The impacts of coal plants relocation on the concentration of fine particulate matter in China

Abstract

The air pollution situation in the northern part of China has become a threat to the current status of China’s energy consumption. This has created pressure to relocate coal plants to the western section of the country. We apply a spatial panel data method to empirically examine the relationship between fine particulate matter (PM_2.5) concentration, precipitation and wind speed by using the sample data of 50 cities over 11 months. Based on the empirical parameters, the paper simulates the potential impacts of the proposed coal plants and the coal plants currently under construction for the PM_2.5 concentrations across all cities once these coal plants are operational. The results show that the new capacities will increase PM_2.5 concentrations in northwest and northeast China, but at acceptable levels and lower than levels in Beijing, Tianjin, Hebei and Shandong.

Keywords

coal plants relocation spatial panel data method power generating technologies emission control

Introduction

Since the winter of 2012, northern China has often been plagued by haze. Beijing’s air quality has attracted international attention and impacted the direction of the government’s energy policy. One of the major energy policy decisions that have been affected by this situation is the projection of growth and location of coal-fired power plants. According to the State Council Air Pollution Prevention and Control Action Plan,^a three regions are expected to have a negative growth of coal consumption, namely, the Beijing-Tianjin-Hebei, the Yangtze River Delta and the Pearl River Delta regions. In key locations in these three areas, new coal-fired power plants will be prohibited (except for combined heat and power (CHP) systems), with an expectation that coal will be replaced by ‘increasing imported electricity (from other areas), natural gas and the use of non-fossil energy’^b. In the Energy Industries Strengthen Airborne Pollution Prevention and Control Work Plan,^c detailed plans are made to reduce the share of coal in total energy consumption to less than 65% by 2017, with net decreases of 13, 10, 40 and 20 million tons for Beijing, Tianjin, Hebei and Shandong, respectively. Increased electricity demand will be met by imported electricity from coal-fired power plants that are located in coal-rich areas such as Xinjiang, Inner-Mongolia, Shanxi, Shaanxi and Ningxia. These regions have been proposed as the ‘power bases’ in China’s 12th Five-Year Plan for Energy and Electricity. This strategic initiative to relocate China’s power base is called the ‘Coal Plants Western Movement’. Although it is suggested that these bases should be built in the environmental and ecological carrying capacity of the locale, it is well known that these areas already exhibit high water stress and air quality issues. It is also well recognized that air pollution concerns are directly impacted by coal-fired power generation. In this paper, the questions that are raised include the following. Can ecologically fragile areas bear the newly added particulate emissions? Will the worsening air quality in the west because of the relocation of power plants impact the air quality in northern China?

Interpolating a detailed map of PM pollution that covers a wide area is challenging because of multiple disruptions. However, it is possible to interpolate the pollution trend by considering the key meteorological factors that affect the spread and deposition of pollutants. This paper constructs a PM_2.5 concentration model and considers emissions that are sourced from coal-fired power plants and meteorological factors such as wind speed and precipitation to examine and predict the change in PM_2.5 concentration across China considering the relocation of coal plants to the western region.

Model and data

Quantifying the relations between PM_2.5 concentration and the key affecting factors

PM_2.5 concentration is influenced by many factors, namely, emission sources and emission scales, meteorological factors and location-specific factors, such as forest coverage and topography conditions. Constrained by the relevant data on emissions from vehicles, this paper focuses on the effects of emissions from coal-fired power plants and the associated meteorological factors of precipitation and wind speed.

Before specifying the model, the quantitative relation between PM_2.5 concentration and these factors should be determined. The sample includes 500 coal-fired power plants that are located in 30 provinces in China. Although these power plants are not evenly distributed in space, the spatial characteristic is obvious. The power plants that surround a given target city under review will be effected by their emissions. Suppose target city i has several power plants around it, and their installed capacities are C1–Cn; then, their distance to i are d1–dn, as revealed in Figure 1. Because these power plants will not affect i equally, the total effective capacity that affects i, which is called ‘coal plant capacity intensity’ in this paper, can be calculated by the inverse distance weighted capacity and is expressed as equation (1)

{CC}_{i} = \sum_{j = 1}^{n} {(1 / d}_{ij}) k C_{j} / \sum_{j = 1}^{n} {(1 / d}_{ij}) k

(1)

Typically, a threshold value is assigned to the distance, and the power plants whose distances surpass the threshold value are not considered. That is, when d_ij surpasses the threshold value, its weight will be taken as 0; otherwise, the weight will be the inverse value of the distance. When there is only one emission source, the pollutant deposition rate is low, and the effect of wind direction is not considered; thus, k = 2 is appropriate, as suggested by Tai et al.¹ Considering the cumulated influence of different power plants, k is assigned the value of 1 in this paper, and the weight matrix is not normalized (the sum of all weights equals 1) to discriminate the different effects that are borne by the different targets. However, as Tai et al.¹ suggested, the research results were not sensitive to power k. Thus, the power plant intensities of all target cities can be written as

{CC}_{t} {= W * C}_{t}

(2)

where CC is a 50 × N (N = 1, 2) vector that depicts the power plant intensities of the 50 target cities, W is the 50 × 500 inverse distance spatial weight matrix for the distances within the threshold value, C_t is the 500 × N matrix, which describes the 500 power plants’ total capacity (N = 1) or the capacities that are classified by technologies (N > 1), and the subscript of t means that the capacities are adjusted according to the electricity generation of the specific provinces for a given month.

Figure 1.

Diagram of calculating capacity intensity.

When the generating unit operates normally and the coal quality is stable, the PM_2.5 concentration of target cities would be proportional to its capacity intensity, which is written as

PM = c + β CC + u

(3)

Because each generating unit has different generating hours at different months, CC should be adjusted according to the electricity that is generated.

Second, the quantitative relation between PM_2.5 concentration and the meteorological variables must be ascertained. When the chemical components are different, their reactions to meteorological factors such as precipitation, temperature, and cloud coverage also differ. Therefore, a model that is designed to apply to all situations may be misleading. To address this issue, the Spearman correlation method has been widely used to study the quantitative relation between pollutant concentration and meteorological factors.² In recent years, the studies that use the elasticity analysis method have been rare.³ Most studies still use the simple correlation method to study the relation between pollutant concentration and the meteorological variables with different units of measurement, for example Li et al.⁴ Based on these studies that use meteorological level data, this paper includes wind speed and monthly precipitation in equation (3) to describe their effects on PM_2.5 concentration. Because of the unavailability of the data on temperature, barometric pressure and cloud coverage, these variables are excluded. Their effects are ascribed to fixed effects as described below. Thus, the PM_2.5 concentration is written as equation (4)

PM = c + β_{1} rain + β_{2} wind + β_{3} CC + u

(4)

In the case of China, there are many sources of PM_2.5 pollution, including coal-fired power plants, coal-consuming factories, vehicles, biomass waste burning, family activities and dust re-suspension, but the statistics on these emission sources are insufficiently available on a monthly basis. This paper therefore ascribes their effects with the effects of forest cover and topography on PM_2.5 concentration in a given city. Because PM_2.5 is highly dispersible, there are mutual affecting impacts among adjacent cities. This effect must also be included in equation (4), which changes it from a panel data model to a spatial panel data model.

Another issue relates to the seasonal effects of PM_2.5 concentration. Considering the seasonal characteristics of rain, wind and electricity generation, the seasonal effects are not included in the model to avoid collinearity, and these data are not seasonally adjusted to avoid information loss.

Spatial panel data method

Following the work of Baltagi et al.⁵ and Elhorst,⁶ the spatial panel data method has achieved significant progress and has been widely applied to many areas. Given a static panel data model, with dependent variable y and independent variable X, and assuming that there is spatial interaction among the dependent variables, the model can be described as

y = λ {(I}_{T} \otimes W_{N}) y + X β + u

(5)

where y is the NT × 1 vector of the stacked dependent variables, N is the number of cross section samples that were observed, T is the period of observation, λ is the spatial coefficient, I_T is the unit matrix, W_N is the spatial weight matrix with diagonal values that equal 0, ⊗ is the Kronecker product, X is the stacked independent variable matrix and β is the coefficient of the independent variable. The error u includes two parts, namely, the fixed effects and residuals, which are written as

u = (ι_{T} \otimes I_{N}) μ + ɛ

(6)

where

ι_{T}

is the T × 1 vector with values that are all equal to 1,

I_{N}

is the N × N unit matrix, μ is the specific fixed effects and ɛ has the characteristic of spatial auto-regression, which is written as

ɛ = ρ (I_{T} \otimes W_{N}) ɛ + ν

, where ρ is the spatial auto-regression coefficient, and

ν_{it} ~ IID (0, σ_{ν}^{2})

. The fixed effect μ can be time-specific or time-invariant.

Data and sources

The data for this empirical study and the simulation include the monthly average PM_2.5 concentration data of 50 cities in mainland China from April 2014 to February 2015, the monthly precipitation and average wind speed of these cities during this time, the detailed installed capacity of 500 main coal-fired power plants and the coordinates (longitudes and latitudes) of the 550 (50 cities plus 500 power plants) samples. The data on the total installed capacities (not in detailed units) of 2311 coal-fired power plants and their efficiency, generating hours and emissions are further used in the discussion section.

The data on PM_2.5 concentration are obtained from the Qingyue Historical Air Quality Database (http://air.epmap.org), which publishes hourly PM_2.5, PM₁₀, CO, NO₂, O₃ and SO₂ concentration data for every observation point in every city of the 33 provinces and municipalities (including Taiwan and Hong Kong). This paper chooses 50 cities (two for provinces and one for municipalities) in mainland China and calculates their monthly average from hourly observations from every observation point of the city.

The data on monthly precipitation of all of the cities are obtained from the website of the National Center for Environmental Information of the National Oceanic and Atmospheric Administration (gis.ncdc.noaa.gov). The wind speed data are obtained from the Weather Underground website (www.wunderground.com), which publishes the daily average wind speed of cities. The monthly averages are calculated based on the daily data. Because cities such as Yinchuan, Luzhou, Zunyi and Jilin have no records, their historical averages are collected from published materials.

The data on coal-fired power plants are obtained from Wikipedia’s List of Power Stations in China (en.wikipedia.org/wiki/List_of_power_stations_in_China) (500 for the empirical study) and Greenpeace’s Coal-fired Power Plants (map.greenpeace.cn/ce-cpp-map) (2311 for further discussion). These data are compared with the data from Sourcewatch (www.sourcewatch.org) and the Compilation of Statistical Materials of the Electric Power Industry to guarantee their reliability. In the data, the total capacity of power plants in operation is 584 GW, which accounted for 75% of the total coal-fired power plants by the end of 2013. The total capacity under construction is 128 GW, and the total capacity of the proposed power plants is 206 GW. The samples are adequate for the study.

Concerning the coordinates of the 550 samples, the 50 cities are represented by the coordinates of their city government, considering that the governments are located almost at the centre of the cities. The coordinates of the power plants in operation and under construction are rather precise, and the proposed power plants are estimated on Google map by locating information such as village name.

At the power plant level, the monthly electricity generation data are not available. This paper uses a comprehensive method to adjust the capacity by adjusting the parameters, namely, the monthly provincial electricity generation data (totals minus hydroelectricity generation). By assuming that all of the power plants in the same province have equal generation hours and making all province parameters for April 2014 equal to 1, the monthly parameters for each province in later months are calculated. We use these parameters to multiply the 500 power plants according to the provinces where they belong and then calculate the CC for the target cities for different months. This assumption may cause some difference of the CC from its real value, but in consideration of the reality that the generation hours of power plants are arranged directly by System Operator who tries to balance the interests of all the power plants under control, and of that the CC in equation (4) is a weighted average and the dependent variable is monthly average, the negative effect of this assumption can be ignored. The monthly provincial electricity generation data are obtained from the China Economic Database of the Census and Economic Information Center and the website of the China Electricity Council (www.cec.org.cn).

Empirical and simulation study of China’s PM_2.5 concentration

Spatial weight matrix

To calculate the spatial weight matrix, the neighbourhood relation should be determined by one of the many standards, such as the threshold value distance, N nearest neighbours, common frontier or nodes, Delaunay triangle, etc., and the weight matrix can be normalized or un-normalized. In this paper, two weight matrices are calculated to determine the connections among the 50 cities as well as the connections between the 50 cities and the 500 coal-fired power plants. Suppose that when the distance is longer, the connection is weaker, the weight matrix is calculated by the inverse distance (under threshold values), and the weight matrix is not normalized to reveal the different effects that the cities bear.

Because the location coordinates are longitude and latitudes, this means different distances at different latitude levels for the same longitude. The real distances among the 550 samples are first calculated, and the inverse distance spatial weight matrix is then calculated. In the 550 × 550 matrix, the upper left 50 × 50 is the weight matrix of the spatial panel data calculation, and the upper right 50 × 500 is the weight matrix of the capacity intensity calculation. Although it is more precise to calculate the spatial weight matrix by real distance, the result of the study is almost the same as the weight matrix if it was calculated by longitude and latitude in a Euclidean space.

Empirical results are sensitive to the threshold value

Choosing the reasonable threshold value is rather challenging because it depends on the normal distance that PM_2.5 can disperse. It is commonly believed that PM_2.5 can stay in the air for several days to several weeks and can transport to distances of several hundred – even thousands of – kilometres. NASA has noticed tons of millions of dust particles that have been transported from the Sahara desert to the Amazon forest every year. Because there is no known study that examines the average transport distance of PM_2.5 in China, this paper tests several threshold values to balance the effects of local emissions and distant transportation of PM_2.5 concentrations.

When the threshold value is small, the number of neighbours is small, and some cities do not even have neighbours. The regression results mainly emphasize the effects of local influencing factors, and the coefficients of precipitation, wind speed, and the weighted capacity intensity are robust. When the threshold value expands, the neighbourhood is defined more widely and the number of neighbours is large. As a result, the spatial panel regression emphasizes more spatial interactions. To balance the effects of the local factors and spatial interactions, the results are best when the distance threshold value ranges between 220 and 230 km.

Empirical study of coal-fired electricity generating units’ effects on PM_2.5 concentration

The effects of total installed generating capacity on PM_2.5 concentration

When the threshold value is 225 km, and the capacity intensity is calculated by the products of the spatial weights matrix and the total installed capacities of all power plants without discriminating the technologies that are used, the regression results are shown in Table 1. The units of measurement are µg for PM, mm for Rain, km/hour for Wind, and MW/km for T-cap.

Table 1.

Regression results when the generating technologies are not discriminated.

	Variable	Coefficient	Standard error	t-statistics	p value
Panel data method	Rain	−0.10092	0.00993	−10.1649	0.000***
	Wind	−1.34375	0.61449	−2.1868	0.029**
	T-cap	0.00057	0.00015	3.7224	0.000***
	C	59.13585	7.71615	7.6639	0.000***
Spatial panel data method	Rain	−0.08439	0.01270	−6.6459	0.000***
	Wind	−1.04496	0.62973	−1.6594	0.097*
	T-cap	0.00043	0.00017	2.5576	0.011**
	W*pm	0.24867	0.11945	2.0818	0.037**
	C	48.54760	9.22752	5.2612	0.000***

Note: ***, significant at 1%; **, significant at 5%; *, significant at 1%.

To prove the reliability of the spatial panel data regression method, the result is compared with the result of the panel data regression with fixed effect results, and the results are shown in Table 1. The results show that the empirical tests of the panel data method and spatial panel data method are very similar because there are no large differences in the coefficients. The differences focused on the residuals, and the panel data method had an obvious auto-regression effect. For all of the target cities that were studied, the effect of precipitation on PM_2.5 concentration is negative and statistically significant. When observed from a single city perspective, the effect of wind speed on the PM_2.5 concentration is unclear because it can both blow away local pollutants and bring pollutants from other areas. This distribution depends on where the pollution emission sources are located and the direction of the wind. However, in general, the wind is helpful in lowering PM_2.5 concentration. The coefficient of capacity intensity is small but statistically significant. This finding proves that coal-fired power plants have a robust effect on PM_2.5 concentration, but it cannot explain all PM_2.5 concentrations because other emission sources, such as vehicles, may also have some important effects. In the spatial panel data regression, the coefficient of W*pm is very high and statistically significant. This result can be interpreted to mean that PM_2.5 from emission sources other than power plants has a significant influence on neighbouring areas.

Empirical test of the effects of different technologies

During the ‘Eleventh-Five Year’ period, the Chinese government closed many small coal-fired power plants (the plants with under 300 MW generating units) under the policy of ‘Replacing small units with large ones’. The ‘small’ power plants are mostly sub-critical or critical technologies, but the ‘large’ power plants are super-critical or ultra-super-critical technologies (with a unit power above 600 MW), and they contribute differently to PM_2.5 emissions. To show the difference, this paper classifies the generating units by unit powers into two groups (the units below 600 MW and the units above 600 MW) and examines their effects on PM_2.5 concentration. The regression results are shown in Table 2.

Table 2.

Spatial panel data regression when discriminating the technologies.

Variable	Coefficient	Standard error	t-statistics	p value
Rain	−0.08426	0.01267	−6.6527	0.000***
Wind	−1.07169	0.62998	−1.7012	0.089*
Tb6	0.00078	0.00041	1.8784	0.060*
Ta6	0.00012	0.00038	0.3103	0.756
W*pm	0.24246	0.11919	2.0342	0.042**
C	47.81972	9.22938	5.1813	0.000***

Note: ***, significant at 1%; **, significant at 5%; *, significant at 1%.

In Table 2, Rain, Wind, W*pm and C have the same meanings as in Table 1. Tb6 represents the capacity intensity of the generating units below 600 MW, and Ta6 represents the capacity intensity of the generating units above 600 MW. Compared with the results in Table 1, the coefficients of precipitation, wind speed, spatial influence and the constant are almost unchanged, but the coefficient of Tb6 is much higher than the coefficient of T-cap in Table 1; the coefficient of Ta6 is much lower and not statistically significant. Comparing the coefficients of Tb6 and Ta6 for the same capacity in operation, the units under 600 MW will emit on average 6.5 times more PM_2.5 than the units above 600 MW. Although all units face de-sulpharization, de-NOx, and smoke and dust emission regulations, the super-critical and ultra-super-critical units are more efficient and clean. Of the 584 GW total capacity in operation, 282 GW are under 600 MW, and 302 GW are above 600 MW; the coefficient of T-cap in Table 1 is the weighted average of Tb6 and Ta6 in Table 2.

The cities’ fixed effects and residuals

As mentioned earlier, many factors affect PM_2.5 concentration in a city, and they include coal-fired power plants, coal-burning factories, vehicles, topography, biomass burning, etc. All of these factors contribute to PM_2.5 concentration and are different across cities, and their effects on local PM_2.5 concentration are ascribed to the local fixed effect, and their effects on other cities are also transmitted by the spatial panel item.

The 50 target cities’ fixed effects are shown in Figure 2. These points are located according to the coordinates of the cities, and when the point has a larger radius, the fixed effects are larger. The largest fixed effect is 50.3 for Yichang, and the smallest is−55.8 for Datong. In general, the fixed effects are mainly affected by other industrial emissions. For the regression, most of the residuals are within ±20, except for the Henan, Heilongjiang and Sichuan provinces, whose inter-month changes are volatile.

Figure 2.

The 50 cities’ fixed effects.

Simulation study of effects of increased capacities on PM2.5 concentration

Of the 500 sampled coal-fired power plants, the total installed capacity in operation is 584.2 GW (282 GW below 600 MW and 302 above 600 MW), the total capacity under construction is 128 GW (15 GW below 600 MW and 113 above 600 MW) and the total capacity of the proposed plants amounts to 206 GW (13 GW below 600 MW and 193 above 600 MW). Most of the newly added capacities are clean, highly efficient, ultra-super critical technologies that claim to remove 95% of soot and smoke. However, when these capacities are put into operation, the total emission by coal plants will inevitably increase, which increases PM_2.5 concentrations. Most of the new capacities will clearly be located in western China, which is rich in coal resources but faces severe water pressures. The newly installed coal-fired capacities may increase PM emissions and PM_2.5 concentrations there, and by spatial transmission, the PM_2.5 concentrations in northern China may also increase. By using the parameters that were obtained in the spatial panel data regression, the effects of newly installed coal-fired power plants can be predicted by simulation.

According to equations (4) and (5), which describe the spatial panel relation between PM_2.5 concentration and the independent factors, the PM_2.5 concentration at a certain time can be written as

PM = λ \cdot W \cdot PM + c + β_{1} \cdot rain + β_{2} \cdot wind + β_{3} \cdot CC + Fix + ɛ

(7)

By transforming the equation, the simulated result is

P \overset{\land}{M} = (I - λ W) - 1 (c + β_{1} \cdot rain + β_{2} \cdot wind + β_{3} \cdot CC + Fix)

(8)

Equation (8) can be used to predict the change in PM_2.5 concentration when some independent variables change. To simulate how the newly increased coal-fired capacities will affect PM_2.5 concentration, the fitted results of April 2014 is used as the base scenario. The precipitation, average wind speed and average generating hours are unchanged. Suppose also that there are no retired capacities and the PM_2.5 concentrations are simulated when the capacities under construction and the capacities that are proposed are put into operation in sequence. The results are shown in Figure 3.

Figure 3.

The simulated PM_2.5 concentrations and changes.

In Figure 3, the PM_2.5 concentrations when the capacities that are under construction and the capacities that are proposed are put into operation in sequence are depicted by (a) and (b), respectively, and their change ratios to the base scenario are depicted by (c) and (d), respectively. When the capacities under construction are put into operation, the 50 cities will be impacted unevenly, both in absolute and relative terms. The largest increase will occur in Jiayuguan city, with 18.2 µm (49.7%) increase, and other cities with large relative increases greater than 10% are Qiqihar (15.4%), Jilin (17.3%), Erdos (16.5%) and Yinchuan (16.1%). There are also other cities with relatively large increases that are higher than 5%, such as Wuhu, Wuhan, Changchun, Nanjing, Jinan, Taiyuan and Shanghai. However, when the proposed capacities are further put into operation, the impact map shows a different view, and the mostly affected areas are northwest and northeast China. The two cities in the Jilin province have large-scale increases, and the other cities with large increase are Erdos and Yinchuan. The cities in southern China such as Hefei, Wuhu, Nanjing and the cities in the southeastern coastal areas will also have significant increases. Thus, it can be concluded that although the total capacity of these new plants is very large, because most of them are highly efficient and clean ultra-super-critical technologies and well equipped with de-sulpharization, de-NOx and smoke and soot removing facilities, the PM_2.5 concentrations in the western areas will still be lower than these concentrations in Beijing, Hebei, Shanxi and Shandong.

Further discussions regarding China’s PM_2.5 concentrations

Less efficient coal-fired power plants may still have high-generating hours

From the unit power aspect, the standard of the policy ‘Replacing small units with large ones’ is 300 MW, and the power plants with lower capacities should be closed. From a technology perspective, the small coal-fired units are sub-critical and critical technologies, less efficient, high coal-consuming and high pollutant-emitting. Figure 4 describes the combinations of total capacity and coal-electricity efficiency of the 2311 coal-fired power plants that were in operation in 2013. Figure 4 shows that there are still many inefficient generating units in operation, and this is unfavourable to the ‘Energy conservation and emission reduction’ policy. Of these inefficient generating units, some are coal waste-fired power units and CHP units, but most are self-supply power plants that are owned by energy-intensive industries. These inefficient coal-fired power units are not used for peak power; in fact, they compete with highly efficient units for generating hours. Because they are self-supply power plants that are meant to satisfy their owner’s electricity demand, data from Greenpeace show that they have rather high annual generating hours.

Figure 4.

Combination of capacity and efficiency.

Inefficient, high pollutant-emitting coal-fired power units have high generating hours, which will increase total PM emissions. The situation can be improved but requires some good adjustments by plant managers.

Emission reduction regulation needs more investment

Because the Hulbert peaks of coal production in the eastern provinces have already been attained, and new coal discoveries have been mainly made in Xinjiang, Inner-Mongolia and Shaanxi, producing electricity in the west and transmitting it to the east is a reasonable choice.⁷ Although China has attempted to lower smoke and soot emissions in the past, the situation concerning coal-fired power plants still requires improvement. For the 2311 coal-fired power plants, their TSP/coal emission ratios are calculated by dividing the total TSP that is emitted by the total raw coal consumed, and the total raw coal consumed is calculated by dividing electricity production with efficiency. The TSP emission volumes and emission ratios are shown in Figure 5(a) and (b), respectively. Figure 5(a) depicts the total TSP emissions of all of the power plants, and the distribution map coincides roughly with China’s haze distribution map. Although coal-fired power plants consume only half of China’s total coal, this distribution can be explained to mean that electricity consumption correlates highly with other industrial activities that emit air pollution. Figure 5(b) depicts these plants’ emission ratios. In general, the coastal area provinces have lower emission ratios. More detailed study shows that the power plants with high emission ratios are the plants with total installed capacities that are below 500 MW. However, this definition is not entirely correct because there are many power plants that are below 500 MW with very low emission ratios, whereas there are some plants that are above 500 MW with high emission ratios. This discrepancy means that these power plants do not consistently control TSP emissions, which is crucial for lowering PM concentrations.

Figure 5.

The TSP emission situations in 2013: (a) total emission distribution and (b) emission ratios situation.

The ‘New Normal’ may give China time for PM_2.5 control

In the past 10 years, China’s high economic growth and increase in electricity consumption has caused the government to neglect PM pollution control because the government continued to build more power plants to meet electricity demands and prevent the adverse effect of electricity shortage on the economy. Since 2012, the economy has transited to a ‘New Normal’ because the economic growth rate has declined, and the key driving factor is shifting from investment to consumption. As a result, coal consumption slowed and power plants’ generating hours declined in 2014, which makes the huge installed capacity of coal-fired power plants at the edge of a ‘bubble bursting’.

Although the coal-fired power plant bubble may imply a waste of investment, it has positive effects in three aspects. First, as the generating hours of power plants plunge, the TSP emission volumes will also decline, which will result in lower PM_2.5 concentrations in northern China. Second, the electricity supply situation becomes relaxed, which gives the government more time to manage inefficient coal-fired power plants and to realize a negative growth rate of coal consumption. Finally, China has more space and time to develop renewable energies without sacrificing electricity grid stability.

Conclusion

Against the backdrop of the ‘Coal Plants Western Movement’, this paper analyses the relations between PM_2.5 concentration and precipitation, wind speed and weighted coal-fired power plant capacity intensity by using a spatial panel data method. PM_2.5 concentration change is simulated when the proposed and under-construction power plant capacities are put into operation. The simulation results show that these new capacities will impact cities unevenly but will predominantly increase PM_2.5 concentrations in north western China. However, these PM_2.5 concentrations will still be lower than the PM_2.5 concentrations of Beijing, Tianjin, Hebei and Shandong. To lower PM_2.5 concentration in northern China, some energy policies should be sustained.

First, the policy that attempts to realize the negative growth of coal consumption in northern China must be strictly implemented. If coal consumption is simply stabilized against additional growth and increased electricity demands are met by imported electricity, coal consumption and TSP emission will only decrease marginally, and it will be difficult to lower PM_2.5 concentrations. Only by achieving a negative growth rate of coal consumption at the regional level will the ‘APEC Blue’^d dissipate.

Second, there is need to continue to implement the ‘replacing small units with large ones’ policy. The closure of small coal-fired generating units should focus especially on the self-supply generating units with high generating hours. The electricity demand can be met by a direct electricity trade among power plants and large consumers at lower prices.

Third, there is need to enhance the de-sulpharization, de-NOx and smoke emission controls by upgrading emission control technologies. Although the five largest groups have existing installed emissions control facilities in their power plants, the many self-supply plants, CHP plants and small power plants have none or a minimum of these emission control facilities. The emission control facilities with low efficiency should be updated because the technologies vary significantly according to the efficiency in removing pollutants.

Fourth, environmental regulation should be strictly implemented in western China. This is especially the case because precipitation in the region is low, which contributes to high water stress and a fragile ecological system. Although the simulation results suggest that there may not be a significant increase in PM_2.5 concentration in the northwest once the planned coal-fired power plants begin operation, these models do not account for the economic growth that is stimulated by these power plants, which will further induce an increase in coal and oil consumption. When these factors are considered, the PM_2.5 concentration will likely be much higher. Thus, stricter environmental regulation is necessary.

Footnotes

Acknowledgements

The authors would like to thank Huw Slater from China Carbon Forum and Lauri Myllyvirta from Greenpeace for data and comments to this paper. The authors would also like to thank the anonymous referees for helpful comments and suggestions, and the authors are responsible for the paper.

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

This paper is supported by the MOE (Ministry of Education in China) Project of Humanities and Social Sciences (Project No. 13YJC790110), the Xiamen University Innovation Team Project (Project No. 20720151039) and the National Natural Science Foundation of China (No. 71573217).

Notes

Dunguo Mou, Ph.D. of Economics, Associate Professor at China Center for Energy Economics Research, School of Economics, Xiamen University. Dunguo's research interests lie in energy supply and demand, environmental impacts of energy consumption, and energy policies in China. Email: billmmu@xmu.edu.cn.

Matthew Herington, Ph.D. student at the UQ Energy Initiative, University of Queensland. Matthew's research interests lie in improving the effectiveness of energy-led development by identifying the drivers of transformational change amongst the energy impoverished. Email: m.herington@uq.edu.au.

Oluwasola E. Omoju, Ph.D. student at China Center for Energy Economics Research, School of Economics, Xiamen University. Oluwasola's research interests lie in energy consumption and economic development, social and environmental impacts of energy consumption in developing countries. Email: shollcy@yahoo.co.uk.

References

Tai

Mickley

Jacob

. Meteorological modes of variability for fine particulate matter (PM2.5) air quality in the United States: implications for PM2.5 sensitivity to climate change. Atmos Chem Phys 2012; 12: 3131–3145.

Jacob

Winner

. Effect of climate change on air quality. Atmos Environ 2009; 43: 51–63.

Dawson

Adams

Pandis

. Sensitivity of PM2.5 to climate in the Eastern US: a modeling case study. Atmos Chem Phys 2007; 7: 4295–4309.

Chen

Zhao

. Variations in PM10, PM2.5 and PM1.0 in an urban area of the Sichuan basin and their relation to meteorological factors. Atmosphere 2015; 6: 150–163.

Baltagi

Song

Koh

. Testing Panel Data Regression Models with spatial error correlation. J Econometrics 2003; 117: 123–150.

Elhorst

. Specification and estimation of spatial panel data models. Int Reg Sci Rev 2003; 26: 244–268.

Lin B and Yao X. Power industry location optimization and integrative energy transportation system. Econ Res J 2009; 44(6): 105–115. (in Chinese).