Home

user manual - ipese

1. 4 5 4 Energy Integration Model oe de Ge ct oh ol oh eae a nds e MM ay dd ne arara tes es Ge ha ae 4 6 1 Model smapshot ee oe A O ol o Dkk RR A A as ee 4 7 Analysingdatal A ae ee ee oy SR ws Le BS See it eo ee Ee oe a EEEE E EERE ae ve ae et eR AO ee ee a ae 5 3 Aspen model definiti0M a o a a e e a a 54 AMPL model definition A D eee eee 83 9 1 Tags d finition op 4204405082 824 4444 a de D D 9 A A E ee ee aes Sue ban de de 5 2 9 AAA de 6 6 des A Me ee NS D 6 4 Sensitivity analysis definition oe amp ds oNN 10 12 13 13 15 15 15 15 16 16 18 19 19 19 21 21 22 22 24 24 25 CONTENTS OLENTI ISE STI EPFL be ee A Sd SU US ee 31 D DS CR a DA Bea an eS 33 6 6 Recomputing a Pareto 53 36 a D E EAE a AA 36 PE aia o 36 7 3 Post multiperiod computation 2 02 0200 57 8 Results Analysis 38 a A A ES 38 8 2 REPO acord we ood ae a ARE A dd AA aa 38 8 2 1 Configuring the frontend 38 PE AN 39 ARAN amp da
2. 10 11 12 13 14 AMPL http www ampl com Belsim http www belsim com R Bolliger D Favrat and F Mar chal Advanced Power Plant Design Methodology using Process Integration and Multi Objective Thermo Economic Optimisation In ECOS 2005 18th International Conference on Efficiency Cost Optimization Simulation and Environmental Impact of Energy Sys tems volume 2 pages 777 784 Trondheim Norway 2005 K Deb Multi Objective Optimization using Evolutionary Algorithms Wiley amp Sons 2001 G Leyland Multi objective optimisation applied to industrial energy problems PhD thesis Ecole Polytechnique Federale de Lausanne 2002 F Mar chal Modelisation et optimisation des syst mes industriels 2006 Lecture notes F Mar chal D Favrat F Palazzi and J Godat Thermo economic modelling and optimization of fuel cell systems In Fuel Cell Research Symposium ETH Z rich 2004 F Mar chal D Favrat F Palazzi and J Godat Thermo economic modelling and optimisation of fuel cell systems Fuel Cells From Fundamentals to Systems 5 1 5 24 2005 MathWorks http www mathworks com A Molyneaux A practical evolutionary method for the multi objective optimisation of complex integrated energy systems including vehicle drivetrains PhD thesis Ecole Polytechnique F d rale de Lausanne 2002 F Palazzi N Autissier F Mar chal and J Van herle A Methodology for Thermo Economic Modelin
3. 1060 gt 2 2 1050 Enable data cursor mode 1040 L i 1 1 1 1 50 100 150 200 250 300 350 400 450 reactor cost 5 This gui interface is brought to you with love by Francesca Palazzi Laboratoire d Energetique Industrielle LENI ISE STI 2005 44 OSMOSE manual CHAPTER 8 RESULTS ANALYSIS 1 Functions examples 1 1 PreComputation example 2 Aspenmodel Figure 7 Aspen Plus calculator for import export variables definition U MEA MEA HeatX apw Aspen Plus 2004 1 aspenONE Calculator C 1 Data Browser Fil e Edit View Data Tools Run Plot Library Window Help DH la BO N 147 QA 544 Ni fe gt gt mom GQE 5 x A ac E 4 24 22 y 38 metas M gt lt a v 128 2 Setup a YDefine Calculate Y Sequence Tears Stream Flash EO Options Components Properties Flowsheet Variable name Info flow Definition Streams z L Utilities TEST Import Kklock Var Block B1 Variable 4REA Sentence PARAM Units sgp2 Blocks Y Variable Definition ES Reactions gt Convergence e Flowsheeting Options Select a variable category and reference Refi C2 Design Spec Variable name TEST v T E is 1 Calculator SS cl Category Block Input Oal Variable Resu Fo Y Blocks Sentence EO Ir O Streams Units Transfer J Model Utility Stream Library Ralanre Vi Mixers Splitt
4. In this example we assume that the Tags CakeRadius and Nb0fCakes have been defined in the front end 4 with the piece of code above the value of these tags have been allocated into local variables called CakeRadius and NbOfCakes The variables retreived from tags is used for computation CakeSurface Pi CakeRadius 2 TotalCakeSurface NbOfCakes CakeSurface Selection of the external model remark All the models have to be defined in the o Model i FileName field This is done in the front end if TotalCakeSurface gt BigProd_LowerBound o Model o ModelID FileSelected BigProductionChain bls elseif TotalCakeSurface lt LocalBakery_UpperBound o Model o ModelID FileSelected LocalBakery bls else o Model o ModelID FileSelected CakeProductionChain bls end New tags definition the new tags can be local variables in that case one can organize information storage as follows nt 0 New tag counter incremented for each new tag definition Define the names and units of the new tags nt nt 1 Tag nt TagName CakeSurface Tag nt DisplayName Surface of one cake A g nt Unit m 2 nt nt 1 Tag nt TagName TotalCakeSurface Tag nt DisplayName Total surface of cakes Tag nt Unit m 2 Bibliography 1 2 3 4 5 6 7 8 9
5. PreComputationFunction vali_tag_assignment m Define a matlab function performing computations after software and energy integration optional Model i PostComputationFunction PostcomputeModelSimple m Specific properties for energy integration Specify if EASY is used to perform the energy integration required ComputeEl 1 1 yes 0 no if yes define the fields below Specify the objective used for the resolution of the heat cascade optional Ex o EasyObjective MER or Operating cost or Exergy or Mechanical power Default is Operating cost Easy Objective MER If you need to perform computations between external software model and energy integration specify the name of the matlab file to run optional Model nm ElPreparationFunction easy_tag_assignment m Specify the version of OSMOSE you use optional Default is 2 0 Model i OSMOSEVersion 2 5 Listing 5 1 Front end example General Model Definition Note the syntax displayed in the fake calculator definition is not always correct especially for block variables In case of doubt export the ASPEN file including the fake calculator in a temporary inp file by clicking on File Export The correct syntax can be found by checking the calculator definition in the source code in the temporary inp file Again do not save the bkp file with the fake calcul
6. SoftwareSpecific Ampl0ption nbo Value 2 73e 12 Definition of ampl solver optional default is minos o Model i SoftwareSpecific AmplSolver cplexamp Defintion of solver options optional nbo 0 nbo nbot1 o Model i SoftwareSpecific SolverOption nbo Name prestats o Model i SoftwareSpecific SolverOption nbo Value 1 Listing 5 5 Code matlab AMPL model definition Define the template files containing the easy problem definition i e the list of hot and cold process streams and the definition of the utility and heat recovery systems required 11 Ex o Model i EIFiles h2o_psal txt h2o_psa2 txt o Model i EIFiles h2o_psa txt 4 Give the tags other than flowrates you want to obtain from easy optional 4 Ex o Model i SoftwareSpecific ToOutputTags BALANCE ADV TOTALCO2 ADV RAMBIANT o Model i SoftwareSpecific ToOutputTags BALANCE ADV TOTALCO2 ADV RAMBIANT ADV O02_ELSYS ADV F105 ADV F113 ADV F118 ADV F227 ADV F238 4 Give the position n column of the ToOutputTags in the easy report file required if ToOutputTags are defined 4 Ex o Model i SoftwareSpecific Tags_position 6 7 7 o Model i SoftwareSpecific Tags_position 6 7 7 7 7 7 7 7 7 4 Give the units of the ToOutputTags optional 4 Ex o Model i SoftwareSpecific ToOutputTagsUnit
7. The seminal publications and 8 present the underlying approach based on process simulation and modeling associated with energy integration see also I 7 te Y Chapter 3 Requirements and Installation OSMOSE runs both on Windows and Linux platforms 3 1 Step by step installation 1 Install SVN For windows Subversion is an open source version control system Download from for windows For linux there is no need to install 8 software Further documentation can be found at http leni epfl ch images procedures svn doc_svn_ leni pdf For linux Use the command sudo apt get install subversion 2 Get OSMOSE package The software is constituted by a set of matlab functions organized in several sub folderd For windows a Create a folder osmose for example on C notice that the path to osmose folders must not contain any space b right mouse click SVN Checkout of repository https lenisvn epfl ch svn osmose trun ont forget the s at the en c URL of repository https lenisvn epfl ch svn trunk don t forget the s at the end of http For linux Create a folder wherever you want traditionally in usr local sudo mkdir usr local osmose Import osmose from SVN cd usr local osmose svn co https lenisvn epfl ch svn osmose trunk 3 Install required software The required software is detailed in section 3 2 For users at LENI the preferred installation is
8. Used by Vali Name of Vali tag with cst or precision status Produced by Vali Name of Vali tag Used by Aspen Name of Aspen tag with cst see Section 5 3 Produced by Aspen Name of Aspen variable Used by AMPL Name of an AMPL parameter or set Produced by AMPL Name of an AMPL variable Used by Easy Name used in the Easy input template file Produced by Easy Name of a quantity written into easy output see Section Table 4 1 Tag names contextual correspondence TagName This field is the tag identifier it must be unique Communication is established between OSMOSE and the model using the TagName field that has to correspond to the name of a variable in one of the sub models Table 4 1 gives tag names correspondences so as to perform this communication DisplayName Optional field the display name is a short description of the tag It is used for graphic axes labeling as well as in automatic reports Unit The unit field is an informative field that stores the unit of the variable OSMOSE does not handle unit conversion Status The status can be constant CST variable 0 a Ampl set SET or a number indicating a precision to be used with Vali for data reconciliation Value The value field is either filled by OSMOSE before entering into the model see Chapter 6 or it is completed during model resolution Each model sub section can fill the value of a previously defined tag as well as define new
9. Evaporation temperature of working fluid Unit C 15 1 28 0 Period p Limits 90 120 Period p Limits 100 110 Listing 7 2 Variables definition in multi period problems 7 3 Post multiperiod computation If further calculations have to be made after all the periods have been simulated it is possible to call a post multiperiod matlab function which has to be declared as illustrated by the example of the listing o Model i PostMultiperiodMFunction geotherm_Postcompute_Total Listing 7 3 Definition of a post multiperiod computation matlab function 37 Chapter 8 Results Analysis 8 1 Results structure Each computation stores results in the o Results structures This structure holds all the results con cerning the models the variables the constants the objective functions and of all the interfaces called during the computation ex El The results structure is displayed in listing 8 1 Results structure o Resutls ih indexes ih 3 defines actual result e m defines actual model p defines catual period o Results i Model m Variables Tags tags computed at the end of period dependend computation TagsList list of tag names contained in Tags field Used to accelerate tag recovering Period p Tags period dependent tags values SEE Constants TagsList Objectives objective functions Clusters appears
10. InputTags i Unit kW o Model nm InputTags i Unit bar Specify the default value of the model tag required Ex o Model nm InputTags i DefaultValue 20000 o Model nm InputTags i DefaultValue 1 15 Specify the status of the model tag required Set it CSE if it is fir or give an accuracy Ex o Model nm InputTags i Status CST o Model nm InputTags i Status 0 5 Continue the list 111 o Model nm InputTags i TagName o Model nm InputTags i DisplayName o Model nm InputTags i Unit o Model nm InputTags i DefaultValue o Model nm InputTags i Status CST Listing 5 7 Code matlab Tags definition leu 27 Chapter 6 Computation definition The present chapter discusses computation definition within OSMOSE through a front end function Once a stable model is defined computations to be performed can be determined see Section 77 The possible computations are recalled here Model snapshot also called one run Sensitivity analysis e Optimization using evolutionary algorithm mono or multi objective Detailed recomputation of optimization results Each of the precedent choices is selected into the main function of the front end file as shown in the following section The definition of the computation parameters is then performed in specific sub functions of the front end Th
11. TO 8 408 USER MANUAL VERSION 2 0 NOVEMBER 22 2010 WRITTEN BY FRANCESCA PALAZZI WITH THE COLLABORATION OF ZO PERIN LEVASSEUR RAFFAELE BOLLIGER MARTIN GASSNER AM ECOLE POLYTECHNIQUE FEDERALE DE LAUSANNE Copyright 2003 2007 by Laboratoire d Energ tique Industrielle LENT All right reserved No part of this publication may be reproduced stored in a retrieval system or transmitted in any form or by any means electronic mechanical photocopying recording or otherwise without the prior consent of LENI For information about permission for use of material from this document as well as OSMOSE program please contact LENI Laboratoire d Energ tique Industrielle ME A2 434 B timent ME Station 9 CH 1015 Lausanne TL 41 21 693 35 06 Fax 41 21 693 73 22 http leni epf1 ch e secretariat leni epf ch This software has been initiated and developed by Francesca Palazzi Francesca Palazzi a3 epfl ch Raffaele Bolliger Raffaele Bolliger epf ch The following people have contributed to OSMOSE developement and documentation Julien Godat Francesca Palazzi Raffaele Bolliger Andrea Fabiano Martin Gassner lrene Ricart Puig Zo Perin Levasseur Damien Muller Luc Girardin Nordahl Autissier Nicolas Borbo n Isabelle Juchli Hubert Thieriot 2001 2002 2003 2007 2003 2003 2004 2004 2004 2005 2005 2005 2007 2005 2005 2006 2007 2006 2007 2007 First code allowing to run Vali
12. dd mmmm yyyy Listing 8 3 Frontend options for reporting Other options can be found in the osmose_report_defaults m function 8 2 2 Reporting results The results of reporting are stored in your results folder You will find two folders pictures and SRC The latter contains all the I4TpXfiles used to build the report You can modify them and recompile the report 8 3 Pareto plots using OsmosePlots Pareto analysis can be performed in the frame of OSMOSE When using OSMOSE to perform two objectives optimization an interactive tool OsmosePlots is available to generate visualisations of the results Different kind of plots can be generated using OsmoseP lots e Plot of the Pareto curve e Projection of variable values on an objective axis e Correlation between two variables 8 3 1 Osmose data storage Within OSMOSE the values of model variables decision variables as well as dependent variables are transmitted and updated troughout the computation using a specific sub structure called Tag Each relevant model variable or parameter is associated to a tag During an optimization the values of tags are used to perform computations and to transmit the objectives values to the optimizer The tags are not stored during the evolution of the optimization After optimization termination the points composing the Pareto front can be recomputed For each recomputed point OSMOSE creates a folder where the user can decide to stor
13. kW kg s kg s o Model i SoftwareSpecific ToOutputTagsUnit kW kg s kg s kg s kg s kg s kg s kg s kg s Define the Matlab function that determines which easy template file to use optional if one file is specified in o Model i EIFiles required otherwise In this function you are also free to perform operations on the tags before running easy 4 Ex o Model i EIPreparationFunction easy_tag_assignment o Model i EIPreparationFunction easy_tag_assignment Call the tags definition defined at the end of this template required for osmose version 2 5 see o Model i OSMOSEVersion o DefineTags o Listing 5 6 Code matlab Energy integration model definition 26 lau OSMOSE manual CHAPTER 5 MODELS DEFINITION function o DefineTags o Definition the tags of the model osmose version 2 5 only This tags might be used in vali easy or other computations i 111 Specify the name of the model tag that is displayed required Ex o Model nm InputTags i DisplayName Nominal thermal capacity o Model nm InputTags i DisplayName Reactor pressure Specify the name of the model tag required Ex o Model nm InputTags i TagName pth o Model nm InputTags i TagName 1 101 Specify the unit of the model tag required Ex o Model nm
14. functions this compromise is represented by the Pareto frontier This curve represents the set of non dominated solutions it delimits the unfeasible domain from the feasible but sub optimal one The book of Kalyanmoy Deb 4 contains exhaustive in formation about multi objective problems and the methods to solve these using evolutionary algorithms as it is the case in OSMOSE Evolutionary algorithm search for the optimal trade offs by comparing the fitness of solutions An initial population of random points is generated in the space of the decision variables x each point is eval uated by computing the values of objectives The algorithm performs then a breeding of the points that are more satisfactory with respect to the objective this breeding generates a new population of points that is closer to the optimal solution These points become the new genitors of the next generation and the process starts all over again There is no termination criterion of the search the user normally set an upper limit of iterations at which he expects to be close enough to the real optimum To perform an optimization OSMOSE is coupled with moo and evaluates the objective function by solving the model at the points generated by moo Figure 4 6 3 The user specifies the search space by 18 ae y OSMOSE manual CHAPTER 4 RUNNING OSMOSE STEP BY STEP Figure 4 8 Optimization organization o Variable i i 1 N LowerBound x Upp
15. group and to permit storage and documentation of the work The OSMOSE platform contributes to these goals with the following aspects e It allows to establish communication between models built with different software It allows to run complex computations like optimization or sensitivity analysis It is able to build automatic reports based on computation results e It is a collection of analysis tools OSMOSE manual CHAPTER 2 OVERVIEW The user of OSMOSE should be familiar with design problematics and master optimization tech niques and purposes Basic knowledge of Matlab programming is required OSMOSE handles external models written in Vali and AMPL languages that shall therefore be familiar to the user 2 3 What is osmose OSMOSE is a platform for the study and design of complex integrated energy systems The software is developed in the Laboratory for Industrial Energy Systems LENT in the Ecole Polytechnique F d rale de Lausanne EPFL Lausanne Switzerland The project is motivated by the need of a flexible and performant research tool to study and design energy systems In this perspective the general scope of the software is to help the user to develop and compute technology models that combine a thermodynamic computations b power and energy integration as well as c economic or environomic aspects OSMOSE exploits the models by performing a sensitivity analysis b optimization and c data analysis
16. item mtl o Moo drawing invert item mtl o GraphicOptions Listing 6 6 a completer 9 DoParetoAnalysis e PlotPareto e PlotDecVar e PlotCorrelations e PlotMatrix 6 5 1 Restarting a failed optimization run 6 6 Recomputing a Pareto curve il n y a rien dans la fonction 6 97 77 Be 33 6 6 RECOMPUTING A PARETO CURVE OLENTI ISE STI EPFL function o DefineMooUptim o Definition of the optimisation problem Number of objectives required o Moo nobjectives 2 Number of maximal iterations required o Moo max_evaluations 30000 Number of clusters required A cluster is a subset of the Pareto population with similar values of the 4 variables o Moo nclusters 4 Size of the initial population i e number of initial points required o Moo InitialPopulationSize 500 Objectives definition i 0 Definition of objectives required i i 1 o ObjectiveFunction i o ObjectiveFunction i o ObjectiveFunction i o ObjectiveFunction i i i 1 o 0bjectiveFunction i o ObjectiveFunction i o ObjectiveFunction i Model AZEP TagName TotalPower MinOrMax max DisplayName System efficiency Model AZEP TagName TotalCost MinOrMax min o ObjectiveFunction i DisplayName System efficiency Define the decision variables required i 0 i iti A A E D
17. or sets that can be modified during OSMOSE computation e Static input parameters or sets that can not be modified trough OSMOSE To organize dynamic input OSMOSE extracts the list of parameters from the AMPL model file completes the values by matching the parameter names with the OSMOSE tag names and writes a data file to be used during AMPL model resolutior The data file written by OSMOSE is called the dynamic data file It can contain dynamic sets as well e g sets that are defined by an OSMOSE tag An OSMOSE tag corresponding to an AMPL set has the Status field equal to SET SET An example of AMPL dynamic set definition trough OSMOSE tag is given in listing 4 3 Static input can either be defined in the AMPL model file and or in a static data file that has to be specified in the model definition see example listing 5 5 Define a set as input of ampl model i nt nt 1 Constants nt ModelTagName mymodel Constants nt TagName tech Constants nt DisplayName Set of technologies Constants nt Unit Constants nt Status SET Constants nt Value 1 2 3 4 Omo O 0 O D Listing 4 3 Tag definition A dynamic AMPL set It is important to notice that as AMPL variables and parameters can have multiple dimensions the corresponding OSMOSE tags will have the same dimensions Furthermore if an AMPL parameter used 5 data_ dyn dat in the OSMOSE
18. perform sensitivity analyses in one or two dimensions In the software this is performed by the osmose_assign_variables function OSMOSE manual CHAPTER 4 RUNNING OSMOSE STEP BY STEP Figure 4 6 Model snapshot organization Value o Constant k Value es Values of al 5 o Model i Tag j DefaultValue o Model i Tag j Value o Constant k Value Figure 4 7 Sensitivity analysis organization One or two variables o Variable 1 xz o Variable 2 x2 Bounds E UpperBound 7 x27 LowerBound x 22 7 lumber of steps Numberofsteps N N2 gt Sensitivity analysis Values of decision variables for point Values of all model tags for point o Model Tag k Value X7 o Model i Tag Value o Model Tag 1 Value Gaj gt Osmose Model The decision variables are defined trough the o Variables structure see Paragraph by giving the lower and upper bounds of the variations as well as the number of steps OSMOSE will then call the model by looping the variables as follows Let x and z2 be the variables on which the analysis is performed Define 11 x the upper bounds al xl the lower bounds and N1 No the number of steps The loop on the first variable is defined by equation 4 1 The loop on the second variable is defined by equation 4 2 U ph t Li 4 1 1 Th 23 5 Lies 4 2 2 These loops build a g
19. that reflect the behavior of human customers Luckily in the coarse world of engineers the relationship between variables is often given by the laws of physics As an example the amount of heat exchanged trough a heat exchanger can be computed knowing the temperatures and heat capacities of the exchanging streams The formalization of the above is that a model is defined by a set of equations capturing the rela tionship between variables One can divide the variables into two categories the degrees of freedom x ae le 9 4 3 OSMOSE MODELS GENERAL STRUCTURE OLENTI ISE STI EPFL or decision variables are determined by the user the unknowns 2 or dependent variables are computed by solving the model equations once the value of x is fixed Add to these the values of the parameters p and you can write the model as a set of equalities h a 2 0 The solution of which will give access to the state of the system Practically engineers develop models within specifically tailored modeling languages and solve the set of equations using numerical solvers In the framework of OSMOSE models can be established using the languages Vali Easy and AMPL as well as matlab Each software has solvers associated to it so as to compute the state of the system Please refer to the specific documentation of each software to establish your models 4 3 Osmose models general structure We have seen that a model can be defined as a set of equations
20. trough the svn server Contact system administrator and request access OSMOSE manual CHAPTER 3 REQUIREMENTS AND INSTALLATION 4 Set your matlab path The OSMOSE folder with all its sub folders have to be part of Matlab s search path To add OSMOSE to the search path use either the addpath command or in the Matlab window select File gt Set Path to open the Set Path dialog box Save the changes to the path for successive sessions either with the savepath command or trough the Set Path dialog box 5 Control path definition to external software OSMOSE can call the software listed in section 3 by locating their main executable file Control or define your paths as explained in section 6 2 Once these steps are completed you are ready to begin experiencing OSMOSE Follow the guide in chapter 3 2 Requirements The modeling tools used with osmose can vary as well as the operating systems thus the requirements may vary for every situation You should control that the software specified in the following list is installed on your machine e Matlab version 7 6 or above e Vali 2 in case of Vali model usage Belsim Vali is only supported by Windows 1 From the server http documents epfl ch groups v va vali private VALI4500 Install Vali client VALI Vali4Client Setup exe 2 Replace localhost par 128 178 144 127 3 Ask somebody from LENT to add an account on the license manager with the same username than for the
21. x axis For commodity the two objectives are separated outside the list Tf the third button Another tag select below is checked a scrolling list containing all the tags is displayed as seen in Figure 8 3 3 e b Selection of the y axis One can select the y axis between all the tags using a scrolling menu e Plot command button After having selected x and y axis push the plot button to obtain the graphic e d Plot area Area containing the resulting plot e e Figure saving commands Once a plot is generated there is the possibility of saving the matlab figure You can enter the name you want for your file and then press save The plot is saved in a subfolder of the run folder called OSMOSE _ plots e f Data cursor command Allows enabling data cursor mode Tf data cursor mode is enabled clicking on a point in the plot area will display the point coordinates see Figure 8 3 3 How do I obtain the Pareto curve To obtain the pareto curve select one of the two objectives in the x axis box In the y axis scrolling menu select the other objective Click Plot An example of Pareto plot is shown on Figure The clusters identified by moo in this case four are displayed with different colors How do I analyse variables evolution along the Pareto curve The evolution of the values of one variable along the Pareto can be visualized by projecting the points on the plane formed by the axis of one of the objectives and the axis of t
22. The model field o Model nm FilesToCopy contains the names of files to be copied in the run folder Using the o Variables structure the value of some tags can be modified For a single run the required fields of the o Variables structure are e ModelTagName reference to the model containing the variable e TagName name of the concerned tag e Value value attributed to the defined tag for this computation 6 4 Sensitivity analysis definition The parameters for the execution of a sensitivity analysis are defined within the DefineSensi front end sub function Sensitivity analyses can be performed up to two dimensions Each of the dimension of the analysis is defined by a 6 Variables i structurd For a sensitivity analysis the required fields of the o Variables structure are 2More precisly o Variables is a vector of one or two rows 30 cr y OSMOSE manual CHAPTER 6 COMPUTATION DEFINITION function o launch_osmose o do not edit this function o DefineModel o if o DoOneRun o DefineOneRun o end if o DoSensi o DefineSensi o end if o DoMoo o DefineMooUptim o end if o DoRestartMoo 0 5 DefineMooUptim o o DefineRestartMoo o end if o DoRecompute 0 5 DefineMooUptim o o DefineRecompute o end if o DoAutoReport o DefineAutoReport o end if o DoCustomReport o DefineCustomReport o end o run_frontend o Listing 6 2 Cod
23. _T define_aspentag o HPin_T C Stream Var Stream RICH HOT Substream MIXED Variable TEMP Heat duty of stripper named HP STRIP define_aspentag o HP_DUTY Watt Block Var Block HP STRIP Variable COND DUTY Sentence RESULTS Listing 5 4 Code matlab Aspen Tags definition necessary to compute the energy integration with EASY can be found put chapter or section Optionally other tags than the multiplication factors of the streams can be extracted from the easy report fid The strings of these variables are defined in the field SoftwareSpecific To0utputTags The according column number in which they occur in the report file and their unit is specified in the fields SoftwareSpecific Tags_position SoftwareSpecific ToOutputTagsUnit respectively 5 5 1 Tags definition listing 5 7 L synep html in the OSMOSE_temp EnergyIntegeration folder Be 25 5 5 ENERGY INTEGRATION MODEL DEFINITION OLENTI ISE STI EPFL Main ampl model file containing parameters and variable definition model equations and constraints required o Model i SoftwareSpecific ModFile Probleme_simple_MOO mod Data file of static parameters optional o Model 3 SoftwareSpecific FixedDataFile Probleme_simple_M00 dat Definition of ampl options optional nbo 0 nbo nbo 1 o Model i SoftwareSpecific Ampl0ption nbo Name presolve_fixeps o Model i
24. ae Bie ee 39 8 3 1 Osmose data storagel 39 8 3 2 Calling OsmosePlots 40 8 3 3 Using OsmosePlots 40 1 Functions 88 9868 5 965 dune eux 45 1 1 PreComputation examplel 45 2 Aspenmodell asama p a 5 Maud a dass 9 98 5442 a pau mu 45 a ma Chapter 1 Documentation update Keep the update of the osmose documentation for users is under the responsability of everyone As soon as somebody made changes or added new features for the main documentation this has to be documented in the following way e Main documentation is to be written in the LaTeX files that are available on svn at the following address https lenisvn epfl ch svn osmose doc Please use separate files and input them then in the main files e Then the LaTeX document will be automatically converted in the LeniWiki Please do not write directly new main documentation on the LeniWiki Only authorized people who have an access to svn can make modifications in the documentation Tf you do not have the access and that you would like to make a modification please contact leda gerber epfl ch For the FAQ and the quicktips which are not part of the main documentation this will be written on the LeniWiki The basic idea is that if everyone discovers or cr
25. ain computations as well as new tags definition Its name is specified in the o Model 1 PostComputationFunction field in the front end function 4 6 Choosing the computation to perform Once the model is defined it is perceived as a black box input output structure that can be called under various conditions thus allowing to perform several kinds of computations For completing tag list use the update_output_tags function of the OSMOSE package mr 15 4 6 CHOOSING THE COMPUTATION TO PERFORM OLENTI ISE STI EPFL Software Possible Variable Vali Any Vali Tag with CST status or with measured precision status Aspen Any Aspen Tag with CST status Easy Tag name used in the input template and not com puted by the previous sub models AMPL AMPL parameter or set not computed by the previ ous sub models and not attributed in the static input Matlab matlab variable not computed by the previous sub models and not attributed in a matlab sub function Table 4 2 Available input variables The list of computations currently handled by OSMOSE is e Model snapshot also called one run e Sensitivity analysis e Optimization using evolutionary algorithm mono or multi objective e Detailed recomputation of optimization results As we have seen the communication within a model is performed trough a structure called Tag By varying the value associated to a tag OSMOSE commands the resolution of the model und
26. ame inp for using it in OSMOSE For the definition of an Aspen model in the front end and the definition of a tag for an import or export Aspen variable see section Ampl model AMPL is a language for large scale optimization and mathematical programming problems A model written in AMPL is generally an optimization problem whose solution is found by connecting the problem with a solver There are several solvers available to use with AMPL I ranging from LP to MINLP In addition to the A MPL web site I information on A MPL can be obtained by reading the user manual 14 In AMPL language distinction is made between parameters and variables Parameters have constant value during optimization whereas variables are computed by solving the problem AMPL allows the definition of sets to perform arrays operations therefore AMPL parameters and variables can be multi dimensional An AMPL model is generally composed by a file with the mod extension called the model file and by one or more files with dat extension called the data file s The model file contains the optimization problem definition sets parameters and variables definition as well as the model equations The data files contains values of sets and parameters The input of an AMPL model is composed of the values of the model parameters and of the values of sets For usage of AMPL with OSMOSE two types of inputs are distinguished e Dynamic input parameters
27. and Easy from Matlab Developer Developer Developer and beta tester Developer and beta tester Creator of the OSMOSE report generator OSMOSE documentation and beta tester OSMOSE documentation and beta tester OSMOSE web interface and web service OSMOSE beta tester Subversion maintainer Improvement of the report generator Contributions in development of El structure and re lated software support The OSMOSE project is made possible thanks to the financial efforts of the Laboratoire d Energ tique Industrielle LENT in the cole Polytechnique F d rale de Lausanne Many thanks to Professor Daniel Favrat Director of LENT Fran ois Mar chal First Assistant at LENI FIFO Fond d Innovation pour la Formation EPFL OFEN Office F d ral de l nergie Contents 1 Documentation update 2 1 What does this document contain 2 2 Who should use 8 086 2 3 18 3811138671 e 5 5 5 5 daa 5 5 LME des 6 BBG a we eee oe Sete RN E oon a eA 2 AAA AA Ads eee ee a ad ls e e rara da E ee 42 Modeling ABQ coe ht a a ee eG ta al 2 22 ee A a eee eb et ee an deta ee ee ee 4 5 Establishing a modell 4 5 1 Pre computation function 4 5 2 External Software Model 4 5 3 Intermediate function
28. art section 5 1 that has to be be completed with specific parts according to the model type see sections 5 2 5 4 and 5 5 These parts are described below For clarity the matlab code is displayed in several pieces with listings A complete front end file template is given in annex faire l annexe 5 1 General model definition In this part of the front end the model is defined by its name location and type For identification pur poses as well as for folders naming each model must have a name defined by the 6 Model i TagName field The name is defined by the user and can be any string of characters without spaces As seen in 4 3 an OSMOSE model is generally composed by several sections each section consisting in one or more files All the files composing a model have to be in the same folder the model repository The path to the model repository is defined in the o Model i Location field The Software field refers to the software of the external model Available options are listed in Table Value Action vali The external model is a Vali model aspen The external model is a Aspen model ampl1 The external model is an AMPL model easy The external model is described according to Energy Technologies syntax eiampl The external model is described according to Energy Technologies syntax matlab The external model is a model composed of Matlab functions Table 5 1 Available e
29. ator in it After Aspen tags are defined import variables must be assigned a value according to the OSMOSE struc ture o Constants or o Variables Import variables must have the CST status and export variables must have the OFF status If the status is not defined ASPEN will automatically assign the CST status to all tags present in o Variables and o Constants and the OFF status to all other tags To make sure that execution errors result in a failed convergence in OSMOSE set the maximum number of errors in Aspen to 1 Go to 13818 Browser Simulations_options Limits In that case ASPEN will quit after the first error encountered even if it could have recovered from it A higher number of maximum errors in Aspen could lead to model convergence however the errors would not be seen in OSMOSE Be 93 5 4 AMPL MODEL DEFINITION OLENTI ISE STI EPFL Define the location of the executable vali file optional o Vali Path C Belsim bin valiauto exe Define the PFD of your vali file you want to execute optional Default is MAIN o Model i SoftwareSpecific PFD MAIN Specify additional lines to use in the vali MEA file optional o Model i SoftwareSpecific AdditionalMEA add_mea mea aS Listing 5 2 Code matlab Vali model definition zx Define the software the model is developed in required o Model i Software aspen Def
30. d in the fields The o GraphicOptions subfields offer several options for displaying graphicals results The following fields take the value 1 if the option is activated 0 else Tableau A COMPLETER ci dessous 32 mt OSMOSE manual CHAPTER 6 COMPUTATION DEFINITION function o DefineSensi o Define the parameter to vary required Define the name of the model to which the tag belongs required Variables 1 ModelTagName SNG 4 Specify the name that is displayed in results analysis required o Variables 1 DisplayName Gasification pressure Specify the name of the model tag required o Variables 1 TagName gp 4 Specify the unit of the model tag required o Variables 1 Unit bar 4 Specify the lowest value of the variable required o Variables 1 LowerBound 1 15 4 Specify the highest value of the variable required o Variables 1 UpperBound 5 15 Specify the number of steps in the interval required o Variables 1 Number0fSteps 3 Define a second parameter to vary for 2D sensitivity optional o Variables 2 ModelTagName o Variables 2 DisplayName o Variables 2 Unit ea o Variables 2 TagName SANS o Variables 2 LowerBound o Variables 2 UpperBound o Variables 2 NumberOfSteps Listing 6 5 Sensitivity analysis definition in the front end function o Moo monitor
31. del to which the tag belongs required Variables i ModelTagName 1 SNG_h20_psa Define the name of the tag required This might be a general model parameter you use as tag i e an index for economic calculation or a variable used in vali or easy Ex o Variables i TagName MS_index for a parameter o Variables i TagName F_101_P for a vali variable Variables i TagName gp Specify the value of the variable for the run required Variables i Value 1 15 Continue the list 1 111 o Variables i ModelTagName o Variables 3 TagName SE o Variables i Value Listing 6 4 One run definition in the front end function The optimization problem is defined by the fields e o Moo nobjectives number of objectives e o 0bjectiveFunction structured variable indicating the tags that define the objectives e o Variables structured variable defining the decision variables of the optimization problem with their bounds The algorithm properties are defined in the fields e o Moo InitialPopulationSize number of points evaluated in the initial population e o Moo max_evaluations Termination criterion for the optimization maximal number of points evaluated iterations e o Moo nclusters number of families of points distinguished by the algorithm Negatives values can be used to allows Moo to automatically handle clusters The display and output formats are define
32. e OSMOSE version can also be specified here By default this option is set to 2 0 5 2 Vali model definition A Vali model is defined by the location of the executable of Vali and by specifiying the process flow diagram PFD of the Vali model to be executed as shown in listing 5 2 OSMOSE recognises the Vali model file to be used by extracting from the o Model i FileName list the files with bls extension If the FileName list contains more than one Vali file the selection of the Vali file has to be performed in the pre computation function The measurement file input of the vali model is generated by OSMOSE The user can specify additional lines to be written in the measurement file by writing them in a file with mea extension The SoftwareSpecific AdditionalMEA field specifies the name of this file 5 3 Aspen model definition An Aspen model is defined by the location of the Aspen executable and by specifying the model as shown in listing During execution OSMOSE will create a temporary copy of the filename inp append some lines to it execute the Aspen engine in text mode extract the variables and save them in o Model Tags From Aspen s view the interface appears as a Calculator block automatically inserted by OSMOSE The new function DefineAspenTags must be created to define a new tag for each import and export variable Each tag must be defined in a separate line by calling the function define_aspentag as outlined i
33. e matlab Launch osmose e ModelTagName reference to the model containing the variable e TagName name of the concerned tag e DisplayName short description used for displaying in results analysis e Unit unit of the concerned tag for display purposes e LowerBound lower bound of the concerned sensitivity axis e UpperBound upper bound of the concerned sensitivity axis 9 Number0fSteps number of points computed between upper and lower bound 6 5 Multi objective optimisation The parameters for the execution of optimization using the evolutionary algorithm moo are defined within the DefineMoo0ptim front end sub function An example is given in Listing 6 7 The information required to perform an optimization is of three kinds e Definition of the optimization problem e Definition of algorithm properties e Definition of output format 31 6 5 MULTI OBJECTIVE OPTIMISATION OLENTI ISE STI EPFL Local path to Vali Vali Path C Belsim textbackslash bin textbackslash valiauto exe Local path to Aspen Aspen Path C Program Files AspenTech Aspen Plus V7 0 Engine Xeq aspen exe Local Cygwin install Cygwin Path D textbackslash Programs textbackslash cygwin textbackslash bin Listing 6 3 Paths definition in the front end function o DefineOneRun o 0 1 Model nm FilesToCopy report_h2o_psa html 0 i 1 Define the name of the mo
34. e specific files generated tt le 39 8 3 PARETO PLOTS USING OSMOSEPLOTS OLENTI ISE STI EPFL by the model for example composite curves plots Moreover the values of all the tags are stored in a specific matlab structure that is used by osmose_ gui_ plot for Pareto analysis OSMOSE creates a results folder OSMOSE _ results at the location of the front end function For each run of the problem a numbered run folder is created run_ xxx Computation results are stored into two matlab tables set_o mat and set_s mat Figures and files are stored in a subfolder OS MOSE_ recompute containing as many folders as there are recomputed points 8 3 2 Calling OsmosePlots OsmosePlots is a Graphical User Interface gui written in matlab The corresponding functions within OSMOSE are stored into osmose trunk gui Generally after a recomputation the plotter is automatically called by OSMOSE You can however also acess results by entering with Matlab into the run folder of your choice folder containing set_o mat and set_s mat and using the command osmose_gui_plot in the command line 8 3 3 Using OsmosePlots OsmosePlots uses information retreived from the model tags to generate the required graphics It also separated the points according to the cluster classification of moo Figure shows the plotter interactive window the areas outlined on the figure are e a Selection of the x axis Here you can select which Tag is going to be used for the
35. eates a new function that can be useful to the others it has to be documented in the quicktips Same for the FAQ even for the very basic questions Chapter 2 Overview 2 1 What does this document contain The present document is a complete user guide to OSMOSE In addition elements of software structure and architecture are exposed This makes therefore the document useful as an introduction for future developers in the project Chapter 1 explains the procedure to follow for the update of the documentation After the present overview the installation procedure is given in chapter 3 along with technical spec ifications Chapter introduces the basic mathematical and engineering concepts used in the frame of OSMOSE The reader not familiar with design and optimization methodologies will find a summary of the important aspects and a list of references for further reading Chapter 4 introduces osmose usage in a step by step approach Chapter 5 is an extensive guide to OSMOSE model setting Osmose is strongly linked to the use of the Energy Technologies database For more information on Energy Technologies have a look at the documentation https lenisvn epfl ch svn Energy Technologies 2 2 Who should use osmose OSMOSE is designed for researchers and engineers that deal with complex technology models and want to extract the more out of them OSMOSE is also conceived as a way to unify modeling philosophies within a
36. efine the name of the model to whom the tag belongs required 4 Ex o Variables i ModelTagName SNG o Variables i ModelTagName Specify the name of the model tag required Ex o Variables i TagName F_101_P o Variables i TagName 14201 Specify the unit of the model tag required Ex o Variables i Unit bar o Variables i Unit Et atte 4 Specify the name that is displayed in results analysis required Ex o Variables i DisplayName Reactor pressure o Variables i DisplayName Specify the domain of the variable required Ex o Variables i Limits 6208 o Variables i Limits 4 Specify the type of variable required 1 integer 0 continuous Ex o Variables i Is_integer 0 4 o Variables i Is_integer Continue the list Sl o Variables i ModelTagName o Variables i DisplayName o Variables i TagName SAT o Variables i Unit SEEN o Variables i Limits o Variables i Is_integer DRE E ER Eee REA Definition the parameters for displaying o Moo monitor 1 o Moo InitialPopulationSize moo_restart_monitor Ce Le o Moo InitialPopulationSize moo_count_monitor population display in the prompt number of eval o Moo InitialPopulationSize moo_speed_monitor
37. er various conditions For each of the above computations the user defines which are the tags to be varied using the structure o Variables Table 4 2 indicates the nature of the variables for each software used by the model OSMOSE allows also to define o Constants a structure defining tags to be kept constant during a computation but which value can differ from the default value of the tag Some remarks can be made at this point e An input defined by a o Variables field or o Constant field do not need to be defined as a o Model i Tag o Variable or o Constant can however correspond to a o Model i Tag In the latter case the value attributed by OSMOSE through the o Variables or o Constant field is the one considered for computation e Any Vali CST tag that is neither defined as a o Model i Tag aS an o Variables or as ao Constant takes automatically the value defined into the Vali or Aspen model 4 6 1 Model snapshot The model snapshot performs a single computation of the model and stores the obtained tags files and energy integration graphs in the computation run directory see Figurel 4 6 1 When performing a model snapshot the values of tags can be made different than the default value by defining the o Constant structure 4 6 2 Sensitivity analysis Sensitivity analyses are used to observe the variation of the model dependent variables under the varia tion of one or more decision variables OSMOSE allows to
38. erBound x o ObjectiveFunction j TagName j 1 N Multi Objective optimization o Model Tag k Value o Model i Tag m Value FO gt Osmose Model defining the decision variables and their bounds So as to be retrieved by the optimizer the objective has to be a tag of the model The result of such an optimization is a Pareto set The points of the Pareto set are defined and stored by their coordinates in the decision variables space OSMOSE stores also the Pareto curve with the values of the objectives However the complete state of the optimal systems is not retrieved by an optimization computation To have a complete insight on the state of the optimal systems the pareto sets points have to be recomputed and the value of all model tags retrieved as explained in the following section 4 6 4 Recomputation The recomputation tool is used after an optimization As we have seen the results of an optimization in OSMOSE do not furnish the complete state of the system at each optimal point but merely the coordinates of these points in the space of the decision variables During a recomputation the optimal points are run once again in the model in order to retrieve complete information on the state of each optimal point see Figure 4 6 4 This information consists in the value of all the model tags Other properties such as energy integration curves or model input files can also be stored for each of the optimal poi
39. ers O Reactions anipulators Solids Us gt padsheet Physical Property Parameters oh Information flow Material STREAMS Mixer 5 Import variable O Export variable O Tear variable For Help press F1 EO input Required Input Incomplete 8 45 2 ASPENMODEL OLENTI ISE STI EPFL function o PreComputationExample o function o PreComputationExample o OSMOSE models This is an example of pre computation function In a pre computation function all operations are possible on the structured variable o Generally two actions need to be performed ih definition of new tags selection of the external model to run The code below is an example on how you can perform these two operations other ways are of course possible and allowed AUTHOR Francesca Palazzi novembre 2006 Local parameters parameters used only in this function their value is not transmitted to the tag structure Pi 3 14 BigProd_LowerBound 500 LocalBakery_UpperBound 10 Use existing tags definition to create local matlab variables All the existing tags are loaded in the local workspace as variables by evaluating the expression TagName TagValue for i 1 size o Model o ModelID Tags 2 eval sprintf s s char o Model o ModelID Tags i TagName num2str o Model o ModelID Tags i Value end
40. es of Vali can be distinguished a modeling b data reconciliation Modeling usage solves the model equations with an equation solver approach thus computing the state of the system Data reconciliation is used to compute the most probable state of a system with uncertain measurements by minimizing the sum of the square residues A Vali model is stored of a file with bls extension Vali software internally defines objects called tags for modeling and computation The correspondence between Vali tags and OSMOSE tags is performed by giving them the same tag name OSMOSE in puts tag information to Vali through a file called measurement file P The OSMOSE tag structure is updated after resolution by reading the output fil For the definition of a Vali model in the front end see section Aspen model Aspen Plus http www aspentech com is a process modeling tool for conceptual design optimization and performance monitoring for chemical and power industries Aspen is used to simulate chemical and thermodynamic systems thus computing the state of the system It can be run in GUI Graphical User Interface mode or in text mode from the command prompt After having drawn the Aspen model and temp _mea mea in the OSMOSE temp folder r6v extension in the OSMOSE_ temp folder EE qe 4 5 ESTABLISHING A MODEL OLENTI ISE STI EPFL having validated it in the GUI mode the Aspen model has to be exported as an input file filen
41. fore modestly the description to what is called a system The first step to establishing a model is thus to carefully select the system to be described To fix the ideas here is a random list of interesting systems an atom an engine a chocolate factory a parking lot in Upper Manhattan the solar system a district heating in Brno the blood vessels of an trout a heat 8 mem OSMOSE manual CHAPTER 4 RUNNING OSMOSE STEP BY STEP Organized User reques OSMOSE3 knowledge Mss Results abstraction J and communication layer Figure 4 4 OSMOSE3 exchanger You will recognize with this list that there are systems that are more easily described than others Variables The second step to the description of a system is to decide what aspects of it are of interest This is done by selecting the variables of the system Consider the example of the parking lot in Upper Manhattan To be able to run the parking smoothly the tenant is interested in knowing the occupation at each moment of the day the frequency of access and the size of the cars But he will not be interested in knowing the brand and color of each car nor how many people occupy each car Therefore the variables used in a model established by the tenant will be time occupation as a function of time frequency as a function of time and size of cars as a function of occupation The landlord of the same parking lot might be interested in other aspects for examp
42. g and Optimization of SOFC Systems Chemical Engineering Transactions 7 13 18 2005 F Palazzi D Favrat F Mar chal and J Van herle Energy Integration and System Modelling of Fuel Cell Systems Technical report CH 3003 Berne Switzerland 2004 A Plus http www aspentech com B K R Fourer D M Gay Ampl a modeling language for mathematical programming 2003 47
43. he variable Select one objective in the x axis box In the y axis scrolling menu select the tag corresponding to the variable of interest An example is shown on Figure 40 cr y OSMOSE manual CHAPTER 8 RESULTS ANALYSIS Figure 8 1 Osmose plotter interactive window Untitled 1 a A Femmes OSMOSE p ots System efficiency beta version of the osmose super plotter tool Carbon deposition risk ref enhance your understanding of the system Another tag select below by plotting your favorite tags bf seecttne y axis Select a tag for y axis Save this plot This gui interface is brought to you with love by Francesca Palazzi Laboratoire s Energetique Industriels LENLSE STI 2005 Figure 8 2 Menu for x axis selection among all model tags osmose_gui_plot Untitled 1 al Select is mo OSMOSE plots System efficiency beta version of the osmose super plotter tool C Carbon deposition risk ref enhance your understanding of the system by plotting your favorite tags Select the y axis sl Select a tag for y axis z Select a tag for x axis 4 inlet burner temp System cost Fuel ejector cost Air ejector cost Plot Air compressor cost Fuel compressor cost Heat exchangers cost Fuel cell cost Perens Combustor cost Carbon deposition risk reformer System efficiency Power entering the system at the fuel processing inlet methane Power enteri
44. hin OsmosePlots osmose_gui_plot Untitled 1 Select the x axis System efficiency Carbon deposition risk ref 0 Another tag select below BE OSMOSE plots beta version of the osmose super plotter tool enhance your understanding of the system by plotting your favorite tags Select the y axis Carbon deposition risk reformer z Pareto curve analysis for tag carbon deposition risk reformer Plot TAS Ty 5 gt 4 Figure name TL LE o t Paretocurvel 35 2 TA s woe Save this plot S we te 3 3 gn 2 25 2 2L a En Enable data cursor mode 15 i 1 1 L 1 1 1 1 y 044 045 046 047 048 049 05 0 51 052 0 53 system efficiency This gui interface is brought to you with love by Francesca Palazzi Laboratoire d Energetique Industrielle LEN ISE STI 2005 42 Len OSMOSE manual CHAPTER 8 RESULTS ANALYSIS Figure 8 5 Variables projection plot whithin OsmosePlots BEBE a OSMOSE plots beta version of the osmose super plotter tool enhance your understanding of the system by plotting your favorite tags osmose_gui_plot Untitled 1 Select the x axis System efficiency C Carbon deposition risk ref Another tag select below Select the y axis Reactor cost gt Pareto curve analysis for tag reactor cost 450 Plot 400 a o Figu
45. ine the location of the executable aspen file optional Default is C Program Files AspenTech Aspen Plus 2004 1 Engine Xeq aspen o Aspen Path C Program Files AspenTech Aspen Plus V7 0 Engine Xeq aspen exe Define the files that are needed for the computation required o Model i FileName filename inp Define the model file to be run required o Model i FileSelected filename inp Define tags for import export variables o DefineAspenTags o aS Listing 5 3 Code matlab Aspen model definition 5 4 AMPL model definition Listing 5 5 shows the front end section used to define an AMPL model The properties of the model are defined trough the o Model SoftwareSpecific substructure The mandatory field for an AMPL model definition is ModFile that contains the name of the model file complete with extension Optional static data files can be specified trough the field FixedDataFile The solver to be used is specified in the AmplSolver field This field is optional as the minos solver is set by default Consult the AMPL manual for a list of solver options these can be specified through the SolverOption field Additional solving and reporting options can be specified through the Amp10ption field consult the manual for a list of options 7 5 5 Energy integration model definition In this section of the frontend template the EASY model is defined Several specification
46. ing and costing Moreover in these models the treatment of heat exchanges has a distinctive role OSMOSE is tailored to in a first step establish such models rapidly see Section 4 5 and in a sec ond step exploit the model by performing computations such as sensitivity analyses see Section 4 6 2 or optimization see Section 14 6 3 OSMOSE has three levels of functionalities as shown in fig Where OSMOSE fig 4 2 represents the model itself The different softwares that may be used to build a model and the way to implement it in OSMOSE will be described OSMOSE 8883 deals with the different computation types that can be performed on the model Each computation type will be discussed and the way to define it in OSMOSE will be explained OSMOSE3 888 4 represents the data extraction and treatment from the computation results The way to extract and interpret results will be discussed 4 2 MODELING ABC C LENTISE STI EPFL Figure 4 1 OSMOSE onion layer scheme State of the system Decision variables NS Figure 4 2 OSMOSE Computation definition SS Results gt Figure 4 3 OSMOSE2 4 2 Modeling ABC Select the system As said a model is the mathematical translation of phenomena of the real world More precisely models allow to describe and make predictions on how things happen Usually the goal is not to establish a precise description of the whole universe Models restrict there
47. is process is detailed in the subsequent sections For clarity the matlab code is displayed in several pieces within listings A complete front end file template is given in annex faire l annexe 6 1 Computation selection The main function of the front end contains computation definition Listing 6 1 presents an example As seen previously the communication within OSMOSE is performed through a structured variable o All the information is stored within o in different fields The first two fields concern display and output environment settings The environment field o Silent indicates if routine messages have to be silenced o Silent 1 or not o Silent 0 As a rule of thumb deactivate silencing when developing a model or debugging a computation OSMOSE will provide you with many useful messages about computation progression Displaying messages is however extremely time consuming therefore enable silencing for long computations The computation field o ComputationMode selects the quantity of output data to be generated and stored The computation selection fields are o DoOneRun o DoSensi o DoMoo and o DoRecompute Their names are self explanatory The attribution of the value 1 to the field commands the computation ex ecution the value O desactivates the calculation The field o DoRestartMoo is used to restart a moo optimization in case of premature termination Each of the computation fields is naturally activated independentl
48. kages distributed with cygwin The installation of gnuplot has to be done when installing Cygwin by selecting the gnuplot package when running Cygwin setup 3For LENI users get it through the svn server 4nttp www cygwin com EME Lea Chapter 4 Running osmose step by step After a general overview of OSMOSE organization Section 4 1 this chapter introduces the main conceptual steps that compose OSMOSE usage Namely a model definition Section 4 5 b possible computations Section 4 6 and c data analysis Section 4 7 You will get used to interact with OSMOSE through a matlab function called the front end Section 4 8 introduces front end usage The implementation of your problem following each of the steps above is fully developed in Chapters 5 land 4 1 Schematic overview The art of modeling refers to the ability of translating phenomena occurring in the real world into mathematical language Modeling is an essential tool of both scientists and engineers as it allows to understand phenomenas and to make predictions upon the real see Section 4 2 OSMOSE users focus on engineering models developed for designing complex energy systems such as production processes heating networks and power plants These models are generally characterized by the fact that they include phenomenological models such as thermodynamical descriptions or chemical reactions models as well as engineering models such as dimension
49. l convert windows end of lines to unix end of lines Then start cygwin again usr bin dos2unix etc profile Enter EASY2 in the cygwin window to test your installation On linux 1 Install GnuPlot sudo apt get install gnuplot 2 Depending if you are using a 32 or 64 bit version svn co https lenisvn epfl ch svn easy linux x86_32 trunk usr local zero Or svn co https lenisvn epfl ch svn easy linux x86_64 trunk usr local zero 3 Testing source etc profile easy2 e AMPL in case of AMPL model usage or if energy integration is performed with the software Eiampl Control also the availability of the solver you want to use with the AMPL model A student version is available under http www ampl com LENI purchased recently licenses but the files are not tested for the moment e GLPK for energy integration GLPK is an open source software for solving large scale linear programming LP mixed integer programming MIP and other related problems It can be downloaded under http www gnu org software glpk In the near future eiampl using glpk will replace easy for energy integration problems e moo for multi objective optimization moo is also a set of matlab functions SVN Checkout from http lenisvn epfl ch svn moo For the installation of the above consult the specific software documentation Under Microsoft Win dows platforms Easy and gnuplot can only run with the linux emulator Cygwin Gnuplot is a Linux pac
50. le the heat generated by the cars the air pollution and the road usury For the same system the landlord could start with the tenant model and add complexity using more variables such as average heat on each floor air pollution distribution and average road usury in each division of the lot Parameters There are aspects of a system that do not change but that influence the system behavior These are referred to as parameters Some people prefer to consider parameters as variables taking a constant value We make however the distinction between parameters and variables as it can be useful particularly when the parameters take uncertain values this is however behind the scope of the current manual State of a system Using the variables one gains the ability to describe what is happening in the system Each possible situation is referred to as a possible state of the system When the value of all the variables is known one says that the state of the system is known As an example the state of the tenant s parking lot at 1 p m on Wednesday could be occupation 256 places taken over 500 one car arriving every 5 minutes one car leaving every 10 minutes 70 berlines 30 SUV Model equations Now the interest of building a model is to try to capture the relationship between variables so as to be able to draw predictions The case of the parking lot model is a little tricky as the relationship between variables is be dictated by empirical laws
51. m as described in 6 By setting up the appropriate utilities like cooling water or refrigeration cycles and hot streams cre ated through combustion of fuels in a boiler gas turbine etc the system is integrated to meet the MER Defining further heat recovery technology like Rankine cycles the optimal heat recovery for combined heat and power production with respect to operating cost mechanical power production exergy losses or CO emissions is determined In OSMOSE the energy integration is performed with EASY which is an advanced energy inte gration program developed by LASSC of the University of Li ge EASY stands for Energy Analysis and SYnthesis of industrial processes It implements the Effect Modelling and Optimisation approach using Mixed Integer Linear programming techniques to target the combined energy and environment optimal integration of industrial processes An overview of the functionalities of the software are avail able at http leni epf1 ch exsys EASY_0n_1ine_manuals and nttp leniviki epf1 ch The activation of the energy integration model is performed in the front end by setting the variable o ComputeEI to 1 The definition of the streams is generally done trough a template file as explained in Section 4 5 5 Post computation function The post computation function closes the sequence of model subsections It is a Matlab function with accepting o as argument and outputting o This function can cont
52. n listing 5 4 The variables unit type param_ 1 param_ 2 and param_ 3 must match the corresponding ASPEN variable definition It can be found in ASPEN by creating a fake calculator From the ASPEN GUL simply go to Data Browser Flowsheeting options Calculator Variable definition Do not save any file in ASPEN with a calculator in it An example screenshot can be found in appendix Table 5 3 shows the supported variable types and keywords and Table gives a few examples of common variables Type Keyword 1 Keyword 2 Keyword 3 STREAM VAR STREAM SUBSTREAM VARIABLE BLOCK VAR BLOCK VARIABLE SENTENCE INFO VAR INFO VARIABLE STREAM STREAM PROP STREAM PROPERTY Table 5 3 Supported Aspen variable types and keywords m me ay OSMOSE manual CHAPTER 5 MODELS DEFINITION function o DefineModel o 0 the index i can be incremented for any new model itt Define a name of the model required Model i TagName namemodel Define the storage location of the model required Model i Location C oc SNG_model_2 5 Define the software the model is developed in required Model i Software vali Define the files that are needed for the computation required Model i FileName valimodel bls easymodel txt matlabmodel m add_mea mea Define a matlab function performing computations before entering in the software optional Model i
53. ng the system with addtional firing methane Total power entering the system methane Total electricity Net Electricity without addtional firing HX total costs number HX pa 0 2 Save this plot Enable data cursor mode 0 L L L 1 0 0 2 0 4 0 6 0 8 1 This qui interface is brought to you with love by Francesca Palazzi Laboratoire d Energetique Industrielle LENLISE STI 2005 ae y 41 8 3 PARETO PLOTS USING OSMOSEPLOTS OLENTI ISE STI EPFL Figure 8 3 Obtain coordinates of a point using the cursor mode osmose_gui_plot Untitled 1 Select the x axis System efficiency Carbon deposition risk ref Another tag select below a OSMOSE plots beta version ofthe osmose super plotter tool enhance your understanding of the system by plotting your favorite tags Select the y axis Carbon deposition risk reformer Pareto curve analysis for tag carbon deposition risk reformer Plot 45 5 4t Figure name S carbondepositionriskreformer x X 0 4933 3 5 e Y 3329 a Save this plot 2 Ea 31 on 3 jo gt 25 2 A 2 A e Enable data cursor mode 15 f 1 i i 1 f 044 045 046 047 048 049 05 051 052 0 53 system efficiency This gui interface is brought to you with love by Francesca Palazzi Laboratoire d Energetique Industrielle LEN ISE STI 2005 Figure 8 4 Pareto plot whit
54. nts OSMOSE organizes the outputs of a recomputation by storing the information in specific folders For complete recomputation description see section 4 7 Analysing data The previous computations generally generate a large amount of data The organisation of the data for further analysis is an essential point OSMOSE provides the opportunity to generate graphical outputs of the computation using an interactive window 4 8 Frontend usage OSMOSE is composed of several matlab functions The user interacts with the software by defining the problem in a matlab function called the the front end or front end function JN md 4 8 FRONTEND USAGE OLENTI ISE STI EPFL Figure 4 9 Recomputation organization Pareto set points Pareto set of optimal configurations with values of all model tags at each point Ke 2 A lt Recomputation LK Values of decision variables for point Values of all tags for point o Model Tag k Value o Model i Tag m Value FO gt Osmose Model 20 Chapter 5 Models Definition As seen in chapter 4 problem setting in OSMOSE can be performed through a matlab function called front end function or front end This function has to contain model definition as well as computation instructions The present chapter details model definition aspects of the front end In a front end models are defined within the define_model sub function Each model definition consists in a general p
55. only for some computation types ModelsList Contains the TagName of the models For faster recovering Period El El results referring to the whole problem Listing 8 1 Reuslts structure 8 2 Reporting OSMOSE is able to create an automatic report of the computation The report is a pdf file contaning the most important details fo the computation Next paragraphs explain how to setup reporting and indicate where to find the report 8 2 1 Configuring the frontend To allow reporting you must activate this option if the first part of the frontend as shown in listing 8 2 38 OSMOSE manual CHAPTER 8 RESULTS ANALYSIS o DoReport 1 acttivate reporting Listing 8 2 Frontend activation of reproting Several options are available to personnalize the report They can be set in the defineReport sub function as shown in listing 8 3 i ees function o DefineReport o A eee DS Soe Ss Saeed eae Definition of the options to produce a report of the computation iy ER RE Re o Report Format PDF PDF or HTML avre available o Report Name Automatic_Report Name of the report file o Report Title AUTOGENERATED REPORT Title of the report o Report Authors Me you and someone else o Report Date datestr now
56. ration Corresponding variable Corresponding sub function Single run o DoOneRun DefineOneRun Sensitivity analysis o DoSensi Def ineSensi Optimisation o DoMoo Def ineMooOptim Pareto points recomputation 0o DoRecompute Def ineRecompute Restart stopped optimization o DoRestartMoo DefineRestartMoo Default paths are set in the osmose_defaults function However you can redefine the paths in the front end function to match your local installation This definition can be added in the main function of the front end described above The default values of the paths are resumed in table 6 2 for Windows environment Under Linux the paths are already set in the system environment This table gives also the variable name associated to each path Name Default path Variable Name Vali v bin valiauto exe o Vali Path Cygwin c cygwin bin o Cygwin Path Table 6 2 Default paths to external software under Microsoft Windows Listing 6 3 gives an example of path redefinition into the front end 6 3 One run definition Parameters for the execution of a single evaluation are defined in the DefineOneRun front end sub function see Listing 6 4 The definition of parameters for a single run computation is optional If nothing is specified the model is run once using default values for the tags Two optional indications can be given for a one run computation 1 additional files to be stored and 2 values of tags differing from the default values
57. re name Isystemefficiency_reactorcost lt o Save this plot 1 a o 200 reactor cost 150 Enable data cursor mode 100 05 1 1 0 52 0 53 This gui intertace is brought to you with lave by Francesca Palazzi Laboratoire d Energetique Industrielle LENI ISE STI 2005 047 048 049 0 51 system efficiency 50 1 f 0 44 045 046 How can I have an idea of existing correlations between variables Existing correlation between tags can be visualized by plotting the points on a plane formed by the two concerned tags Select the first tag in the x scrolling menu and the second tag in the y scrolling menu An example is shown on Figure In this example the green cluster of points shows a strong linear correlation Be 43 8 3 PARETO PLOTS USING OSMOSEPLOTS OLENTI ISE STI EPFL Figure 8 6 Correlations plot whithin OsmosePlots osmose_gui_plot Untitled 1 a a OSMOSE plots System efficiency beta version of the osmose super plotter tool Carbon deposition risk ref enhance your understanding of the system Another tag select below by plotting your favorite tags Select the y axis Reactor cost Retormer temperature Correlation plot for tag reformer temperature 11005 1090 ee 5 1080 Forcost_retarmetempersture y 2 1070 m Save this plot 2 5
58. rid The model is computed at each node of this grid For each computed point OSMOSE retreives tag information and organizes it into a matrix Energy integration graphics and model files can also be saved to be analyzed in a second time see Figure 4 6 2 tt a 17 4 6 CHOOSING THE COMPUTATION TO PERFORM OLENTI ISE STI EPFL 4 6 3 Optimization Another main feature of OSMOSE is the opportunity to perform optimization on the considered model For this purpose OSMOSE is coupled with an optimizer that allows multi objective optimization of black box models The algorithm used is evolutionary based and is implemented in the software moo moo stands for Multi Objective Optimizer It is a Matlab based application which has been devel oped at LENI by G Leyland 5 and A Molyneaux 10 Its search for optima inspired from genetics is perfectly adapted for solving optimization problems on energy systems which are frequently non linear and non continuous In this context an optimization problem is generally defined as follows min f a z gt s t 4 3 2 IA 0 2 2 lt 0 mt i 1 x 2 N Otherwise stated the problem is to minimize function f x 2 of the decision variables x x1 n and dependent variables 2 21 24 The search is submitted to the model equality constraints h x z 0 and inequality constraints 2 lt 0 and is limited to the space determined by lower bounds
59. s are required to achieve the definition of the EASY model as shown in listing 5 6 Don t forget that the selection of EASY to perform the energy integration and the specification of the objective function to be minimized for the resolution of the heat cascade have been done earlier in the model definition section with the fields Easy Objective and ComputeEI A template of the text files specified as Model i EIFiles and containing the list of hot and cold streams and the definition of the utility and heat recovery systems type param_ 1 param_2 param _3 description STREAM VAR stream name MIXED TEMP temperature STREAM VAR stream name MIXED MOLE FLOW mole flow Table 5 4 Example of common Aspen import export variables 24 ce y OSMOSE manual CHAPTER 5 MODELS DEFINITION function o DefineAspenTags o Definition the tags for import export from the aspen model To get a list of possablilites type help define aspentag in the matlab command window o define_aspentag o name unit type paraml param2 param3 TO ASPEN H20 partial massflow in stream GAS HOT in kg hr named in_H20 define_aspentag o in_H20 kg hr Mass Flow Stream GAS HOT Substream MIXED Component H20 Compressor COMPR pressure named Pressure define_aspentag o Pressure bar Block Var Block COMPR Variable PRES Sentence PARAM From ASPEN Temperature of stream RICH HOT in C named HPin
60. s i TagName cold_t_in p 0 p p 1 o Constants i Period p Value 8 p p 1 o Constants i Period p Value 10 Listing 7 1 Constants definition in multi period problems 7 2 Variables definition It is possible as well possible to use multi period variables to perform a multi objective optimization Moo and to generate randomly a different value per period for each variable with different limits for all the periods if necessary The syntax explained by listing is quite similar to the one used for the constants It has to be noticed that one new variable is generated for each period which multiplies the number of decision variables by the number of period for the optimization problem For example if 7 decision variables are defined for 3 periods this is equivalent to 21 variables in a mono period problem The advantage of this approach is that it is more robust and easier to treat the results afterwards than generating randomly a vector of values However it requires a higher initial population and a higher number of total iterations to converge to a stable Pareto and therefore more computing time 36 OSMOSE manual CHAPTER 7 MULTI PERIOD PROBLEM DEFINITION O O O 522 i 1 Variables i Variables i Variables i Variables i Variables i p 0 p p 1 o Variables i p p 1 o Variables i ModelTagName orc_simple TagName orc_wf3_t DisplayName
61. session windows e Aspen IJ in case of Aspen Plus model usage Aspen Plus is only supported by Windows 1 EPFL licenses for Aspen Plus are hold by the Chemistry and Chemical engineering section 2 To get the installation CD ask someone from LENI for the contact of the person in charge at ISIC e Easy if energy integration is performed in the model with Easy On windows 1 Install Cygwin http www cygwin com Choose Base and GnuPlot also ssh if you want to use pleiades2 2 Prepare the environment The goals here is to set environment variables that are used by easy On windows systems be sure you are using a smart text editor Wordpad and Notepad are not In case of doubts just install Notepad Edit C cygwin etc profile file Add following lines at the end of the file Easy environment variables export SCRLIBTERM tel export EASY_TDF usr local zero dat tdf export EASY_MSG usr local zero dat belsim export EASY_SYNEP usr local zero dat synep export PATH PATH usr local zero bin 2To add osmose folder and all the subfolders combine addpath with genpath in the following way addpath genpath Your0smosePath ae le 5 3 2 REQUIREMENTS OLENTI ISE STI EPFL 3 Open folder C cygwin usr local On Windows restart cygwin If some silly messages appear in the terminal something went wrong with the text editor used to edit profile To solve the problem type the following command which wil
62. speed display in the prompt number of eval 28 A A A O A AO A A RIA O re 97 A A E A pe 0 MN OSMOSE manual CHAPTER 6 COMPUTATION DEFINITION Specify the name of the results directory o Paths MooResultsDirectoryName Listing 6 8 Code matlab Restart moo Definition of files and folders for recomputing the points on the Pareto front Listing 6 9 Code matlab Recompute a Pareto curve 55 Chapter 7 Multi period problem definition The present chapter explains how to define multi period problems in OSMOSE For the moment it is only possible to define periods that have no links between them and storage problem solving is not yet implemented Therefore multi period problems are currently more equivalent to scenarios and each period is simulated sequentially by OSMOSE independently of the others 7 1 Constants definition For any computation type OneRun Sensi Moo Recompute constants can be defined for the required number of periods These constants are tags of the models taking a different value for each period defined There are no limitations in the number of constants that can be defined in multi period but the number of periods has to be the same for every constant defined with the multi period syntax The syntax to be used is described by listing 7 1 i 0 i i 1 o Constants i ModelTagName 1 cold_source o Constant
63. tags to be sent to output and or to be used by successive sub sections Define a tag as input of model i nt 0 Tag index to be incremented at each new tag definition nt nt 1 o o o o Model i Tag nt TagName ZFP_T Model i Tag nt DisplayName Temperature of furnace Model i Tag nt Unit K Model i Tag nt Status CST o Model i Tag nt Value This field is attributed during computation Listing 4 1 Tag definition An example This structure for tags allows also to extract easily tags values For example if the tag Hot utility has been defined in the frontend its value can be recovered as described in the listing 4 5 Establishing a model In this section are grouped the descriptions of the possible sub models you are going to use 2Within OSMOSE this operation is performed by the update_output_tags function It is also the function to call within a user defined matlab function 12 EME Le OSMOSE manual CHAPTER 4 RUNNING OSMOSE STEP BY STEP List of all tags defined in the model o Model Tags TagName Find the position of a given Tag find strcmpi hot_utility o Model Tags TagName Recover the value of a given tag based on its position here 5 o Model Tags 5 Listing 4 2 Tag value extraction An example 4 5 1 Pre computation function The pre computation function is a matlab function performing comp
64. ted for computation trough its numeric identifier stored in the field o ModelID The values of variables decision variables as well as dependent variables are transmitted and updated throughout the computation using a specific sub structure of o Model called Tag Each relevant model variable or parameter is associated to a tag The tag structure is composed of several fields as shown in Listing For an introduction to matlab structured variables refer to matlab user guide 9 10 ce y OSMOSE manual CHAPTER 4 RUNNING OSMOSE STEP BY STEP Figure 4 5 Osmose model general structure Model i Energy flow model i Model preprocessing matlab o Model i PreModelMFunction Model external software or matlab vali aspen matlab o Model i FileSelected ampl o Model i SoftwareSpecific ModFile Model postprocessing matlab o Model i PostModelMFunction Energy integration model i El Model preprocessing matlab o Model i PreEIMFunction El Model easy o Model i EIFiles Re fant El Model postprocessing matlab o Model i PostEIMFunction Legend Tane and modal informatian fags ana model inrormation gt Model sub section osmose variable defining sub section EE 11 4 5 ESTABLISHING A MODEL OLENTI ISE STI EPFL Context Corresponding Name Used or produced by Matlab Name of the Matlab local variable
65. temp folder 14 ce y OSMOSE manual CHAPTER 4 RUNNING OSMOSE STEP BY STEP in dynamic input is an array the sets defining the array dimensions have to be numeric sets starting with one and with an increase of one Example 1 2 3 4 5 is a valid set whereas 4 berkley tomato is an invalid set in OSMOSE input context 4 5 3 Intermediate function The intermediate function is a matlab function performing tags operations and selection of models be fore entering the energy integration model It has the same format as the pre computation and post computation functions and is called o Model 1 ElPreparationFunction in the frontend 4 5 4 Energy Integration Model One important feature of OSMOSE is to allow integration of energy systems involving heat exchange This integration leads to a better efficiency of the considered system by determining the minimum energy requirements MER from external utilities The Energy Integration Model is therefore an optimization problem solved in a sub section of the main model After computation of the thermal and mechanical energy requirements by the external software model and preparing the information in the intermediate function it is possible to perform the energy integra tion of the whole system For this purpose the list of hot and cold process streams is defined which allows to formulate the heat cascade and to determine the minimum energy requirements MER of the syste
66. that allow to compute a system state knowing the value of decision variables In OSMOSE a model is organized as an input output entity The general structure of an OSMOSE model is shown on Figure As the scope of OSMOSE is to allow communication between different software and to include energy integration a model is composed of several subsections A model can contain up to five subsections e Pre computation function External software model e Intermediate computation function e Energy integration model e Post computation function All the model sub sections are linked to the input and to the output of the model Inside the model the subsections are called sequentially as shown by the arrows in Figure 4 3 In the following paragraph we introduce the OSMOSE communication structure by defining the Tag structure The description of each model sub section follows 4 4 The Tags Structure The communication within OSMOSE is performed through a matlab structured variable o All the information is stored in o within different fields The field o Model contains all information defining models OSMOSE provides the option of defining several models to be computed in the same time The variable o Model is thus a vector containing as many elements as there are models defined o Model i for 1 1 Nmodels The model field is itself a structured variable organized into several subfields Section 5 1 details the model definition fields A model is selec
67. ustom report of the computation required o DoCustomReport 0 1 yes 0 no Call of the subfonction handling computations o launch_osmose o Listing 6 1 Front end main function computation definition Each of the computation selection fields has a sub function counterpart containing computation pa rameters definition The definitions sub functions are named DefineOneRun DefineSensi DefineMoo 0 DefineRecompute and DefineRestartMoo Table 6 1 gives computation variables and sub functions connections To call the required sub function and thereafter launch computation front end execution passes through the launch_osmose o sub function This sub function is normally placed at the end of the front end and does not have to be edited Listing 6 2 shows the launch_osmose sub function The last line o run_frontend o calls the OSMOSE program 6 2 Software location Before running a model the path and installation of the software used has to be checked OSMOSE can communicate with the software listed in section 3 by locating their main executable file OSMOSE lo cates automatically Easy and AMPL whereas paths to Vali Aspen and cygwin executables have to be inserted For moo the path has to be added to the Matlab path following the same procedure as for the OSMOSE package see Paragraph 3 1 lil 29 6 3 ONE RUN DEFINITION OLENTI ISE STI EPFL Table 6 1 Computations in OSMOSE Ope
68. utations before calling the external model Its call is defined in the front end by the variable o Model i PreComputationFunction In this function tags can be created and other fields of o Model can be modified Notably the pre computation function can perform the task of selecting the external model next to be called among several files listed The pre computation function argument as well as its output is the structured variable o An example of pre computation function for a Vali model is given in Appendix 1 4 5 2 External Software Model This section of the model calls an sub model established either in Vali Aspen or in AMPL To perform the call OSMOSE uses the values of already defined tags After the external model resolution the list of tags is updated and completed to include external model variables values As seen in section 4 5 1 several files thus several models can be defined as available for the external model step In this case the selection of the external model to call is part of the main model and is usually a function of the decision variables Therefore the pre computation function has to be written so as to select the appropriate model according to the value of the input variables of the problem Vali model Vali is a software to assess validate monitor and optimize process plant performances It includes thermodynamic states computations as well as chemical reactions and equilibrium resolutions Two us ag
69. x for i 1 and upper bounds for i 1 N on the decision variables The nature of the objective function and of the model equations influences the way to solve an op timization problem The following four situations can occur in ascending order of difficulty When f x 2 and the model equations are linear the problem is a linear programming LP model When some of the variables are integer representing for example yes or no decisions the problem is a mixed integer linear programming MILP problem If there are non linear equations the problem is known as being a non linear programming NLP problem A NLP problem is a mixed integer non linear programming MINLP problem if integer variables are involved When f x 2 has more than one dimension the problem is known as being a multi objective problem This is the kind of problem that often arise in the design of industrial systems A simple example is the problem of maximizing the production of a plant while minimizing its installed cost These two objectives are competing when the installed cost is low the production will also be low when the production is high the cost will be high However some solutions will offer a better compromise between the two objectives as others this are the solutions we are going to find The solution of a multi objective problem is a set of solutions that express the possible compromise between the objectives In the domain of the objective
70. xternal softwares The FileName field contains the names of the files composing the model The list should contain all the files that are necessary for model computation Files specified in another o Model field do not need to be mentioned as the list is internally completed by OSMOSE For example the matlab func tions used for pre computation intermediate computation and post computations do not need to be in the FileName list as they are defined by PreComputationFunction ElPreparationFunction and PostComputationFunction Table 5 2 gives the list of fields containing file names 21 5 2 VALI MODEL DEFINITION OLENTI ISE STI EPFL Field Description PreComputationFunction Matlab function performing pre computation EIPreparationFunction Matlab function preparing energy integration PostComputationFunction Matlab function performing post computation ElFiles List of files used for energy integration SoftwareSpecific ModFile AMPL model name of the model file SoftwareSpecific FixedDataFile AMPL model eventual fixed data files list FileName All other model files not defined in above fields Table 5 2 o Model fields defining model files If the model contains an energy integration section the o ComputeEl field has to be set to 1 o Easy Objective contains the name of the objective function used for the resolution of the heat cascade The energy integration properties of the model are further detailed in section 5 5 Th
71. y as it makes for example little sense to perform an optimization right after a sensibility analysis One exception is the advisable recomputation of optimization results right after an optimization For OSMOSE model structure see Section 4 5 For model definition through the front end refer to Chapter 5 28 OSMOSE manual CHAPTER 6 COMPUTATION DEFINITION Customizing options for reporting are also defined in the main part of the front end their description is given in chapter function o My_Front_End_Function_Name Environement and output settings Reduce the routine messages to the minimum required o Silent 0 A 1i yes 0 no Decide about the output of results optional o ComputationMode simple simple no details details detailed results which include composite curves Computation setting Perform a single evaluation of your model required o DoOneRun 1 a E EE 08 ano Perform a sensibility analysis required o DoSensi 05 1 yes 0 no Perform an optimisation with moo required o DoMoo 0 A le yeso 0 no Restart an optimisation with moo required o DoRestartMoo 0 1 yes 0 no Recompute the points on the Pareto front to get the details required o DoRecompute 0 1 yes 0 no Reporting setting Generate a report of the computation required o DoReport 0 1 yes 0 no Generate a c

user manual - ipese

Contents

Download Pdf Manuals

Related Search

Related Contents