Adjustment of the nonlinear parameters in dynamic simulations of steam generators

Transcription

1 Adjustment of the nonlinear parameters in dynamic simulations of steam generators Dipl.-Ing. Peter Tusche Institute of Process Technique, Process Automation and Measuring Technique IPM) University of Applied Sciences Zittau/Goerlitz, Germany Prof. Dr.-Ing. habil. Tobias Zschunke Department of Mechanical Engineering Chair of Power Plant Technology and Power Engineering University of Applied Sciences Zittau/Goerlitz, Germany Prof. em. Dr.-Ing. habil. Rainer Hampel University of Applied Sciences Zittau/Goerlitz, Germany KEYWORDS dynamic simulation, Oxyfuel, load change operation mode, modelling ABSTRACT In the long term, due to political decisions and changes in the market of power generation, the position of German coal-fired power plants as base load plants is in question. This fact requires further consideration for the use as highly dynamic power plant in the range of peak load coverage. Dynamic behavior assessment of steam generators related to load change operation, start-up and shut-down processes is in the focus of power plant operators also with regard to possible wear and tear of the components. Moreover, an innovative technology is being investigated. In the near future, the oxyfuel process, depicting a climate-friendly combustion technology of lignite and hard coal, will be ready for production and applied in large power plants. The use of simulation tools is a low-cost option to gain knowledge from analyses about the behavior of steam generators in dynamic operation or untypical situations without any component damage. The use of dynamic simulations is also reasonable if in addition a new combustion technology is used that is not applied in large power plants yet. Content of this paper is the modelling of power plant components for the use in complex dynamic systems and the simulation of these highly dynamic processes. The way of modelling is explained by means of an example. NOMENCLATURE t s time m mass p bar pressure ṁ s massflow kj h enthalpy ξ mass concentration ϑ C temperature ρ v g c m 3 density m 3 specific volume m s acceleration of gravity 2 J K specific heat capacity m 3 volume V z m heigth d m diameter L m length s m wall thickness r m radius A m 2 flow cross-section α W m 2 K heat-transfer coefficient ζ pressure loss coefficient Q W heat flow T s time constant K amplification factor x output data u input data Proceedings 26th European Conference on Modelling and Simulation ECMS Klaus G. Troitzsch, Michael Möhring, Ulf Lotzmann Editors) ISBN: / ISBN: CD)

2 in out W F inlet outlet wall fluid shall finally be stored in the underground. The oxyfuel process proceeds according to the scheme depicted in figure fig: 1) and is described as follows: In INTRODUCTION Dynamic Simulations The use of dynamic simulations has rapidly increased during the last years in many fields of science and technology. The temporal transient behavior of many instationary processes is analyzed in computer models and used for optimization. Interventions like setpoint controlled operation or process control procedures often provide solutions for transient processes which are economically more efficient and also more sparing for the component. The dynamic behavior of the overall system which consists of a combination or overlap of several instationary sub-processes, can simply be depicted by the application of dynamic simulation tools which was usually impossible in the conventional way by expert knowledge. In power plant technology the focus is placed on the analysis of accident situations or simulated load changes as well as the start-up and shut-down of complete plants. The analysis may avoid possible destructions of particular plant components by incorrect parameters. Thus, the occurring economic costs caused by the blackout of large plants e.g. power plants can be minimized or even avoided. Another point is efficiency improvement in power plant operation and the consideration of aspects during the design phase of large plants. Innovative solutions are investigated and tested by means of dynamic simulation models. The low cost option for checking the dynamic behavior during the dimensioning of power plants is indispensable at vast investment costs like new building of large power plants. In the near future, part-load operation will mainly be used in conventional steam generators. For the operators, this change will raise several questions in terms of availability and economic efficiency which can simply be analyzed by dynamic simulations. Oxyfuel Process Being an innovative combustion technology, the oxyfuel process can be categorized into the line of combustion processes of lignite and hard coals. Carbon dioxide CO 2, which is harmful for the climate and which arises from the combustion with pure oxygen and from the exhaust gases during distilling separation of water in high concentration, is being prepared and compressed and Figure 1: OxyFuel process 6) the air separation facility, pure oxygen is provided as an oxidant. The coal is combusted with a material mixture of oxidant and recirculating flue gas. Due to the recirculation of dust and ash-less flue gas, the combustion temperature decreases to parameters permitted for steam generators with regard to component wear and tear of the heat transfer heating surfaces. The inert nitrogen missing in the oxidant is replaced by the recirculation of flue gas so that the mass flow of the combustion gas changes only marginally compared to the conventional combustion with air. Here, the composition of flue gas and its associated specific values and properties are determining for the transmission properties in the steam generator. After dust separation the flue gas is partly recirculated. The remaining exhaust gas consists basically of carbon dioxide and is cooled for water separation. It is then compressed and stored underground. Simulation Software Dynstar With the help of Dynstar, an in-house developed simulation tool of Zittau/Goerlitz University, simple control sections up to complex systems can be modeled. On a graphical surface, function blocks are placed and interconnected in a way the process simulation becomes possible. The various function block libraries contain many different function blocks which offer multiple possibilities for process modelling. For the simulation of power plant processes plant components like heat exchanger, pump, turbine stage, condenser, tube, valve, mixing or distribution elements are modeled. Special property libraries have been implemented for

3 the input and output data for the model. Constructive data, geometric variables and material data depict the parameters. Related to the modelling of thermodynamic ṁ in ṁ out p in model p out h in h out ξ in ξ out Figure 2: Simulation Software Dynstar 5) modelling thermodynamic processes and the application of special media like water and combustion gases. Library LibIF97 1) is applied for water or steam and for ideal gases is used library LibIDGas 2). MODELLING Structure Of The Models For the modelling of a combustion process using a new innovative technology no measuring values are available for the user. Black-box modelling is thus excluded. The lack of expert knowledge or operating point dependent design data of the steam generator components for various load cases and dynamic switching operation of a start-up process of the plant with air and a later switching to oxyfuel process makes the modelling by means of the grey box model impossible as well. Thus, to the modeler must be given complete constructive and geometric data to simulate the physical laws, just as in the glass box model in figure fig: 3) 3). The function blocks u glass box T ẋ + x = K u T = f a, b, c) K = f a, b, c) a b c Figure 3: glass box model for imaging thermodynamic processes in Dynstar are designed according to the scheme in figure fig: 4). Flow variables like mass flow, and state variables like pressure, enthalpy and the composition of substances are x V, d, L, z,... Figure 4: modelling scheme at Dynstar processes, Dynstar has a special feature in bidirectional processing of input and output data. Contrary to the other variables mass flow, enthalpy and composition of substances, the pressure has a reverse direction of data processing. The reasonable interconnection of function block inputs and outputs as indicated in figure fig: 5) produces the image of the instationary process which is going to be analyzed. ṁin pin hin ξin ṁ p h ξ model model model Figure 5: simulation assembly Model of a tube bundle heat exchanger As a transfer function, the modelling of a tube bundle heat exchanger fig: 6) using the given constructive data from a reference power plant of one of the leading German power generating companies, contains an equation system of non-linear differential equations with distributed parameters. The system is reasonably dissected and flow principles are particularly considered. The following figures fig: 7) depict this subdivision. Due to the different flow course of the hot and cold fluid within the heat exchanger, the submodels are indexed according to a matrix design. The shifting of the flow-specific complex heat exchanger into a circuit of heat exchangers having an ideal flow-through provides advantages for system calculation. The heat exchange without calculating the logarithmic temperature difference by axial discretization allows ṁout pout hout ξout

4 Ė2in Ė2out Ė1out Ė1in a) flow scheme of a tube bundle heat exchanger Ė2in Ė1out Ė1in Ė2out b) energy flow at the tube bundle Figure 6: flow scheme of a tube bundle heat exchanger fluid friction on the walls eq:6). ) z p in p out = g out v out zin ṁ2 out vout v in + 2 A 2 out p friction = ṁ2in vin + p 2 A 2 friction 5) in ζ L d ṁ 2 v 2 A 2 6) The calculation rules for pressure loss coefficients ζ = f ṁ, p, h, ξ, ϑ W, d, L, R z ) are calculated according to the VDI-heat atlas 4). The nonlinear system of ξ in ξ out p in h in Q p out h out r 0,1 0,0 1,1 1,0 2,1 2,0 3,1 3,0 4,1 4,0 0,1 0,0 1,1 1,0 2,1 2,0 3,1 3,0 4,1 4,0 ṁ in p h ξ ṁ out x a) flow scheme of the cold fluid b) flow scheme of the hot fluid Figure 7: flow scheme in matrix structure the calculation with the ideal heat transfer through the superheater tubes, as in figure fig: 6). The finite volume method is here applied. Submodel Fluid A fluid flows through a control volume and releases or absorbs diffusive energy through the outer surfaces in the form of heat, depending on the amount of temperature difference. The thermodynamic state of the fluid at the output is calculated by means of the energy flow balances eq:2), mass flow balances eq:1), and material flow balanceseq:3). ṁ in ṁ out = V dρ dt = V v 2 dt V dp dt + h ṁ in ṁ out ) ) h in + ṁ2in v2 in + g z 2 A 2 in in V v dh = P + Q + ṁ in ṁ out dv dt 1) ) h out + ṁ2out v2 out + g z 2 A 2 out out 2) dξ dt V v + ξ ṁ in ṁ out ) = ṁ in ξ in ṁ out ξ out 3) Q = α A W ϑ F ϑ W ) 4) The flow-caused pressure loss is calculated with the use of Bernoullis Theorem eq:5) under consideration of Figure 8: thermodynamic quantities at the pipe section differential equations requires the description of average values of the control volume. The static energy and material balances for a flow-through system are the calculation basis of yet unknown average state variables eq:7), eq:8),eq:9). p = p in A in A in+a out + p out A out A in+a out 7) h = Ain vout hin+aout vin hout A in v out+a out v in 8) ξ = Ain vout ξin+aout vin ξout A in v out+a out v in 9) The specific volume can be obtained from the already mentioned specific property tables via functions v = f p, h, ξ) 1),2). The heat flux over the outer surfaces is the coupling variable with another basic model: Instationary heat conduction through a tube wall. Submodel Wall In this submodel also, the energy flow balance eq:10) is the focus of considerations. Having solid walls, a linear equation system develops which is locally discretized according to the finite volume method and with the linearization of the material constant for the storing effects. Related to average wall temperatures as in figure fig: 9), the calculation rule for static calculation of temperatures in a cylinder wall is applied eq:11),eq:12). Q in Q out = V v c dϑ dt 10) ϑ r) = 1 x) ϑ i + x ϑ a 11)

5 ϑ1 C ϑ2 C The following example depicts an instationary transition process of a heat exchanger. ri r ra Figure 9: scheme of the nonlinear discretisation method for a cylinder wall ϑi ϑ r) x r = r ) i + r ln 1 + s a = ) 12) 2 ln 1 + sri ϑa 2 r i ) Direct Coupling Fluid with Wall Both submodels are directly coupled through the diffusive energy flow of the heat transfer, as it is shown in figure fig: 10). But the energy flow has nonlinear values as in the heat transfer coefficients α = f ṁ, p, h, ξ, ϑ W, d, L) in calculation rule according to 4). The calculation of the coefficient is indirectly back-coupled with the submodels via the temperature of the fluid on one hand and via the wall on the other hand. Flow conditions and substance constants do influence the alpha-value as well. The iterative Ėin F luid Ėout ϑf ϑf P arameter α = f ϑf, ϑw, F luid) QF W = α A ϑf ϑw ) QF W QF W Figure 10: scheme of the direct coupling between Fluid and Wall adjustment of nonlinear parameters to the current conditions and sequential processing of the submodel calculations require high effort in calculation which, however, provides the topicality of nonlinear parameters. This advantage plays a significant role in load change operation of power plants which will be gone into explicitly in the conclusion. ϑw ϑw W all SIMULATION AND RESULTS A sudden load demand from full-load to 70% also causes a change of the material flow in both fluids tab: 1). The complexity of such a load change operation is not supposed to be the subject of this simulation. A ramp-shaped reduction of the flue gas flow and steam mass flow in an interval of 10s is sufficient to show how non-linear parameters of the heat transfer coefficient fig: 14) behave within the model coupling during changes on the inputs of the overall model and which effects does it have on the transferred heat flux fig: 13) or on the wall fig: 12) and fluid temperatures fig: 11). quantity unit value ṁ 1 s s) to s) s) to s) ṁ 2 s ϑ1 in C 842 ϑ2 in C p 1 bar p 2 bar 174 ξ N 2 = 0.64, CO 2 = 0.18, H 2 O = 0.15, O 2 = 0.04 Table 1: input values t s DISCUSSIONS fluid temperature on the output ϑ1_out ϑ2_out Figure 11: fluid temperature diagramm A fast simulated ramp-shaped load change caused by the reduction of mass flow of the cold and hot fluid also has a direct impact on the working point of the superheater

6 α1 W/m²K α2 W/m²K Q MW ϑ0,0) C ϑ1;0) C ϑ_wa 0;0) ϑ_ra 0;0) ϑ_ri 0;0) ϑ_wa 1;0) ϑ_ra 1;0) ϑ_ri 1;0) wall temperature of the section 0;0) and 1;0) t s Figure 12: wall temperature diagramm heat flux t s Q1 Q2 Figure 13: heat flux diagramm heat transfer coefficient of the section 0;0) and 1;0) α1 0;0) α1 1;0) 60 α2 0;0) α2 1;0) t s Figure 14: heat transfer coefficient diagramm The transferred power decreases by ca. 20% as it can be seen in figure fig: 13). The instationary transition process into the new working point is finished after about 5min. The adjustment of interior T Ri ) and exterior tube temperatures T Ra ) and the temperature of the debris layer on the hot side T W a ) in fig: 12) related to matrix structures in fig: 7) to the new values is finished and no difference between the hot medium onto the wall Q 1 ) and the heat flux from the wall onto the cold medium Q 2 ) could be found - the energy storing processes in the wall are completed. The change of heat transfer coefficient fig: 14) is enormously visible in the range of mass flow change but continues marginally due to the change of fluid temperatures fig: 11). The adjustment process to the new load-change working point requires a similar period of time as the instationary transfer process described before. CONCLUSION The system dynamic during load change operation is depicted in the diagrams. The instationary process of the tube bundle heat exchanger during jump-like changes takes several minutes until stationary behavior of the heat transfer does exist in the new working point with new wall temperatures of the tube. In each step of the calculation, the nonlinear coupling functions of the heat transfer are adjusted to the current conditions as well. OUTLOOK The topicality of the nonlinear parameters is an essential point in modelling the oxyfuel process. During start-up the steam generator is initially working in air operation.after reaching its operation temperature it is switched to the innovative combustion technology fig: 15). The inert nitrogen which due to air combustion is yet contained in the recirculated flue gas will be completely replaced by carbon dioxide after some time elapsed. For this process showing permanent changes in material composition and thus also nonlinearities in all thermodynamic state variables, the topicality of the parameters to the current conditions by iterative adjustment and sequential calculation of all submodels is indispensable for correct analysis of the switching process. Linearization is possible only to a limited extent and may lead to distortions. To allow the optimization of the switching process at a later point in time it is essential to act to the process as close as possible. The application of this kind of modelling is thus required. FUTURE PROSPECTS Goal of the work is to design a complete steam generator model which is able to simulate the start-up process containing the switching process which has already been referred to. Heat exchangers, tubes for the recirculation line and combustion-specific components have been produced. The interconnection of these components and the function test of the overall model are further

7 air LZA?? O2 + coal combustion recirculation Figure 15: start-up process scheme actions and the focus of the author. The results offer the opportunity to start the oxyfuel process stable and optimal. Moreover, they help to avoid critical states which may lead to component damage. ACKNOWLEDGEMENTS This work is supported by the Free State of Saxony and the European Social fonds ESF. REFERENCES 1 University of Applied Sciences Zittau/Goerlitz, Department of Technical Thermodynamic in Water and steam, calculated from the industrial formulation IAPWS-IF97 and all supplementary standards, University of Applied Sciences Zittau/Goerlitz, Department of Technical Thermodynamic in Combustion gases, calculated as an ideal mixture of ideal gases from the VDI Guideline University of Applied Sciences Zittau/Goerlitz, Department of Technical Thermodynamics, R.Isermann in Mechatronische Systeme 2.Auflage Springer Verlag, Verein Deutscher Ingenieure VDI-Gesellschaft Verfahrenstechnik und Chemieingenieurwesen GVC) in VDI-Waermeatlas 10.Auflage Springer Verlag, livecms/ cmsdocs/ipm/de/data/artikel/ Software1/dynstar.html ) 6 de/ib/site/documents/media/ a6a06fe0-d cec d3241. pdf/vortrag_jentsch.pdf ) AUTHOR BIOGRAPHIES Peter Tusche is working as a PhD student at the Institute of Process Technique, Process Automation and Measuring Technique IPM) of the University of Applied Sciences Zittau/Goerlitz. The author studied from 1999 until 2003 mechatronics at this university, finishing with the diploma. After graduation he was employed by the RTT GmbH for 3 years in manufacturing automation. Since 2006 he has been involved in research and industry projects in the section power plant, boiler and fuel technology at the IPM. His is ptusche@hs-zigr.de. Tobias Zschunke was born in Dresden in The author is married and has three children. At TU Dresden from 1982 to 1990, he completed his studies and his doctorate in the field of thermodynamics and fluid mechanics. From 1990 to 1993 he worked as a project engineer at the engineering company Energietechnik Dresden und Stuttgart tasks in terms of energy technology and energy economics). He worked as a scientific fellow in the field of combustion and energetic bio-mass utilization as well as on the preparation and execution of national and international research projects at TU Dresden from 1993 to At TU Dresden from 2007 to 2009, Professor Zschunke completed his post-doctoral dissertation in the field of energy process engineering concerning aspects of coupled energy and heat supply gained from biomass. In 2007 he became professor for power plant and energy technology at Zittau/Goerlitz University of Applied Sciences. Moreover, he became deputy to rector for research at the university in 2010 and he is involved in cooperative research projects with companies from the fields of energy technology and institutes of the Fraunhofer Organization, the organization Biomasseforschungszentrum DBFZ) in Leipzig and at the TU Dresden. His is tzschunke@hs-zigr.de. Rainer Hampel was born in He studied in the field of power engineering at TU Dresden from 1964 to After his graduation he went to Zittau/Goerlitz University of Applied Sciences. There he finished his doctorate in 1975 and his post-doctoral dissertation in 1984 in the field of power plant automation. In 1987 he became professor for power plant automation at Zittau University of Technology. In 1992 the fields of measuring technology and process automation were added to his chair. From 2003 to 2010 he was the rector of Zittau/Goerlitz University of Applied Sciences. His is r.hampel@hs-zigr.de.