11 20.11 21.11 22.11 23.118

9

10

11

12

13

14

15

16

Time[dd.mm]

[°C

]ϑ

ret Sim

ϑret

Mes

Figure 7: Comparison of the return temperature

19th and 20th were colder days ( � 0 � C) in contrastto the 22nd which had a higher ambient temperature(4 � C). Hence, it can be suggested that the unknownaverage room temperature was not constant duringthe measurement. Changing the temperature by 1 or2 � C gives the right mass flow rate.

17.11 18.11 19.11 20.11 21.11 22.11 23.115

10

15

20

25

30

35

40

Time[dd.mm]

[kg/

s]

17.11 18.11 19.11 20.11 21.11 22.11 23.110

100

200

300

400

500

600

[kW

]

mdot Simmdot MesQdot

Figure 8: Cooling requirement and comparison of thewater mass flow rate

It should be pointed out that the consumer model pre-sented in this paper though it is based on a very simpleapproach can predict the mass flow and the return tem-perature with a sufficient accuracy. In addition to this,it enables a very fast model generation and simulationwith computation times of a few seconds for a wholeyear. The model is also equipped with an overload rou-tine which enables stable simulations when the heatingand cooling capacity is too small for peak loads. Insuch an event the lack of energy will be compensatedwhen the demand decreases again. A flag variable in-dicates that the plant’s capacity is not sufficient.

4 Simulation of a Complex IndustrialEnergy Supply System

Within the development of an innovative energysupply system for a production facility the simulationtool HKSim was used to analyse the efficiency ofthe design concept and the running costs. Afterthe simulation of the original layout optimisationmeasures were developed to increase the economicaland technical efficiency.

The plant produces heat for the production lines, theheating system and the domestic hot water supply.Cold water is needed for industrial cooling processesand for air conditioning. The total cooling requirementreaches a maximum level of 2.6 MW in summer. Themain idea behind the given plant schematic in Fig. 9is to save primary energy by reusing as much wasteheat as possible. Therefore, the heat needed for theperiodic production processes is recovered and the su-perheat of the refrigerant after the compressor and thesuperheat of the compressed air is also transferred tothe heating system. To ensure a minimum feed tem-perature a steam heat exchanger is implemented as abackup heat source. For cooling purposes, the con-tinuous fresh water supply for steam production anddomestic use is treated as a heat sink.Since the continuous operation of the production lineshas the highest priority it has to be ensured that theproduction lines are supplied with a sufficient amountof cooling water independent of the actual heat de-mand of the associated heating system. Hence, cool-ing towers with a large capacity were installed as abackup. Two of the three existing cooling towers canbe switched between cooling of production lines, aircompressors and chillers or free cooling of returningcold water to increase the average utilisation. The wa-ter of the cooling system is stored in two parallel tankswith a capacity of 700 m3.The following questions and tasks were identified andshould be clarified by means of the system simulation:

1. How much heat can be recovered to decrease theadditional heat input for the facilities heating? Acoverage of 70% by waste heat would allow acheaper building insulation with regard to Ger-man regulations.

2. Which inexpensive optimisation measures couldhelp to realise a further reduction of runningcosts?



Figure 9: Simplified schematic of the original plantconcept

Starting with the modelling of the original schematicdisplayed in Fig. 9 the plant model is divided intothree supermodels (modules): the “heating module”(shown in Fig. 10), the “cooling module” integratingthe cooling towers and the “cold water module”.The model of the heating and the production heatrecovery system consists of eight different consumers(1), a boiler model representing the backup steamheat exchanger (2), four heat sources standing forthe production heat recovery (8) and a number ofhydraulic interfaces to the two other modules (4, 6,7). The modules are connected in a supermodel. Thereason for dividing the plant into subsystems is thatthe schematic is much clearer and the graphical updateof Dymola works quicker, too.

After the model generation all necessary boundaryconditions have to be determined. This data set in-cludes the ambient temperature and humidity which isprovided by a test reference year of the correspondingregion in Germany. This fundamental input is alsoused in combination with the known rated loads toderive the thermal demand profiles for the air condi-tioning and heating with a simple linear approach. Bymeans of measurements carried out on the existingproduction processes the possible heat recovery wasdetermined and implemented using table interpolationmodels. The production processes can be describedby a characteristic, transient heat output (Fig. 11).As a result from that fact, the outlet temperature ofthe heat exchanger’s cold side varies between 45 and95 � C during a period of a few minutes.

Based on a simulation of the original plant layout, po-tential modifications for an improved performance aredetermined, regarding the hydraulic circuit and gov-erning control of the plant and its components:

0

200

400

600

800

1000

1200

1400

1600

0,0 2,5 5,0 7,5 10,0 12,5

Time [h]

[kW

]

Figure 11: Periodic heat output of one production pro-cess

� Modification 1: All heat exchangers connectedto the return side of the cold water storage tankare connected to the return side of the cold waterconsumers, whereby the heat exchanger’s inputtemperature is increased since the mass flow ofthe chiller pumps is always higher than the massflow of the main pumps.

� Modification 2: A fixed periodical day / nightswitch of the cooling towers is replaced by atemperature dependent control to account for achanging level of primary cooling demand of theproduction lines. This measure shall increase thefree cooling capacity, when more heat is trans-ferred to the heating system.

� Modification 3: Cooling water with a tempera-ture level below the desired value of the feed tem-perature is hydraulically connected to the coolingheat exchangers.

� Modification 4: The set point of the heating feedtemperature is dependent on the actual heat de-mand and can be decreased if high temperaturesare not necessary.

The simulations of the actual and the modified layoutare carried out with identical boundary conditionsand except from the modified parts the models areidentical. The reference system for the economicalanalysis is a non-coupled heating and cooling sys-tem without heat recovery. All modifications areinvestigated separately as well as combined, sincetheir impact may show a compensation effect in thecoupled system. The investigation is focused on thepossible savings of natural gas, which result fromthe demand of heating and domestic hot water withvariable heat recovery.



6 Hyd. interface to module “cooling water”7 Hyd. interface for service water from

module “cold water”8 Heat recovery from production processes and

air compressors

1 Heat consumers2 Backup steam heat exchanger (boiler model)3 Hot water storage tank4 Hyd. interface to module “cold water”5 Service water heat exchangers

Legend:

1

2

3

4

5

6

7

8

Figure 10: Modelica model of the heating module

Fig. 12 reveals the development of the primary energyconsumption for the conventional system (the steamexchanger covers the total load), the original layoutand the optimised system. Since the total heat demandis 9,546 MWh and the original concept is only able tocover it with not more than 38% waste heat the firstrequirement is not satisfied with this approach. Thesituation changes when the plant layout is modifiedin the way described above. Now, the heat recoverycontributes more than 80% to the heat load which issufficient enough to reduce the compulsory thermalinsulation of the associated buildings. The totalsavings of gas needed for backup heating by the steamheat exchanger sum up to 148,000 EUR in one yearregarding the difference between the original and finallayout. This corresponds to 4,000 MWh (-68%) lessheat supply at a heat price of 37 EUR/MWh. Thismoderate price results from experiences of the plantowner and also considers costs for maintenance andamortisation of the devices.

Comparing the remaining cooling load of the chillersbefore and after the optimisation one can determinean increased coverage of the cooling demand (15,445kWh) by the cooling towers from 3.1% to 15.8%

compared to a separated conventional heating andcooling plant. This result can be explained by theenhanced free cooling, especially during winter,spring and autumn (Fig. 13 and 14) since the returntemperature is higher and the operation time of thecooling towers in free cooling mode is prolonged.

Conventional system Original concept Optimised concept0

2

4

6

8

10

12

14

16

[MW

h/a]

15.44514.967

13.0061,158,375 Eur/a1,122,525 Eur/a

975,450 Eur/a

9.546

5.888

1.877

353,202 Eur/a

217,856 Eur/a

69,449 Eur/a

heating loadchiller cooling load

Figure 12: Development of the primary energy con-sumption for heating and cooling and reduction of run-ning costs for the conventional separated system with-out heat recovery, the original concept with heat re-covery and the optimised plant



It has to be emphasised that this study in this depthwas only possible by means of transient simula-tion which enables the display of the temperaturesthroughout all components under consideration of thethermal capacities. This unique feature of a dynamicsimulation is one advantage when the temperatureshave a large impact on the efficiency of the process(i. e. the COP of chillers, denoting the quotient of thecooling capacity to the electric power consumption)and dominating capacities are characterising theenergy system (i. e. storage tanks). A static simulationis not able to consider these important effects.

6.01 8.01 10.01 12.01 0

50

100

150

200

250

300

350

400Free cooling in winter

Time [dd.mm]

[kW

]

originaloptimised

5.07 7.07 9.07 11.07 13.07 −50

0

50

100

150

200

250

300

350Free cooling in summer

Time [dd.mm]

[kW

]

originaloptimised

Figure 13: Cooling impact of the heat exchanger forfree cooling in winter (t.) and summer (b.) - optimisedplant including all modifications

The remaining cooling load of each calculation ismultiplied with a fixed cooling price of 75 EUR/MWhresulting from experiences of the plant owner, again.The original concept could already save 36,000EUR/a. Taking all modifications of the optimisationinto account the costs could be drastically decreasedby 183,000 EUR/a. These cost reductions are mainlydue to the optimisation of the free cooling heat

exchanger position. The plot of the heat, transferredby the free cooling heat exchanger, is shown in Fig. 14.

Regarding the economical effect of the simulation, itis evident, that the optimised plant layout can save295,000 EUR/a in comparison to the original conceptand more than 465,000 EUR/a if the plant wouldhave been built in a conventional way without heatrecovery. Apart from the costs for the changed pipingthe modifications of the optimisation do not requireexpensive components.

0 1 2 3 4 5 6 7 8 9 10 11 120

0.5

1

1.5

2

2.5x 10

6 Free Cooling

Time [m]

[kW

h]

originaloptimised

Figure 14: Amount of free cooling - optimised plantincluding all modifications

The computing time for a year time simulation of theoriginal model and the optimised model are in therange of 10 to 24 hours depending on the installed pro-cessor.

5 Conclusions

This article is dedicated to the transient simulationof complex energy systems like they appear in largebuildings and industrial plants. For this purpose asimulation tool, called HKSim [1, 4], was developedby Imtech Deutschland GmbH & Co. KG and theDepartment of Technical Thermodynamics of theTechnical University of Hamburg–Harburg. It waspointed out, that such a tool is capable to simulateeven complex systems. The computational effort canbe justified by the prevention of possible failures insystem layouts and estimation of possible savings,which are of economic interest as could be shownin the described optimisation. In the carried outsimulation of the industrial plant the savings will payback the investment in a short period of time.



In addition, a model of a thermal consumer was pre-sented, which is necessary to integrate the heating orcooling demand from an external file into the systemsimulation. The model is designed for fast model gen-eration and simulations of whole years and it predictsthe mass flow rate and the return temperature with agood agreement to measurement data.

References

[1] B. Lüdemann, St. Wischhusen, O. Engel, G.Schmitz: Optimierte Energiesysteme, BWK, Bd.55, Nr. 9, Springer VDI-Verlag, Düsseldorf, Ger-many, 2003.

[2] St. Wischhusen, G. Schmitz: NumericalSimulation of Complex Cooling and Heat-ing Systems, Proceedings of the 2nd Int.Modelica–Conference, Modelica Association,Oberpfaffenhofen, Germany, 2002.http://www.modelica.org/confer-ence2002/papers.shtml

[3] Dynasim AB, Dymola 5.1, Lund, Sweden, 2003.

[4] St. Wischhusen, G. Schmitz: Abschlussberichtzum Projekt “Entwicklung eines Simulations-werkzeuges zur wirtschaftlichen und energe-tischen Planung von Prozessen zur Wärme– undKälteerzeugung – Entwicklung der Komponen-tenmodelle”, Department of Technical Thermo-dynamics, Technical University of Hamburg–Harburg, Hamburg, Germany, 2003.

[5] Modelica–Association: Modelica StandardLibrary, Linköping, Sweden, 2002.http://www.modelica.org/librar-ies.shtml

[6] B. Lüdemann: Auslegung, Energiebedarfund Komfort von Anlagen zur Heizung undWarmwasserbereitung im Niedrigenergiehausbei Berücksichtigung des Nutzerverhaltens,PhD thesis at the Department of Techni-cal Thermodynamics, Technical UniversityHamburg–Harburg, Books on Demand, 2002.http://www.bod.de



stefan wischhusen, bruno lüdemann, gerhard schmitz … · 2010. 10. 23. · s. wischhusen, b....

Documents