Emission Characteristics of Gas-Fired Boilers in Beijing City, China: Category- Specific Emission Factor, Emission Inventory, and Spatial Characteristics

Gas-fired boilers are the main stationary sources of NOx in Beijing. However, the understanding of gas-fired boilers is limited. In the present study, the emission characteristics of NOx, SO2, and CO from gas-fired boilers in Beijing were established using category-specific emission factors (EFs) from field measurements. To obtain category-specific EFs, boilers were classified through influence analysis. Factors such as combustion mode, boiler type, and installed capacity were considered critical for establishing EFs because they play significant roles in pollutant formation. The EFs for NOx, CO, and SO2 ranged from 1.42–6.86 g m, 0.05–0.67 g m and 0.03–0.48 g m. The emissions of NOx, SO2, and CO for gas-fired boilers in Beijing were 11121 t, 468 t, and 222 t in 2014, respectively. The emissions were spatially allocated into grid cells with a resolution of 1 km × 1 km, and the results indicated that top emitters were in central Beijing. The uncertainties were quantified using a Monte Carlo simulation. The results indicated high uncertainties in CO (–157% to 154%) and SO2 (–127% to 182%) emissions, and relatively low uncertainties (–34% to 34%) in NOx emission. Furthermore, approximately 61.2% and 96.8% of the monitored chamber combustion boilers (CCBs) met the standard limits for NOx and SO2, respectively. Concerning NOx, low-NOx burners and NOx emission control measures are urgently needed for implementing of stricter standards. Adopting terminal control measures is unnecessary for SO2, although its concentration occasionally exceeds standard limits, because reduction of its concentration can be achieved thorough control of the sulfur content of natural gas at a stable low level. Furthermore, the atmospheric combustion boilers (ACBs) should be substituted with CCBs, because ACBs have a higher emission despite lower gross installed capacity. The results of this study will enable in understanding and controlling emissions from gas-fired boilers in Beijing. Keyword: Classification of boilers; Uncertainty analysis; GIS-based approach; Operating load; Monte Carlo simulation.


INTRODUCTION
Beijing, the capital city of China, experiences severe air pollution, making it rank among the most polluted megacities globally (Gurjar et al., 2008;Wu et al., 2016;Zhang et al., 2016).A series of air pollution control measures have been implemented to improve air quality.Among these measures, the adjustment of energy structure was the most crucial and effective.Hence, the demand of natural gas has increased drastically and is likely to increase rapidly in the future because it is an environmental friendly and efficient fuel.According to the planning for natural gas consumption from the Beijing Gas Group Co. Ltd, the demand for natural gas will grow from 7 billion m 3 in 2011 to 20 billion m 3 in 2020.Among the end-users, the natural-gas-fired boiler has contributed substantially to the growth in demand.
However, the combustion of natural gas produces NO x , which is the major contributor to PM 2.5 pollution due to the increase of [NO 3 -/SO 4

2-
] ratio (1.05) in Beijing according to the source apportionment published by the Beijing Municipal Environmental Protection Bureau (BMEPB) (BMEPB, 2014).In addition, NO x acts as a precursor of tropospheric ozone (O 3 ) and fine particulate matter (PM 2.5 ) in the atmosphere, NO x not only poses a serious threat to humans and the environment but also causes severe local or regional air quality deterioration; therefore, it has become the priority pollutant to be monitored and controlled in Beijing (Jaffe et al., 2004;Pope III, 2004;He et al., 2011;Simmons and Seakins, 2012).As stated previously, the gas-fired boiler is the main stationary source of NO x in Beijing.Therefore, the monitoring and control of NO x should receive attention in the future, for improving air quality.However, limited studies are available on the emission characteristics of gasfired boiler (Pulles et al., 2004).The emission characteristics of gas-fired boilers in the Netherlands and Asia were established by Pulles et al. (2004) through field measurement and by Kato et al. (1992) through theoretical analysis, respectively.Regarding the emission characteristics of gasfired boilers in China, studies have focused on the emission of NO x (Yao et al., 2009;Xue et al., 2014).Yao et al. (2009) established a NO x emission factor based on the field measurements of 64 gas-fired boilers in China.Xue et al. (2014) compiled a NO x emission inventory in Beijing through field measurements of five gas-fired boilers.However, no set of systematic and integrated emission factors (EFs) for various species compiled for different types of gas-fired boiler in Beijing is currently available.In the present study, the emission characteristics were obtained by establishing category-specific EFs of NO x , SO 2 , and CO based on field measurements for gas-fired boilers and the emission inventory, determining spatial characteristics with high resolution, and calculating uncertainty using a Monte Carlo simulation.To obtain the EFs, the gas-fired boilers were classified through influence analysis.

Field Measurements
To investigate emission characteristics of NO x , SO 2 , and CO from various types of boilers, 128 gas-fired boilers were selected.The following factors affecting NO x , SO 2 , and CO formation and emission were considered: operating load, combustion mode, boiler type, installed capacity, control measures, installation year, pressure of natural gas and control mode of burner.Combustion mode included atmospheric combustion (AC) and chamber combustion (CC), and the latter was subdivided into horizontal internal-combustion (HIC) and others according to boiler type.The boiler capacity of HIC ranged from 1 MW to 16 MW, which was divided into two grades, namely < 10 MW and ≥ 10 MW by using cluster analysis.The operating load was controlled in the range 15%-90% in this investigation to clarify its role in the pollutants formation.The range of installation year of the existing gas-fired boilers was 1999-2015.The pressure of natural gas ranged from 101.3 to 137kPa.The control mode of burner in gas-fired boiler included mechanical and electrical control modes.
A flue gas analyzer (testo 350, Testo, USA) was used to analyze the O 2 content, flue gas temperature, and gaseous pollutant emissions, including NO x , CO and SO 2 , at the outlet of the gas-fired boilers.Simultaneously, a pilot tube was applied to monitor flue gas velocity.For quality assurance and quality control, the gas analyzers were calibrated with zero gas and targeted standard gases (NO, NO 2 , CO, SO 2 and O 2 ) prior to the first test of the day.

Methods for Estimating Emission
Emissions from gas-fired boilers were established using the EF method.The emission from gas-fired boilers in individual categories was calculated separately by combining the category-specific EFs and the corresponding natural gas consumption.Then the aggregate was calculated to obtain emission at the district and province levels.The basic formula can be expressed as follows: where E i is the emissions of the specific pollutant i in 2014 (t); EF i is the EFs measured (g m -3 ), which is calculated based on Eq. (2) as described in section 2.2.1,A represents the natural gas consumption (m 3 ); i represents the type of pollutants, namely NO x , SO 2 , or CO; j represents the combustion mode; k represents the type of boiler; l represents the installed capacity of boiler; and m represents the specific district.In the present study, the emission was estimated without considering control efficiency, because of the absence of measures adopted to reduce NO x , SO 2 , or CO concentration as stated in section 3.1.5.

Determination of EFs
To establish the category-specific EFs, gas-fired boilers were classified through influence analysis.The EF of a specific category was calculated using Eq.(2).
where C ijkl is the average concentration of pollutant i corresponding to combustion mode j, boiler type k and installed capacity l (mg Nm -3 ), S is the feeding rate of natural gas (Nm 3 h -1 ), ν is the velocity of dry flue gas under standard condition (m s -1 ), and D is the diameter of chimney (m).

Activity Level and Data Source
Typically, the emission of boiler was treated as point source.However, obtaining the details of emission from each gas-fired boiler in Beijing was not possible.A database (called database of gas-fired boilers) containing detailed information (e.g., consumption of natural gas, installed capacity, boiler type, burner patterns, installation year, emission control devices and geographical location) of 10293 gas-fired boilers with a gross capacity of 50738 t h -1 has been established and the separate emissions for each boiler were calculated in the present study.These boilers accounted for more than 65.9% of the total boilers and 67.0% of the gross installed capacities in Beijing, which is assumed to be representative of various boilers.Except for the emissions from the point sources of gas-fired boilers, the other sources are all treated as "regional area source" in each district.Thus, the natural gas consumption of a "regional area source" was calculated by subtracting the natural gas consumption of the point source from that of industry and heating at the district level in Beijing.The natural gas consumption from industry and heating is available in a statistical yearbook (BSB, 2015).

Methods for Spatial Allocation
In this study, an approach based on a geographic information system was adopted for allocating the emission from gas-fired boilers in Beijing by using administrative districts data as a basal map.Considering the "regional area source" in each district was identified on the basis of previous statement.The allocation factor of "regional area source" for a grid cell is the area percentage of the grid cell in the total area of a district.The spatial distributions of gridded emission were determined by the integrated consequence of point source and regional area source in each grid cell.

Uncertainty Analysis
Quantitative uncertainty analysis on the total emission of various pollutants was conducted with Monte Carlo simulation using a MATLAB software package.Normal distribution with a coefficient of variation (CV, the standard deviation divided by the mean) of 30% was assumed for natural gas consumption.The variation of emission factors was assumed as normal distribution by using means and standard deviations based on the field measurement.The details of the Monte Carlo simulation are available in previous studies (Cullen and Frey, 1999;Frey and Zheng, 2002).

Classification of Gas-Fired Boilers Based on Influence Analysis
The formation of gaseous pollutants by boilers, which is closely associated with the flame temperature and O 2 content, are affected by the operating load, combustion mode, boiler type, installed capacity, installation year, pressure of natural gas, and control mode of burner.Furthermore, the control measure was a key factor determining gaseous pollutants emission.

Operating Load
Studies have reported that the operating load plays a significant role in NO x , SO 2 , and CO formation during combustion (Tanetsakunvatana and Kuprianov, 2007;Ling et al., 2014;Ma et al., 2016).Typically, NO x is produced mainly through three pathways including (a) the reaction of N 2 with O 2 , known as thermal-NO x , (b) the oxidation of fuel containing N, known as fuel-NO x , and (c) prompt-NO x (Flagan and Seinfeld, 1988;Li et al., 2011;Korpela et al., 2015;Ma et al., 2016).Considering that the contribution of prompt-NO x is less than 5% and the fuel-nitrogen included in natural gas is in elementary form (N 2 ), which behaves similarly as N 2 in atmospheric, the thermal-NO x is the dominant formation pathway.Therefore, chamber temperature and O 2 content are the key factors determining NO x formation during natural gas combustion (Korpela et al., 2015).However, detecting chamber temperature is difficult.In this article, we assumed that the variation in the trend of flue gas temperature is similar to that of chamber temperature for a specific boiler, and thus, is an indicator of NO x emissions.
To clarify the role of chamber temperature, the relationship between the outlet concentration and operating loads of the gas-fired boiler was investigated with capacities of 10.5 MW, 14 MW, and 16 MW, whereas the O 2 content was not varied considerably, as shown in Figs.1(a)-1(c).The NO x concentration increased as the operating load increased for all investigated capacities, which can be attributed to the increase in thermal-NO x formation at higher chamber temperatures.The effects of operating load on CO and SO 2 concentrations were negligible because their concentrations are maintained at a relatively low level for the investigated capacities, namely ≤ 2.5 mg m -3 .
By contrast, the concentrations of NO x , SO 2 , and CO were investigated when the O 2 content increased and decreased with an increase in the operating load, respectively, in order to clarify the role of O 2 content (as shown in Figs.1(d)-1(f)).The results depicted in Fig. 1(d) indicate that as the load increases, the chamber temperature and O 2 content increase, thereby resulting in a higher NO x concentration.By contrast, the NO x concentration either decreased or remained constant although the chamber temperature increased with increase in operating load, which was attributed to the decrease of O 2 content, as illustrated in Figs.1(e) and 1(f).Thus, it was concluded that the effects of the operating load on NO x formation were the integrated consequences of O 2 content and chamber temperature.
Considering that the characteristics of NO x differed obviously among operating loads, operating loads above 70% were used in the following studies because they reflect the actual situation of emission levels of gaseous pollutants.

Combustion Mode
In terms of combustion mode, the boilers were classified into atmospheric combustion boilers (ACBs) and chamber combustion boilers (CCBs), and the latter comprised a majority, accounting for 86.46% of gas-fired boilers from the database of gas-fired boilers.The emissions of SO 2 , CO and NO x from ACBs and CCBs are depicted in Fig. 2. The NO x emissions of two CCBs were significantly higher than those of the other CCBs (143 mg m -3 on average), namely 374 mg m -3 and 304 mg m -3 , resulting from the higher excess air ratios during ineffective operation (Li et al., 2011).In terms of NO x , 47 boilers, accounting for 38.8% of the monitored CCBs, exceeded the emission limit of current standards because of ineffective operation or insufficient supervision from the department of environmental protection (BMEPB, 2007).Thus, low-NO x or a control measure for NO x emission is urgently required because of the implementation of considerably stricter emission control standards that recommend a limit of 80 mg m -3 for existing boilers (BMEPB, 2015).By contrast, SO 2 concentration in flue gas ranges from 0 to 16.05 mg m -3 with an average value of 2.35 mg m -3 .The SO 2 concentrations of 120 CCBs, which accounted for 96.8% of the monitored CCBs, met the current standards for boiler emission (50 mg m -3 for existing boilers and 20 mg m -3 for newly-built boilers), (BMEPB, 2007).Adopting terminal control measures for Fig. 1.Effect of operating load on pollutants emissions at (a) 10.5 MW, 3% ± 0.3% O 2 ; (b) 14 MW, 3% ± 0.3% O 2 ; (c) 16 MW, 3% ± 0.3% O 2 ; (d) 7 MW, 1.56%-3.49%O 2 ; (e) 7 MW, 2.88%-4.51%O 2 ; (f) 7 MW, 2.75%-6.75%O 2 .

Fig. 2.
The effect of combustion mode on the NO x , SO 2 and CO emission.SO 2 is unnecessary although its concentration occasionally exceeds standard limits because reduction of its concentration can achieved thorough control of the sulfur content of natural gas at a stable low-level.In addition, the emission of CO was dependent on the combustion efficiency and the average value was 4.32 mg m -3 within a range of 0-93.29 mg m -3 , which completely met the emission limit of 95 mg m -3 in the Safety Technical Regulation for Oil and Gas Burners (GAQSIQPRC, 2008).
Only four ACBs were found because of their small proportion (13.54% of the entire boilers), as depicted in Fig. 2. The results indicated that the NO x and SO 2 emissions of ACBs were higher than those of CCBs, at 180-268.36 mg m -3 and 3.24-33.45mg m-3 , respectively.The disparity in NO x emission was attributed to higher excess air ratio and burning temperature of ACBs.The combustion modes of ACBs and CCBs are partially premixed combustion and diffusion combustion, respectively.It has been proved that the excess air ratio and burning temperature of partially premixed combustion are higher than that of diffusion combustion (Tongji University, 2000).SO 2 concentration in ACBs was higher than that in CCBs because of high excess air ratios of ACBs.The results indicated that the ACBs, which exceeded the emission standards for NO x and SO 2 , accounted for 75% and 25%, respectively.Therefore, the ACBs must be replaced by CCBs because of the issue of new standards (BMEPB, 2015).The CO emission of ACBs lies in the range 5.06-41.25 mg m -3 with an average concentration of 19.26 mg m -3 , which is higher than that for CCBs despite higher excess air ratios than in the CCSs because of higher premixed temperatures in the ACBs than in the CCBs.

Boiler Type
The emission of gaseous pollutants is affected by boiler type.In the present study, the investigation of boiler type was based on CCBs because of the small quantity and variety of ACBs.The gross capacity of HIC boilers comprises 60% of the total capacity for CCBs, thus, the HIC is the priority boiler type to be monitored.The other boilers were classified as "others" because of the absence of detailed information on boiler type or the small numbers of the specific type.Fig. 3 depicts the emission profiles of HIC boilers and the other CCBs.The results indicated that the average NO x concentration of the HIC boilers was 152.74 mg m -3 , which is higher than that of the others with an average of 123.87 mg m -3 .This phenomenon was due to the higher temperature of HIC boilers compared with other CCBs, while the slightly lower in the O 2 content of HIC cannot influence the NO x formation, as shown in Fig. 3.The SO 2 concentration of the HIC boiler and the other CCBs were similar, namely 2.31 mg m -3 and 2.50 mg m -3 on average, respectively.A similar phenomenon was found in the CO concentrations, at 4.28 mg m -3 and 4.48 mg m -3 for the HIC boilers and other CCBs on average, respectively.There was no significant difference in SO 2 and CO concentration for HIC and other CCBs, which is attributed to their similar O 2 concentration.Thus, the effect of boiler type on SO 2 and CO formation is likely insignificant, whereas the boiler type plays a crucial role in NO x formation.

Installed Capacity
The effects of installed capacity on gaseous pollutants were analyzed for HIC boilers.The HIC boilers were classified into two grades according to the installed capacity, namely < 10 MW and ≥ 10 MW by cluster analysis with SPSS software.The < 10 MW units had a higher NO x concentration (averagely 160.87 mg m -3 ) than the ≥ 10 MW units (averagely 118.41 mg m -3 ) (Fig. 4 (a)), indicating smaller boilers produced more NO x than larger ones do.To explore the influence mechanism of installed capacity, the relationship between the gas flue temperature and O 2 content with installed capacity was investigated, as depicted in Fig. 4(b).Because the temperature and O 2 content of the < 10 MW units were higher than those of the ≥ 10 MW units, smaller units produced more thermal-NO x than did large ones.The average SO 2 concentrations of the < 10 MW and the ≥ 10 MW units were similar at 2.35 mg m -3 and 2.14 mg m -3 , respectively, implying that the formation of SO 2 is minimally affected by installed capacities.Notably, the average CO concentration of the < 10 MW units was 3.36 mg m -3 , which was lower than that of the ≥ 10 MW units (8.12 mg m -3 ).This phenomenon could be due to the higher O 2 content of the < 10 MW units, thereby leading to the further oxidization of larger amounts of CO to CO 2 in the <10 MW units than in the ≥ 10 MW units.

Other Factors
In gaseous pollutants generation, control measure is a significant factor.According to the database of gas-fired boilers, currently, more than 95% burners used in existing gas-fired boilers in Beijing are provided by European enterprises.Regarding the burners' NO x grading, more than 80% of the burners adopt the level 1 and level 2 standard products, namely EN676, i.e., NO x ≤ 119.88 mg m -3 , which is adequately meets the NO x emission limits of standard (BMEPB, 2007), thus neither low-NO x burners nor NO x emission control measures were used in the existing boilers.
In addition, the average NO x concentration of CCBs is 1.19 times higher than that of standard in the EN676, which probably because of the small dimensions of the chambers in Beijing compared with those in Europe.
The characteristics of gaseous pollutants were affected by installation year, pressure of natural gas and control mode of burner, as depicted in Fig. 5.Although burner annual deterioration occurs naturally, the average concentrations of NO x , SO 2 and CO remained constant during 1999-2005 (Fig. 5(a)), implying that the formation of gaseous pollutants was not affected by installation year.Fig. 5(b) illustrates the correlations between the concentrations of gaseous pollutants and pressures of natural gas were not significant.Furthermore, the results illustrated in Fig. 5(c) indicated that the concentrations of NO x , SO 2 and CO in boilers with mechanical control mode are similar to those with electrical control mode; thus, the control mode did not basically affect the gaseous pollutants formation.
In conclusion, operating load, combustion mode, boiler type, and installed capacity are the priority factors affecting pollutants formation, whereas the other factors, namely control measure, installation year, pressure of natural gas, and control mode of burner, did not affect gaseous pollutants formation.Thus, the boilers were classified on the basis of combustion mode, boiler type, and installed capacity.

Estimate of EFs
On the basis of the classification, the averages of the EFs of various gas-fired boilers were calculated and summarized in Table 1.The category-specific EFs of NO x , CO, and SO 2 range were as follows: 1.42-6.86g m -3 , 0.05-0.67g m -3 and 0.03-0.48g m -3 , respectively.The variation characteristics of the EF values are summarized as follows.
First, the ACBs have high EFs, which were more than 3.2 times higher than those of CCBs.This phenomenon indicated that combustion mode is the most crucial factor affecting gaseous pollutant formation.Second, boiler type also significantly affected EF NO x and EF CO .The averages of EF NO x and EF CO in HIC boilers were higher than those in other CCBs by more than 19% and 33%, respectively.Third, there were differences among various installed capacities for EF NO x and EF CO in HIC boilers.When the capacity increased from < 10.5 MW to ≥ 10.5 MW, the EF NO x decreased from 2.13 g m -3 to 1.42 g m -3 , whereas the EF CO increased from 0.05 g m -3 to 0.10 g m -3 .Finally, the EF SO 2 remained constant although EF NO x and EF CO vary significantly with boiler types and installed capacities.
There is not a set of systematic and integrated emission factors (EFs) for various species compiled for different types of gas-fired boiler in Beijing is currently available.Therefore, EFs of gas-fired boilers from other countries or districts were compared with those of Beijing in the present study as illustrated in Table 2 (Kato et al., 1992;Pulles et al., 2004;EPA, 2006;MEPPRC, 2007;Yao et al., 2009;Xue et al., 2014).To make the companion more scientific and clear, the EFs of CCBs were selected from our study, because those of ACBs were absent in the AP-42, HPPEI, and previous studies.EF NO x in this study is lower  , 2006), Pulles et al. (2004, 2.24 g m -3 ), Yao et al. (2009, 2.78 g m -3 ) and Xue et al. (2014, 2.19  g m -3 ), whereas it is 0.79-1.18times as that in HPPEI.The EF NO x theoretical value calculated by Kato et al. (1992) was 1.90 g m -3 , which was 0.89-1.34times as that in the present study.Thus, classification based on the field measurements was more reasonable and more suitable for Beijing.
The EF SO 2 value in our study is 3.85 times and 3.12 times higher than the EF SO 2 values in the study of Kato et al. (1992) and that published by AP-42, respectively.However, it is 0.04 times lower than that in the HPPEI.
The value of EF CO in our study is lower than the AP-42 (in the range of 0.56-0.98g m -3 ) and research of Pulles et  al. (0.78 g m -3 ) by approximate one order of magnitude.The reasons for this disparity cannot be identified with our present data; therefore, further work is required to identify the causes of this disparity.

Compilation of Emission Inventory
The emissions of gas-fired boilers from Beijing in 2014, based on the EFs in the present study and the consumption of natural gas was estimated.The emissions were 11121 t, 468 t, and 222 t for NO x , CO, and SO 2 , respectively.A detailed emission inventory is presented in Table 1.The emissions of NO x , CO, and SO 2 from ACBs are 487 t, 48 t, and 34 t, contributing 4%, 10%, 15% of the gross emission, respectively, although their capacities account for only 1.12% of the entire capacities.Thus, the substitution of ACBs by CCBs is urgently required.Additionally, the SO 2 , NO x, and CO emissions of gas-fired boilers accounted for 0.30%, 7.07% and 0.047% of total emission from Beijing in 2014, which was obtained from the emission inventory report released by the BMRIEP (BMRIEP, 2015).

Spatial Characteristics
The emission inventory was spatially allocated into 1 km × 1 km grid cells, as illustrated in Fig. 6.At the district level, the top five unit emitters are Xicheng, Dongcheng, Haidian, Chaoyang and Fengtai, respectively.All five districts are in central Beijing.This phenomenon was consistent with the implementation of the program to substitute coalfired boilers with gas-fired boilers, and the central of Beijing has been given high priority.Moreover, these five districts are densely populated and are residential areas; therefore, a high number of boilers are needed to produce heat, thereby resulting in high emission.Shijingshan, Changping, Daxing, Tongzhou, and Shunyi are less densely populated and are residential area, resulting in low emission of NO x , SO 2 , and CO.The other districts have the lowest emission because they have the lowest population density, do not have many residential areas, or have the limitation of gas pipelines.

Uncertainty Analysis
The estimated means and associated uncertainty ranges for pollutant-based emission are presented in Table 3.The means established were calculated by averaging 10000 simulated values form Monte Carlo sampling.The uncertainty in CO and SO 2 emissions are relatively high with an approximate range -157% to 154% and -127% to 182% relative errors, respectively.Compared with the uncertainties in CO and SO 2 emission estimation, the uncertainty in NO x emission estimated is relatively small, at -34% to 34% relative errors at 95% confidence intervals.Because the uncertainty of an activity level is fixed for varied pollutants, the EFs are dominant uncertainty sources, thereby leading to higher uncertainty in CO and SO 2 estimates because of their low concentrations; thus it is evitable.

CONCLUSIONS
Gas-fired boilers were classified through influence analysis; thus, category-specific EFs were established.An emission inventory was compiled through the categoryspecific EFs and specific consumption of natural gas.Finally, the spatial variable characteristics with high resolution (1 km × 1 km) were proposed.The main findings can be summarized as follows: (1) Gas-fired boilers were classified on the basis of combustion mode, boiler type, and installed capacity, because these factors play significant roles in NO x , SO 2 , and CO formation, whereas the installation year, pressure of natural gas, and control mode of burner did not significantly affect the formation of gaseous pollutants.
(2) Category-specific EFs were established through classification, and the values for NO x , CO, and SO 2 lie within the following ranges 1.42-6.86g m -3 , 0.05-0.67g m -3 , 0.03-0.48g m -3 , respectively.(3) On the basis of category-specific EFs and specific consumption of natural gas, the emissions in 2014 were established at 11121 t, 468 t, and 222 t for NO x , CO, and SO 2 , respectively.Additionally, ACBs should be substituted with CCBs because they have higher  emission than do CCBs, although their gross installed capacities are lower than those of CCBs.(4) At the district level, the top five unit emitters are Xicheng, Dongcheng, Haidian, Chaoyang and Fengtai, respectively.All these units are in central Beijing.
(5) The uncertainties in NO x emission were lower, ranging from -34% to 34%, and those in CO and SO 2 emissions were relatively higher (ranging from -157% to 154 and -127% to 182%), which are evitable.(6) Considering that stricter standards will be implemented in 2017, low-NO x burner or NO x emission control measures are urgently required.Furthermore, supervision should be increased.Adopting terminal control measures for SO 2 is unnecessary although its concentration occasionally exceeds the standard limit because the reduction of its concentration can be achieved thorough effective control of the sulfur content of natural gas at a stable low-level.

Fig. 3 .
Fig. 3.The effect of boiler type on the gaseous pollutants emission (a), O 2 content (b) and temperature (b).

Fig. 4 .
Fig. 4. The effect of installed capacity on the gaseous pollutants emission (a), temperature (b) and O 2 content (b) (Error bars represent standard deviations).

Fig. 5 .
Fig. 5.The effect of installed year (a), pressure of natural gas (b), control mode of burner (c) on the emission of NO x , SO 2 and CO.

Fig. 6 .
Fig. 6.Spatial variable characteristics of NO x (a), SO 2 (b) and CO (c) from gas-fired boilers in Beijing.

Table 1 .
Summary of the emission factors and emission inventory for various gas-fired boilers.

Table 2 .
Comparison of emission factors reported for the current and previous studies.

Table 3 .
Uncertainties in emission estimates.