Emission Characteristics of Particulate Matter from RuralHousehold Biofuel Combustion in China

Xinghua Li, Lei Duan, Shuxiao Wang, Jingchun Duan, Xingming Guo, Honghong Yi,Jingnan Hu, Chao Li, and Jiming Hao*

Department of EnVironmental Science and Engineering, Tsinghua UniVersity, Beijing 100084, China

ReceiVed April 6, 2006. ReVised Manuscript ReceiVed October 9, 2006

Field measurements in rural households were conducted in three Chinese municipalities/provinces (Beijing,Chongqing, and Henan) to determine the emission characteristics of particulate matter from biofuel combustion.The selected biofuel types and stove types are representative of local rural areas. Particle number concentration,size distribution, and mass size distribution were determined. Both the particle number and mass of theseemissions were dominated by submicrometer particles. The emission factor of PM2.5 from combustion isbetween 1.80 and 7.00 g/kg of biofuel (dry basis) and 0.84-2.40 g/MJ of delivered energy and is averaged to4.21 g/kg of biofuel (dry basis) and 1.46 g/MJ delivered energy. In this study, it appears that particle emissionscan be correlated with combustion conditions and stove configuration. Particle emissions are the highest duringthe high power phase. Unfortunately, the more thermally efficient stove has higher per kilogram fuel particulatematter (PM) emissions than the less thermally efficient stoves, that is, the increase in thermal efficiency cannotoffset the increase in particle emissions.

Introduction

With the rapid development of the economy and society,commercial energy, such as electricity and liquid propane gas(LPG), is becoming more popular in China’s rural households.However, biofuel, such as crop waste and wood, still dominatesthe rural energy supply in China. In 2000, 288 million tons ofagricultural biomass and 141 million tons of firewood weredirectly burned by rural households for cooking and heating,contributing 57% of total rural household energy use (353million tons of coal equivalent).1,2 Patterns of energy use inChina’s rural households, which are dominated by biofuels, willnot change significantly for a long time. Unfortunately, biofuelcombustion is mainly carried out in small household stovesunder poor combustion conditions and without any emissioncontrol. This results in high levels of particulate matter, CO,polycyclic aromatic hydrocarbons (PAHs), and other air pol-lutant emissions, which cause high levels of indoor air pollution,3

local air pollution,4 and regional and global climate impacts.5,6

The particulate matter contains a large carbonaceous fraction,i.e., organic carbon (OC) and black carbon (BC). BC is thoughtto absorb solar radiation and contribute to global warming.6

Street et al. estimated that BC emissions from biofuel combus-

tion in China were 512 Gg in 1995, comprising 38% of thetotal national emissions.7 OC contains a multitude of organiccompounds, some of which are carcinogenic and mutagenic,such as polycyclic aromatic hydrocarbons (PAHs).8,9

In view of the adverse effect of particulate matter emittedfrom biofuel combustion, it is necessary to know the emissioncharacteristics so as to reduce the emissions. A few studies onemissions from household stoves in developing countries havebeen conducted. For these small combustion devices, three majortypes of methods are used to determine emissions: the chambermethod, hood method, and the carbon balance approach.10 Joshiet al. adopted the chamber method to study CO and totalsuspended particle (TSP) emissions from burning biofuels inmetal cook stoves and found that the more efficient stoves havehigher emission factors of both CO and TSP for all three biofuelstested.11 Ballard-Tremeer et al. used the hood method to compareefficiencies and emissions of five rural, wood-burning cookingdevices and observed that the average emissions of TSP werelowest for the improved open fire and the two-pot ceramicstove.12 Venkataraman et al. adopted the hood method and adilution sampler to measure CO, size-resolved aerosols, andPAH emissions from biofuel combustion in India.13,14However,studies of the emission characteristics of particulate matter from

* Author to whom correspondence should be addressed. Telephone:+86-10-62782195. Fax: +86-10-62773650. E-mail address:hjm-den@tsinghua.edu.cn.

(1) Ministry of Agriculture.P. R.C. China agriculture statistical report2000; China Agriculture Press: Beijing, China, 2001.

(2) Department of Industry and Transport Statistics, National Bureau ofStatistics, P. R. C. and Energy Bureau, National Development and ReformCommission.P. R. C. China energy statistical yearbook 2004. ChinaStatistics Press: Beijing, China, 2005.

(3) Bruce, N.; Neufeld, L.; Boy, E.; West, C.Int. J. Epidemiol.1998,27, 454-458.

(4) Guo, H.; Wang, T.; Simpson, I. J.; Blake, D. R.; Yu, X. M.; Kwok,Y. H.; Li, Y. S. Atmos. EnViron. 2004, 38, 4551-4560.

(5) Jacobson, M.Nature2001, 409, 695-697.(6) Menon, S.; Hansen, J.; Nazarenko, L.; Luo, Y.Science2002, 297,

2250-2253.

(7) Street, D. G.; Gupta, S.; Waldhoff, S. T.; Wang, M. Q.; Bond, T. C.;Bo, Y. Atmos. EnViron. 2001, 35, 4281-4296.

(8) Menzie, C. A.; Potoki, B. B.; Santodonato, J.EnViron. Sci. Technol.1992, 26, 1278-1284.

(9) Pedersen, D. U.; Durant, J. L.; Taghizadeh, K.; Hemond, H. F.;Lafleur, A. L.; Cass, G. R.EnViron. Sci. Technol.2005, 39, 9547-9560.

(10) Mitra, A. P.; Morawska, L.; Sharma, C.; Zhang, J.Chemosphere2002, 49, 903-922.

(11) Joshi, V.; Venkataraman, C.; Ahuja, D. R.EnViron. Manage.1989,13, 763-772.

(12) Ballard-Tremeer, G.; Jawurek, H.Biomass Bioenerg.1996, 11, 419-430.

(13) Venkataraman, C.; Rao, G. U. M.EnViron. Sci. Technol.2001, 35,2100-2107.

(14) Venkataraman, C.; Negi, G.; Sardar, S. B.; Rastogi, R. J.AerosolSci.2003, 33, 503-518.

845Energy & Fuels2007,21, 845-851

10.1021/ef060150g CCC: $37.00 © 2007 American Chemical SocietyPublished on Web 03/07/2007

rural household biofuel combustion in China are limited. Smithet al. studied greenhouse gas and other airborne pollutions(including TSP) emissions from household stoves in China andother developing countries using the carbon balance approach.15-17

The purpose of this paper is to provide an initial assessment ofparticulate matter emissions from typical rural householdbiofuels and stove combinations used in China. Therefore, theobjectives of this study were to (a) measure particle numberconcentration, size distribution, and mass size distribution; (b)quantify PM2.5 emission factors, and (c) relate these to stoveconfiguration, fuel, and combustion conditions.

Methods

Selection of Biofuels and Stoves to Test.In China’s ruralhouseholds, crop waste which is used as fuel to burn in thehousehold stoves for cooking and heating include the following:rice straw, wheat residue, maize residue, bean straw, cotton stalkand kaoliang stalk, and so on. Fuel wood and branches are alsooften used as rural household fuel. Improved stoves have becomepopular in rural households. These are stoves with an enclosedcombustion chamber and a flue. About 190 million improved stoveshave been installed, accounting for 80% of rural households.1 Inthe northern rural areas, stove/kang is used commonly for cookingand heating. Kang is a bed made of brick or clay, with passagewaysinside it, and flue gas, after leaving the stove, passes through thepassageways and transfers heat to the kang.

Three Chinese municipalities/provinces (Beijing, Chongqing, andHenan), located in the north, southwest, and the middle of China,respectively, were chosen for field measurement. Tests were carriedout in local rural households. Before determining the fuel/stovecombination to be tested, investigations were carried out. Theselected fuels and stove types are representative of the local ruralarea. In Chongqing, an improved stove and five typical biofuel

types, rice straw, maize residue, bean straw, fuel wood, andbranches, were selected for this study. The stove has two combus-tion chambers, two fuel inlets, two main pots, one auxiliary pot,and a flue. One combustion chamber burns biofuel, and the otherburns coal. The former has no bottom grate. Fuel is burned in thecombustion chambers, and heat is transferred to the main pot whilethe flue gas is drawn through the duct to the flue. The auxiliarypot is located in the duct and uses the remaining heat energy offlue gas. Fuel could be burned in both combustion chamberssimultaneously or only in one chamber. When one chamber is used,the duct of the other chamber connecting it to the flue is closed.This study is conducted only with the combustion chamber thatburns biofuel. In Henan, three different biofuels typical of the area,wheat residue, kaoliang stalk, and cotton stalk, were chosen to burnin a local typical improved stove. The stove has almost the sameconfiguration as stoves in Chongqing. Both combustion chambersburn biofuel and have a bottom grate. The experiment was carriedout in one combustion chamber. In Beijing, branches and a stove/kang were selected as the fuel/stove combination. The stove hasone combustion chamber, one fuel inlet, and one main pot.

All the selected stoves were made of brick, and their configu-ration parameters are shown in Table 1. The proximate and ultimateanalyses of the tested fuels are listed in Table 2. According to localcooking habits, the tested crop wastes were wrapped into batchesof 20-25 cm in length and woody fuel was cut into pieces about20 cm long.

Definition of the Burning Cycle. Cooking is not a steadyprocess causing emissions from the biofuel combustion to varyduring the cooking.16 Therefore, it is necessary to choose a burningcycle similar to the common cooking practice in the field. The mostcommon cooking practices include high power and low powerphases. High power phase means heating a quantitative amount ofwater from the ambient water temperature to the boiling temperatureas rapidly as possible. A low power phase would involve the watersimmering at the lowest power. The “water-boiling test”18 wasadopted with slight modification according to a Chinese standardmethod19 to define a burning cycle. The water-boiling test isdescribed as follows:

(15) Zhang, J.; Smith, K. R.; Ma, Y.; Ye, S.; Jiang, F.; Qi, W.; Liu, P.;Khalil, M. A. K.; Rasmussen, R. A.; Thorneloe, S. A.Atmos. EnViron.2000, 34, 4537-4549.

(16) Smith, K. R.; Khalil, M. A. K.; Rasmussen, R. A.; Thorneloe, S.A.; Manegdeg, F.; Apte, M.Chemosphere1993, 26, 479-505.

(17) Smith, K. R.; Uma, R.; Kishore, V. V. N.; Lata, K.; Joshi, V.; Zhang,J.; Rasmussen, R. A.; Khalil, M. A. K.Greenhouse gases from small-scalecombustion deVices in deVeloping countries, phase IIa: household stoVesin India; Office of Research and Development, US EPA: Washington, DC,1999.

(18) VITA (Volunteers in Technical Assistance, Inc.).Testing theefficiency of wood-burning cookstoVes; International Standards: Arlington,VA, 1985.

(19) Standardization Administration of the People’s Republic of China(SAC). Testing method for the heat characteristics of firewood stoVes;SAC: Beijing, China, 1984.

Table 1. Stove Configuration Parameters

location description

combustchamber

volume (m3)

grate topot-bottomdistance (m)

grateair-inletarea (m2)

fuelinlet

area (m2) fuel used

Chongqing improved stove, one main potand one auxiliary pot, no grate

0.060 0.23 no grate 0.045 rice straw, maize residue,bean straw, fuel wood branch

Henan improved stove, one main pot andone auxiliary pot, bottom grate

0.034 0.13 0.016 0.041 wheat residue, kaoliang stalk,cotton stalk

Beijing stove/kang, one pot, no grate 0.039 0.20 no grate 0.042 branch

Table 2. Tested Fuels Proximate and Ultimate Analysis

fuelrice

strawmaizeresidue

beanstraw

fuelwood branch (I)

cottonstalk

wheatresidue

kaoliangstalk branch (II)

Proximate (as Received, mass %)moisture 6.24 6.49 10.07 7.51 7.90 8.40 5.42 5.47 7.06volatile matter 62.50 70.33 71.35 76.14 73.16 70.83 68.74 72.27 75.33fixed carbon 16.56 16.87 15.59 14.41 17.14 17.54 18.07 17.98 16.71ash 14.70 6.31 2.99 1.94 1.80 3.23 7.77 4.28 0.90

Ultimate (Dry Basis, mass %)C 41.42 46.32 47.54 50.68 48.96 49.50 44.64 47.76 49.90H 5.57 6.09 6.35 6.37 6.23 6.28 5.85 6.12 6.36N 0.94 1.62 1.15 0.29 0.76 1.25 0.56 0.59 0.70S 0.16 0.15 0.16 0.018 0.053 0.010 0.20 0.078 0.037O (by difference) 39.43 39.07 41.48 40.54 42.05 39.43 40.53 40.92 42.03low heating value(dry basis, MJ/kg)

15.67 17.54 17.86 19.11 18.21 18.71 16.61 17.89 18.69

846 Energy & Fuels, Vol. 21, No. 2, 2007 Li et al.

• Fill the stove pot (including the main and auxiliary pots) to2/3of its capacity with room temperature water.

• Weigh out a quantity of biofuel such that its weight is about1/3-1/2 of the water weight in the main pot.

• Cover the pot with a lid and a thermometer is inserted throughthe lid into the water.

• Start the fire at high power to bring the water in the main potto boil.

• Continue the test at low power using the remaining biofuel.• Terminate the test when the water temperature in the main pot

dropped by about 1°C.• Bail out the water in the pot and weigh it.• Extinguish the ash and any unburned fuel residue by shutting

off access to air, and weigh them after they are cool.Prior to the planned sampling for each fuel/stove combination,

trial tests were performed to standardize the burning cycle.Preliminary tests were carried out until a satisfactory methodprecision (relative standard deviation less than 20%) for the mainparameters, i.e., time to boil, time of burning cycle, and thermalefficiency, was obtained. The time for completion of the burningcycle ranged from 30-50 min for most of the test.

During the whole burning cycle, biofuels were manually fed intothe stove in batches. The time, temperature, and weight of waterwere recorded at the beginning and end of the burning cycle, andthe amount of biofuel burned was also recorded. Stove thermalefficiency can be determined according to these data. The samplingperiod covered the whole burning cycle from the moment of thebeginning of the burn (biofuel had been ignited) to the end of theburning process (the water temperature in the main pot droppedby about 1°C).

Sampling Approach. An outline of the sampling system isshown in Figure 1. It includes a dilution sampling system, particlemeasurement, and flue gas monitoring. The system is described inmore detail in the following subsection.

Dilution Sampling System.The dilution sampling system simu-lates the cooling and dilution processes after the hot flue gas leavesthe stack and is widely used for characterizing emissions from

stationary combustion sources.20 A compact dilution samplingsystem was developed for this field study that consists of four mainparts: sampling inlet, dilution part, residence chamber, and sampler.

In the sampling inlet part, flue gas is withdrawn from the flue,put through a cyclone separator, and then advanced to the firstdilution. The cyclone, which removes particles larger than 10µm,is installed outside the flue because the flue is not large enough toaccommodate it. The sampling inlet was heated to about 150°C toreduce particle thermophoresis losses. Because the flue gas velocityin the flue is unstable, isokinetic sampling cannot be achieved. Thesampling error arising from the velocity mismatch can be neglectedfor particles with diameters less than 2.5µm.21

The dilution part consists of two stage diluters. The operationprinciple of the first diluter is based on an ejection type dilution(Dekati Ltd, Finland). The ejector diluter is used to keep the dilutionratio constant at about 10. About 0-50 L/min can be drawn fromthe outlet of the diluter to the second diluter according to researchrequirements. The second diluter is a cylindrical enclosure with aperforated plate inside. The sample flow from the first diluter isintroduced inside the enclosure, and the dilution air is forced throughthe apertures of the plate into the enclosure where it mixes withthe sample flow. Two vortex flow meters record sample flow fromthe first diluter and the dilution air flow rate in the second diluter.The second diluter can supply a dilution ratio from 1 to 10. Thetotal range of the dilution ratio is from 10 to 100. An oil-free aircompressor and a pump supply the first and second diluters dilutionair, respectively; the air must be purified before entering into thedilutor.

All of the diluted sampling gas is transferred to the residencechamber. The temperature, relative humidity, and pressure in thechamber are monitored.

The sampler is attached to the end of the residence chamber andhas eight sampling ports for connecting with particle measurementinstruments. In this study, an electrical low-pressure impactor,(ELPI, Dekati Ltd., Finland)22 and three parallel PM2.5 cycloneswith filter packs were used to collect particles. The gas in thechamber is under a small positive pressure, and extra gas can beautomatically discharged from the unused sampling ports. Thewhole dilution sampling system has shown considerable stability.

The sampler is made entirely from stainless steel, copper, andTeflon. A dilution air ratio of about 20 and an aging time of about80 s were applied in the study.

Particle Measurement.ELPI was used to measure in the real-time particle number concentration and size distribution. Operatingat a flow rate of 9.89 L/min, the ELPI has 50% cut-pointaerodynamic diameters of 0.028, 0.056, 0.095, 0.157, 0.263, 0.382,0.613, 0.948, 1.600, 2.390, 4.000, 6.680, and 9.920µm on stages1-13. Our focus was on the lower ELPI stages 1-9 whichcharacterized the particle size distribution range of 0.028-2.390µm, the typical particle distribution range of biofuel combus-tion.13,23-25 Greased and baked aluminum foils of 25 mm in diameterwere used as collection foils.

A low pressure impactor (LPIswhen ELPI is used without theelectrical charger) was used to determine particle mass sizedistribution over the range of 0.028-2.390µm. In the upper fourstages, aluminum foils were coated with high vacuum grease toprevent particle bounce. For further size resolved chemical analysis,the lower nine stages were not greased to avoid interference. Particlebounce within the lower impactor stages was not a problem.24

(20) Hidemann, L. M.; Cass, G. R.; Markowski, G. R.Aerosol Sci. Tech.1989, 10, 193-204.

(21) Hinds, W.C.Aerosol technology: properties, behaVior, and mea-surement of airborne particles; John Wiley & Sons, Inc.: New York, 1999.

(22) Keskinen, J.; Pietarinen, K.; Lehtima¨ki, M. J. Aerosol Sci.1992,23, 353-360.

(23) Purvis, C. R.; McCrillis, R. C.; Kariher, P. H.EnViron. Sci. Technol.2000, 34, 1653-1658.

(24) Kleeman, M. J.; Schauer, J. J.; Cass, G. R.EnViron. Sci. Technol.1999, 33, 3516-3523.

(25) Hays, M. D.; Smith, N. D.; Kinsey, J.; Dong, Y.; Kariher, P. J.Aerosol Sci.2003, 34, 1061-1084.

Figure 1. Outline of the sampling system.

Emission Characteristics Household Biofuel Energy & Fuels, Vol. 21, No. 2, 2007847

Before and after sample collection, all substrates were conditionedfor 24 h at about 40% RH and 25°C in an air-conditioned roomand weighed on a microbalance with a resolution 1µg.

All number and mass concentrations measured by ELPI and LPIwere back-calculated out to the dilution air ratio at each measure-ment and normalized to 3% CO2 dry gas at normal temperature (0°C) and pressure (101.3 kPa).

PM2.5 was also collected by three parallel PM2.5 cyclones withfilter packs operated at 16.7 L/min. The first filter pack consistedof a 47 mm Teflon-membrane filter for mass by gravimetric andelemental analysis. The other two filter packs consisted of a 47mm quartz-fiber filter for carbon, ions, and speciated organiccompound analysis. The Teflon-membrane filters were conditionedfor 24 h at about 40% RH and 25°C in an air-conditioned roomand weighed on a microbalance with a resolution 10µg. Thechemical speciation results will be presented in a companion paper.

Flue Gas Monitoring.The concentrations of gaseous air pollut-ants (including CO2, CO, SO2, O2, and NOx) and temperature inthe flue were continuously monitored by a flue gas analyzer (ModelKM9106, Keison). The data were recorded every 10 s by a datalogger and transmitted to a notebook PC. The instrument wascalibrated before each field study.

Determination of Emission Factors. The carbon balanceapproach was used to calculate the emission factors.15 The carbonbalance equation for combustion is based on the total carbon massburned being equal to the total mass of carbon emitted by both asparticles and gases. The carbon content of the tested fuels, ash,unburned fuel residue, and PM2.5 was analyzed. The averageconcentrations of CO2 and CO over the whole burning cycle in theflue were calculated using the data monitored by the flue gasanalyzer. The amount of total hydrocarbon was not measured inthis study. Estimation based on previous study15 shows that errorcaused by neglecting total hydrocarbon is less than 5%stherefore,the emission factors in this study are credible.

The emission factors are reported on a dry fuel mass basis (gramper kilogram of fuel) and on an energy basis (gram per megajouleof delivered energy). Dry fuel mass can be converted from fuelmass burned and fuel moisture content. The emission factor based

on delivered energy can be derived from the emission factor basedon dry fuel along with the stove thermal efficiency and fuel heatingvalue. For the stove-burned woody fuel, the unburned fuel residue,i.e., char, is often used later and produces pollutants. In our study,the pollutants generated from char secondary combustion are notcalculated in determining the emission factors.

In this paper, thermal efficiency is the ratio of energy absorbedby the water in the pots (including the main pot and the auxiliarypot) to the energy content of the fuel consumed. For the stove-burned woody fuel, when calculating its thermal efficiency, theenergy content of the fuel consumed was subtracted by the energycontent of char. For stove/kang, thermal efficiency just refers tothe stove; the heat delivered to the kang is not considered.

Results and Discussion

Particle Number Concentration and Size Distribution.Particle number concentrations for all tested biofuel combustionduring the whole burning cycle, measured by the ELPI, togetherwith geometric mean diameters (GMDs) are summarized inTable 3. The average particle number concentrations within therange of 0.03-2.39 µm were between 1.0× 107-5.0 × 107

particles/(N cm3) and GMDs were between 0.11 and 0.21µm.The particle number concentration varied 1-2 orders ofmagnitude in the whole burning cycle. The results of the particlenumber concentration were somewhat consistent with someprevious studies. For example, Hueglin et al.26 reported particlenumber concentrations from residential wood stoves varyingfrom 7.8× 106-4.4× 107 particles/(N cm3), and Johansson etal.27 showed particle number concentrations from domesticheating devices varying between 1.4× 107-13.4 × 107

particles/(N cm3).In this study, the number concentrations in the Chongqing

stove are generally less than those measured in Henan. Apossible explanation is the difference in stove configuration.The stove in Chongqing has the largest combustion volume,

Table 3. Particle Number Concentrations and GSD during the Whole Burning Cycle

stove Chongqing Henan Beijing

fuel rice straw maize residue bean straw fuel wood branch (I) cotton stalk wheat residue kaoliang stalk branch (II)

avg number concentration(particles/(N cm3))

1.2× 107 2.9× 107 1.6× 107 1.3× 107 1.9× 107 3.3× 107 4.6× 107 5.0× 107 1.0× 107

varied range of numberconcentration

(particles/(N cm3))

1.1× 106 to4.6× 107

1.6× 106 to2.5× 108

1.3× 106 to7.0× 107

2.5× 106 to1.5× 108

2.1× 106 to4.3× 108

3.3× 106 to2.6× 108

5.4× 106 to1.0× 108

1.2× 107 to1.3× 108

1.2× 106 to7.1× 107

GMD (µm) 0.11 0.14 0.17 0.15 0.12 0.13 0.16 0.11 0.21

Figure 2. Particle average number size distribution during the whole burning cycle. (Note: SC, SH, and SB denote the stoves in Chongqing,Henan, and Beijing, respectively.)

848 Energy & Fuels, Vol. 21, No. 2, 2007 Li et al.

which can provide a relatively long time for the combustion ofvolatiles released from pyrolysis of biofuel and, therefore,decreased particle formation from incomplete combustion. TheBeijing stove/kang has the least number concentration and thehighest GMD, which may be related to its peculiar configuration.Before being drawn to the flue, flue gas, after leaving the stove,passes through the passageways inside the kang and transfersheat to the kang thereby losing temperature. Flue gas residencetime in the kang is over 100 ssenough time for particlecoagulation and condensation growth that could decrease particlenumber concentration and shift the GMD toward a high value.

Particle number size distributions are given in Figure 2. Theyare mainly unimodal during the whole burning cycle, with apeak between 0.12 and 0.32µm. However, they are bimodalfor branches burned in the Chongqing stove and wheat residueand kaoliang stalk in the Henan stove, with a nucleation modepeak less than 0.04µm and an accumulation mode peak near0.12 µm.

Particle number concentrations and size distributions duringthe whole burning cycle were strongly related to the phase ofthe cooking practice. Typical particle number concentrationsand size distributions during the different phases of the burningcycle are shown in Figures 3 and 4, respectively.

In this study, the lower power phase was divided into twoparts: the earlier portion of the lower power phase (lower powerphase I) and the later portion (lower power phase II). In lowerpower phaseΙΙ, no fuel was fed and combustion occurs in theburn-out phase. This causes particle emissions to be significantlydifferent from that of lower power phase I. Particle numberconcentrations were 4.8× 107 particles/(N cm3) for the highpower phase, 5.8× 106 particles/(N cm3) for lower power phaseΙ, and 3.6× 106 particles/(N cm3) for lower power phaseΙΙ.Total particle number concentrations decreased significantly

along with the cooking procedure, which is correlated with therate of fuel burning and combustion condition. From Figure 4,it can be observed that a prominent mode was less than 0.04µm, other than an accumulation mode peak near 0.20µm inlower power phaseΠ, which suggests that particle numberemissions were largely in the nucleation mode. This observationmay be attributed to the lower particle number concentrationduring lower power phaseΙΙ, which limited the condensationand coagulation process and, hence, particle growth.28,29

Although particle number concentrations and size distributionsvaried to a certain extent in the repeated runs, a general trendwas found. The results of three repeated measurements ofparticle emissions from stoves in Beijing were shown in Figure5, and these indicated that the test data were satisfactorilyreproducible. Moreover, the relative standard deviation of theparticle average number concentration was less than 11%.

(26) Hueglin, Ch.; Gaegauf, Ch.; Ku¨nzel, S.; Burtscher, H.EnViron. Sci.Technol.1997, 31, 3439-3447.

(27) Johansson, L. S.; Tullin, C.; Leckner, B.; Sjo¨vall, P. BiomassBioenerg.2003, 25, 435-436.

(28) Lipsky, E.; Stanier, C. O.; Pandis, S. P.; Robinson, A. L.EnergyFuels2000, 16, 302-310.

(29) Chang, M. C. O.; Chow, J. C.; Watson, J. G.; Hopke, P. K.; Yi, S.M.; England, G. C.J. Air Waste Manage. Assoc.2004, 54, 1494-1505.

Figure 3. Particle number concentration during the different phasesof the burning cycle.

Figure 4. Particle average number size distribution during the differentphases of the burning cycle.

Figure 5. Reproducibility of test data.

Figure 6. Particle mass size distribution of particles emitted from (a)crop waste and (b) woody fuels combustion. (Note: SC, SH, and SBdenote the stoves in Chongqing, Henan, and Beijing , respectively.)

Emission Characteristics Household Biofuel Energy & Fuels, Vol. 21, No. 2, 2007849

Particle Mass Size Distribution.Particle mass size distribu-tions for all tested biofuel combustion during the whole burningcycle, measured by LPI, are given in Figure 6. Particle masssize distributions from crop waste combustion show a singlemode with the peak at approximately 0.20-0.48µm. However,woody fuels show a bimodal size distribution; one prominentmode peaks between 0.12 and 0.32µm, and the other weakmode peaks at 0.76µm. It is supposed that soot contributes tothe mode at 0.76µm. A possible explanation is connected tothe fuel itself. Woody fuel has a higher lignin concentrationcompared to the crop wastes. It was thought that a high ligninconcentration could increase the soot yield.30 Unoxidized sootmay have undergone agglomeration and then growth to arelatively large size. This needs to be further investigated. Theparticle mass median aerodynamic diameter (MMD) is between0.21 and 0.45µm. For all the biofuel combustion cases,submicrometer particles (less than 1µm) contributed to over90% of the mass of PM in the range of 0.03-2.39 µm.

PM2.5 Emission.The emission factors of PM2.5 from ruralhousehold biofuel combustion are between 1.80 and 7.00 g/kgof biofuel (dry basis) and 0.84-2.40 g/MJ delivered energy;the average emission factor is 4.21 g/kg of biofuel (dry basis)and 1.46 g/MJ delivery energy for all of the biofuels tested(Table 4). The results are consistent with previous studies.13,31-33

The stove thermal efficiencies range from 13.4% to 22.1%.The data are in accord with those measured from Chinese stovesthat burned biofuel in the laboratory.15 Within that range, stoveswith different configurations have various thermal efficien-cies: 13.4-16.1% for stoves in Chongqing, 18.0-22.1% forstoves in Henan, and 14.8% for stoves in Beijing (Table 4).The thermal efficiencies of the Henan stoves are generallygreater than the others. The stove in Henan has the leastcombustion chamber volume, a grate, and the lowest distancefrom the grate to the pot-bottom. All of those design featurescontribute to high thermal efficiencies. However, the Henanconfiguration can cause high emissions, described in thefollowing sections.

The mean emission factor of CO is between 29.3 and 134.0g/kg of biofuel (dry basis) and 7.9-57.3 g/MJ delivered energy,and the averaged value is 76.0 g/kg of biofuel (dry basis) and

29.4 g/MJ delivered energy over all the biofuel types tested(Table 4). The results are also consistent with other re-search.11-13,15,32,34

PM2.5 emission factors in Henan are higher than thosemeasured in Beijing and Chongqing. This may be attributed todifferent stove configurations. As mentioned above, the stovein Henan has a small combustion volume, which leads to highthermal efficiency, but the small volume may cause incompletecombustion. Incomplete combustion of emitted organic mattersignificantly enhances the formation of particles. The air supplythrough the grate increases the temperature in the fuel bed and,thereby, enhances vaporization of ash that would result in highparticle emissions.35,36Ash entrainment in the fuel bed escapesto the flue gas as air flows through the grate. On the basis ofour findings, an increased thermal efficiency stove does notimply reduced PM emissions (Figure 7). Other researchers alsoreported this phenomenon.11,13

Figure 8 shows the relationship between PM2.5 emissionfactors on an energy basis and stove thermal efficiency, whichimplies that the increase in thermal efficiency cannot offset theincrease in particle emissions.(30) Wiinikka, H.; Gebart, R.Combust. Sci. Technol.2005, 177, 741-

763.(31) McDonald, J. D.; Zielinska, B.; Fujita, E. M.; Sagebiel, J. C.; Chow,

J. C. ; Waston, J. G.EnViron. Sci. Technol.2000, 34, 2080-2091.(32) Sheesley, R. J.; Schauer, J. J.; Chowdhury, Z.; Cass, G. R.; Simoneit,

E. R. T. Characterization of organic aerosols emitted from the combustionof biomass indigenous to South Asia.J. Geophys. Res.2003, 108 (D9),4285; doi:10.109/2002JD002981.

(33) Fine, P. M.; Cass, G. R.; Simoneit, E. R. T.EnViron. Sci. Technol.2001, 35, 2665-2675.

(34) Purious, C. R.; Mccrillis, R. C.; Kariher, P. H.EnViron. Sci. Technol.2000, 34, 1653-1658.

(35) Jensen, P. A.; Frandsen, F. J.; Dam-Johansen, K.; Sander, B.EnergyFuels2000, 14, 1280-1285.

(36) Wiinikka, H. High temperature aerosol formation and emissionminimisation during combustion of wood pellets. Ph.D. Thesis, TechnicalUniversity of Denmark, Lyngby, Denmark, 2005.

Table 4. Thermal Efficiencies and Emission Factors of Biofuel Combustion

PM2.5 emission factors CO emission factors

stove fuel thermal efficiency (%) Em (g/kg)a Ee(g/MJ) Em (g/kg) Ee(g/MJ)

Chongqing rice straw 13.7( 1.1b 1.66-1.94c 0.79-0.90 108.9-128.6 51.9-59.5maize residue 13.4( 1.0 2.45-3.85 1.03-1.68 111.5-156.5 48.5-66.1bean straw 14.3( 0.9 3.28( 0.87 1.30( 0.35 77.0( 13.9 30.5( 5.5fuel wood 16.1( 1.1 2.21-4.58 0.79-1.53 47.2-65.8 16.9-21.9branch (I) 13.8( 1.1 2.95-3.97 1.17-1.58 62.7-93.3 24.8-37.1

Henan cotton stalk 18.7( 0.9 6.04( 0.52 1.75( 0.11 68.4( 23.8 19.7( 6.4wheat residue 18.0( 0.9 5.61-8.39 1.86-2.94 62.5-68.7 21.9-22.7kaoliang stalk 22.1( 0.7 6.27-7.19 1.63-2.02 27.9-30.7 7.3-8.6

Beijing branch (II) 14.8( 2.8d 3.04( 0.85 1.11( 0.37 56.4( 22.6 20.4( 8.5

a Dry fuel mass basis.b Three or more repeated tests, the results are given as means( standard deviations (x ( s). c Two repeated tests, the results aregiven as a range.d Just stove thermal efficiency, kang thermal efficiency is not included.

Figure 7. Relationship between stove thermal efficiency and PM2.5emission factor on fuel mass basis. (Note: SC, SH, and SB denote thestoves in Chongqing, Henan, and Beijing, respectively.)

850 Energy & Fuels, Vol. 21, No. 2, 2007 Li et al.

CO is a product of incomplete combustion. For low temper-ature combustion in the household stove, organic matter emittedfrom incomplete combustion enhances the formation of particles.Gupta et al. found that CO emissions are correlated to those ofrespirable suspended particulates (RSP).37 However, in ourstudies, a negative correlation between particle and CO emis-sions was observed (Figure 9). Venkataraman et al. also foundopposing trends in PM and CO emissions in their study.38 Theypointed out that CO emissions cannot be used as a surrogatefor PM emissions across stoves and fuels. This would indicatethat PM formation is complicated and needs to be studiedfurther.

Conclusions

The total average PM number concentrations for biofuelcombustion during the whole burning cycle was between 1.0× 107 and 5.0× 107 particles/(N cm3). Particle number sizedistributions are mainly unimodal, with a peak between 0.12and 0.32µm. In some cases, bimodal size distributions are alsoobserved, with a nucleation mode peak less than 0.04µm and

an accumulation mode peak near 0.12µm. Particle mass sizedistributions from crop waste combustion show an obvioussingle mode peak at 0.20-0.48µm; however, woody fuels showa bimodal size distribution: one prominent mode peaks between0.12 and 0.32µm, and the other weak mode peaks at 0.76µm.Both with regard to number and mass, particle emissions frombiofuel combustion were dominated by submicrometer particles,which imply adverse health concerns.

The emission factor of PM2.5 from biofuel combustion isbetween 1.80 and 7.00 g/kg of biofuel (dry basis) and 0.84-2.40 g/MJ delivered energy. The averages were 4.21 g/kg ofbiofuel (dry basis) and 1.46 g/MJ delivered energy.

For household stoves, advanced gas cleaning devices are notan economic option. The feasible way to reduce particleemissions is, therefore, to decrease the formation of particlesin the combustion process. In this study, it appears that particleemissions are correlated with combustion conditions and stoveconfiguration. Particle emissions are highest during the highpower phase. A possible way to reduce emissions in this phaseis by feeding the fuel at a moderate pace. Stove configurationimpacts both thermal efficiency and emissions. As noted, adilemma arises in that the more thermally efficient stove hashigher per kilogram fuel PM emissions and the increase inthermal efficiency cannot offset the increase in particle emis-sions. It is a stimulus for developing a stove with high efficiencyand low emissions.

Acknowledgment. The authors acknowledge the financialsupport provided by the National Key Basic Research and Develop-ment Program of China (Grant No. 2002CB211600) and theNational Natural Science Foundation of China (Grant No.20521140077).

EF060150G

(37) Gupta, S.; Saksena, S.; Shankar, V. R.; Joshi, V.Biomass Bioenerg.1998, 14, 547-559.

(38) Venkataraman, C.; Joshi, P.; Sethi, V.; Kohli, S.; Ravi, M. R.AerosolSci. Tech.2004, 38, 50-61.

Figure 8. Relationship between stove thermal efficiency and PM2.5emission factor on energy basis. (Note: SC, SH, and SB denote thestoves in Chongqing, Henan, and Beijing, respectively.)

Figure 9. Relationship between CO and PM2.5 emission factor.

Emission Characteristics Household Biofuel Energy & Fuels, Vol. 21, No. 2, 2007851