Tests on co-firing of municipal solid waste and coal in a circulating fluidized bed

Tests on co-firing of municipal solid waste and coal in a circulating fluidized bed

Energy Conversion and Management 43 (2002) 2189–2199 www.elsevier.com/locate/enconman Tests on co-firing of municipal solid waste and coal in a circul...

349KB Sizes 0 Downloads 45 Views

Energy Conversion and Management 43 (2002) 2189–2199 www.elsevier.com/locate/enconman

Tests on co-firing of municipal solid waste and coal in a circulating fluidized bed Changqing Dong *, Baosheng Jin, Zhaoping Zhong, Jixiang Lan Education Ministry, Key Laboratory on Clean Coal Power Generation and Combustion Technology, Thermoenergy Engineering Research Institute, Southeast University, Nanjing 210096, China Received 1 May 2001; accepted 7 September 2001

Abstract Energy recovery from municipal solid waste (MSW) is a feasible method through various processes, such as combustion, pyrolysis and gasification. Tests on the co-firing of MSW and coal were conducted in a 0.2 MWth circulating fluidized bed, and the emissions of NO, N2 O, HCl and SO2 were studied. A three layer feed forward neural network was constructed and trained by the BP method with experimental data. The impacts of mixing ratio and bed temperature on the gaseous emissions were considered. The model predicted gaseous pollutions emissions were consistent with experimental data.  2002 Elsevier Science Ltd. All rights reserved. Keywords: Circulating fluidized bed; Municipal solid waste; Feed forward neural networks

1. Introduction The rising prices of raw materials and the energy crisis have resulted in an increasing concern for material recovery and reuse from both management and technical aspects. On the other hand, thermal treatment has been proven as an attractive method of waste disposal due to the primary advantages of hygienic control, volume reduction and energy recovery. Fluidized bed combustion allows clean and efficient combustion of various solid fuels. Several studies on the co-combustion of fuel mixtures have shown that the blends of low rank coal and biomass can be successfully combusted in a circulating fluidized bed (CFB) and that increasing the proportion of biomass improves the combustion efficiency and the environmental impact [1–4]. Desroches-Ducarne and

*

Corresponding author. Tel.: +86-25-3794744; fax: +86-25-7714489. E-mail addresses: [email protected], [email protected] (C. Dong).

0196-8904/02/$ - see front matter  2002 Elsevier Science Ltd. All rights reserved. PII: S 0 1 9 6 - 8 9 0 4 ( 0 1 ) 0 0 1 5 7 - 1

2190

C. Dong et al. / Energy Conversion and Management 43 (2002) 2189–2199

Marty [5] have also revealed that the co-combustion of municipal solid waste (MSW) and coal mixture in a CFB are feasible. This study presents the results of co-combustion tests of MSW and coal in a 0.2 MWth CFB. A three layer feed forward neural network (FFNN) was proposed to predict the gaseous pollutions emissions, such as HCl, NO, N2 O and SO2 .

2. Experimental 2.1. Fuel and their composition The MSW used in this experiment was collected in a living area of Nanjing (China). Its main components were classified in 19 categories and ground to obtain a mean diameter of 5 mm. All the categories were finally mixed to achieve a fuel with constant composition corresponding to typical China MSW [6]. The coal used in this experiment was Xuzhou bituminous coal. The properties of the fuels are given in Table 1 in the form of average values from several samples. 2.2. The circulating fluidized bed The tests were conducted in a 0.2 MWth CFB (shown in Fig. 1). The experimental system is composed of a riser of 23 cm i.d. and 7 m height, fuel (MSW, coal) feeding systems, a line for fly ash and bed materials circulation and a fumes cooling and filtration system. The CFB is preheated to the coal ignition temperature by a start-up burner. Coal is fed into the bed by a screw feeder, and MSW is fed into the bed by a rotary feeder. The total combustion air is divided into two streams: primary air is preheated to about 400 C and distributed at the bottom of the bed; secondary air is injected through the airtight end of the MSW feeding path-way to prevent MSW blocking the path-way. At the top of the riser, a cyclone allows the recovery of entrained particles. 2.3. Analysis The CFB is equipped for continuous measurement of temperature and pressure drops at different heights in the reactor. The major gas components of extractive samplings of the exhaust gas Table 1 Coal and MSW properties

Xuzhou bituminous coal MSW

Proximate analysis (wt.%)

Ultimate analysis (wt.% daf)

Moisture

Ash (dry matter)

Volatiles (daf)

C

H

O

N

S

Cl

Ca

Ca=ðSþ 0:5ClÞ

Lower heating value (kJ/kg)

8.42

16.75

31.42

81.9

5.68

9.67

1.44

1.16

0.12

0.1

0.097

23 362

43.7

37.5

92.2

54.7

6.89

34.7

1.38

0.27

1

1.02

1.7

4197

C. Dong et al. / Energy Conversion and Management 43 (2002) 2189–2199

2191

Fig. 1. The CFB combustor.

are analyzed by a multi-function flue gas analyzer. HCl in the flue gas is sampled and measured by the silver nitrate volume method. Bottom ash and fly ash samplings are taken for researching the combustion efficiency. All the tests were performed under the same condition: with excess air of 30 and 100 vol % of primary.

3. Results and disscusion 3.1. Combustion efficiency Fig. 2 shows the temperatures varying from the air distributor to the outlet of the CFB at different mixing ratios. The temperature in the dilute region is about 100–300 C higher than that with only firing coal. It is due to the combustion of volatile matter in the dilute region. MSW has a

Fig. 2. Temperature draft vs relative height.

2192

C. Dong et al. / Energy Conversion and Management 43 (2002) 2189–2199

Fig. 3. Combustion efficiency vs mixing ratio R.

higher volatile matter that dominates the processes of the MSW combustion. The carbon consumption is faster during MSW added. Therefore, the concentration of char in the bed is lower, and the efficiency increases (as shown in Fig. 3). 3.2. Gaseous pollution emission 3.2.1. SO2 emission As shown in Table 1, the sulfur contents mainly come from coal. SO2 emission is clearly lower at high mixing ratio as shown in Fig. 4. The reason must be the high Ca and low sulfur contents in MSW. Previous study results concerning coal or waste combustion have shown that a significant retention of sulfur can be obtained when sorbents are introduced into the CFB [7,8]. Other studies have noted that calcium, potassium or sodium, which are present in relatively large amounts in biomass fuel ashes, may also act as sorbents and be active for acid gas emissions reduction [9,10]. In a general bubbling fluidized bed, the formation rate of SO2 increases with the temperature increasing, but during the co-firing of MSW and coal in the CFB, SO2 emission remains constant with temperature increasing (Fig. 5). The probable reason is attributed to the effects of higher freeboard temperature and secondary air supply.

Fig. 4. The effect of mixing ratio on SO2 concentration in the flue gas at bed temperature 973 C.

C. Dong et al. / Energy Conversion and Management 43 (2002) 2189–2199

2193

Fig. 5. The effect of bed temperature on SO2 concentration in the flue gas at mixing ratio R ¼ 4.

3.2.2. HCl emission HCl may cause corrosion of the water-walls of the boiler through the following reactions: Fe þ 2HCl ! FeCl2 þ H2 FeO þ 2HCl ! FeCl2 þ H2 O Fe2 O3 þ 2HCl þ CO ! FeO þ FeCl2 þ H2 O þ CO2 Fe3 O4 þ 2HCl þ CO ! FeO þ FeCl2 þ H2 O þ CO2 and 2Cr2 O3 þ 4Cl2 þ O2 ! 4CrOCl2 Cr2 O3 þ 4HCl þ H2 ! 2CrCl2 þ 3H2 O NaCl and PVC may be the source of Cl during the MSW incineration. HCl can be formed through the following reactions: NaCl þ H2 O ! NaOH þ HCl 2NaCl þ H2 O þ SO2 ! Na2 SO3 þ 2HCl 2NaCl þ H2 O þ SO3 ! Na2 SO3 þ 2HCl 2NaCl þ H2 O þ SiO2 ! Na2 SiO3 þ 2HCl PVC ) L þ HCl þ R þ HC Here L is condensable organic matter, R is solid char and HC is volatile organic matter. It can be seen that the formation of HCl is promoted when Cl, S, H2 O and O2 co-exist. Fig. 6 shows that HCl emission increases with the mixing ratio R increasing. The increasing of H2 O and Cl contents in the fuel mixture causes it. The effect of temperature on HCl emission is not obvious (Fig. 7). It must be that HCl has been formed under lower temperature. Therefore, the HCl emission concentration only increases slightly with temperature increasing. 3.3. Nitric and nitrous oxides As shown in Figs. 8 and 9, when MSW is fed into the CFB, the NO and N2 O decrease abruptly. The reasons are supposed to be

2194

C. Dong et al. / Energy Conversion and Management 43 (2002) 2189–2199

Fig. 6. The effect of mixing ratio on HCl concentration in the flue gas at bed temperature 973 C.

Fig. 7. The effect of bed temperature on HCl concentration in the flue gas at mixing ratio R ¼ 4.

Fig. 8. The effect of mixing ratio on NO in the flue gas at bed temperature 973 C.

• the volatile products that surround the small MSW particles combust. The diffusion of oxygen to the surface of carbon is suppressed, and the reduction reactions that cause the decrease of nitric and nitrous oxide are promoted,

C. Dong et al. / Energy Conversion and Management 43 (2002) 2189–2199

2195

Fig. 9. The effect of mixing ratio on N2 O in the flue gas at bed temperature 973 C.

• the nitrogen content in wood chip and rice husk is lower and volatile nitrogen is generally released as NHi compounds rather than HCN, leading to the formation of N2 O being suppressed, • radicals like H and OH coming from combustion of biomass volatile matter reduce N2 O through the following reactions: N2 O þ H ! N2 þ OH N2 O þ OH ! N2 þ HO2 • MSW ash contains relatively some calcium, potassium and sodium, which have catalytic effects on N2 O decomposition. With the ratio of MSW to coal increasing, the N2 O increases slightly. It may be caused by the lower location temperature zone that surrounds the particles. Fig. 10 shows the effect of bed temperature on NO and N2 O emissions. NO increases, whereas N2 O decreases with bed temperature increasing. According to the results of coal firing tests, it is

Fig. 10. The influence of bed temperature on NO and N2 O emissions at R ¼ 4.

2196

C. Dong et al. / Energy Conversion and Management 43 (2002) 2189–2199

expected that high NO emission is due to the effect of higher temperature in the dilute region and lower unburned carbon content in the furnace. • The temperature in the dilute region is much higher than that when only firing coal (as shown in Fig. 3). It is the cause of the volatile content combusting in the dilute region. Higher temperature promotes the reduction of N2 O and the oxidation of NO. • With the bed temperature increasing, the concentration of char in the furnace decreased. The reduction reaction of nitric oxide on the char is suppressed.

3.4. Prediction of gaseous pollutant emission Models that describe the combustion process during CFB incineration have already been proposed to predict the impact of operating parameters on the gaseous pollutant emissions [11– 13]. Desroches-Ducarne [5] also proposed a simplified prediction model for NO emissions during the co-combustion of MSW and coal in a CFB. Because the composition of MSW is very complicated, many elements may influence the formation and destruction of gaseous pollutants, such as operating parameters and the composition of the fuel mixture. The impacts of mixing ratio and temperature are considered in this paper. Neural networks have been used in many areas, such as automatic control, computer science and chemistry [14]. In principal, a neural network has the power of a universal approximation [15]. The main advantage of neural networks is the fact that they are able to use some a priori unknown information hidden in the data. A three layer FFNN is constructed to predict the gaseous pollutant emissions in this paper. As seen in Fig. 11, the neural network is composed of an input layer, a hidden layer and an output layer. The mixing ratio R and temperature T are used as input signals, and gaseous pollutants are used as output signals. Each neuron in a particular

Fig. 11. Feed forward neural network.

C. Dong et al. / Energy Conversion and Management 43 (2002) 2189–2199

2197

layer is connected with all neurons in the next layer. The connection between the ith and jth neurons is characterized by the weight coefficient wij . The output value of xi is determined by Eqs. (1) and (2). They hold that xi ¼ f ðni Þ ni ¼

X

ð1Þ ð2Þ

wij xj

Here, ni is the potential of the ith neuron and function f ðni Þ is called the transfer function. It holds that f ðnÞ ¼

1 1 þ expðnÞ

ð3Þ

The weight coefficients wij are revised to minimize the sum of the squared differences between the computed and required output values. This is accomplished by minimization of the object function E: E ¼ 12ðx0  x0r Þ2

ð4Þ

Here, x0 and x0r are the computed and required activities of the output neuron. A back propagation training algorithm is used to vary the weight coefficients. It holds that  ðkÞ oE ðkþ1Þ ðkÞ wij ¼ wij  k ð5Þ owij Here, k is the learning rate. The training mode begins with random numbers of the weights and proceeds iteratively. The crucial problem in the model selection is how to determine the numbers of hidden units. There is no way to determine a good network topology just from the number of inputs and outputs. It depends critically on the number of training cases, the amount of noise and the complexity of the function or classification. The optimal number of hidden units is determined by experiments in this paper.

Fig. 12. The prediction of SO2 emission.

2198

C. Dong et al. / Energy Conversion and Management 43 (2002) 2189–2199

Fig. 13. The prediction of HCl emission.

Fig. 14. The prediction of NO emission.

Fig. 15. The prediction of N2 O emission.

C. Dong et al. / Energy Conversion and Management 43 (2002) 2189–2199

2199

When the hidden units are 7 and the learning rate is 0.9, the network converges faster. The calculated results are showed in Figs. 12–15. The impact of R and T on each gaseous pollutant can be seen. The predicting results are consistent with the experimental data. 4. Conclusion Combustion tests performed at different mixing ratios of MSW to coal show that the additions of municipal refuses lead to • higher combustion efficiencies due to the increased volatile content of the mixing fuel, • lower NO and SO2 emissions. On the other hand, HCl concentrations increase with the amount of waste added, • N2 O emissions decrease rapidly when MSW is fed into the CFB. Increasing the ratio of MSW to coal, the N2 O emissions increase slowly. When the fuel-mixing ratio remained constant, NO emissions increase, SO2 and HCl remain constant, but N2 O decreases with temperature increasing, • A FFNN model was proposed to predict gaseous pollutant emissions with variation of the mixing ratio and bed temperature. The predicted results are consistent with experimental data.

Acknowledgements This work has been supported by the foundation of Key Laboratory of Education Ministry (China) and Key Science and Technology Project of Education Ministry (China). The support of these organizations is gratefully acknowledged. References [1] Leckner B, Karlsson M. Biomass Energy 1993;4:379–89. [2] Roesler JF, Yetter RA, Dryer FL. Combust Flame 1995;100:4950504. [3] Andries J, Verloop M, Hein K. Proceedings of the 14th International Conference on FBC, Vancouver, Canada, 1997. p. 313–20. [4] Armesto L, Cabanillas A, Bahillo A, Segovia JJ, Escalada R, Martinez JM, Carrasco JE. Proceedings of the 14th International Conference on FBC, Vancouver, Canada, 1997. p. 301–12. [5] Desroches-Ducarne E, Marty E, et al. Fuel 1998;77(12):1311–5. [6] Chinese Statistics Yearbook, 1997. [7] Jozewicz W, Chang JCS, Seman CB. Environ Progr 1996;51(11):137–42. [8] Mstsukata M, Tajeda K, Miyatani T, Ueyama K. Chem Eng Sci 1996;51(11):2529–34. [9] Nordin A. Fuel 1995;74(4):615–22. [10] Nordin A. Biomass Energy 1994;6:339. [11] Mukadi L, Guy C, Legros R. Fuel 2000;79:1125–36. [12] Kilpinen P, Hupa M. Combust Flame 1991;85:94–100. [13] Andanez J, Diego LF, Garcia-Labiano F, Gayan P. Coal Sci Technol 1995;2:1839–42. [14] Svozil D et al. Chemom Intell Lab Syst 1997;39:43–62. [15] Zupan J, Gasteiger J. In: Neural networks for chemists: an introduction. Weinheim: VCH; 1993. p. 5–10.