a model on co2 emission reduction in integrated steelmaking by optimization methods

INTERNATIONAL JOURNAL OF ENERGY RESEARCHInt. J. Energy Res. 2008; 32:1092–1106Published online 17 July 2008 in Wiley InterScience(www.interscience.wiley.com). DOI: 10.1002/er.1447

A model on CO2 emission reduction in integrated steelmaking byoptimization methods

C. Wang1, M. Larsson1, C. Ryman1,2,�,y, C.-E. Grip2,3, J.-O. Wikstrom1,2,

A. Johnsson1 and J. Engdahl4

1Centre for Process Integration in Steelmaking, MEFOS—Metallurgical Research Institute AB, P.O. Box 812, SE-97125 Lulea, Sweden2Lulea University of Technology, SE-97187 Lulea, Sweden

3SSAB Tunnplat AB, SE-971 88 Lulea, Sweden4SSAB Tunnplat AB, SE- 78184 Borlange, Sweden

SUMMARY

The iron and steel industry is a large energy user in the manufacturing sector. Carbon dioxide from the steel industryaccounts for about 5–7% of the total anthropogenic CO2 emission. Concerns about energy consumption and climatechange have been growing on the sustainability agenda of the steel industry. The CO2 emission will be heavilyinfluenced with increasing steel production in the world. It is of great interest to evaluate and decrease the specific CO2

emission and to find out feasible solutions for its reduction. In this work, a process integration method focusing on theintegrated steel plant system has been applied. In this paper, an optimization model, which can be used to evaluate CO2

emission for the integrated steel plant system, is presented. Two application cases of analysing CO2 emission reductionpossibilities are included in the paper. Furthermore, the possibility to apply the model for a specific integrated steelplant has been discussed. The research work on the optimization of energy and CO2 emission has shown that it ispossible to create a combined optimization tool that is powerful to assess the system performance from several aspectsfor the steel plant. Copyright r 2008 John Wiley & Sons, Ltd.

KEY WORDS: process integration; modelling; CO2 emission; optimization; steel industry

1. INTRODUCTION

The iron and steel industry is the largest energy-consuming manufacturing sector in the world.

Therefore, concerns about energy consumptionand climate change have been growing on thesustainability agenda of the steel industry. Theworld’s annual steel production has been steadily

*Correspondence to: C. Ryman, Centre for Process Integration in Steelmaking, MEFOS—Metallurgical Research Institute AB,P.O. Box 812, SE-97125 Lulea, Sweden.yE-mail: [email protected]

Contract/grant sponsor: Swedish Governmental Agency for Innovation SystemsContract/grant sponsor: Knowledge FoundationContract/grant sponsor: Swedish Foundation for Strategic ResearchContract/grant sponsor: Luossavaara-Kirunavaara ABContract/grant sponsor: SSAB Tunnplat ABContract/grant sponsor: Rautaruukki Oyj

Received 15 October 2007

Revised 18 February 2008

Accepted 18 March 2008Copyright r 2008 John Wiley & Sons, Ltd.

increasing during the past decades, and a particu-larly rapid increase is noticed since 2000. Accord-ing to IISI [1,2], the world’s crude steelproduction was 1058Mt in 2004, exceeding1000Mt for the first time in the steel productionhistory. The BF/BOF (blast furnace/basicoxygen furnace) route and the electric arcfurnace (EAF) route are the two dominatingprocess routes. The share of the BF/BOF andEAF-based production of crude steel in 2005 was65.4 and 31.7%, respectively. The CO2 emissionfrom the steel industry links to the productionprocess. As for the BF/BOF route, the reductionand melting of iron ore to hot metal (HM) in theBF is almost entirely based on coal. Conse-quently, the steel production industry emits largeamounts of carbon dioxide, accounting for about5–7% of total anthropogenic CO2 emission [3].The CO2 emission will be heavily influenced byincreasing steel production in the world. It canbe anticipated that CO2 emission from thesteel industry will increase with the increase incrude steel production in the near future unlesssignificant changes in the current processroute shares or significant energy/productionefficiency can be made, or some effective CO2

emission reduction technologies, e.g. carboncapture and storage, can be employed widely inthe iron and steel industry. It is of greatsignificance to develop a method to analysepotential CO2 reduction possibilities in the steelindustry. Some studies [4–6] have analysed CO2

emission reduction options within the iron andsteel industry. Most of these studies applieda simple top–down econometric approach,neglecting complex interactions of differentprocess units for the steelmaking. In this study,a process integration (PI) method focusing onthe integrated steel plant system (the conven-tional system of the BF/BOF) has been applied.An optimization model, which can be used toevaluate CO2 emission by optimizing ferrousburden material use in the BF–BOF system, ispresented. The study also covers carbon-tradingschemes in order to find out the lower abatementcost option(s). Finally, a possibility of applyingthe model for a specific integrated steel plant isdiscussed.

2. MODEL DESCRIPTION

The model developed is based on a PI technique,mathematical programming, to analyse CO2 emis-sion by optimizing material and energy systems inthe steel industry. A survey on mathematicalprogramming applications indicates that a broaderapplication of optimization has been focusing onchemical and petroleum engineering. For themetallurgical industry it has been mainly restrictedto the application of linear programming forinventory control, blending, scheduling and simi-lar purposes [7]. Deo et al. [8] described thepossibilities to use either mathematical program-ming or genetic algorithms to find the optimumoperating conditions in integrated steelmaking.However, till now unexpectedly few reports onhow to solve the complex steelmaking by PI toolsare available. In this paper, the method describedis based on the mixed integer linear programming(MILP). The method uses a graphical interfaceequation editor ReMIND, which was developedin cooperation between two Swedish Universitiesof Linkoping University and Lulea University ofTechnology, to generate the mathematical equa-tions to be optimized. Figure 1 shows the flowchart of the model structure. There are severalnumerical solvers available, which can be used foroptimization. In the presented work, the ILOGCPLEX linear programming solver is used.Microsoft Excel is used to analyse the modellingresults with some MACRO commands.

The principle of ReMind model is presented inFigure 2. The model is represented by nodes andbranches where the branches represent energy ormaterial flows and a node may represent a process

GUI/Equation editor

Equation Solver

Spreadsheet

Model design tool

Optimizing tool

Analyzing tool

Results

Figure 1. Flow chart of the optimization model.

A MODEL ON CO2 EMISSION REDUCTION IN INTEGRATED STEELMAKING 1093

Copyright r 2008 John Wiley & Sons, Ltd. Int. J. Energy Res. 2008; 32:1092–1106

DOI: 10.1002/er

unit as well as a production line or a whole factory.Each process node has its own energy demand inthe form of electricity and/or heat demand. Thesedemands depend on the amount of materialprocessed in the unit and may be described bylinear or piecewise linear relations. The variationsare described in the system with boundaryconditions, for instance, production capacity,limited availability for various resources such as

fuels, electricity or raw materials. Each system isadjusted to the situation in each individual case.The adjustment is made to answer the questions inthe individual case and to make the model asefficient as possible.

In this work, ReMIND has been used for theintegrated steelmaking system, which coversprocesses of coke oven plant (COP), limefurnace, BF, BOF, ladle metallurgy, continuouscasting (CC) and combined heat and power(CHP). The model includes four kinds of nodes:material flow nodes, energy flow nodes, processnodes and end product nodes. Material and energyflow nodes are the input nodes for the model. Thecore nodes for the model are the process nodesthat contain the basic metallurgy processes.Processes are described by mass and energybalance to link ingoing material and energyflows, thereby connecting the different processes.An example process node, the BF node, is shownin Figure 3. The end nodes include the mainproduct from the processes, for instance, slabs forthe whole system, HM or liquid steel if we are onlylooking at the BF or the BF1BOF, etc. The otherend nodes could be heat and power generation, gasto flare, etc.

Production demand node

Process nodes

Energy supply node

Energy flow

Material flow

Material supply node

1

2

3 4

5 6

7 8

9

Figure 2. Schematic description of the principle of theReMind model.

Figure 3. An example of the function editor in the BF node.

C. WANG ET AL.1094


DOI: 10.1002/er

2.1. Objective function(s)

There is a possibility of defining several objectivesin the model depending on the objective problemstudied. These can either be analysed one a time,i.e. single objectives, or combined, i.e. as multi-objective function. Generally, the objective can beexpressed in mathematical terms as follows:

min zðx; yÞ ¼X

cjxj þ bjyj ; j ¼ 1; . . . ; n ð1Þ

s:t:

A1x� b1

A2xþ By� b2

x 2 Rn; y 2 f1; 0g or integer

where z is the objective function for minimizingCO2 emission, x represents the studied variables(xi means the ith variable), y represents the binaryvariables, cj is the coefficient for the jth variable inthe objective function and bj is the coefficient forthe jth binary variable in the objective function.

Engineering design often deals with multiple,possibly conflicting, objective functions or designcriteria. For instance, one may want to maximizethe performance of a system while minimizing itscost. Such design problems are the subject ofmulti-objective optimization. Thus, the multi-objective function is needed when optimizingmore objectives at a time is required. It is usefulto find out an optimum solution with a lowerproduction cost and at the same time with a lowerCO2 emission. There are several different

approaches for multi-objective optimization,e.g. weighted sum, e-constraint and goalprogramming. A more detailed description oneach approach can be found in [9]. In this work,e-constraint method is used for multi-objectiveoptimization. For the e-constraint method, onlyone objective is optimized, whereas the otherobjectives are bounded by some constraints.

In this study, for the multi-objectiveoptimization problem based on Cost and CO2

emission minimization, it can be expressed by thefollowing equation:

minX

n

anbnX

t

X

m

ðCm;t;nxm;tÞ ð2Þ

where n denotes objective, e.g. the objective will bethe cost when n5 1, and CO2 emission when n5 2,etc., xm,t is the flow m for the time step t, cm,t,n isthe coefficient for the flow m of objective type n intime step t, an is a coefficient making it possible tonormalize each objective function n, whereas bn isa coefficient making it possible to weight theobjectives (note that one constant could also havebeen used, but two constants were used tofacilitate the study). The constants also providethe possibility to exclude any objectives from theoptimization by setting them to zero. an and bncorrespond to K1 and K2 in Figure 4.

The studied objective is bounded according tothe following equation:

X

t

X

m

Cm;t;nxm;t � Cn;t 8n ð3Þ

Figure 4. Scope and time step of the model.



DOI: 10.1002/er

where Cn,t is a constraint for the objective, eithercost or CO2 emission, during the time step t.

As shown in Figure 4, the scope of the CO2

emission can be defined locally for direct emissionfrom a specific plant or globally including bothupstream and downstream emissions. The lattercan be used when doing a life cycle assessment(LCA) for the studied system. The model cansimulate CO2 emission for a fixed time or during atime span; therefore, a time-step function isneeded, see Figure 4. For example, the time-stepfunction is needed when analyzing the CO2

emission for different periods for the steel plantsin the emission-trading program.

In connection with the multi-objectiveoptimization, it is possible to find Pareto-optimalsolutions [10]. A Pareto-optimal solution is asolution where no objective can be improvedwithout another deteriorating. The plot of theobjective functions is called the Pareto front, anexample of a Pareto front is shown in Figure 5. Asfor the bi-objective optimization problem, thePareto front curve represents all the solutionsfrom minimizing one objective with upper-levelconstraints bounded by the other objective, andvice versa. This allows the decision maker tochoose an acceptable trade-off between the twogoals by considering the different solutions alongthe Pareto front.

2.2. System definition

Figure 6 shows the system boundary. At the firststep, the model boundary covered the mainprocess units of the BF and the BOF, i.e. SystemI. The model was further extended to cover COP,CC and CHP in System II. Finally in System III asub-model of a rolling mill (RM) is included; thus,the model boundary has covered a fully integratedsteel plant, i.e. COP-BF-BOF-CC-RM.The model can be used to analyze the CO2

emission either for the whole system jointly orfor one or a few sub-models separately dependingon the research interests. Two application casescovered by this paper correspond to differentsystem boundaries in Figure 6, optimizing ferrousburden materials in BF–BOF [11] and emission-trading schemes’ (ETS) influence on CO2 emissionreduction [12]. A customized model for a Swedishsteelmaker, SSAB Tunnplat AB, with two inte-grated production sites of steelmaking and RM, asan example of a fully integrated steel plant, will bediscussed in the paper as well (System III).

2.3. Validation

The model used in this work is based on anexisting model that was initially developed foranalysis of the energy use for an integrated steelplant, and the model has been validated by usingactual production data [13]. This model has beensuccessfully used in several studies mainly focusingon material, energy use and production cost

Figure 5. Example of a Pareto front for a bi-objectiveminimization problem.

Coke oven

CC

BF

BOF

System I

CHP Rolling Mill

System II System III

Figure 6. Scheme of the model development layout.Note: Material and energy flows between and withinprocesses/sub-models are not included in the figure.

C. WANG ET AL.1096


DOI: 10.1002/er

[14,15]. In each paper one reference simulation hasbeen performed where the model has beendelimited in production practice accordingly.Agreement between the model result and thereported data has been good.

3. APPLICATION FOR OPTIMIZINGFERROUS BURDEN MATERIALS IN THE

BF–BOF SYSTEM

This modelling work corresponds to System I (seeFigure 6), which consists of the BF and the BOFmodules. These two processes are interconnectedto each other by HM. Two sub-models can beoptimized separately or combined together.

This model has been used to analyse howconversion costs and CO2 emissions can beinfluenced by use of different ferrous burdenmaterials; for instance, iron ore pellets, steelscrap or direct reduced iron/hot briquette iron(DRI/HBI) when producing crude steel. In thisstudy, the use of DRI/HBI has not beenseparately analysed as they have a similarbehaviour as scrap in the BF–BOF system. Thecoefficients for the objective functions and somebase constraints set for the main processes arepresented in Tables I and II. A crude steel demandof 500 t h�1 has to be satisfied for all cases. In theBF, the HM silicon content has been allowed tovary in the range 0.2–1.0% to extend the feasibleoperating range for the model. The scrap use in theBF process has been restricted to 20% of the Fe

input for HM production. In a general study, someparameter combinations cannot be used in everyplant (e.g. scrap charging in the BF).

3.1. BF1BOF baseline optimization

Table III shows the modelling results for the twosub-models of the BF and the BOF combined andseparately. For the combined optimization, theresults are related to optimization of the objectivefunctions in relation to the produced steel leavingthe BOF. In general terms, the most cost-efficientsolution, with the given cost values, is to produce aHM with low silicon content on a 100% pelletburden in the BF, and to use iron ore pellets ascoolants in the BOF process. The strategies toproduce crude steel with low CO2 emissions andlow energy use are completely different from thecost-optimized solutions. To minimize CO2 the

Table I. Coefficients used for different objective function [11].

Unit Energy (GJ) CO2 emission (ton) Cost (USD)

Iron ore pellet (KPBO) (ton) — — 90Scrap (�97% Fe) (ton) — 0.0147 230Purchased coke (ton) 28.05 3.035 250Pulverized coal injection (PCI) (ton) 27.21 2.468 50Natural gas (GJ) 1 0.0565 5Lime (ton) — — 60Quartz (ton) — — 10Limestone (ton) — 0.44 10Dolomite (ton) — 0.477 10Oxygen (1000m3n) — — 25Power (MWh) 3.6 — 50

Table II. Base model constraints for the BF andthe BOF.

BF BOF

Production (t h�1) — 500Pellet use (%) — —Scrap use (%) o20 —% C in product (%) 4.5 0.05% Si in product (%) 0.2–1.0 0Coal injection (kg t�1 HM) 160 0Slag volume (kg t�1 HM) 165 —Slag CaO/SiO2 — 1 3.3Tap temperature (1C) 1468 1675

(—) means that the variable is unconstrained.



DOI: 10.1002/er

model prescribes that the scrap addition to the BFis maximized and that the HM should have thehighest possible silicon content to allow massivescrap melting capacity in the BOF. The strategy forenergy minimization is similar. It is noticeable thatthe cost-optimized practices cause more than 45%higher CO2 emission compared with the CO2 andenergy-optimized practices. On the other hand, theCO2 and energy-efficient practices are more costly.

As for only the BF optimization, the resultsare interesting because the way of minimizingCO2 and energy is different compared with the

former combined optimization. Now the strategyis to produce a low silicon HM, as low as allowed,in order to keep the specific coke use as low aspossible. However, when looking at the combinedBF1BOF system, it is more beneficial toallow a higher specific coke consumption in theBF to gain a higher scrap melting capacityin the next process step. This result demons-trates the benefits that can be gained byusing a system-oriented analysis approachcompared with the optimization of each processseparately.

Table III. Optimization results for systems of BF1BOF and BF.

BF1BOF combined system

No opt. Min. CO2 Min. energy Min. cost

Objective valueCO2 emission (t t�1 LS) 1.25 0.99 0.99 1.43Energy (GJ t�1 LS) 12.56 9.95 9.95 14.29Cost (USD t�1 LS) 246 256 256 238

BFPellets (kg t�1 HM) 1425 1124 1124 1431Scrap (t t�1 HM) 0 197 197 0HM quality (% Si) 0.60 1.0 1.0 0.20Coke1PCI (kg t�1 HM) 475 425 425 468Fluxes (kg t�1 HM) 117 149 149 109Slag volume (kg t�1 HM) 165 165 165 165BOFPellets (kg t�1 LS) 24 0 0 56Scrap (kg t�1 LS) 170 296 296 0Oxygen (m3n t�1 LS) 48 50 50 48Fluxes (kg t�1 LS) 53 76 76 24Slag volume (kg t�1 LS) 110 147 147 62

BF system only

No opt. Min. CO2 Min. energy Min. cost

Objective valueCO2 emission (t t�1 HM) 1.25 1.07 1.07 1.23Energy (GJ t�1 HM) 13.97 12.13 12.13 13.78Cost (USD t�1 HM) 224 228 228 223

BFPellets (t t�1 HM) 1425 1146 1146 1431Scrap (t t�1 HM) 0 197 197 0HM quality (% Si) 0.60 0.20 0.20 0.20Coke1PCI (t t�1 HM) 475 412 412 468Fluxes (kg t�1 HM) 117 132 132 109Slag volume (kg t�1 HM) 165 165 165 165

Note: The bold figures indicate the optimization objective values.

C. WANG ET AL.1098


DOI: 10.1002/er

3.2. The effect of scrap distribution between BF andBOF

Scrap offers effective means to lower the CO2

emissions in the BF–BOF system, however theprevious calculations have delivered inconsistentsolutions on which combination of the BF andBOF practice that is the most effective. Scrap ispossible to charge to both the BF and the BOFprocesses.

Figure 7 shows the different scrap distributionsbetween the two considered processes. Constantquantities of scrap (50, 100, 150, 200 and 225 t h�1)have been added to the system and have beenallocated in different proportions to the BF andthe BOF. The propagation of each of the filledlines in the figure corresponds to the feasibilityregion of the defined system. The minimum CO2

objective of the system is 0.99 t t�1, which wasgiven earlier in Table II. This corresponds to asingular point in the diagram situated directlybelow the 200 t h�1 line. The minimum CO2

objective when there will be no scrap charged tothe system is 1.43 t t�1, corresponding to thesingular point situated on the right upper side ofthe diagram. The dotted line in Figure 7 representsthe distribution that corresponds to the minimumCO2 objective for different scrap addition levels tothe system. It can be seen that the CO2 emission isdecreasing with adding more scraps to the system.When the scrap addition level is lower than

100 t h�1, the optimized solution will alwayschoose to add scraps into the BOF in order tohave a lower CO2 emission. The minimum CO2

objective when 100 t h�1 (200 kg t�1 LS) of scrap isavailable to the system is 1.20 t t�1, whichcorresponds to the right end point of the100 t h�1 line. When the addition level is above100 t h�1, the scraps to system will be distributedbetween the BF and the BOF for the minimumCO2 emission. Thus, when seeking a lower CO2

emission by increase of the scrap additions, it ispossible to find an optimum distribution betweenscrap charging in the BF and the BOF for eachscrap-charging level.

3.3. Pareto front analysis

A Pareto-optimal solution is a solution where noobjective can be improved without another dete-riorating. The two objectives of Cost and CO2 canbe weighed versus each other as shown in Figure 8,where the Pareto front defined by minimum cost atdifferent CO2 emission levels have been drawnwith a line between A1 and A2. The point’s A1 andA2 represent the solutions Min CO2 and Min Costfrom the optimization of the BF–BOF system.A simplistic description of the conditions A1

and A2 is that the use of scrap is maximized inA1 and the use of iron ore pellets is maximized inA2. There are several breakpoints for the Paretofront illustrated in Figure 8, which relates to the

0.95

1.00

1.05

1.10

1.15

1.20

1.25

1.30

1.35

1.40

1.45

0

Scrap allocated to BOF (%)

t C

O2

/ t L

S 100 t/h (≈20%)

225 t/h (≈45%)

200 t/h (≈40%)

150 t/h (≈30%)

BF BOF

50 t/h (≈10%)

10 20 30 40 50 60 70 80 90 100

Figure 7. CO2 emissions at different scrap distributions between the BF and BOF.



DOI: 10.1002/er

different regions of a– c in Figure 8. Thesebreakpoints clearly show the borderlines for thetechnical solutions for the different Pareto frontlines. For purposes of illustration and betterunderstanding, the solutions outside the Paretofront are also shown with a dotted line.

4. APPLICATION FOR ANALYSING CO2

EMISSION REDUCTION IN THE STEELINDUSTRY BY USING ETS

To comply with Kyoto Protocol (KP) commit-ments, the EU decided to introduce a cap andtrade program, the so-called ETS, to curb Eur-ope’s industrial emissions. EU ETS is an internalmarket within EU countries to trade carbondioxide emissions, enabling companies exceedingindividual CO2 emission targets to buy allowancesfrom ‘greener’ ones. It is permissible to useCertified Emission Reductions (CERs) gainedfrom CDMz projects to meet the CO2 emission

allowance for EU countries. This practice has notyet been fully accepted for Swedish conditions.Instead, a general study, using the optimizationmodel on a given example, has been carriedout to evaluate how steel plants in Europeancountries can meet their emission reductioncommitment [12].

As shown in Figure 6 (System II), the modelboundary was extended to cover the ETS (in thiscase, they are CDM and EU ETS). In the model,the function of the time step is used as both CDMand ETS are time-step-based schemes. The timesteps set in the model are the following: before theKyoto Protocol (BKP), the KP, and the postKyoto Protocol (PKP), as shown in Table IV. Thetable also presents the production forecast andassumed CO2 emission allowance during the timesteps.

The following cases are simulated in the model:

� Reference case—business as usual (BAU): Thisscenario is a projection based on a series ofconsistent assumptions. In this scenario, nomeasures (internal or external) were taken toreduce CO2 emissions at the steel plant. Thedriving force in the model is the projectedproduction during time steps.

� Case 1—ETS simulation: In this simulation, theEU ETS is used to fill up the emission gaps. Themodel was bounded by the CO2 emissions

a b c

235

240

245

250

255

260

0.90

CO2 emission (t/t LS)

Co

st (

US

D/t

LS

)

A2

A1

(0.99) (1.43)1.00 1.10 1.20 1.30 1.40 1.50

Figure 8. Optimization of Cost and CO2 for the studied BF–BOF system. The Pareto front is the solution on thethicker line.

zCDM is one of the three so-called ‘flexible mechanisms’defined in the Kyoto Protocol enabling Parties to access cost-effective opportunities to reduce emissions or enhance carbonsinks in other countries. It is generally considered that thesemechanisms have the capacity to lower the overall costs ofachieving the emission targets of the Protocol. In addition toCDM, the other two mechanisms are the InternationalEmissions Trading (EIT) and the Joint Implementation (JI).

C. WANG ET AL.1100


DOI: 10.1002/er

allowance, i.e. the steel plant needs to buy theexcess emission via the emissions-trading mar-ket within the EU. An average carbon permitprice of 29.6 US$ t�1 CO2 indicated in Hidalgoet al.’s study [5] has been used in the model.

� Case 2—ETS and CDM optimization: In thisscenario, the emission gap will be filled up byeither buying allowance permits via ETS orpurchasing CERs via CDM. The types of CDMprojects in the study are recovery of BF gas,injection of natural gas, pulverized coal injec-tion system for BF and waste gas recovery fromBOF.

� Case 3—Optimization scenario: The optimizedcost objective strives to decrease the productioncost for the system to its minimum whilesatisfying the CO2 emissions limitation, andhence minimizing the CO2 reduction cost. Be-sides the EU ETS and CDM, internal changeswithin the steel plant are included. The exam-

ples of internal changes are coking coal mixingin the COP; different coal injection rates, BOFslag charging and flue dust injection into theBF; HM/scrap rate and decreased iron orepellet charging into BOF; back pressure/con-densing operation in CHP, etc. The model wasset free to optimize among the different alter-natives.

The simulation results of CO2 emission arepresented in Figure 9 indicating lower predicatedCO2 emission than the emission allowanceallocated for the first 2 years in the BKP period.However, the predicted CO2 emission will exceedthe allocated emission from the last year (2007) inthe BKP period through the entire time step.

Figure 9 also shows the CO2 emission gapsduring the different time steps and the cost forCO2 emission reduction in the different cases. Theabatement cost shown in the figure is calculated

Table IV. Time steps used in the model and steel production forecast in the studied system.

Time step BKP KP PKP

Year span 2005–2007 2008–2012 2013–2020Production projection (%)� �107 108 108CO2 emission allowance (kt year�1) �4000 �3800y (�4%) �3600y (�10%)

�Production forecast change is based on the production for the reference year with an assumed increased production by 8% at the endof each period. For the year of 2007 in the BKP, the production forecast is assumed as a 7% increase compared with the first 2 years.Note that the increased production is only a calculation scenario and not a decided production plan.yAssumed emission levels for the KP (�4%) and the PKP (�10%) of the BKP level.

Figure 9. CO2 emission allowance, calculated CO2 emission (BAU scenario) and abatement cost during differentperiods at the studied system.



DOI: 10.1002/er

based on the assumed permit price from EU ETSand CDM, and the amount of CO2 emission gapduring different periods.

In case 1 (ETS simulation), the EU ETS is usedto fill up the emission gaps. The steel plant needsto buy the excess emission via the emission-tradingmarket within the EU with the price per unitallowance of 29.6US$ t�1-CO2.

In case 2 (ETS and CDM optimization),purchasing allowance permits via ETS or CERsvia CDM will fill the emissions gap. Comparedwith case 1, the abatement cost for the differenttime steps decreases to 15.4 US$ t�1-CO2, onaverage.

In case 3 (optimization scenario), all possiblealternatives are included in the model, i.e. internalmeasures, ETS and CDM scenarios. The modelwas set free to optimize among these differentalternatives. The result from the optimizationshows that through internal changes, thecalculated CO2 emissions are reduced for allperiods. Consequently, the studied system willnot make use of CDM and ETS during the firstperiod (including the year of 2007), when the CO2

saved through the internal changes will be enoughto fill up the gap. However, from the KP period,the calculated CO2 emissions will exceed theemission allowance allocated if the plant onlymakes internal changes. Thus, other measures arenecessary. When further analyzing the modellingresults, it was found out that ETS will not be usedto fill up the emission gap even for the last twoperiods; instead the model will choose thealternatives from the CDM scenario due to itslower abatement cost. The resulting abatementcost in case 3 is the lowest (9.8 US$ t�1-CO2 onaverage) compared with the other two cases.

It should be pointed out that for the studiedcase, internal changes can play a major role inreducing the abatement cost. When the internalchanges are taken during the whole BKP period,there will be no emission gap at all; instead there isan allowance surplus, which can either be used tofill up future gap or bank them for the futuretrade. Consequently, in the optimization case, thecost for CO2 reduction is further lowered to 9.6US$ t�1-CO2 during the period of the KP and to13.6 US$ t�1-CO2 during the period of the PKP.

5. A TWO-SITE MODEL OF SWEDISHSTEELMAKER SSAB TUNNPLAT AB

SSAB Tunnplat AB is one of Europe’s leadingmanufacturers of high-strength strip steels. Thecompany has ore-based steel production and stripsteel manufacture. Compared with a conventionalintegrated steel plant, SSAB has a unique featurein that the steel and sheet/strip production arelocated at the two different geographic locations,Lulea and Borlange, approximately 800 km apart.The slabs produced from the steel work (Lulea)have to be transported by train to the RM(Borlange) to produce hot-rolled and cold-rolledproducts. This creates several challenges for thesteelmaker:

� Owing to the geographical situation it isnecessary to extend the energy-saving meth-odologies compared with the situation at anormal integrated plant;

� A holistic view is needed to economize the useof resources, and to evaluate and incorporatenew technologies and methods, in terms of asustainable development.

As shown in Figure 10, the integrated steelplant in Lulea includes coke ovens, an ironmakingplant with one BF, a steelmaking plant withtwo BOFs, and a CC plant with 100% CC ofslabs. The RM in Borlange includes both hotand cold rolling. Depending on the customizedproducts, the other process units such aspickling, annealing, aluzinkline and galvline areincluded. Both the sites provide hot water tocommunities via district heat system. Unlike thecommon integrated steel plant, in which someparts of process gases generated duringsteelmaking is used in the RM, the excessprocess gases are transported to a CHP plantfor electricity production for both internaland external use. The excess electricity istransmitted to the power grid. Thus, the RM isconnected to the steel plant to some extent asthe electricity consumed at the RM is from thepower grid.

Recently, a specific model for the RM has beencreated to analyse the energy system. It will be veryinteresting to link this two-site model to analyse

C. WANG ET AL.1102


DOI: 10.1002/er

the possibilities of reducing CO2 emission from anintegrated point of view.

6. DISCUSSIONS

The optimization model developed for the CO2

emission analysis for integrated steelmaking can beused in different ways. The system boundary can bechosen depending on interests of the research work;correspondingly, CO2 emission can be simulatedfor different process units, the whole steel plant orfrom a global point of view (e.g. LCA).

The model used for analyzing scrap additioninto the BF/BOF system shows that different

technical solutions have been chosen to minimizeCO2 emission. When looking at the combinedsystem, it is more beneficial to allow higher cokeconsumption in the BF and higher silicon contentin HM in order to gain a higher scrap meltingcapacity in the BOF; thus, lower CO2 emission willbe achieved. However, if only looking at the BFoptimization, the solution will tend to a lowersilicon content HM production and to keep alower consumption instead. It can be seen fromthis analysis that it is important to actually have asystematic view in order to avoid a sub-optimalsolution by just adding scrap, which will make itpossible to decrease CO2 emission by changing theraw materials in a clever way. However, it should

Figure 10. Schematic diagram of steel and sheet production line at SSAB Tunnplat AB



DOI: 10.1002/er

be pointed out that the prerequisites for scrapcharging in the processes are different, and thereare also a number of other factors to consider,among other things charging technology,productivity effects, scrap availability and trampelement contamination.

The model can be used as an assistant tool tohelp the decision maker choose an acceptabletrade-off between two goals by considering thedifferent solutions when looking at the Paretofront. The use of the Pareto front for the BF/BOFsystem shows that the solution range is quiteextensive. This means that there is a wide range ofpossibilities to operate the production systemconsidering the trade-off between the twocriteria’s cost and CO2 emission. The choice ofsolution will, of course, vary, depending on thedecision maker’s preference. This approachsupports the insight that optimization can beused as a means to help the decision makers tomake their decisions, especially for the futureemission trading.

With the extension of the model boundary, theoptimization model was used to investigate theopportunities of meeting the emission allowancewith a lower cost for the studied steel plant viacarbon-trading schemes, in this case EU ETS andCDM. The results show that compared with EUETS, a lower CO2 reduction cost could beachieved by use of CERs generated from CDMprojects. The internal changes within the plant willalso play an important role to help the studiedsteel plant to meet the emission-trading allowanceand the further emission reduction comments,indicating the importance of the internal changesfor the steel plant independent of carbon-tradingschemes. Therefore, internal abatement should beencouraged as they can further improve theefficiency and promote the discovery of newtechnologies for creating a more sustainableenergy supply both from an economic and anenvironmental point of view. It should be pointedout that the carbon prices from different tradingschemes have been fluctuating. A sensitivityanalysis would show the influence of carbonprices on potential CO2 emission reductionoptions. However, the analysis shows that byusing this kind of analysis it is possible to

evaluate different measures for CO2 reductionand their effects on the whole operation system.It should also be pointed out that this studyis based on a Swedish steel plant as a calculationexample. However, the model developedcan with little modification be used in anysimilar steel plant within the EU countriesand beyond.

As a specific integrated steel plant, it will be ofgreat interest to investigate some energy-savingpotentials within SSAB Tunnplat AB. Consideringthe fact that these two sites are located in twodifferent geographical locations, it is impossiblefor the RM to directly utilize process gasesgenerated from the steel plant as an energycarrier. However, it would be possible if someprocess gases, e.g. coke oven gas, could beliquefied or transformed to other kinds of fuel.At the moment, at the steel plant there is excesscoke oven gas for potential energy use. Two recentreports have studied the possibilities of coke ovengas liquification and methanol production fromcoke oven gas [15,16]. As a fuel that could betransported by using the current existing traffictools between two sites, the possibility ofsubstituting parts of fuels used at the RM, i.e. oiland LPG, will increase, which is worthinvestigating in the future.

7. CONCLUSIONS

A model on CO2 emission reduction in integratedsteelmaking is described in this paper. A fewapplication cases have also been presented. Themain conclusions drawn in this paper are asfollows:

� A PI method has been used to analyze CO2

emission for the steel industry with considera-tion of the material and the energy system. Thismodel has a friendly interface, easy to bemanipulated by non-programming persons andto make the analysis.

� The optimization model has the generality andflexibility to be extended to cover more pro-cesses, and it can be used to analyze CO2

emission for a small, large or global integrated

C. WANG ET AL.1104


DOI: 10.1002/er

steelmaking system depending on the researchinterest.

� The optimization of the BF process and thecombined BF1BOF system resulted in differentstrategies to minimize CO2 and energy use. Theinteraction between the processes can show acomplex behaviour with several counteractingmechanisms that can be considerably influencedwhen constraints are introduced to the system.This demonstrates the benefits that can begained by using a system-oriented analysisapproach, and thus possibly avoid sub-optimi-zation of the individual processes.

� The aids of the Pareto fronts analysis provide acomprehensive view of the trade-offs betweenthe objectives of the Cost and the CO2 emission,which can provide useful information fordecision makers to generate strategies, forinstance, their stance in the future emissiontrading.

� The case study of ETS’ influence on CO2

emission shows that internal changes andCDM scenario will both contribute to help thesteel plant to meet the emission-trading allow-ance and future emission reduction commit-ments. The model developed can serve as abenchmark for the future emission-tradingsimulations purpose by steel plants withinEuropean countries and beyond.

� Some benefits regarding CO2 emission reduc-tion by integrating a two-site model could beachieved for the case of the specific integratedsteel plant.

NOMENCLATURE

BF 5 blast furnaceBOF 5 basic oxygen furnaceCC 5 continuous castingCDM 5 clean development mechanismCER 5 certified emission reductionCHP 5 combined heat and power

plantCOP 5 coke oven plantDH 5 district heat

EAF 5 electric arc furnaceEl. 5 electricityEU ETS 5European Union emission-

trading schemeHM 5 hot metalIISI 5 International Iron and Steel

InstituteKP 5Kyoto ProtocolLCA 5 life cycle assessmentMILP 5mixed integer linear program-

mingPI 5 process integrationRM 5 rolling mill

ACKNOWLEDGEMENTS

We are grateful to the Centre for Process Integration inSteelmaking (PRISMA). The Centre for PRISMA is anInstitute Excellence Centre (IEC) supported by theSwedish Agency for Innovation Systems, the KnowledgeFoundation, the Foundation for Strategic Research andby the industrial participants Luossavaara-KiirunavaaraAB, SSAB Tunnplat AB and Rautaruukki Oyj.

REFERENCES

1. IISI (International Iron and Steel Institute). 2005. Editionworld steel in figures. Available from: www.worldsteel.org(accessed on November 30, 2006).

2. IISI (International Iron and Steel Institute). 2006. WorldSteel in Figures 2006. Available from: www.worldsteel.org(accessed on December 1, 2006).

3. Birat JP. The challenge of global warming to the steelindustry—a European viewpoint. POSCO Conference,Pohang, Korea, September 2002.

4. Gielen D, Moriguchi Y. CO2 in the iron and steel industry:an analysis of Japanese emission reduction potentials.Energy Policy 2002; 30:349–363.

5. Hidalgo I, Szabo L, Ciscar C, Soria A. Technologicalprospects and CO2 emission trading analyses in the ironand steel industry: a global model. Energy 2005;30:583–610.

6. Cai W, Wang C, Chen J, Wang K, Zhang Y, Lu X.Comparison of CO2 emission scenarios and mitigationopportunities in China’s five sectors in 2020. Energy Policy2008; 36:1181–1194.

7. Dutta G, Fourer R. A survey of mathematical program-ming applications in integrated steel plants. Manufacturing& Service Operations Management 2001; 3:387–400.

8. Deo B, Kalyanmoy K, Jha S, Sudhakar V, Sridhar NV.Optimal operating conditions for the primary end of an



DOI: 10.1002/er

integrated steel plant: genetic adaptive search and classicaltechniques. ISIJ International 1998; 38 (1):98–105.

9. Andersson J. A survey of multi-objective optimisation inengineering design, Technical Report LiTH-IKP-R-1097,Linkoping Institute of Technology, Linkoping, Sweden,2000.

10. Kim IY, de Weck OL. Adaptive weighted-sum methodfor bi-objective optimization: Pareto front generation.Structural and Multidisciplinary Optimization 2005; 29:149–158.

11. Wang C, Ryman C, Dahl J. Potential CO2 emissionreduction for BF–BOF steelmaking based on optimizeduse of ferrous burden materials. International Journalof Greenhouse Gas Control 2008, doi: 10.1016/j.iggc.2008.06.005.

12. Wang C, Larsson M, Yan J, Dahl J. Impacts on CO2

emission allowance of EU emission trading scheme(ETS) in a Swedish steel plant by clean development

mechanism (CDM). International Green Energy ConferenceI, Waterloo, Ont., Canada, June 12–16, 2005.

13. Larsson M, Dahl J. Reduction of the specific energyuse in an integrated steel plant—the effect of anoptimization model. ISIJ International 2003; 40(10):1664–1673.

14. Ryman C, Larsson M, Niemi T, Grip CE. Modelling of theblast furnace process with a view to optimize the steel plantenergy system. Second International Conference on NewDevelopments in Metallurgical Process Technology, Riva delGarda, Italy, September 2004.

15. Nordgren S. Potential for alternativ anvandning avprocessgaser vid SSAB i Lulea. Master Thesis Report,Lulea University of Technology, 2006:205 CIV. ISSN:1402–1617.

16. Asp B. Biomass and coke oven gas based methanolproduction. Master Thesis Report, Lulea University ofTechnology, 2007: 189 CIV. ISSN: 1402–1617.

C. WANG ET AL.1106


DOI: 10.1002/er

a model on co2 emission reduction in integrated steelmaking by optimization methods

Documents