Development of high-temperature corrosion risk monitoring system in pulverized coal boilers based on reducing conditions identification and CFD simulations

Applied Energy xxx (2017) xxx–xxx

Contents lists available at ScienceDirect

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

Development of high-temperature corrosion risk monitoring system in pulverized coal boilers based on reducing conditions identification and CFD simulations Norbert Modlinski ⇑, Tomasz Hardy Department of Boilers, Combustion and Energy Processes, Faculty of Mechanical and Power Engineering, Wroclaw University of Technology, 27 Wybrzeze Wyspianskiego St, 50-370 Wroclaw, Poland

h i g h l i g h t s
Corrosion risk monitoring system was demonstrated based on CO and O2 measurements.
Four global gas phase kinetic mechanisms were implemented into CFD.
Global mechanisms were compared with detailed one in a perfectly stirred reactor.
CFD predictions were compared with measurements received from the monitoring system.

a r t i c l e

i n f o

Article history: Received 16 January 2017 Received in revised form 19 April 2017 Accepted 27 April 2017 Available online xxxx Keywords: Corrosion monitoring Boilers Combustion CFD Pulverized coal

a b s t r a c t Low-emission combustion (for example the use of low-NOx burners and air staging) contributes to formation of a reducing atmosphere in the furnace, that is accompanied by oxygen depletion and excess of CO in the vicinity (boundary layer) of waterwalls. Corrosion of boiler tubes is often caused by reducing atmosphere. O2 and CO measurement in the boundary layer of evaporators can be a good indicator of corrosion risk assessment. System based on the on-line measurement of the O2 and CO concentration in the boundary layer of the industrial scale boiler walls was described. To improve the functionality of the monitoring system Computational Fluid Dynamics may appear helpful. A validated CFD model capable of properly predicting the CO and O2 concentration in the vicinity of the combustion chamber walls may help to adjust the monitoring system during variable boiler operating conditions or different fuel properties without the necessity to repeat the measurements for new conditions. The scientific part of the current research is concentrated on volatiles combustion simulation with the emphasis on CO burnout. Four popular global mechanisms have been implemented into CFD code and their CO and O2 predictive capabilities are demonstrated. Additionally global mechanisms have been compared to detailed one in Perfectly Stirred Reactor model. It appears that the choice of global mechanism has significant influence on CO and O2 prediction. The measurements of the CO and O2 in the waterwalls boundary layer have been extracted from the monitoring system and compared to simulation results. One of the tested mechanisms demonstrated acceptable qualitative agreement with the measurement in terms of O2 predictions. The quantitative accuracy of CFD-based oxygen prediction in the boundary layer was described as moderate. CFD-based CO prediction was less satisfactory. Ó 2017 Published by Elsevier Ltd.

1. Introduction Deployment of renewables, need to improve coal plant efficiency and increase in natural gas utilization tends to decrease coal consumption in OECD (Organization for Economic Co-operation and Development) member countries. However according to International Energy Agency [1] coal will long remain a key energy ⇑ Corresponding author. E-mail address: [email protected] (N. Modlinski).

fuel for electricity generation in a number of developed countries. Utility industry face the challenge of adopting their boilers to new requirements and conditions. One of the most crucial matters of the conventional coal energy sector are unprecedented environmental requirements for pollutant emissions. More restrictive environmental regulations are being enforced and power generation companies are forced to maintain the agreement between efficient and clean combustion. The implementation of the low-emission combustion technologies for NOx emission reduction often results in the intensification

http://dx.doi.org/10.1016/j.apenergy.2017.04.084 0306-2619/Ó 2017 Published by Elsevier Ltd.

Please cite this article in press as: Modlinski N, Hardy T. Development of high-temperature corrosion risk monitoring system in pulverized coal boilers based on reducing conditions identification and CFD simulations. Appl Energy (2017), http://dx.doi.org/10.1016/j.apenergy.2017.04.084

2

N. Modlinski, T. Hardy / Applied Energy xxx (2017) xxx–xxx

of high-temperature corrosion of the boiler heating surfaces. The corrosion of boiler tubes increases the operational costs and has strong negative effect on the reliability and availability of the whole power unit [2]. Low-emission combustion (for example the use of low-NOx burners and air staging) contributes to formation of a reducing atmosphere in the furnace, that is accompanied by oxygen depletion and excess of CO in the vicinity of waterwalls. The process of high temperature corrosion is affected by many factors such as the fuel chemical composition (chlorine and alkali content), composition of the flue gas (O2, CO, H2S, HCl), deposition process, chemical composition of ash deposits formed on the surfaces, flue gas and tube surface temperature [3]. Most of the corrosion monitoring systems are based on off-line analysis. Ultrasonic tube thickness, electrical resistance and coupon weight measurement are usually carried out periodically and are labor intensive. Plant operator obtains historical corrosion data that has already occurred and may undertake only corrective actions. To prevent corrosion a monitoring system operating in on-line (real-time) manner is needed with the measurement data updated every few minutes. Ideally corrosion measurements should be integrated within the plant Distributed Control System. However on-line high-temperature corrosion monitoring is still a developing technology. Market review showed that there is no on-line monitoring solution widely applied in the industry. The real-time corrosion monitoring system adapt either electrochemical approach or identify the reducing atmospheres were corrosion process usually occurs. The first one is often based on electrochemical noise (EN) measurements. The principal of operation of the instrument is associated with spontaneous fluctuation of the current and potential of a corroding electrode. This phenomena is commonly described as electrochemical noise. The fluctuations are converted to a digital signals and supplied to a computerized data acquisition unit. The EN based corrosion monitoring technique has been traditionally exploited in industry at low temperature. In the work of [4] researchers demonstrate attempt to continuously monitor both fouling and corrosion of a convective heat exchanger tube. The paper shows situations where EN variations correlate with changes in corrosion and fouling. This technology was applied recently in high temperature radiant sections of large scale utility boilers. In [5] the EN probe was installed in a 125 MW boiler. The probe operated in the temperature range 300–1000 °C. The corrosion rate results showed good agreement with previously measured rates using long-term single-coupon probes. Electrochemical corrosion sensors ability to respond to changing combustion conditions were conducted in a pilot-scale combustion test facility [6]. Large scale tests were arranged in the 600 MW, supercritical boiler. Two locations have been selected for probe installation. A good correlation was observed between boiler load and indicated corrosion rate. In general, as load increased, corrosion rate also increased. It was also discussed that testing of an electrochemical corrosion monitoring system difficult to identify with certainty because operating conditions other than firing rate also vary with load. Quantitative accuracy of on-line techniques by means of precision metrology. Satisfying results were presented. A multiple EN sensors were applied in a 1300 MW coal-fired boiler [7]. Six electrochemical sensors were prepared for installation on two walls. The probe locations were chosen on the basis of prior knowledge of plant corrosion behavior. The research also demonstrated good correlation between boiler load variation and corrosion rate. It was concluded that the local corrosion behavior also can be influenced by ash and slag deposition. The presence of molten slag might have facilitated corrosion by dissolution of iron into the slag melt. More recently the realtime electrochemical measurements were used to evaluate the impact of oxy-fuel combustion conditions on corrosion rate of three superheater materials and one boiler waterwall material

[8]. Firing tests were conducted in a 1.5 MW furnace while firing bituminous, high-sulfur bituminous and subituminous coals. Waterwall corrosion rates decreased consistently when converting from air- to oxy-firing while superheater corrosion rates generally increased, although they were less than twice the air-fired rate under most conditions. The electrochemical noise based monitoring system was validated in numerous research studies in pilot as well as large scale plants. It has proven both qualitative and quantitative reliability. Despite encouraging results this technology has not become a widely-applied solution for boiler waterwalls high-temperature monitoring system. Because EN requires monitoring of very small signal fluctuations, this approach to corrosion monitoring is also affected by extraneous sources of signal noise in the plant. One of its limitations is related to fact that electrochemical sensors provide only local information. The measurement are usually conducted in few furnace locations. The probes are located in sites determined from boiler corrosion experience. Change in boiler operating conditions or firing system modification might result in original probe location choice being inadequate. The real-time corrosion monitoring system demonstrated in the current paper is based on the identification of reducing conditions zones. There is very limited literature [9] on this type of corrosion prevention technology applied in large scale furnaces. It is manifested that the more reducing the conditions, the higher is the corrosion rate. One of the dominant paths of corrosion is the attack of reduced sulfur species (e.g. H2S) which are formed under reducing conditions (O2 < 0.5% and CO > 2%), especially for high chlorine coals [10]. The system detects atmospheres with low oxygen content and high levels of carbon monoxide. Knowing these components the corrosion hazard to the boiler wall can be identified based on assumed criteria and appropriate measures can be taken. The advantage of the system is that it generates two-dimensional maps covering most of the boiler wall and not just selected locations. Additionally it is less complicated in terms of instruments used and cost effective. It needs to be emphasized that reducing atmosphere monitoring system describes corrosion hazard while electrochemical based system monitors corrosion rate. The practical part of the research resulted in the implementation of a corrosion risk monitoring system in industrial scale unit. The system was developed and tested in 160 MW tangentially fired boiler operating in Combined Heat and Power (CHP) plant in Wroclaw (Poland). Detailed description of the monitoring system and operational experience of its application to 380 MW tangentially fired boiler was presented in previous work [9]. Some researchers demonstrate development of CFD and field instrumentation as a complementary approach to corrosion management. In [4–7] expressions defining corrosion rate as a function of tube temperature, steel composition, and H2S concentration were developed. Correlations were based on the published data and the experiments at a pilot-scale test furnace. These functions were implemented into CFD code and applied in utility boiler simulations. Model predictions show good overall agreement with furnace observations and field measurements (ultrasonic tube wall thickness analysis). It was suggested that this tool can be applied to a range of firing configurations, firing rates and fuel types. A validated CFD model capable of properly predicting the CO and O2 concentration in the vicinity of the combustion chamber walls may help to adjust the demonstrated monitoring system based on the identification of reducing atmospheres. The characteristic feature of presented system is the requirement of it to learn after introducing major changes affecting boiler operating conditions [9]. Reliable CFD model might substitute the expensive learning process. Simulation of pulverized coal flow with combustion brings the necessity to employ three major sub-models: char burnout,

Please cite this article in press as: Modlinski N, Hardy T. Development of high-temperature corrosion risk monitoring system in pulverized coal boilers based on reducing conditions identification and CFD simulations. Appl Energy (2017), http://dx.doi.org/10.1016/j.apenergy.2017.04.084

3

N. Modlinski, T. Hardy / Applied Energy xxx (2017) xxx–xxx

implement four global gas phase kinetic mechanisms, (3) compare four global mechanisms with the detailed mechanism [13] in a Perfectly Stirred Reactor computations, (4) correlate CO and O2 CFD predictions with measurements received from the monitoring system. 2. Corrosion risk monitoring system Boiler operator needs to maintain compromise between efficient and clean combustion. The occurrence of reducing zones with high corrosion hazard in the boundary layer of waterwalls to large extent depends on fuel properties and combustion process organization associated with low-NOx air staging and boiler load. At the same time optimized air/fuel distribution is crucial for combustion quality and efficiency as it effects CO and NOx emission and unburned carbon in ash level. Information on the type of atmosphere in the vicinity of furnace walls provided continuously allows the boiler operator to react instantaneously. A decision can be made on altering specific mill activity, secondary air distribution or equivalence ratio in burner belt. The monitoring system measures oxygen continuously with the zirconia analyzer. It converts gas into electric signal directly. No such analyzers exist for CO which makes its measurements more challenging. For this reason the individual O2 concentration measurement seems to be more robust and easier to perform on-line. Additionally CO concentrations are unstable in the boiler environment on the contrary to oxygen. Gas measurements in different locations of the considered 160 MW industrial pulverized coal boiler indicate direct correlation between O2, CO and H2S (Fig. 1). Oxygen depletion promotes CO and H2S production. Tube thickness was measured after specific operation period and compared with oxygen concentrations in those location. It can be noticed that spots of high corrosion rates correspond with low O2 levels. The general conclusion was made that on-line corrosion hazard monitoring can be estimated from O2 measurement only. However for better accuracy the CO measurement was recently included in the system. In recent years a stationary corrosion hazard monitoring system was installed and tested in 160 MW tangentially fired boiler operating in Combined Heat and Power (CHP) plant in Wroclaw (Poland). To achieve a good compromise between complexity and accuracy it was decided to use five zirconium probes located in preselected places on the boiler’s wall (Fig. 2) from the 26 total available measurement ports. The five selected probes operate continuously. Locations of selected probes was chosen in points that guarantee variable oxygen concentrations for different boiler operating conditions. Analysis of previous series of measurements using all available ports preceded the choice of probe locations. This solution allows to reduce the amount of instruments of the system. The measurement results are visualized on the computer screen. The system comprised a temperature-controlled measuring 10

0,8

8

0,6

0,4

0,2

0 0

2

4

6

8

Concentraon of O2, %

10

12

14

500

400 CO H2S

6

300

4

200

2

100

0

Concentraon of H2S, ppm

1

Concentraon of CO, %

Corrosion rate, mm/year

devolatilization and gas phase combustion. The scientific part of the current research is concerned with the effect of volatiles combustion simulation. The emphasis was put on CO burnout simulations since CFD predictions of this product are usually encumbered with significant error. In most cases the gas phase kinetics are represented by global mechanisms. Four popular global mechanisms have been implemented into CFD code and their CO and O2 predictive capabilities are demonstrated. Additionally global mechanisms have been compared to detailed one [11] in a Perfectly Stirred Reactor (PSR) model. It appears that the choice of global mechanism has significant influence on CO and O2 prediction. CFD modelling of pulverized coal combustion has been extensively applied [12–14] in utility boilers simulations. Majority of the CFD validation studies utilize measurements provided by the Distributed Control System. However, the useful information is limited to average species concentrations at the outlet. Xu et al. [15] predicted flow with combustion in a front fired boiler. Measured (estimated) data of furnace outlet temperature, unburned carbon, O2, CO2, CO and NOx concentrations at the furnace outlet as well as heat absorbed by walls and platen super heaters have been used for model validation. Yin et al. [16] investigated a furnace and part of the rear pass in the tangentially fired boiler. The simulation have been validated with global design parameters including O2 at the furnace outlet, heat transfer in the furnace and furnace exit temperature. Site operation data was used to verify NOx predictions. Pallares et al. [17] simulated a front fired boiler. The work concentrated on char burnout predictions. A limited furnace modelling validation included only O2 plant measurements. Karampinis et al. [18] have evaluated the effect of cocombustion of cardoon with lignite in a 300 MWe boiler. Validation of the simulations was performed using plant data for the reference case of pure lignite combustion (furnace outlet temperature, O2 and NOx concentrations). Choi et al. [19] have used various measured and design values to validate tangentially-fired furnace. Local temperatures at different furnace locations, total heat flux to the furnace walls, O2, CO2, and NOx concentrations at the boiler exit have been compared with computed values. The novelty demonstrated in the current paper is related to fact that two-dimensional maps of CO and O2 measured by the corrosion monitoring system in the vicinity of the boiler wall are compared to CFD simulation results and not just the average values. To the authors knowledge such comparison has not been previously described in the literature. The demonstrated research can be divided into two major parts. The first practical part describes the corrosion risk monitoring system installed on industrial scale boiler. The second scientific part is devoted to CFD simulations of the boiler. The main objectives of this paper are: (1) to demonstrate the high-temperature corrosion risk monitoring system based on CO and O2 measurements, (2) to formulate a general tangentially fired boiler CFD model, (2)

0 0

2

4

6

8

10

12

14

16

18

20

Concentraon of O 2, %

Fig. 1. Correlation between O2 concentration and corrosion rate. Dependence of CO and H2S concentration upon O2 level in the waterwall boundary layer of the considered 160 MW boiler.

Please cite this article in press as: Modlinski N, Hardy T. Development of high-temperature corrosion risk monitoring system in pulverized coal boilers based on reducing conditions identification and CFD simulations. Appl Energy (2017), http://dx.doi.org/10.1016/j.apenergy.2017.04.084

4

N. Modlinski, T. Hardy / Applied Energy xxx (2017) xxx–xxx

Fig. 2. Corrosion risk monitoring system with arrangement of 26 measuring ports on the front wall and locations of the 5 probes in tangentially fired 160 MW boiler.

probe, signal converter and data acquisition modules, a gas filter, cooling air supply and a computer for data processing. The instrumentation data acquisition modules were enclosed in a rugged dust-free metal cabinet. An example of O2 and CO continuous measurements is demonstrated in Fig. 3. One of the significant feature are large fluctuations of instantaneous CO measurements and moderate fluctuations of O2. CO fluctuations increase in reducing conditions and the O2 fluctuations stay low. In oxidizing conditions the O2 fluctuations become stronger while CO becomes more stable. To maintain the operational reliability of the measuring system approximately 10–20 min operating cycle is typically used, with the oxygen measurement time not exceeding 5 min. Finally the measuring signal may be delivered to the plant operator control room to visualize actual results of measurements and indicate areas with high corrosion hazard. This solution was applied and tested in two other boilers in Opole and Belchatow Power Plants (Poland) [9]. The boiler operator may decide about altering operating conditions while maintaining low NOx emission. The monitoring system may appear helpful when deciding about protective air installations, anti-corrosion coating investments or combustion process optimization activities. In order to obtain information about the distribution of oxygen in the boundary layer over the whole wall (and not just at 5 selected points in Fig. 2) neural network technique can be used [9]. However for the system to be fully operative learning procedure needs to be conducted in the first instance. For this purpose,

it is necessary to perform measurements of oxygen concentration and carbon monoxide in the boundary layer for many different boiler operating conditions (load, mills activity and burners configurations, air distribution etc.). The visualization program use the obtained data to interpret the results and create maps of actual corrosion hazards. After collecting enough data of flue gas composition and the most important operating parameters for many configurations the so-called genetic optimizer [9] is capable of optimizing the firing system set-up. The system is dedicated to limit NOx emission by considering Loss on Ignition, CO emission and fireside corrosion risk. Unfortunately the genetic optimizer used in developed on-line monitoring system is not a fuel depended model, since it works on the basis of gas sensors’ indications. Significant changes in fuel characteristics, operating conditions or modifications in the design of the furnace requires to repeat the learning process by artificial neural networks used in optimizer. In this case the measurements need to be carried out again by using all available ports which is the biggest weakness of the system. This makes the neural network learning an expensive and laborintensive process. To improve the functionality of the monitoring system Computational Fluid Dynamics may appear helpful. Instead of doing measurements for different modes of boiler operation the information data could be obtained from simulation results instead. CFD model needs to be capable of properly predicting the CO and O2 concentration in the boundary layer of waterwalls. 3. Numerical simulations 3.1. Case study boilers and operating conditions

Fig. 3. Sample of continuous O2 and CO measurements in selected measuring port.

The simulations were performed for a drum type radiant tangentially fired unit installed at the EDF CHP plant in Wroclaw (Poland) utilizing bituminous coal. Thermal power at nominal load is 160 MW. Steam capacity of the boiler is 230 t/h. Main steam temperature is 540 °C. The boiler is approximately 28 m high, 8.5-m wide, 7.5 m deep. Boiler is a dual pass type with twentyfour burners arranged in an array of six burners disposed in four different locations. Boiler is also supplied with four dual air zone over-fired air ports for reducing NOx formation. The region of study comprises only the furnace: ash pit, burners belt, section of radiant superheater. Radiant superheaters geometry has been included for the better development of gas outflow.

Please cite this article in press as: Modlinski N, Hardy T. Development of high-temperature corrosion risk monitoring system in pulverized coal boilers based on reducing conditions identification and CFD simulations. Appl Energy (2017), http://dx.doi.org/10.1016/j.apenergy.2017.04.084

5

N. Modlinski, T. Hardy / Applied Energy xxx (2017) xxx–xxx

The boiler model is demonstrated in Fig. 4. The circular OFA inlets have been replaced by rectangular ones to ease the meshing process. The platen super heaters have been modelled as zerothickness horizontal planes, since it is practically impossible to model actual tubes. A mesh independence analysis was performed for the initial mesh consisting of 3.5 million control volumes. Reduction of cell number to about 2.8 million did not substantially changed the predicted temperature field and this number of cells was selected to obtain compromise between solution accuracy and computational time. 3.2. Mathematical model. Gas phase kinetics The mathematical model used here is based on commercial CFD code FLUENT [20]. Simulation of the following processes takes place in the furnace: turbulent flow, coal combustion, gas phase combustion, particle transport and radiative transport. The gas phase is modelled assuming an Eulerian approach, whilst for the solid phase, the Lagrangian approach is more suitable. The details about modelling approach can be find in [21]. The pulverized coal combustion process can be divided into two parts: devolatilization and char combustion. The devolatilization consists of primary and secondary step. As the coal particle is heated in the absence of oxygen, tar and light gases are initially driven out of the particle and a solid residue remains. These light gases are generally oxides (CO2, CO, H2O) and light hydrocarbons (C1–C4) [22]. Tar is generally defined as those volatiles that condense to a solid or liquid at room temperature. All these species have different devolatilization rate constants. A single-rate kinetic devolatilization model [23] was used to predict the volatiles yield rate, which assumes that the rate of devolatilization depends on the amount of volatiles remaining in the particle via a first-order reaction. This model was extended by using FG-DVC network devolatilization model [24,25] as a pre-processor. The particle heating rate was assumed to be 105 K/s. The model predicts the rate of production and high temperature yields of char, tar, volatiles and the composition of key species during devolatilization of any coal. The results as well as the proximate and ultimate analysis for the used coal are given in Table 1. In the CFD modelling of turbulent flow with combustion it was assumed that volatiles are produced as a single compound that instantaneously breaks up into reaction into tar, light hydrocarbons, CO, CO2 and H2O.

FG-DVC calculates devolatilization rate of tars and other species. The most significant mass drop of fuel particle occurs when tar is evolved. Tar is produced as a first species. For this reason tar release rate is used in devolatilization model. Knowing the volatile fraction of dry ash free (daf) coal (fvolatile) and assuming that residual char is pure carbon with known lower heating value LHVchar, we can calculate lower heating value of volatiles (LHVvolatile).

LHV v olatile ¼

LHV coaldaf  f char  LHV char f v olatile

ð1Þ

Assuming that lower heating value of light hydrocarbons is approximately equal to that of methane (LHVgas = 50 MJ/kg) we can easily calculate lower heating value of tars from the instantaneous break-up reaction of volatiles:

LHV tar ¼

LHV v olatile  ðygas  LHV gas þ yCO  LHV CO Þ ytar

ð2Þ

where ygas, ytar, yCO stand for mass fraction in volatiles and LHVCO is the known lower heating value of CO. Secondary devolatilization is the further break-down and reorganization of pyrolysis tars. Some researchers incorporate that phenomena in the CFD simulations of pulverized coal combustion by assuming that tar decomposes to soot, light hydrocarbons, hydrogen, carbon monoxide [26]. An assumption can be made that rate of secondary tar decomposition is equal to the rate of primary devolatilization. The produced light hydrocarbons (CmHn) are considered to have the same composition as those produced with primary volatiles. Soot is assumed to be pure carbon. The pathway of devolatilization model is demonstrated in Fig. 5. The current research is focused on the gas phase modelling approach. In non-premixed reacting turbulent flows the local time-dependent mixing, chemical reaction of the species and heat transfer away from the reaction determine the combustion process. The key gas phase combustion modelling issue is the necessity to calculate source terms in reactive species transport equations, which are the average values of strongly non-linear reaction rates. The most simple approach is to assume infinitely fast chemistry which originates from the observation that most species produced during combustion rapidly reach chemical equilibrium at high temperatures. Conclusion is made that the mixing process of

Radiant superheaters

28 m

7.5 m

Front wall

OFA ports

Burner belt

Ash pit

Fig. 4. CFD model geometry of a tangentially fired 160 MW boiler.

Please cite this article in press as: Modlinski N, Hardy T. Development of high-temperature corrosion risk monitoring system in pulverized coal boilers based on reducing conditions identification and CFD simulations. Appl Energy (2017), http://dx.doi.org/10.1016/j.apenergy.2017.04.084

6

N. Modlinski, T. Hardy / Applied Energy xxx (2017) xxx–xxx

turbulence model. Mass fraction occupied by fine structures is modeled as:

Table 1 Coal analysis and FG-DVC output. Ash

Volatile Matter

Moisture

Proximate Analysis (wt%, as received) 15.1 34.7 C

H

Ultimate Analysis (wt%, daf) 84.1 4.03

c ¼ 2:13  

9

41.2

N

S

O

0.84

0.58

10.49

Volatiles

Char

FG-DVC High Temperature Yield (wt%, daf) 42.5

57.5

H2O

CO

CO2

Volatile Composition from FG-DVC (wt%) 4.23 4.8 2.85 CmHn (gas)

"

Fixed Carbon

CmHn

CxHyOz

6.12

24.5 CxHyOz (tar)

Empirical formula for light hydrocarbons (CmHn) and tar (CxHyOz) m = 1, n = 6.28 x = 7, y = 4.26, z = 0.6

large eddies with the time scale k/e controls the rate at which chemical reactions proceed [27]. No kinetic information is needed. In case of CO prediction the mixed is burned assumption might be oversimplified. CO does oxidize rapidly at high temperatures with oxygen supply, but does not oxidize so well at the cooler temperatures or less intensive mixing conditions. Such conditions are common in some low-NOx control technologies. Additionally at high temperatures the dissociation reactions are promoted. For such systems it is necessary to use more detailed and complex approaches, which usually require the adoption of a finite rate chemistry. In this study the Eddy Dissipation Concept (EDC) [28] is used to model the influence of turbulence on chemical reactions. With the EDC it is possible to take into account detailed kinetics of reactions. In this approach the total space is subdivided into fine structures and the surrounding fluid. Small scale structures can be pictured as a part of the control volume, where Kolmogorov sized eddies containing combustion species are located so close together, that mixing on the molecular level is taking place [28]. All reactions of the reactive components are assumed to react only in these spaces which are locally treated as a Perfectly Stirred Reactors (PSR) with a residence time:

s ¼ 0:41 

rffiffiffi

m e

ð3Þ

where m is the kinematic viscosity, e denotes turbulent kinetic energy dissipation rate. These parameters are calculated from

me k

2

ð4Þ

The reaction rates of all species are calculated on a mass balance for the fine structure reactor. Denoting fine structures quantities with asterisk, the conservation equation of species i can be defined:

s

q

 ð1

 c Þ

ðmi  mi Þ ¼ M i  -i

ð5Þ

where mi is the average mass fraction of the species i, Mi is the molecular weight of the species i, x⁄i denotes the chemical reaction rate calculated from Arrhenius equation. The mean net mass transfer rate of species i between the fine structures and the surrounding fluid can be expressed as:

Ri ¼

Fig. 5. Devolatilization modelling approach.

 0:25 #2

s

qc

 ð1

 c Þ

ðmi  mi Þ

ð6Þ

The EDC model is implemented into CFD code by solving the non-linear system of equations for the fine structure reactor in each control volume and finding Ri, which is the source term in species i transport equation. Modeling turbulent reactive flow requires extensive computational resources. The cost rises with the number of chemical species involved in kinetic mechanism. The detailed chemistry is adopted for combustion systems only in case of simple geometries and fuels that incorporate small number of species. Solid fuels devolatilization products are composed of hydrocarbons for which a comprehensive mechanism is not available. For this reason simplified kinetic mechanisms were developed. The global reaction mechanisms are the most convenient and hence considered in this paper. In the current paper four often used in pulverized coal combustion simulation global mechanisms have been investigated (Table 2). The first two mechanisms [29] (A and B) account for primary pyrolysis and do not consider tar combustion. In these cases the composition of the volatiles provided by Table 1 has been modified due to lumping of tar (CxHyO) and light gases (CmHn) into single hydrocarbon species. Mechanism A is mainly based on the work of [30] and is often employed for combustion modelling. It is capable to describe fuel-rich as well as fuel lean conditions with higher accuracy than more simple models. With the original model of Jones and Lindstedt [30] the convergence is quite poor mainly due to inclusion of reaction orders lower than unity. Hydrogen oxidation is considered as a reversible reaction. To achieve correct equilibrium properties the reaction order of reversible reaction becomes lower than zero ([H2O]1). In this paper a very stable implementation of the mechanism from [30] in conjunction with the EDC has been tested. The only reversible reaction is the water-gas shift reaction. Mechanism B [29] works with the assumption of a water-gas shift equilibrium. A linearly independent equilibrium relation is introduced. The assumption is used to calculate parameter a and equilibrium composition in the fuel rich case. For equivalence ratio higher than one the amounts of H2 and H2O concentration are frozen at the values given by this equilibrium calculation. The only active reaction is the one describing the oxidation of CO. Mechanism C is mainly based on the one described in [31]. It is similar to mechanism B but one additionally tar combustion reaction is included [32]. As previously, parameter a is computed from water-gas shift equilibrium. The only mechanism that considers secondary pyrolysis is the 9-step Mechanism D [33]. After being released, the volatile matter

Please cite this article in press as: Modlinski N, Hardy T. Development of high-temperature corrosion risk monitoring system in pulverized coal boilers based on reducing conditions identification and CFD simulations. Appl Energy (2017), http://dx.doi.org/10.1016/j.apenergy.2017.04.084

7

N. Modlinski, T. Hardy / Applied Energy xxx (2017) xxx–xxx Table 2 Investigated global reaction mechanisms. Reaction

Rate equation

A (s-m-kmol)

Tb

Ea (J/kmol)

Mechanism A Volatiles ? mCmHnCmHn + vCO2CO2 + vCOCO + vH2OH2O CmHn + m/2O2 ? mCO + n/2H2 CmHn + mH2O ? mCO + (n/2 + m)H2 CO + H2O = CO2 + H2 H2 + 0.5O2 ? H2O

k1[volatiles] k2[CmHn]0.5[O2]1.25 k2[CmHn][H2O] k4[CO][O2]0.5[H2O]0.5 k5[H2][O2]0.5

1.00e13 0.44e12 0.30e09 0.275e10 0.10e07

0 0 0 0 0

0 1.2552e08 1.2552e08 8.368e07 8.368e06

Mechanism B Volatiles ? mCmHnCmHn + vCO2CO2 + vCOCO + vH2OH2O CmHn + (m/2 + n/4a)O2 ? mCO + (1-a)n/2H2 + n/2aH2O CO + 0.5O2 ? CO2 H2 + 0.5O2 ? H2O

k1[volatiles] k2[CmHn][O2] k3[CO][O2]0.5[H2O]0.5 k4[H2][O2]0.5

1.00e13 7.28 + 09 5.42e + 09 1.00e07

0 0.5 0 0

0 1.6736e08 1.2552e08 8.368e06

Mechanism C Volatiles ? mCxHyOzCxHyOz + mCmHnCmHn + vCO2CO2 + vCOCO + vH2OH2O CxHyOz + (x/2 – z/2)O2 ? xCO + y/2H2 CmHn + (m/2 + n/4a)O2 ? mCO + (1-a)n/2H2 + n/2aH2O CO + 0.5O2 ? CO2

k1[volatiles] k2[CxHyOz][O2] k2[CmHn][O2] k3[CO][O2]0.5[H2O]0.5

1.00e13 3.80e07 2.33e11 1.30e11

0 0 0.5 0

0 5.5463e07 1.6737e08 1.252e08

k1[volatiles] k2[CxHyOz] k3[CxHyOz] k4[CxHyOz][O2] k5[CmHn][O2] k6[CO][O2]0.5[H2O]0.5 k7[H2][O2]0.5 k8[CmHn][H2O] 2 K1exp ðEa1 RT ÞpO2 2 As  ½C 1þK2exp ðEa2 RT ÞpO2

1.00e13 0 5.02e08 0 5.42e04 0 3.80e07 0 2.33e11 0.5 5.42e09 0 1.00e07 0 4.40e11 0 2 K1 = 3.05e-03 (kg/m s Pa2) Ea1 = 2.4112e+08 K2 = 3.1 Pa2 Ea2 = 2.436e+08

0 1.9872e08 1.0060e08 5.5457e07 1.6712e08 1.2552e08 8.3680e06 1.51e+04

Mechanism D Volatiles ? mCxHyOzCxHyOz + mCmHnCmHn + vCO2CO2 + vCOCO + vH2OH2O CxHyOz ? mCC + vH2H2 + vCOCO CxHyOz ? mCC + mCmHnCmHn + vCOCO CxHyOz + (x/2  z/2)O2 ? xCO + y/2H2 CmHn + (m/2 + n/4a)O2 ? mCO + (1  a)n/2H2 + n/2aH2O CO + 0.5O2 ? CO2 H2 + 0.5O2 ? H2O CmHn + mH2O ? mCO + (n/2 + m)H2 C + 0.5O2 ? CO2

undergoes secondary reactions of tar decomposition. The secondary reactions of tar are modeled by considering three competing reactions yielding soot, hydrogen, light hydrocarbons, and carbon monoxide. Soot was treated as a gas phase species in combustion according to observation that no boundary layer exists on the soot particle surface [34]. The kinetic rate of soot combustion (reaction 9) is taken from [34]. A constant soot specific particle surface area of 1.2e05 m2/kg corresponding to a particle diameter of 25 nm was assumed [33]. Dependence on H2O in CO oxidation reaction was determined empirically or estimated based on kinetic arguments. The inclusion of H2O in the rate expression can be explained because most CO is consumed by reaction with OH that, to a first approximation may be assumed to be in equilibrium with water. Mechanisms B, C and D were implemented into CFD code via the user define interface. 3.3. Ideal reactor study Reduced mechanisms are used to decrease the computational cost. Their usage is often associated with losing accuracy and therefore making the simulations unpractical. A good practice is to compare global reaction mechanisms with either experimental data from combustion flame or validated detailed kinetic mechanisms. The global hydrocarbon combustion mechanisms were compared with the detailed GRI-Mech 3.0 model (325 elementary chemical reactions, 53 species) [11]. The point of this section is to characterize the behavior of global mechanisms at different conditions and high temperature. An attempt was made to indicate which mechanism has the biggest accuracy. The chosen detailed mechanism was validated in various conditions. We can assume that mechanism which results will be closest to GRI-Mech 3.0 brings the highest level of confidence. The calculations have been carried out with an in-house homogenous Perfectly Stirred Reactor (PSR) model. The calculations utilized GSL (GNU Scientific Library) libraries to solve the system of

stiff differential equations. The basic assumption of PSR is the instantaneous mixing of components caused by turbulence. The rate of reactions depends only on chemical kinetics and not mixing process. Isothermal reactor conditions were assumed. The temperature was set to 2000 K for all considered cases because this is the point where H2O dissociation gains significance. Fig. 6 presents CO and O2 calculation results at the PSR outlet as a function of equivalence ratio. The residence time for all five cases was set to 0.01 s. We can observe that O2 mole fractions computed by all four mechanism correlate well with the GRI-Mech 3.0. The highest level of similarity can be noticed for Mechanism A. In case of CO computations only Mechanism D was capable of qualitatively representing the CO variation for fuel-rich conditions. The higher the equivalence ratio the more CO is produced while O2 decreases. Fig. 7 demonstrates the influence of residence time in reactor on CO and O2 prediction. Fuel rich conditions were considered and the equivalence ratio was 1.25. It can be observed that mechanism A shows the best accuracy in O2 computations in terms of quantitative and qualitative correlation with the detailed mechanism. All four mechanisms show poor CO predictive capabilities in fuel rich atmosphere and only a general trend of CO decrease for longer residence times can be observed. A similar observations can be made regarding fuel lean conditions demonstrated in Fig. 8. For the equivalence ratio equal 0.8 Mechanism A again proves best capability in O2 prediction. Almost perfect correlation with detailed GRIMech 3.0 mechanism was presented for high residence times. For long residence times Mechanism D and B correspond well with detailed kinetics in terms of CO results. The reasons of discrepancy between detail and global mechanisms have a source in their generality. The global mechanisms describing hydrocarbon combustion account for simplified breakdown of fuel to CO and oxidation of CO to CO2. No intermediate steps and species are considered. In the absence of hydrogen species the oxidation of CO is very slow. Dry CO - O2 mixture (mixture without hydrogen containing additives) includes only two steps:

Please cite this article in press as: Modlinski N, Hardy T. Development of high-temperature corrosion risk monitoring system in pulverized coal boilers based on reducing conditions identification and CFD simulations. Appl Energy (2017), http://dx.doi.org/10.1016/j.apenergy.2017.04.084

8

N. Modlinski, T. Hardy / Applied Energy xxx (2017) xxx–xxx

0,07

0,09

GRIMech3.0 Mechanism A Mechanism B Mechanism C Mechanism D

CO mole fraction

0,07 0,06 0,05

GRIMech 3.0 Mechanism A Mechanism B Mechanism C Mechanism D

0,06 0,05

O2 mole fraction

0,08

0,04 0,03

0,04 0,03 0,02

0,02

0,01

0,01 0 0,7

0 0,9

1,1

1,3

1,5

0,7

0,9

Equivalence ratio

1,1

1,3

1,5

Equivalence ratio

Fig. 6. CO and O2 mole fractions as a function of equivalence ratio. Residence time = 0.01 s.

0,045

0,07

0,04

0,05 0,04 0,03

GRIMech3.0 Mechanism A Mechanism B Mechanism C Mechanism D

0,035

GRIMech3.0 Mechanism A Mechanism B Mechanism C Mechanism D

O2 mole fraction

CO mole fraction

0,06

0,02

0,03 0,025 0,02 0,015 0,01

0,01

0,005

0

0 0

0,002

0,004

0,006

0,008

0

0,01

0,002

Residence time (s)

0,004

0,006

0,008

0,01

Residence time (s)

Fig. 7. CO and O2 mole fractions as a function of residence time. Equivalence ratio = 1.25.

0,1

0,03

0,015

0,08

O2 mole fraction

CO mole fraction

0,02

0,01 0,005 0

GRIMech3.0 Mechanism A Mechanism B Mechanism C Mechanism D

0,09

GRIMech3.0 Mechanism A Mechanism B Mechanism C Mechanism D

0,025

0,07 0,06 0,05 0,04

0

0,002

0,004

0,006

0,008

0,01

Residence time (s)

0,03

0

0,002

0,004

0,006

0,008

0,01

Residence time (s)

Fig. 8. CO and O2 mole fractions as a function of residence time. Equivalence ratio = 0.8.

CO + O2 = CO2 + O and CO + O + M = CO2 + M. In case of hydrocarbon fuels CO production is a part of longer reaction sequence. The CO oxidation should be described by: (1) (2) (3) (4)

CO + O2 = CO2 + O + H2O = OH + OH CO + OH = CO2 + H H + O2 = OH + O

The first step is relatively slow and do not contribute significantly to CO destruction but it is a chain initiating reaction. Step two and four are chain branching reaction and CO oxidation

proceeds mainly through third reaction. If H2 is present, then following steps need to be involved: (1) + H2 = OH + H (2) OH + H2 = H2O + H (3) CO + HO2 = CO2 + OH These reactions continue to occur until an equilibrium is met between all species. The most important oxidation reaction occurs between OH and CO. This is due to the fact that OH radicals are rapidly consumed where as HO2, H2O and O2 require chain branching before these reactions can continue. Generally in high

Please cite this article in press as: Modlinski N, Hardy T. Development of high-temperature corrosion risk monitoring system in pulverized coal boilers based on reducing conditions identification and CFD simulations. Appl Energy (2017), http://dx.doi.org/10.1016/j.apenergy.2017.04.084

N. Modlinski, T. Hardy / Applied Energy xxx (2017) xxx–xxx

temperatures the entire H2-O2 reaction system should be included to describe CO oxidation due to significance of dissociation reactions. Even small detailed mechanism of hydrocarbon combustion involves description of carbon monoxide oxidation with 30 steps among 11 species [35]. A conclusion can be made that Mechanism A shows best O2 predicative capabilities in all considered test cases due to closest correlation with detailed mechanism. Global mechanisms perform much worst in CO calculations. None of the mechanisms proved its usefulness in CO prediction. The disparity of CO results in comparison to detailed mechanism is significant especially in fuel rich conditions. 3.4. CFD simulation results The previously evaluated global mechanisms (Table 2) were implemented into CFD code. The purpose of 3-D simulation study was to indicate which combustion kinetics provide the most accurate information on CO and O2 distribution in the furnace. Four simulation tests were conducted for nominal boiler load. The only difference between each test was the gas phase kinetic mechanism used. All other modelling aspects were identical. Fig. 9 presents large effect of gas phase global mechanism choice on simulated CO distribution in the boiler. CO distribution in a vertical plane crossing boiler axis of symmetry and two horizontal planes was demonstrated. Area of CO measurements covered by corrosion monitoring system on the front wall was also indicated. Substantial qualitative and quantitative differences

9

among different mechanisms can be seen. It was noticed that Mechanism A predicts relatively slow CO production in the burner region and its slowest burnout. Significant CO levels are visible close to the boiler axis above OFA ports. This is associated with high temperature region in the middle of the combustion chamber above burners where the reverse rate of water-gas shift reaction gains significance (Table 2). Mechanism B shows highest concentration of CO in the burner region and ash pit. On the other hand relatively fast CO burnout was predicted with this mechanism. The concentration of CO rapidly decreases after Over Fire Air injection. Mechanisms C and D simulate lowest CO concentration across combustion chamber with its maximum in the vicinity of the burners. Maximum local CO mole fraction did not exceed 2% in all considered cases. Analysis of Fig. 9 brings to a general conclusion that the choice of global gas phase mechanism has significant influence on CO prediction in CFD simulations. The simulated distribution maps of CO and O2 as well as temperature contours in the vicinity of front boiler wall are presented in Fig. 10. No special treatment of boundary layer was implemented. At wall, no slip condition were applied and the values of velocity components were set to zero. The flow near the wall was influenced by molecular viscosity rather than by turbulence. The wall function method of [36] which uses algebraic formulations to link quantities at the wall to those further away. Significant prediction differences were noticed between each numerical test. For low temperature combustion applications, CO kinetics become an important consideration. For these cases, most of the

Fig. 9. Comparison of simulated CO mole fractions (ppm) in a vertical plane crossing boiler axis of symmetry and two horizontal planes in the burner region.

Please cite this article in press as: Modlinski N, Hardy T. Development of high-temperature corrosion risk monitoring system in pulverized coal boilers based on reducing conditions identification and CFD simulations. Appl Energy (2017), http://dx.doi.org/10.1016/j.apenergy.2017.04.084

10

N. Modlinski, T. Hardy / Applied Energy xxx (2017) xxx–xxx

reactions still occur in a mixing limited flame zone. However, there will be zones in the domain where unburned CO must react with oxygen at low temperature regions like boundary layer of waterwalls. Fig. 10 indicates that the burner design promotes oxygen rich zone close to the waterwalls and intensive mixing. Above the burner belt O2 concentration decreases. All four mechanisms

show the occurrence of CO-rich regions despite oxygen presence. On the other hand Fig. 9 exposed relationship between CO and temperature. Zones with relatively low temperatures correlate with CO-rich atmosphere. These regions are far from chemical equilibrium and low CO oxidation rates correspond with moderate temperatures. These feature is noticeable for all the cases. Size of

Fig. 10. Calculated mole fractions of CO (ppm), O2 (%) and temperature (K) distribution in the boundary layer of the front waterwall.

Please cite this article in press as: Modlinski N, Hardy T. Development of high-temperature corrosion risk monitoring system in pulverized coal boilers based on reducing conditions identification and CFD simulations. Appl Energy (2017), http://dx.doi.org/10.1016/j.apenergy.2017.04.084

N. Modlinski, T. Hardy / Applied Energy xxx (2017) xxx–xxx

11

Fig. 11. Comparison of O2 and CO mole fractions (%) measurement with CFD modelling results.

these zones is different for each mechanism. Mechanism B again predicts highest CO levels in the burner region while mechanism D depicts its lowest level. The CO and O2 measurements the boundary layer of waterwalls have been extracted from the monitoring system and compared with simulation results. The measuring ports were located on the front wall of the boiler (Fig. 11). The measurements presented below have been done by utilizing all of the 26 ports. The neural network model was not used to generate the distribution maps. Obtained data was then implemented into a grid based contour program to visualize the 2-D CO and O2 distribution. By analogy the local values of CO and O2 were calculated by the CFD code in points corresponding to probe position. Consequently they have been processed with the same interpolation method. The comparison is shown in Fig. 11. The experimental data of O2 and CO concentrations in each of 26 measurement ports with the CFD simulation results have been demonstrated in Appendix A. It can be noticed that O2 predictions produced by Mechanism C demonstrates the best qualitative agreement with the measurement (Fig. 11). Oxygen depletion zone is visible in the upper part of front waterwall. Other mechanisms reproduce similar trend of oxygen distribution. None of the mechanisms was capable of properly predicting CO concentration increase in the upper part of the furnace as shown in Fig. 11. It is worth noticing that Mechanism A does predict late CO burnout in the upper furnace (Fig. 9) but the CO-rich zone does not reach front wall as shown in Figs. 10 and 11. Comparison of 3-D simulation results with measurements proves that it is difficult to assess the fidelity of the combustion mechanism applied in large scale boiler 3-D simulation based on 0-D ideal reactor analysis despite the fact that gas phase modelling approach used in CFD (EDC model [28]) utilizes the concept of Perfectly Stirred Reactor. The ideal reactor analysis indicates that global mechanisms perform better in O2 calculation at high temperature and relatively long residence time. Generally both CFD and ideal reactor modelling show better O2 predictive capabilities in comparison to CO computations. However more detailed juxtaposition of CFD results with 0-D homogenous reactors calculations indicate that individual global mechanisms behave differently in the EDC

model implemented into CFD code and Perfectly Stirred Reactor model. For example Mechanism C shows best accuracy in CFD oxygen prediction (Fig. 11) but not in PSR calculations. An approach to quantitatively evaluate the CFD accuracy was presented in Fig. 12. Absolute difference between measured mole fractions of O2 (%) and CO (ppm) and their calculated values in 26 measurement points was presented for global mechanisms. In case of oxygen the average difference over all measurements fluctuate between 6% for Mechanism C and 10.3% for Mechanism D. Differences between calculations and measurements in specific local points are in the range of 1 to 13%. The locally measured O2 concentration vary between 0 and 20%. The quantitative accuracy of CFDbased oxygen prediction in the boundary layer should be described as moderate, with Mechanism C being closest to measurements. In case of CO the smallest average difference is also calculated with Mechanism C (3292 ppm) while the largest differences were presented by Mechanism B (13,274 ppm). The locally measured CO concentration vary between 100 and 20,000 ppm. The minimum local difference between CO calculations and measurements was 100 ppm while the highest one almost 30,000 ppm. Neither quantitative nor qualitative (Fig. 11) CFD-based CO predictions can be described as satisfactory.

4. Conclusions The on-line monitoring system of corrosion risk based on O2 and CO measurements was demonstrated. Since the correlation between O2 and CO can be estimated (Fig. 1) the system can operate basing on O2 measurement only. There are 26 available measurement ports on the front wall of the case study boiler. As a compromise between cost and accuracy only five ports are used continuously. In order to obtain information about the distribution of oxygen in the boundary layer over the whole wall (and not just at selected points) neural network technique was used [14]. The process of neural network learning after design, fuel characteristics or significant operating conditions changes is expensive and labor-intensive. To improve the functionality of the monitoring system Computational Fluid Dynamics may appear helpful.

Please cite this article in press as: Modlinski N, Hardy T. Development of high-temperature corrosion risk monitoring system in pulverized coal boilers based on reducing conditions identification and CFD simulations. Appl Energy (2017), http://dx.doi.org/10.1016/j.apenergy.2017.04.084

N. Modlinski, T. Hardy / Applied Energy xxx (2017) xxx–xxx

Difference from O 2 measurment (%)

12

Mechanism A, average difference = 7.8% Mechanism B, average difference = 7.3% Mechanism C, average difference = 6% Mechanism D, average difference = 10.3%

18 16 14 12 10 8 6 4 2 0

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

16

17

18

19

20

21

22

23

24

25

26

Difference from CO measurment (ppm)

Measurement point Mechanism A, average difference = 3470 ppm Mechanism B, average difference = 13274 ppm Mechanism C, average difference 3292 ppm Mechanism D, average difference = 4003 ppm

30000 25000 20000 15000 10000 5000 0

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

Measurement point Fig. 12. Absolute difference between measured mole fractions of O2 (%) and CO (ppm) and their CFD-based calculated values in 26 measurement points.

An effort has been made to check whether CFD simulations are capable of properly predicting O2 and CO. The scientific part of the current research was concentrated on volatiles combustion simulation with the emphasis on CO burnout. Four popular global mechanisms have been implemented into CFD code. The global mechanisms were compared with the detailed mechanism [11] in a Perfectly Stirred Reactor model. It was concluded that the models perform better with O2 prediction than CO especially in fuel-rich conditions. Selection of the global mechanism has significant influence on CO and O2 prediction. The measurements of the CO and O2 in the waterwalls boundary layer were extracted from the monitoring system and compared to simulation results. Both CFD and ideal reactor modelling show relatively better O2 predic-

tive capabilities in reference to CO in addition to other intrinsic CFD model inaccuracies. Mechanism C demonstrates acceptable qualitative agreement with the measurement in terms of O2 predictions. The quantitative accuracy of CFD-based oxygen prediction in the boundary layer was described as moderate. Mechanism C was closest to O2 measurements. CFD-based CO predictions were described as unsatisfactory. Future research on development of CFD-based tool for O2 and CO prediction in the boundary layer will require implementation of more detailed gas-phase kinetic mechanism. It is also necessary to verify he influence of char combustion model on simulation results.

Table A O2 measurements and calculations. Measurement port number

O2 (%) Measurement

O2 (%) Mechanism A

O2 (%) Mechanism B

O2 (%) Mechanism C

O2 (%) Mechanism D

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26

0.1 0.5 0.6 0.1 0.32 1.5 0.2 0.1 9.5 7.2 0.8 0.9 3.15 10.5 9.5 6 7.2 10.4 10.85 11.78 12.2 7 8.5 10 10.5 5.2

9.8047399 9.86E+00 9.42E+00 9.84E+00 9.41E+00 1.51E+01 1.06E+01 4.33E+00 1.09E+01 1.55E+01 1.60E+01 1.29E+01 1.25E+01 1.70E+01 1.68E+01 1.65E+01 5.31E+00 1.71E+01 1.53E+01 1.56E+01 17.460561 1.24E+01 1.48E+01 1.48E+01 16.42429 17.154696

7.26 10.49 5.30 5.98 9.89 11.05 7.54 8.23 17.58 13.19 9.74 10.15 13.26 15.82 15.09 17.83 17.31 16.63 17.35 17.40 17.25 9.91 15.24 15.70 15.27 12.85

3.91 9.12 2.53 2.85 7.64 10.86 6.54 6.88 14.91 12.39 11.32 11.56 13.31 16.05 15.11 13.32 13.67 14.90 15.86 16.88 17.01 11.53 12.37 14.26 15.30 3.91

11.18 14.43 11.94 10.07 15.42 15.31 13.07 13.02 18.00 17.90 16.32 15.91 16.10 18.00 18.21 18.64 18.06 19.57 17.21 18.79 14.06 13.71 19.34 15.65 17.50 16.68

Please cite this article in press as: Modlinski N, Hardy T. Development of high-temperature corrosion risk monitoring system in pulverized coal boilers based on reducing conditions identification and CFD simulations. Appl Energy (2017), http://dx.doi.org/10.1016/j.apenergy.2017.04.084

13

N. Modlinski, T. Hardy / Applied Energy xxx (2017) xxx–xxx Table B CO measurements and calculations. Measurement port number

CO (%) Measurement

CO (%) Mechanism A

CO (%) Mechanism B

CO (%) Mechanism C

CO (%) Mechanism D

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26

1.55000 0.10000 0.09000 2.00000 0.40000 0.12000 0.50000 1.40000 0.01300 0.00760 0.13000 0.10400 0.01900 0.01600 0.00700 0.01700 0.01600 0.05000 0.01000 0.01500 0.03500 0.01000 0.00800 0.02000 0.04800 0.01300

0.00392 0.00051 0.00040 0.00822 0.00124 0.00085 0.01513 0.04201 0.00301 0.00183 0.00020 0.00031 0.01538 0.00066 0.01314 0.01460 0.16804 0.14237 0.00286 0.11885 0.55363 0.00000 0.00004 0.10061 1.03734 0.70748

0.12002 0.11865 0.10123 0.15769 0.14931 0.17160 0.06738 0.33629 1.25143 1.77463 0.94323 1.24475 1.90873 2.76683 2.88351 0.80071 1.18734 2.34421 0.77087 0.65093 0.56055 5.13509 2.38489 1.28368 1.02728 1.03140

0.03730 0.01557 0.03211 0.03555 0.03359 0.00735 0.01182 0.03722 0.00238 0.00020 0.00279 0.00457 0.00989 0.13290 0.12677 0.00875 0.00229 0.01298 0.40994 0.44676 0.25544 0.00035 0.00031 0.43936 0.45950 0.17219

0.00957 0.00501 0.00524 0.00828 0.00394 0.00738 0.00000 0.00124 0.00113 0.00097 0.00000 0.00283 0.00444 0.00301 0.06079 0.02005 0.00494 0.03714 0.51438 0.69152 0.00120 0.00032 0.41086 0.97826 0.93956 0.46670

Appendix A Tables A and B demonstrate the time-averaged O2 and CO concentration measurements in specific ports (Fig. 2) and the CFD calculated values in these locations. Port number 1 is situated in the top left part of the front wall, while the 26th port is located in the bottom right as visualized in Fig. 2. Numerical simulations have been conducted after implementation of four global gas phase reaction kinetic mechanisms described in Table 2.

References [1] World Energy Outlook 2014. International Energy Agency Report IEA, Paris; 2014. [2] EPRI. Waterwall wastage mechanisms in coal-fired boilers: the effect of coal chemistry on Waterwall wastage, TR-1004021; 2001. [3] Bryers R. Fireside Slagging, fouling and high-temperature corrosion of heattransfer surface due to impurities in steam-raising fuels. Prog Energy Combust Sci 1996;22:29–120. [4] Winters M, Stokes P, Zuniga P, Schlottenmeier D. Real-time performance monitoring of fouling and under-deposit corrosion in cooling water systems. Corr Sci 1993;35:1667–75. [5] Farrell D, Mok W, Pinder L. On-line monitoring of furnace-wall corrosion in a 125 MW power generation boiler. Mater Sci Eng 1989;121:651–9. [6] Linjeville T, Valentine J, Davis K, Harding N, Cox W. Prediction and real-time monitoring techniques for corrosion characterization in furnaces. Mater High Temp 2003;20:175–84. [7] Davis K, Linjewile T, Valentine J, Swensen D, Shino D, Letcavits J. A multi-point corrosion monitoring system applied in a 1 300 MW coal-fired boiler. AntiCorr Methods Mater 2004;51(5):321–30. [8] Fry Andrew, Adams B, Davis K, Swensen D, Munson S, Cox W. An investigation into the likely impact of oxy-coal retrofit on fire-side corrosion behavior in utility boilers. Int J Greenhouse Gas Control 2011;2011(5):179–85. [9] Kakietek S, Hardy T, Golec T, Kordylewski W. Neural networks and genetic algorithms as a tool for monitoring of low emission corrosion and optimisation of utility boiler performance. In: 10th Conference on energy for a clean environment, 7–10 July 2009, Lisbon, Portugal. [10] Harb J, Smith E. Fireside corrosion in PC-fired boilers. Prog Energy Combust Sci 1990;16:169–90. [11] GRI-Mech 3.0. The University of California at Berkeley, Stanford University, SRI (Stanford Research Institute) International and sponsored by the Gas Research Institute. [12] Yacheng L, Weidong F, Yu L. Numerical investigation of air-staged combustion emphasizing char gasification and gas temperature deviation in a large-scale, tangentially fired pulverized-coal boiler. Appl Energy 2016;177:323–34.

[13] Hu Y, Li H, Yan J. Numerical investigation of heat transfer characteristics in utility boilers of oxy-coal combustion. Appl Energy 2014;130:323–34. [14] Yin C, Yan J. Oxy-fuel combustion of pulverized fuels: combustion fundamentals and modeling. Appl Energy 2016;162:742–62. [15] Xu M, Azevedo J, Carvalho M. Modeling of a front wall fired utility boiler for different operating conditions. Comput Methods Appl Mech Eng 2001;190:3581–90. [16] Yin C, Caillat S, Harion J, Baudoin B, Perez E. Investigation of the flow, combustion, heat-transfer and emissions of a 609 MW utility tangentially fired pulverized-coal boiler. Fuel 2002;81:997–1006. [17] Pallares J, Arauzo I, Diez L. Numerical prediction of unburned carbon levels in large pulverized coal utility boilers. Fuel 2005;84:2364–71. [18] Karampinis E, Nikolopoulos N, Nikolopoulos A, Grammelis P, Kakaras E. Numerical investigation Greek lignite/cardoon co-firing in a tangentially fired furnace. Appl Energy 2017;97:514–24. [19] Choi C, Kim C. Numerical investigation on the flow, combustion and NOx emission characteristics in a 500 MWe tangentially fired pulverized-coal boiler. Fuel 2009;88:1720–31. [20] ANSYS FLUENT, . [21] Modlinski N, Madejski P, Janda T, Szczepanek K, Kordylewski W. A validation of computational fluid dynamics temperature distribution prediction in a pulverized coal boiler with acoustic temperature measurement. Energy 2015;92:77–86. [22] John C, Niksa S. Coal devolatilization during rapid transient heating. 1. Primary devolatilization. Energy Fuels 1992;6:254–64. [23] Williams A, Backreedy, Habib R, Jones J, Pourkashanian M. Modeling coal combustion: the current position. Fuel 2002;81:605–18. [24] Advanced Fuel Research Inc., Usurers guide for FG-DVC, East Hartford, CT; 2004. [25] Williams A, Backreedy, Habib R, Jones J, Pourkashanian M. An extended coal combustion model. Fuel 1999;78:1745–54. [26] Bradley R, Marc A, Dave H. Modeling non-equilibrium co oxidation in combustion systems. In: International mechanical engineering congress & exposition, Orlando, Florida, November 5–10; 2000. [27] Spalding DB. Mixing and chemical reaction in steady confined turbulent flames. In: 13th Symp (Int’l.) on combustion. The Combustion Institute; 1971. p. 649–57. [28] Magnussen B. On the Structure of Turbulence and a Generalized Eddy Dissipation Concept for Chemical Reaction in Turbulent Flow. Nineteenth AIAA Meeting, St. Louis, 1981. [29] Ruckert F, Sabel T, Schnell U, Hein K, Risio B. Comparison of different global reaction mechanisms for coal-fired utility boilers. Progr Comput Fluid Dyn 2003;3:130–9. [30] Jones W, Lindstedt R. Global reaction schemes for hydrocarbon combustion. Combust Flame 1988;73:233–49. [31] Hautmann D, Dryer F, Schug K, Glassman I. A multiple-step overall kinetic mechanism for the oxidation of hydrocarbons. Combust Sci Technol 1981;25:219–49. [32] Shaw D, Zhu X, Misra M, Essenhigh R. Twenty-third symposium (international) on combustion. Pittsburgh: The Combustion Institute; 1990. p. 1155–62.

Please cite this article in press as: Modlinski N, Hardy T. Development of high-temperature corrosion risk monitoring system in pulverized coal boilers based on reducing conditions identification and CFD simulations. Appl Energy (2017), http://dx.doi.org/10.1016/j.apenergy.2017.04.084

14

N. Modlinski, T. Hardy / Applied Energy xxx (2017) xxx–xxx

[33] Fortsch D, Kluger F, Schnell U, Spliethoff, Hein R. Kinetic model for the prediction of no emissions from staged combustion of pulverized coal. In: Twenty-Seventh Symposium (International) on Combustion. The Combustion Institute; 1998. p. 3037–44. [34] Magnussen B. Thirteenth symposium (international) on combustion. Pittsburgh: The Combustion Institute; 1971. p. 869–77.

[35] Petrova M, Williams F. A small detailed chemical-kinetic mechanism for hydrocarbon combustion. Combust Flame 2006;144:526–44. [36] Launder B, Spalding D. The numerical computation of turbulent flows. Comput Methods Appl Mech Eng 1974;3:269–89.

Please cite this article in press as: Modlinski N, Hardy T. Development of high-temperature corrosion risk monitoring system in pulverized coal boilers based on reducing conditions identification and CFD simulations. Appl Energy (2017), http://dx.doi.org/10.1016/j.apenergy.2017.04.084