Ecological network analysis of the virtual water network within China’s electric power system during 2007–2012

Applied Energy 168 (2016) 110–121

Contents lists available at ScienceDirect

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

Ecological network analysis of the virtual water network within China’s electric power system during 2007–2012 Ruipeng Guo a,⇑, Xiaojie Zhu a, Bin Chen b,⇑, Yunli Yue c a

College of Electrical Engineering, Zhejiang University, Hangzhou 310027, PR China School of Environment, Beijing Normal University, Beijing 100875, PR China c State Grid Jibei Electric Power Company Limited Economic Research Institute, Beijing 100053, PR China b

h i g h l i g h t s
Virtual water flows within China’s electric power system are tracked.
Important grids are identified through information theory.
The impact on water stress mitigation of China’s electric power system is analyzed.

a r t i c l e

i n f o

Article history: Received 27 May 2015 Received in revised form 12 December 2015 Accepted 22 January 2016

Keywords: Virtual water Ecological network analysis Electric power system China

a b s t r a c t Substantial virtual water concurrent electricity is transferred among six grids in China’s electric power system. An in-depth understanding of this virtual water is essential considering the increasingly severe water deficiency in China. Using ecological network analysis (ENA), we investigate the virtual water network within China’s electric power system (VWNCEPS) during 2007–2012, including (1) tracking the virtual water flows from the power generators to the power consumers and analyzing the tendency of the track results, (2) identifying the important grids that largely influence both the magnitude and diversity of the VWNCEPS, and (3) evaluating the overall performance of the VWNCEPS. Additionally, a new indicator is proposed that incorporates the concept of the water stress index (WSI) to measure the impact of the VWNCEPS on national water stress mitigation. The results show that during 2007–2012, the northern and central grids were always the most important input-oriented and output-oriented grids, respectively. Furthermore, the input-oriented impacts of the six grids are similar, while the output-oriented impacts exhibit substantial variations among the different grids. Regarding the overall performance, the VWNCEPS exhibits a high level of system efficiency, whereas its flexibility is relatively low. Moreover, the VWNCEPS has a rapidly increasing positive effect on national water stress mitigation in China during 2007–2012, and the virtual water connection between the central grid and the eastern grid is found to be the main contributor to this positive effect. Ó 2016 Elsevier Ltd. All rights reserved.

1. Introduction With rapidly growing electricity generation, the electric power industry began to pose a significant threat to the normal life of human populations due to its substantial water consumption [1–4], especially in China. China has the largest electric power system in the world [5–7], which consumed 13.18% of total water consumption in China in 2008 [8]. However, China’s annual

⇑ Corresponding authors at: 38 Zheda Street, Hangzhou 310027, PR China. Tel./fax: +86 571 87951542 (R. Guo). 19 Xinjiekouwai Street, Beijing 100875, PR China. Tel./fax: +86 10 58807368 (B. Chen). E-mail addresses: [email protected] (R. Guo), [email protected] (B. Chen). http://dx.doi.org/10.1016/j.apenergy.2016.01.063 0306-2619/Ó 2016 Elsevier Ltd. All rights reserved.

availability of renewable water resources per capita is only 25% of the world average, indicating a severe water deficiency in China. Accordingly, it is essential to examine the water issues within China’s electric power system to make it more water-efficient, thereby mitigating water scarcity problems in China. Development in China is very different between regions. Coastal areas are generally more developed, such as in eastern China, whereas inland areas have experienced much slower development, such as in the northwest. The demand for electricity in the coastal areas has increased faster than in the inland areas due to rapid development. However, the primary energy sources (i.e., coal and hydro) for power generation are located far from the coastal areas of China. Coal is primarily located in the northwest, whereas

R. Guo et al. / Applied Energy 168 (2016) 110–121

111

Nomenclature ENA ecological network analysis VWNCEPS virtual water network within China’s electric power system TEVWF total efficient virtual water flux TSTP total system throughput AMI average mutual information A ascendancy C development capacity O overhead TD total distribution NE northeast NW northwest N north E east C central S south T/H/W/N thermal/hydro/wind/nuclear L liter

hydroelectric power is primarily generated in the southwest [9]. To address this problem, electricity transmission is believed to be a sensible choice because energy resource transmission by train or ship is expensive and results in excess pollution [10]. Moreover, with the rapid increase in the development of China’s electric power system, the capacity of interregional electricity transmission is expected to be much larger in the future [11]. The increasing interregional electricity transmission is expected to largely enhance the interactions among regions including the virtual water interactions [12]. An in-depth understanding of these virtual water interactions is very important when researching the water issues within China’s electric power system. The term ‘virtual water’ was first proposed in 1994 by Allan. It is defined as the water used to produce food crops that are traded internationally [13]. Then, the concept further evolved to represent the volume of water required to produce a commodity or service [14,15]. Based on this concept, regional virtual water accounting and interregional virtual water flows are widely studied using the input–output (IO) [16,17] and multi-region input–output (MRIO) [18–21] frameworks. An advantage of these analyses is that both direct and indirect water consumption is considered. However, no indicator is provided in these analyses to evaluate the results. More specifically, the information embedded within the virtual water results is not well depicted. This gap can be filled using ecological network analysis (ENA) [22]. ENA is a useful tool to holistically analyze the structure and interactive flows in ecosystems [23–25]. Furthermore, both direct and indirect interactions are identified and quantified in ENA [26,27]. The most important feature of ENA is that many processes, such as throughflow analysis and ascendancy analysis, are represented to reveal the mutual relationships among different components [28–30]. ENA has been successfully applied to many types of systems, such as urban systems [31,32], wetlands [33], energy systems [34,35], and water systems as well [36–40]. Sustainable water use in the Yellow River Basin during 1998–2006 has been examined using ENA [41], in which total system throughput intensity (TSTI), which incorporates environmental, social and economic factors, was proposed for a more accurate sustainability analysis. Except for real water systems, virtual water systems have also been investigated using ENA. Mao et al. [42] utilized ENA to investigate the virtual water trade among different sectors in the Baiyangdian Basin in northern China. The boundary inputs and outputs and the

kW h F f ij T j T i Gout Gin Nout Nin NGD NDG EIO EII S

kilowatt hour virtual water transmission matrix virtual water flow from the jth node to the ith node sum of the virtual water flows flowing out from the jth grid sum of the virtual water flows flowing into the ith grid output-oriented virtual water proportion flows through direct paths input-oriented virtual water proportion flows through direct paths total output-oriented virtual water proportion flows total input-oriented virtual water proportion flows output-oriented track matrix input-oriented track matrix impacts of regional power generators on the VWNCEPS impacts of regional power consumers on the VWNCEPS vector of regional water stress indexes

contribution of each sector were detected using the unit environ and final contribution ratio methods, providing feasible ways to optimize the virtual water trade structure by adjusting the relationships among compartments. Moreover, global virtual water trade has also been studied using ENA [43]. The indicator called the integral control intensity was proposed, and the global food trade market was found in a competitive environment. Previous results have shown a strong guiding significance in policy making in different countries to increase the water efficiency of global virtual water trade. These successful applications have demonstrated that ENA is a suitable and powerful approach for the study of complex virtual water systems. This paper aims to analyze the virtual water network within China’s electric power system (VWNCEPS). Interregional virtual water flows are calculated according to the method proposed in [12]. It should be noted that these virtual water flows represent the embodiment of virtual water of power generation in China’s electric power system; the virtual water network composed of these virtual water flows is considered to represent the VWNCEPS in this paper. Then, ENA is used to reveal the information embedded within the VWNCEPS, including (1) tracking the virtual water flows from the power generators to the power consumers and analyzing the tendency of the track results, (2) identifying the important grids that largely influence both the magnitude and diversity of the VWNCEPS, and (3) evaluating the overall performance of the VWNCEPS. Moreover, a new indicator called the total efficient virtual water flux (TEVWF), which incorporates the concept of the water stress index (WSI), is proposed to measure the impact of the VWNCEPS on national water stress mitigation.

2. Method and data 2.1. Study area China’s electricity generation increased by a factor of 11.2 from 1980 to 2009 [44]. In 2010, China became the largest source of electric power in the world. However, China’s rapid development has been primarily dependent on continuous expansion at the expense of production efficiency because the focus was on quantity rather than quality, leading to many environmental impacts including vast amounts of water consumption. With the rapid development of the electric power industry, water consumption

112

R. Guo et al. / Applied Energy 168 (2016) 110–121

Fig. 1. Power transmission pattern between China’s six grids during 2007–2012.

has also increased sharply, which is believed to be an important reason for the severe water scarcity in China. The electric power system in China is divided into six grids: the northern, northeastern, northwestern, central, eastern and southern grids. The power transmission pattern among these six grids during 2007–2012 is depicted in Fig. 1. Tibet and Taiwan are not included in this study because their generation capacity and electricity consumption are small compared to other areas [45] and their connections with other regions are very weak [46]. 2.2. Data sources The power transmission data for six grids during 2007–2012 are derived from the State Grid Corporation of China (SGCC) Power Exchange Annual Report [47] and showed in Table 1. Moreover, the regional power generation data during 2007–2012 are collected from the China electric power yearbook [46] and showed in Table 2. Due to data availability, the water intensities of various types of power generation are assumed to be unchanged during 2007–2012, and the water intensities in [12] (see Table 3) are used as the water intensities during 2007–2012.

2.3. Construction of the virtual water network model The virtual water network model of China’s electric power system is shown in Fig. 2. Each grid is divided into three components, including a ‘‘G part”, which represents the grid’s power generators, a ‘‘D part”, which represents the grid’s power consumers, and a ‘‘grid part”, which represents a hub for the virtual water transmission without any virtual water consumption or virtual water generation. Thus, virtual water flows from the power generators to the power consumers can be analyzed and tracked more intuitively. Moreover, the directional lines among the components in Fig. 2 represent the interregional virtual water flows. For convenience, every component is represented by a numbered node from 1 to 18, and the virtual water flow from the jth node to the ith node is represented by f ij . 2.4. Flow track based on throughflow analysis Throughflow analysis is similar to input–output analysis [48]. In this paper, throughflow analysis is used to track the total virtual water flows (including indirect and direct flows) from the power

Table 1 Power transmission data of China during 2007–2012. Source: [47]. Path

2007 (108 kW h)

2008 (108 kW h)

2009 (108 kW h)

2010 (108 kW h)

2011 (108 kW h)

2012 (108 kW h)

NE to N N to NE NW to N N to E NW to C C to NW C to N N to C C to E E to C C to S

24.65 8.96 53.19 131.4 29.6 0 0 0 350.65 11.65 145.14

52.47 0 129.87 169.04 31 0 0 0 384.64 8.32 139.65

69.73 0 149.85 166.26 35.98 0 33.15 55.48 379.27 3.11 150.33

87.99 0 213 165.48 128.51 16.67 45.51 72.35 395.3 17.36 136.96

100.24 0 486.41 157.7 157.93 14 13.76 56.84 368.17 1.41 137.32

108.98 0 495.37 169.8 215.68 34.52 30.39 102.72 538.66 0 145.43

Note: NE represents the northeastern grid; NW represents the northwestern grid; N represents the northern grid; E represents the eastern grid; and C represents the central grid.

113

R. Guo et al. / Applied Energy 168 (2016) 110–121 Table 2 Regional power generation data of China during 2007–2012. Source: [46]. Region

2007 T/H/W/N (108 kW h)

2008 T/H/W/N (108 kW h)

2009 T/H/W/N (108 kW h)

Northeast Northwest North East Central South

2186/112/13.3/0 1893/533/8.6/0 8381/52.1/24.2/0 6729/465/7/327 4084/2243/0/0 3936/1294/4/302

2264/104/35.4/0 2127/577/16.5/0 8679/48/56.4/0 7099/510.3/15.6/379 4072/2633/0.42/0 3787/1768/6.44/313

2302/102/68.4/0 2221/687/30.5/0 9230/52.1/136.7/0 7647/460/27.5/382 4504/2716/2.4/0 4211/1686/10.7/318

2010 T/H/W/N (108 kW h)

2011 T/H/W/N (108 kW h)

2012 T/H/W/N (108 kW h)

2494/182/113.3/0 2769/822/52.2/0 10396/70.1/266.1/0 8646/729/41.9/414 5127/3089/4.1/0 4733/1961/16.5/334

2683/131/150/0 3612/882/113.03/0 11335/67.74/375.5/0 9992/482/61/447 5978/3041/7.1/0 5399/2052/32.4/425

2708/161/174/0 3612/1042/178.8/0 11767/91.4/516.8/0 10068/744/83.7/508 5686/3855/12.4/0 5201/2646/62.4/474

Northeast Northwest North East Central South

Note: T/H/W/N represents thermal/hydro/wind/nuclear; NE represents the northeastern grid; NW represents the northwestern grid; N represents the northern grid; E represents the eastern grid; and C represents the central grid.

Table 3 Regional water intensities of various types of power generation of China. Source: [12]. Type

Northeast

North

Northwest

East

Central

South

Hydropower (L/kW h) Thermal (L/kW h) Nuclear (L/kW h) Wind (L/kW h)

14.82 3.83 3.10 0.56

18.69 3.53 3.10 0.56

25.92 3.46 3.10 0.56

15.53 2.60 3.10 0.56

14.87 4.08 3.10 0.56

16.96 2.18 3.10 0.56

Fig. 2. Virtual water network model within China’s electric power system. (Note: D represents power demand; G represents power generation; NE represents the northeastern grid; NW represents the northwestern grid; N represents the northern grid; E represents the eastern grid; and C represents the central grid).

generators to the power consumers. The virtual water transmission matrix, F, is composed of the direct virtual water flows as follows:

F ¼ ff ij g1818

ð1Þ

To calculate the indirect virtual water flows, we transform the direct virtual water flows into ratios between 0 and 1. Because both of the two terminals can be chosen as the denominator, there are two forms (output-oriented and input-oriented forms):

114

R. Guo et al. / Applied Energy 168 (2016) 110–121

Gout ¼ fg ij g where g ij ¼ f ij =T j

ð2Þ

Gin ¼ fg 0ij g where g 0ij ¼ f ij =T i

ð3Þ

where T j represents the sum of the virtual water flows flowing out from the jth grid; T i represents the sum of the virtual water flows flowing into the ith grid; and matrices Gout and Gin represent the output-oriented and input-oriented virtual water proportion flows through paths with length 1 (direct flows), respectively. Based on Gout and Gin , the paths with lengths exceeding 1 (indirect flows) can be easily calculated. For example, g 12 and g 23 represent the output-oriented virtual water proportion flows through the paths with length 1 (2 ? 1 and 3 ? 2, respectively). Then, the output-oriented virtual water proportion flows through the path with length 2 (3 ? 2 ? 1) can be calculated by multiplying g 12 by g 23 . Thus, using the characteristics of matrix multiplication, the output-oriented and input-oriented virtual water proportion flows through paths with length m can be expressed as Gm out and Gm , respectively. As a result, the total output-oriented and inputin oriented virtual water proportion flows from one component to another can be calculated as follows: 1 Nout ¼ G0out þ G1out þ G2out þ G3out þ . . . þ Gm out ¼ ðI  Gout Þ

Nin ¼

G0in

þ

G1in

þ

G2in

þ

G3in

þ ... þ

Gm in

¼ ðI  Gin Þ

1

the system [51]. The upper limit to ascendency is the development capacity (C). The difference between A and C is defined as the overhead (O), which represents the flexibility of the system [50]. When every component has only one flow leaving and one entering, A is equivalent to C, indicating that A has reached its maximum and O is zero. However, at the same time, every component has only one channel to transfer its virtual water out or in, leading to a level of flexibility that is very low; more specifically, the network is vulnerable to collapse when disturbed [51]. Thus, A is not necessarily better when it is larger. A good development of a system also requires adequate amounts of overhead, and the ratio between A and C (A/ C) is often used to represent the development level of a system [50]. These traditional indicators can reflect most characteristics of the VWNCEPS. However, the individual impacts on the VWNCEPS are not well represented. Based on information theory, we calculate the information entropy of each grid to investigate the impact of each grid on the VWNCEPS. Large information entropy corresponds to large impacts on the VWNCEPS. The output-oriented and input-oriented information entropies of each grid can be calculated as follows:

ð4Þ ð5Þ

EIOj ¼ 

6 X ðf ij =TSTPÞ  log2 ðf ij =TSTPÞ ðj ¼ 1; 2; 3; 4; 5; 6Þ

ð8Þ

i¼1

There are 18 rows and 18 columns in both N out and N in . However, only 6 rows and 6 columns are needed to track the virtual water flows from the power generators to the power consumers, which can be expressed as:

EIIi ¼ 

6 X ðf ij =TSTPÞ  log2 ðf ij =TSTPÞ ði ¼ 1; 2; 3; 4; 5; 6Þ

ð9Þ

j¼1

where N GD is the output-oriented track matrix, which represents the virtual water flow track from the power generators to the power consumers (a production perspective); and N DG is the inputoriented track matrix, which represents the virtual water flow track from the power consumers to the power generators (a consumption perspective).

EIO represents the impacts of regional power generators on the VWNCEPS; and EII represents the impacts of regional power consumers on the VWNCEPS. Moreover, the same amount of water consumption may yield very different impacts in different grids [52]. The WSI concept defined and advanced by Pfister et al. is adopted in this paper to obtain more accurate information related to the VWNCEPS [53]. A novel indicator called the total efficient virtual water flux (TEVWF) is proposed:

2.5. Overall evaluation based on ecological information theory

TEVWF ¼

NGD ¼ Nout ð7 : 12; 1 : 6Þ

ð6Þ

NDG ¼ Nin ð7 : 12; 1 : 6Þ

ð7Þ

18 X 18 X ððSðj  12Þ  Sði  12ÞÞ  f ij þ ðSði  12Þ i¼13 j¼i

From the throughflow analysis, the virtual water interactions between the power generators and the power consumers can be quantitatively analyzed. However, the overall performance of the VWNCEPS is not well analyzed, which may provide different information. Based on information theory, which was proposed by Shannon [49], Ulanowicz proposed several indicators to measure the characteristics of an entire system [50]. Regarding the virtual water system in this paper, the indicators can be calculated as shown in Table 4. The system size can be quantified using the total system throughput (TSTP), and the system complexity can be measured using the average mutual information (AMI). Ascendancy (A), combined with the TSTP and AMI, is used to represent the efficiency of

 Sðj  12ÞÞ  f ji Þ

ð10Þ

where S represents the vector of regional water stress indexes. In [52], the exporter’s WSI is used to generate virtual scarce water flows. The focus of Feng et al. was on modifying the magnitude of virtual water flows. However, in this paper, we use both the exporter’s WSI and the importer’s WSI to modify the traditional virtual water flows, which is shown in Eq. (10). The modification is not only applied to the magnitude of virtual water flows but also to their directions. The sum of these modified virtual water flows, TEVWF, can be used to determine the impact of the VWNCEPS on national water stress mitigation. 3. Results and discussion

Table 4 Four traditional indicators used in the ecological network analysis. Average Mutual Information (AMI) Total System Throughput (TSTP) Ascendency (A) Development Capacity (C) Note: F is equal to TSTP.

AMI ¼

P18 P18

TSTP ¼

i¼1

j¼1 ðf ij =FÞ

 log2 ðf ij  F=ðT i  T j ÞÞ

P18 P18 i¼1

j¼1 f ij

A ¼ AMI  TSTP P18 P18 i¼1 j¼1  f ij  log2 ðf ij =FÞ

C¼

3.1. Virtual water flow track The annual average output-oriented track matrix and inputoriented track matrix are shown in Tables 5 and 6, respectively. Each column in Table 5 represents the output-oriented track results in one grid; thus, the sum of each column is 1. Similarly, each column in Table 6 represents the input-oriented track results in one grid, and the sum of each column is 1. From a production perspective, Table 5 shows that from 2007 to 2012, most of the virtual water from the regional power generators was consumed by

115

R. Guo et al. / Applied Energy 168 (2016) 110–121 Table 5 Annual average output-oriented track matrix during 2007–2012.

Northeast Northwest North East Central South Sum

Northeast (%)

Northwest (%)

North (%)

East (%)

Central (%)

South (%)

97.30288 0.00002 2.64321 0.04115 0.01251 0.00022 100

0.00038 91.08322 6.31092 0.21890 2.34445 0.04213 100

0.01731 0.00069 98.03434 1.55620 0.38461 0.00684 100

0 0.00005 0.00018 99.92781 0.07047 0.00149 100

0 0.10142 0.24679 5.08828 92.73171 1.83180 100

0 0 0 0 0 100 100

Table 6 Annual average input-oriented track matrix during 2007–2012.

Northeast Northwest North East Central South Sum

Northeast (%)

Northwest (%)

North (%)

East (%)

Central (%)

South (%)

99.94611 0.00077 0.05312 0 0 0 100

0.00000 99.76085 0.00090 0.00007 0.23817 0 100

0.77710 4.88143 93.92813 0.00015 0.41319 0 100

0.01306 0.18243 1.59719 89.13915 9.06817 0 100

0.00228 1.16395 0.22995 0.03846 98.56536 0 100

0.00005 0.02796 0.00555 0.00112 2.64657 97.31874 100

the local power consumers. Additionally, from a consumption perspective, Table 6 shows that the majority of the virtual water consumption was from local power generators during 2007–2012, which was due to the interregional power transmission being small compared to the total power generation and power consumption. Although power transmission had increased by 273% from 2007 to 2012, only 4.1% of the generated national power was transferred within the electric transmission system in 2012. It can be seen in Table 5 that all of the virtual water from the power generators in the south was consumed by local power consumers, indicating that the southern grid never transferred virtual water to the other grids during 2007–2012. Regarding the northeastern grid, 2.6% of the virtual water from the power generators in the northeast flowed into the northern grid, accounting for 98% of the total virtual water outflow of the northeastern grid. Considering the northwestern grid, the northern and central grids accounted for 70.8% and 26.3% of its total virtual water outflow, respectively. For the eastern grid, more than 99.9% of the virtual water from its power generators was consumed by local power consumers; its largest virtual water proportion outflow was to the central grid, which was only 0.07%. The central grid was the only grid that had four virtual water proportion outflows exceeding 0.1%, of which the largest two were the flows into the eastern (5.1%) and southern (1.8%) grids. Moreover, Table 6 shows that in the northeastern and northwestern grids, more than 99% of the virtual water consumption was from the local power generators. Additionally, more than 6% of the virtual water consumption from power generators in other grids was found in the northern and eastern grids. Considering the northern grid, only 93.9% of the virtual water consumption was from the local power generators, whereas 4.9% of the virtual water consumption was from the northwestern grid, indicating a large power dependence on the northwestern grid. For the eastern grid, the power generators in other grids contributed approximately 11% of its virtual water consumption, 97.62% of which was from the power generators in the central grid. Regarding the southern grid, 97.3% and 2.6% of its virtual water consumption were from the local power generators and the power generators in the central grid, respectively. The tendencies in the output-oriented track results and inputoriented track results during 2007–2012 are shown in Figs. 3 and

4, respectively, based on the time series data. It is very difficult and meaningless to show and analyze the tendencies of all elements in the output-oriented track matrix and the input-oriented track matrix; therefore, the results were classified and filtered. The detailed classification criterion is shown in Table 7. External virtual water flow means the sum of virtual water proportion flows from (into) other grids, i.e., the local virtual water proportion flows (level 0) are excluded. The water proportion flows accounting for 10–100% of the external virtual water flows will be identified as level 1, while 0.1–10% level 2 and less than 0.1% level 3. Only track results belonging to level 0 and level 1 are depicted in Figs. 3 and 4. Because the southern grid never transferred virtual water to other grids during 2007–2012, it is excluded in Fig. 3. According to Fig. 3, the northeastern, northwestern and northern grids exhibited a decrease in their output-oriented track results in level 0 during 2007–2012. Additionally, the output-oriented track results in level 0 of the northern grid exhibited a fluctuation during 2007–2012, which was mainly due to the nearly opposite tendencies in the virtual water impacts in level 1 from the eastern and central grids. Moreover, the eastern and central grids exhibited an increase in their output-oriented track results in level 0 during 2007–2012. The output-oriented track results in level 0 of the eastern grid decreased substantially in 2010, although it rebounded quickly in 2011. Regarding the central grid, it exhibited a large decrease in 2012 primarily due to the large increase in the virtual water impact in level 1 from the eastern grid and the virtual water impact in level 1 from the southern grid was nearly unchanged in 2012. Similarly, according to Fig. 4, the northwestern, northern and central grids exhibited a decrease in their input-oriented track results in level 0 during 2007–2012. Additionally, the northeastern and southern grids exhibited an increase in their input-oriented track results in level 0 during 2007–2012. Besides, three of the six grids (the northern, eastern and central grids) have two grids in their output-oriented track results in level 1. Moreover, according to Figs. 3 and 4, all of the grids belonging to the track results in level 1 of one grid are directly connected with it, and meanwhile, all of the grids directly connected with it belong to level 1 (see Fig. 1). For example, four grids, the northern, northwestern, eastern and southern grids, were directly connected to the central grid, and these four regions are all belonged to the

116

R. Guo et al. / Applied Energy 168 (2016) 110–121

1.00

NE

NW

N

E

C

Level 0

0.99

0.032

0.98

0.024

0.97

0.016

0.96

0.008

0.95

0.000

1.00

0.125

0.96

0.100

0.92

0.075

0.88

0.050

0.84

0.025

0.80

0.000

0.987

0.020

0.984

0.016

0.981

0.012

0.978

0.008

0.975

0.004

0.972

0.000

1.0000

0.0020

0.9995

0.0016

0.9990

0.0012

0.9985

0.0008

0.9980

0.0004

0.9975

0.0000

0.945

0.060

0.938

0.048

0.931

0.036

0.924

0.024

0.917

0.012

0.910

Level 1

0.040

0.000

2007 2008 2009 2010 2011 2012 Northeast Northwest North

2007 2008 2009 2010 2011 2012 East Central South

Fig. 3. Virtual water track results from the power generators to the power consumers during 2007–2012. (Note: NE represents the northeastern grid; NW represents the northwestern grid; N represents the northern grid; E represents the eastern grid; and C represents the central grid).

track results in level 1 of the central grid during 2007–2012. The results show conformity between the physical connections and the virtual water connections. To obtain a more intuitive understanding, the output-oriented virtual water proportion flows listed in Table 5 are plotted in Fig. 5. The boldness of the line is used to distinguish the different levels. It can be found that the output-oriented virtual water proportion flows belonging to level 1 are mainly from the inland areas to the coastal areas, whereas most output-oriented virtual water proportion flows that belong to level 2 and level 3 exhibit the opposite direction. In [52], Feng et al. investigated the virtual scarce water in China, demonstrating that the water consumption in developed coastal provinces is heavily dependent on water resources in the water scarce northern provinces that are manly located in inland regions. The reason maybe that coastal regions are generally more developed and in need of electricity than inland regions, and inland regions generally have ample electricity that

coastal regions desperately need. Thus, virtual water will be transferred from coastal regions to inland regions concurrent with electricity. It is believed that with rational responsibility allocation of water consumption based on quantitative accounting and analysis, forthcoming water consumption mitigation policies such as water cap schemes and water tax will help incline both the energyproducing and consuming regions follow a coordinated development roadmap. The output-oriented track results from a production perspective may provide valuable information for water-related cost allocation within the electric power system. For example, the water tax can be priced according to the different levels. Regions in level 1 should have the highest price, while those in level 2 lower and in level 3 the lowest. Additionally, the input-oriented track results from a consumption perspective may provide valuable information for the water-related benefit distribution within the electric power system. For example, the economic benefit brought by using

117

R. Guo et al. / Applied Energy 168 (2016) 110–121

Level 1

Level 0

NE

NW

N

E

C

1.001

0.00375

1.000

0.00300

0.999

0.00225

0.998

0.00150

0.997

0.00075

0.996 1.002

0.00000 0.0080

1.000

0.0064

0.998

0.0048

0.996

0.0032

0.994

0.0016

0.992 1.00

0.0000 0.10

0.98

0.08

0.96

0.06

0.94

0.04

0.92

0.02

0.90 0.92

0.00 0.120

0.91

0.096

0.90

0.072

0.89

0.048

0.88

0.024

0.87 1.000

0.000 0.025

0.995

0.020

0.990

0.015

0.985

0.010

0.980

S

0.975

0.005

0.970

0.000

0.980

0.036

0.976

0.032

0.972

0.028

0.968

0.024

0.964

0.020 2007 2008 2009 2010 2011 2012

2007 2008 2009 2010 2011 2012

Northeast

East

Northwest

Central

North

South

Fig. 4. Virtual water track results from the power consumers to the power generators during 2007–2012. (Note: NE represents the northeastern grid; NW represents the northwestern grid; N represents the northern grid; E represents the eastern grid; and C represents the central grid). Table 7 Classification criterion of virtual water proportional flows. Level

Definition

Level 0 Level 1

Local virtual water proportion flows 10–100% (exclude 10%) of the external virtual water proportion flows 0.1–10% (exclude 0.1%) of the external virtual water proportion flows 0–0.1% (exclude 0) of the external virtual water proportion flows

Level 2 Level 3

Note: External virtual water proportion flow = 1  local virtual water proportion flow.

the electricity can be allocated according to the different levels. Regions in level 1 should have the biggest benefit, while those in level 2 smaller and in level 3 the smallest.

3.2. Identification of the important grids Fig. 6 shows the input-oriented impacts of six grids on the entire system. The northern was always the most important input-oriented grid and has the largest input-oriented information entropy during 2007–2012. The input-oriented impact of the central grid was initially less than that of the southern grid at first; however, in 2012, the impact exceeded that of the southern grid. Moreover, the northwestern grid had zero input-oriented impact during 2007–2010 because there was no virtual water flowing into it. Moreover, during 2010–2012, this grid began receiving virtual water from other grids, although its input-oriented impact remained small compared to other grids. Fig. 7 shows the output-oriented impacts of the six grids on the entire system. The central grid was always the most important output-oriented grid during 2007–2012. The output-oriented

118

R. Guo et al. / Applied Energy 168 (2016) 110–121

Level 3 Level 2 Level 1 Fig. 5. Annual average output-oriented virtual water proportion flows in China’s electric power system during 2007–2012. (Note: NE represents the northeastern grid; NW represents the northwestern grid; N represents the northern grid; E represents the eastern grid; and C represents the central grid).

2.8 2.6 2.4 2.2 2.0 1.8 1.6

Northeast Northwest North East Central South

1.4 1.2 1.0 0.8 0.6 0.4 0.2 0.0 2007

2008

2009

2010

2011

2012

Fig. 6. Input-oriented impacts of the six grids during 2007–2012.

2.8 2.6 2.4 2.2 2.0

Northeast Northwest North East Central South

1.8 1.6 1.4 1.2 1.0 0.8 0.6 0.4 0.2 0.0 2007

2008

2009

2010

2011

2012

Fig. 7. Output-oriented impacts of the six grids during 2007–2012.

impact of the northwestern grid increased substantially from 2007 to 2012, whereas the output-oriented impacts of the northeastern and northern grid exhibited only slight changes from 2007 to 2012. In 2012, there was no virtual water flowing from the eastern grid; thus, the output-oriented impact of the eastern grid was zero. Moreover, the southern grid never transferred virtual water to the other grids during 2007–2012; thus, its output-oriented impact was always zero. The results of regional input-oriented impacts and regional output-oriented impacts identified the important grids from consumption perspective and production perspective, respectively. The power consumption of more important input-oriented grids has more influence on the VWNCEPS. Also, the power consumption in the northern had the largest impact on the VWNCEPS, which is consistent with the results of [7]. Besides, the power generation of more important output-oriented grids has more influences on the VWNCEPS, showing that the power consumption in the central had the largest impact on the VWNCEPS. Three Gorges Dam may be the main reason, which had contributed an installed capacity of 18.2 GW with power generation of 79.85 TW h in 2009 and most of the power generation are transferred to other regions [54]. Thus, the power consumption of most important input-oriented grids and the power generation of most important output-oriented grids could be the most effective way to change the VWNCEPS. The sum of the input-oriented impacts and the sum of the output-oriented impacts are identical; this quantity is called the total distribution (TD). The TD reflects the characteristics of the virtual water flow distribution. Both an additional virtual water transmission path and a more evenly distributed virtual water flow can increase the TD. Table 8 shows detailed information regarding the virtual water flows during 2007–2012. From 2007 to 2008, the number of paths decreased, and the virtual water flow became more evenly distributed, leading to an increase in the TD. From 2010 to 2011, the number of paths remained unchanged, while

119

R. Guo et al. / Applied Energy 168 (2016) 110–121 Table 8 Virtual water flows during 2007–2012. Path

2007 (108 L)

2008 (108 L)

2009 (108 L)

2010 (108 L)

2011 (108 L)

2012 (108 L)

NE to N N to NE NW to N N to E NW to C C to NW C to N N to C C to E E to C C to S

107.01 32.68 445.03 479.25 247.66 0 0 0 2770.32 42.21 1146.68

223.39 0 1065.72 619.69 254.39 0 0 0 3196.89 30.55 1160.69

292.37 0 1300.83 610.49 312.34 0 268.69 203.72 3074.13 10.96 1218.48

388.41 0 1807.30 608.62 1090.41 134.95 368.42 266.10 3200.08 65.08 1108.73

416.02 0 3740.11 583.42 1214.36 107.67 105.82 210.28 2831.36 4.70 1056.04

460.38 0 4060.94 631.15 1768.10 289.30 254.69 381.81 4514.37 0 1218.81

Note: NE represents the northeastern grid; NW represents the northwestern grid; N represents the northern grid; E represents the eastern grid; and C represents the central grid.

the TD decreased substantially, indicating a very uneven distribution in 2011. 3.3. Overall evaluation The results of the four traditional indicators introduced in Section 2 are shown in Fig. 8. The VWNCEPS functioned well during 2007–2012 when only considering the increasing ascendency (A). The annual growth rate of the ascendency was approximately 10.1% during this study. Similarly, the TSTP also exhibited an upward tendency during 2007–2012, with an annual growth rate of 10.5%. However, the AMI index fluctuated during 2007–2012, and its general trend was downward with an annual decrease of 0.4%. To better understanding its changes, the AMI can be considered to represent the logarithm of the number of roles in the VWNCEPS. A role represents a group of nodes that obtains its inputs from one source and passes them to a single destination [55]. More roles mean that more nodes worked independently. Thus, the decreased AMI index means that the mutual virtual water relationship between the various grids became increasingly strong. Moreover, the A/C ratio increased during 2007–2012, with an annual decrease 0.5%. In 2007, the A/C ratio was 0.94, indicating that the ascendancy was very high and the overhead was relatively low in the development stage. More specifically, the efficiency of the VWNCEPS was very high, whereas the flexibility of the

3.360

VWNCEPS was very low, indicating that the VWNCEPS was vulnerable to disturbances. Moreover, the decreased A/C ratio demonstrates that the VWNCEPS developed toward being more robust during 2007–2012 [51]. Fig. 9 shows the TEVWF during 2007–2012. The TEVWF was always negative during 2007–2012, which indicates that the VWNCEPS tended to transfer virtual water from grids with low WSI to those with high WSI, yielding a positive effect on national water stress mitigation. Moreover, this negative value increased during 2007–2012 at an annual growth rate of 16.4%. Furthermore, the growth rate of the negative value rapidly increased each year during 2009–2012. In 2012, the growth rate exceeded 33.7%, i.e., a fourfold increase in three years. Moreover, Fig. 9 also shows the contribution of different virtual water connections to the TEVWF during 2007–2012. Note that the virtual water connections correspond to the virtual water relationships between two grids and have no specific direction. The results show that virtual water connection ‘‘C–E” was the main contributor to the negative TEVWF values, especially in 2007, when more than 87.7% of the negative TEVWF component was derived from virtual water connection ‘‘C–E”. Although this percentage decreased during 2007–2012, it was still as high as 57.6% in 2012. The continuous increase in virtual water connection ‘‘N– NW” was the primary cause for the proportional decrease in virtual water connection ‘‘C–E”. In 2007, virtual water connection

600000

3.344

540000

AMI

3.328

480000

3.312

420000

3.296

360000

3.280

2007

2008

2009

2010

2011

2012

300000

2000000

0.948

1800000

0.942

1600000

2007

2008

2009

2010

2011

2012

2010

2011

2012

0.936

A

1400000

TSTP

A/C

0.930 0.924

1200000

0.918

1000000 2007

2008

2009

2010

2011

2012

2007

2008

Fig. 8. Four traditional indicators of the VWNCEPS.

2009

120

R. Guo et al. / Applied Energy 168 (2016) 110–121

S-C C-NW

C-E N-NW

C-N N-NE

E-N TEVWF

2000 1000 0 -1000 -2000 -3000 -4000 -5000 -6000 -7000 -8000

2007

2008

2009

2010

2011

2012

Fig. 9. Contributions of various virtual water connections to the TEVWF during 2007–2012. (Note: NE represents the northeastern grid; NW represents the northwestern grid; N represents the northern grid; E represents the eastern grid; and C represents the central grid).

effect on mitigating national water stress in China during 2007– 2012, and the virtual water connection between the central grid and the eastern grid was found to be the main contributor to this positive effect during 2007–2012. It is expected that electric power systems will incorporate an increasing amount of renewable energy in the future [54,56]. According to the Asia Pacific Economic Cooperation (APEC) in 2014, the share of renewables in the APEC energy mix is expected to double by 2030. However, in China, hydropower resources are mainly located in the southern grid, and wind and solar power resources are primarily located in the northern, northwestern, and northeastern grids, while electricity consumption is concentrated in developed areas, such as the eastern and central grids. Additionally, renewable resources are largely dependent on the terrain and climatic conditions and cannot be transported in manners typically used for fossil fuels. Electric power transmission is inevitable when these renewable energy resources are utilized. Accordingly, the magnitude of interregional electric power and virtual water transmission will grow rapidly in the future, and the structure of the virtual water network may become substantially different. This paper provides a feasible way to discover the valuable information embedded within these changes, which may help construct a more water-efficient electric power system. Acknowledgements

‘‘N–NW” only accounted for 10.3% of the negative TEVWF component, whereas in 2012, it accounted for more than 37.6%. Moreover, during 2007–2012, there were three virtual water connections (‘‘E–N”, ‘‘C–S” and ‘‘C–NW”) that exhibited positive contributions to the TEVWF. The contribution of virtual water connection ‘‘C– NW” largely exceeded the other two virtual water connections; thus, the majority of the positive TEVWF component was derived from virtual water connection ‘‘C–NW”. Although, the contribution of virtual water connection ‘‘C–NW” increased by more than 497.1% from 2007 to 2012, it remained relatively small compared to virtual water connections ‘‘N–NW” and ‘‘C–E”, directly resulting in the terrible development tendency of the TEVWF that was discussed above. TEVWF reflects the impact of the VWNCEPS on national water stress mitigation. The results show that during 2007–2012, the VWNCEPS shows rapidly increasing positive effect on national water stress mitigation in China. To maintain and further increase the positive effect of the VWNCEPS, policy can be made according to the contribution of different virtual water connections. The large contributors of the positive and negative TEVWF components such as virtual water connections ‘‘C–NW” and ‘‘C–E”, could be the key objects when making decisions. 4. Conclusions The present study investigated the VWNCEPS by applying ENA. Using throughflow analysis, virtual water was tracked from the power generators to the power consumers, providing a valuable reference for water-related cost allocation and the water-related benefit distribution. Moreover, important grids that largely influence both the magnitude and diversity of the VWNCEPS were identified through ecological information theory, facilitating more efficient and targeted policy making. Furthermore, a novel indicator that incorporates the WSI concept was proposed to measure the VWNCEPS’s impact on national water stress mitigation. The results demonstrated that during 2007–2012, the northern and central grids were always the most important input-oriented and output-oriented grids, respectively. Regarding the overall performance, the VWNCEPS exhibited high efficiency and low flexibility. Additionally, the VWNCEPS exhibited a rapidly increasing positive

This work was supported by the Fund for Creative Research Groups of the National Natural Science Foundation of China (No. 51121003), National Natural Science Foundation of China (Nos. 71573021, 41271543), Specialized Research Fund for the Doctoral Program of Higher Education of China (No. 20130003110027), and National Science and Technology Supporting Program (No. 2013BAA01B02). References [1] Baum E, Chaisson J, Evans L, Lewis J, Marshall D, Thompson J. The last straw: water use by power plants in the arid west. Land and Water Fund of the Rockies; 2003. [2] Pfister S, Saner D, Koehler A. The environmental relevance of freshwater consumption in global power production. Int J Life Cycle Assess 2011;16:580–91. [3] Torcellini P, Long N, Judkoff R. Consumptive water use for US power production. Tech. Rep. NREL/TP-550-33905; Golden, CO; 2003. [4] Macknick J, Newmark R, Heath G, Hallett KC. Operational water consumption and withdrawal factors for electricity generating technologies: a review of existing literature. Environ Res Lett 2012;7:045802. [5] NBS-National Bureau of Statistics of China. China statistical yearbook 2013. Beijing: China Statistics Press; 2013 [in Chinese]. [6] Jiang Y. China’s water scarcity. J Environ Manage 2009;90:3185–96. [7] Zhang C, Anadon LD. Life cycle water use of energy production and its environmental impacts in China. Environ Sci Technol 2013;47:14459–67. [8] NBS-National Bureau of Statistics of China. China economic census yearbook 2008. Beijing: China Statistics Press; 2011 [in Chinese]. [9] CEP-Editorial Board of the China Electric Power Yearbook. China electric power yearbook 2031. Beijing: China Electric Power Press; 2013 [in Chinese]. [10] Chen G, Chen B, Zhou H, Dai P. Life cycle carbon emission flow analysis for electricity supply system: a case study of China. Energy Policy 2013;61:1276–84. [11] Zhu F, Zheng Y, Guo X, Wang S. Environmental impacts and benefits of regional power grid interconnections for China. Energy Policy 2005;33:1797–805. [12] Zhu XJ, Guo RP, Chen B, Zhang J, Hayat T, Alsaedi A. Embodiment of virtual water of power generation in the electric power system in China. Appl Energy 2015;151:345–54. [13] Al Alawi J, Abdulrazzak M. Water in the Arabian Peninsula: problems and perspectives. In: Water in the Arab World: perspectives and progress division of applied sciences. Cambridge, Mass: Harvard University; 1994. [14] Allan JA. Water stress and global mitigation: water food and trade. Arid Lands Newsl 1999;45. [15] Hoekstra A, Weitsel A, Starink M. Perspectives on water: an integrated modelbased exploration of the future. Utrecht, the Netherlands: International Books; 1998. [16] Zhao X, Yang H, Yang Z, Chen B, Qin Y. Applying the input-output method to account for water footprint and virtual water trade in the Haihe River Basin in China. Environ Sci Technol 2010;44:9150–6.

R. Guo et al. / Applied Energy 168 (2016) 110–121 [17] Zhao X, Chen B, Yang ZF. National water footprint in an input–output framework—a case study of China 2002. Ecol Model 2009;220:245–53. [18] Lenzen M. Understanding virtual water flows: a multiregion input–output case study of Victoria. Water Resour Res 2009;45:W09416. [19] Guan D, Hubacek K. Assessment of regional trade and virtual water flows in China. Ecol Econ 2007;61:159–70. [20] Chen Z-M, Chen GQ. Virtual water accounting for the globalized world economy: national water footprint and international virtual water trade. Ecol Ind 2013;28:142–9. [21] Zhang C, Anadon LD. A multi-regional input–output analysis of domestic virtual water trade and provincial water footprint in China. Ecol Econ 2014;100:159–72. [22] Leontief W. Input–output economics. Oxford: Oxford University Press; 1986. [23] Patten BC, Bosserman RW, Finn JT, Cale WG. Propagation of cause in ecosystems. Syst Anal Simul Ecol 1976;4:457–579. [24] Ulanowicz RE. Growth and development: ecosystems phenomenology. New York: Springer-Verlag; 1986. [25] Fath BD, Scharler UM, Ulanowicz RE, Hannon B. Ecological network analysis: network construction. Ecol Model 2007;208:49–55. [26] Fath BD, Patten BC. Review of the foundations of network environ analysis. Ecosystems 1999;2:167–79. [27] Fath BD, Halnes G. Cyclic energy pathways in ecological food webs. Ecol Model 2007;208:17–24. [28] Ulanowicz RE. Quantitative methods for ecological network analysis. Comput Biol Chem 2004;28:321–39. [29] Fath BD. Distributed control in ecological networks. Ecol Model 2004;179:235–45. [30] Lu Y, Su M, Liu G, Chen B, Zhou S, Jiang M. Ecological network analysis for a low-carbon and high-tech industrial park. Sci World J 2012;2012:305474. [31] Chen S, Fath BD, Chen B. Information indices from ecological network analysis for urban metabolic system. Procedia Environ Sci 2010;2:720–4. [32] Zhang Y, Yang Z, Fath BD. Ecological network analysis of an urban water metabolic system: model development, and a case study for Beijing. Sci Total Environ 2010;408:4702–11. [33] Mao X, Yang Z, Chen B, Chen H. Examination of wetlands system using ecological network analysis: a case study of Baiyangdian Basin, China. Procedia Environ Sci 2010;2:427–39. [34] Chen S, Chen B. Urban energy consumption: different insights from energy flow analysis, input–output analysis and ecological network analysis. Appl Energy 2015;138:99–107. [35] Zhang Y, Yang Z, Fath BD, Li S. Ecological network analysis of an urban energy metabolic system: model development, and a case study of four Chinese cities. Ecol Model 2010;221:1865–79. [36] Chen S, Fath B, Chen B, Su M. Evaluation of the changed properties of aquatic animals after dam construction using ecological network analysis. Procedia Environ Sci 2011;5:114–9.

121

[37] Chen S, Chen B, Su M. The cumulative effects of dam project on river ecosystem based on multi-scale ecological network analysis. Procedia Environ Sci 2011;5:12–7. [38] Chen S, Fath BD, Chen B. Ecological risk assessment of hydropower dam construction based on ecological network analysis. Procedia Environ Sci 2010;2:725–8. [39] Li Y, Chen B. Study on sustainable water use of the Haihe River Basin using ecological network analysis. Front Earth Sci China 2009;3:419–30. [40] Bodini A, Bondavalli C. Towards a sustainable use of water resources: a wholeecosystem approach using network analysis. Int J Environ Pollut 2002;18:463–85. [41] Li Y, Chen B, Yang ZF. Ecological network analysis for water use systems—a case study of the Yellow River Basin. Ecol Model 2009;220:3163–73. [42] Mao X, Yang Z. Ecological network analysis for virtual water trade system: a case study for the Baiyangdian Basin in Northern China. Ecol Inform 2012;10:17–24. [43] Yang Z, Mao X, Zhao X, Chen B. Ecological network analysis on global virtual water trade. Environ Sci Technol 2012;46:1796–803. [44] Liu Z. Electric power and energy in China. Beijing: China Electric Power Press; 2012. [45] NBS-National Bureau of Statistics of China. China statistical yearbook 2008– 2013. Beijing: China Statistics Press; 2008–2013 [in Chinese]. [46] CEP-Editorial Board of the China Electric Power Yearbook. China electric power yearbook 2008–2013. Beijing: China Electric Power Press; 2008–2013 [in Chinese]. [47] State Grid Corporation of China. State grid corporation of China power exchange annual report; 2007–2012 [in Chinese]. [48] Zhang Y, Yang Z, Yu X. Ecological network and emergy analysis of urban metabolic systems: model development, and a case study of four Chinese cities. Ecol Model 2009;220:1431–42. [49] Shannon CE. A mathematical theory of communication. ACM Sigmoble Mobile Comput Commun Rev 2001;5:3–55. [50] Ulanowicz RE, Patten BC. Ecology, the ascendent perspective. Estuaries 1999;22:339–41. [51] Kharrazi A, Rovenskaya E, Fath BD, Yarime M, Kraines S. Quantifying the sustainability of economic resource networks: an ecological informationbased approach. Ecol Econ 2013;90:177–86. [52] Feng K, Hubacek K, Pfister S, Yu Y, Sun L. Virtual scarce water in China. Environ Sci Technol 2014;48:7704–13. [53] Pfister S, Koehler A, Hellweg S. Assessing the environmental impacts of freshwater consumption in LCA. Environ Sci Technol 2009;43:4098–104. [54] Liu W, Lund H, Mathiesen BV, Zhang X. Potential of renewable energy systems in China. Appl Energy 2011;88:518–25. [55] Zorach AC, Ulanowicz RE. Quantifying the complexity of flow networks: how many roles are there? Complex 2003;8:68–76. [56] Sayigh A. Renewable energy — the way forward. Appl Energy 1999;64:15–30.