You are on page 1of 13

Journal of Hydrology 595 (2021) 126018

Contents lists available at ScienceDirect

Journal of Hydrology
journal homepage: www.elsevier.com/locate/jhydrol

Research papers

Determination method for influence zone of pumped storage underground


cavern and drainage system
Cheng Cao a, Zengguang Xu a, *, Junrui Chai a, b, Yuan Qin a, Jing Cao a
a
State Key Laboratory of Eco-hydraulics in Northwest Arid Region, Xi’an University of Technology, Xi’an 710048, China
b
School of Civil Engineering, Xijing University, Xi’an 710123, China

A R T I C L E I N F O A B S T R A C T

This manuscript was handled by Huaming Guo, The construction and operation of underground cavern and drainage system for pumped storage are threatened
Editor-in-Chief, with the assistance of Mingjie by groundwater seepage. Numerical simulation methods are commonly used for seepage analysis of cavern and
Chen, Associate Editor drainage system. However, the method for determining the influence zone of the cavern and drainage system is
unclear. The model range is difficult to confirm, and the accuracy of the analysis results is unwarranted. In this
Keywords:
study, the engineering and geological characteristics of 56 operational pumped storages were summarised.
Influence zone
Numerical simulation experiments were conducted based on steady seepage analysis of simplified pumped
Numerical simulation
Model range storage with cavern and drainage system. The influences of the model range on the groundwater level and flow
Seepage analysis rate of the cavern and drainage system were investigated, and a method for determining the influence zone was
Cavern and drainage system proposed for the cavern and drainage system. The results indicated that unclear influence zone of the cavern and
drainage system existed in numerous pumped storages. Moreover, a small model range strongly overstated the
results regarding the flow rate and groundwater level, as the groundwater drawdown at the computation
boundary was ignored. Increased influence factors, including the right and left bank groundwater surface slope
ratio, maximum water head depth above the cavern, and volume of the cavern, delayed the stable tendencies of
the changes of the flow rate and groundwater level with an increased model range. With a receivable flow rate
threshold, the proposed method could be used to predict the influence zone for the cavern and drainage system,
and the optimal model range could be confirmed to improve the calculation accuracy of seepage analysis.

1. Introduction 2020; Trautz and Wang, 2002; Zhang and Goh, 2015; Goh and Zhang,
2012). Furthermore, the ecological environmental pressure caused by
Pumped storage (PS) is necessary for renewable energy and emer­ reducing the groundwater level is inestimable (Javadi et al., 2016;
gency backup power. The planning and construction of PSs are impor­ Bodeux et al., 2017; Xu et al., 2020).
tant for energy structure adjustments and environmental protection Numerical simulation methods are commonly used in seepage
(Bonalumi et al., 2012; Xu et al., 2019; Li et al., 2014). The construction analysis of PS cavern and drainage system with the aid of computers.
of PSs is gradually increasing worldwide, especially in China. By the end However, the computation results for the flow rate and groundwater
of 2019, the installed capacity of PSs had reached 158 GW. China’s level frequently show significant differences from the actual results. One
installed capacity of PSs accounts for 19% of the world’s total, ranking of the main reasons is the improper simulation model range. Darcy’s law
first worldwide. By 2035, the installed capacity of PSs in China is pre­ is the basis of most numerical simulation methods (Chen et al., 2010;
dicted to reach 120 GW. The underground power station is the core of Dang et al., 2020; Himi et al., 2018; Zhang and Goh, 2012). According to
the PS, it includes underground cavern and drainage system. Equipment Darcy’s law, the flow rate of a computation section is determined by its
and operation sites are placed in the cavern. The drainage system, which area, hydraulic conductivity and hydraulic gradient. The computation
mainly includes drainage hole arrays, is usually distributed around the section area and hydraulic conductivity of the stratum are constant and
cavern of the PS to drain the groundwater. Similar to other underground can be confirmed from engineering characteristics and hydrogeologic
engineering problems, the seepage problems caused by groundwater in data. The hydraulic gradient is typically determined by the distance and
PSs are disastrous, and difficult to forecast (Cao et al., 2019; Goh et al., water head difference between the computation section and

* Corresponding author.
E-mail address: xuzengguang@xaut.edu.cn (Z. Xu).

https://doi.org/10.1016/j.jhydrol.2021.126018
Received 23 December 2020; Received in revised form 12 January 2021; Accepted 15 January 2021
Available online 29 January 2021
0022-1694/© 2021 Elsevier B.V. All rights reserved.
C. Cao et al. Journal of Hydrology 595 (2021) 126018

Fig. 1. Section and planform sketches of groundwater level with a different location of constant water head boundary: (a) pumping latent-water well with the
influence radius (Bresciani et al., 2020), (b) PS cavern and drainage system with the influence zone (the solid line represents the cavern, and the dotted line indicates
the drainage system; hereinafter they are the same).

computation boundaries. The water head difference is affected by the is moved from #1 to #2, and the other parameters remain constant.
measured water level of the computation boundaries, and the distance is Thus, the values of the groundwater drawdown and environmental
influenced by the location of computation boundaries. The computation stress caused by pumping latent-water well are underestimated. As
boundaries can be located in a study area possessing a specific water­ shown in Fig. 1(b), as the boundary conditions of the wells of the cavern
shed, valley, reservoir and river. When the geological conditions above and drainage system are generally potential seepage boundaries, the
do not exist, the model range and locations for computation boundaries groundwater depression cone still exists when we replace the pumping
are usually confirmed by experts. latent-water well with the PS cavern and drainage system. In contrast to
Earlier studies concerning the influences of the model range on the pumping latent-water well in Fig. 1(a), the PS study area usually
seepage analysis results can be found in the field of well hydraulics possesses an upstream reservoir and a downstream reservoir or river,
(Bresciani et al., 2014, 2020). As shown in Fig. 1(a), in the study of these act as stable water sources for replenishing groundwater. The
Bresciani et al (2020), when a pumping latent-water well is in operation shape of the groundwater depression zone caused by the cavern and
with homogeneous stratum, a groundwater depression cone is formed drainage system is close to a rectangle when the topography of the right
after the groundwater swarms into the well. Ordinarily, the shape of the and left banks extends horizontally (Fig. 1(b)). Additionally, the influ­
groundwater depression cone in the planform is circular. Moreover, an ence zone for the PS cavern and drainage system is formed which is
influence radius clearly exists in the pumping latent-water well (John­ similar to the situation of the influence radius. Thus, for the computation
son, 1988; Bresciani et al., 2020). The phenomenon of the influence boundaries of the PS cavern and drainage system simulation model, the
radius in the pumping latent-water well can be described by the Dupuit constant water head boundary in the right and left banks should be
solution (Dupuit, 1863) as follows: located at or beyond the influence zone of the cavern and drainage
system (#1 constant water head boundary in Fig. 1(b)). Therefore, for
Qw Rw
(2H0 − sw )sw = ln (1) pumping latent-water wells or PS cavern and drainage system, the model
π K rw
range (R) can be confirmed as equal to the known influence zone.
where H0 is the depth of the aquifer, sw is the groundwater draw­ Thus, it is necessary to confirm the value of R for seepage analysis of
down in the pumping well, Qw is the pumping flow rate, K is the aquifer the PS cavern and drainage system. However, the determination method
hydraulic conductivity, Rw is the influence radius of the pumping well for the influence zone of the cavern and drainage system and influence
and rw is the radius of the pumping well. mechanism of R on the seepage analysis results are not clear, and related
When the seepage analysis simulation model is used for the pumping research is scarce. Based on the definition of the influence radius for the
latent-water well in Fig. 1(a), the constant water head boundary should pumping latent-water well, the groundwater drawdown at the influence
be located at or beyond the influence radius (#1 constant water head radius should be infinitesimal and approach zero. Hence, the value of
boundary in Fig. 1(a)), so as to avoid the influence of the groundwater the influence radius was determined to be infinite. For ease of applica­
drawdown in the computation boundaries. According to Eq. (1), the tion, the drawdown threshold of the groundwater level was defined as
values of Rw and sw are reduced when the constant water head boundary shown in Fig. 1(a). The influence radius of the pumping latent-water

2
C. Cao et al. Journal of Hydrology 595 (2021) 126018

Table 1
Characteristic statistics of it of the right and left banks, Hymax and size of the cavern of 56 operational PSs.
Number Pumped storage Country it of right bank it of left bank Hymax Size of cavern
(m)
Width Height Length V (m3)
(m) (m) (m)

1 Kuhtai Austria Rise to ridge with it = 0.629, Rise to ridge with it = 0.787, 400 / / / /
Lp = 0.74 km Lp = 0.85 km
2 Limberg II Austria Rise to ridge with it = 0.633, Rise to ridge with it = 0.546, 290 34 50 100 170,000
Lp = 2.20 km Lp = 1.50 km
3 Coo – Trois-Ponts Belgium Down to valley with it = Down to valley with it = 270 27 40 128 138,240
0.372, Lp = 1.20 km 0.401, Lp = 1.90 km
4 Chaira Bulgaria Rise to ridge with it = 0.321, Down to valley with it = 650 22 43 111 105,006
Lp = 2.50 km 0.211, Lp = 3.80 km
5 Bailianhe China Down to flat with it = 0.121 Down to valley with it = 210 22 51 147 164,934
0.102, Lp = 2.50 km
6 Guangzhou China Down to valley with it = Down to valley with it = 514 20 44 149 131,120
0.435, Lp = 3.70 km 0.328, Lp = 3.30 km
7 Heimifeng China Down to valley with it = Down to valley with it = 295 27 60 136 220,320
0.481, Lp = 0.88 km 0.263, Lp = 0.91 km
8 Minghu China Down to flat with it = 0.090 Down to valley with it = 316 21 46 127 122,682
0.224, Lp = 4.30 km
9 Taian China Down to flat with it = 0.078 Down to flat with it = 0.069 225 25 54 280 378,000
10 Tianhuangping China Down to valley with it = Down to valley with it = 610 20 46 194 178,480
0.158, Lp = 2.53 km 0.188, Lp = 3.51 km
11 Xilongchi China Down to valley with it = Down to valley with it = 624 22 49 150 161,700
0.329, Lp = 3.07 km 0.503, Lp = 2.85 km
12 Yixing China Down to valley with it = Down to flat with it = 0.295 353 22 52 155 177,320
0.189, Lp = 2.41 km
13 Zhanghewan China Down to valley with it = Down to valley with it = 300 24 50 152 182,400
0.163, Lp = 5.21 km 0.284, Lp = 2.63 km
14 Huhehaote China Down to valley with it = Down to valley with it = 521 24 50 152 182,400
0.040, Lp = 2.33 km 0.356, Lp = 0.91 km
15 Xianju China Down to valley with it = Extend gently 544 26 55 176 251,680
0.402, Lp = 3.22 km
16 Pushihe China Down to valley with it = Rise to ridge with it = 0.222, 330 23 54 140 173,880
0.060, Lp = 3.31 km Lp = 1.30 km
17 Xianyou China Rise to ridge with it = 0.053, Down to valley with it = 470 24 67 162 260,496
Lp = 0.74 km 0.076, Lp = 1.31 km
18 Hongping China Down to valley with it = Down to valley with it = 540 22 49 157 169,246
0.439, Lp = 2.44 km 0.178, Lp = 2.81 km
19 Xiangshuijian China Rise to ridge with it = 0.097, Down to valley with it = 218 25 56 175 245,000
Lp = 0.71 km 0.133, Lp = 1.51 km
20 Baoquan China Extend gently Rise to ridge with it = 0.719, 510 23 50 101 116,150
Lp = 0.25 km
21 Dlouhe Strane Czech Down to valley with it = Down to valley with it = 545 27 52 120 168,480
0.394, Lp = 3.05 km 0.340, Lp = 1.93 km
22 La Coche France Down to valley with it = Down to valley with it = 930 12 30 64 23,040
0.526, Lp = 1.82 km 0.295, Lp = 2.84 km
23 Le Cheylas France Down to valley with it = Down to valley with it = 490 / / / /
0.277, Lp = 6.03 km 0.331, Lp = 3.11 km
24 Revin France Down to valley with it = Down to valley with it = 240 17 41 114 79,458
0.333, Lp = 1.12 km 0.035, Lp = 4.12 km
25 Grand Maison France Rise to ridge with it = 0.485, Rise to ridge with it = 0.474, 905 16 40 160 102,400
Lp = 2.25 km Lp = 2.11 km
26 Super Bissorte France Rise to ridge with it = 0.459, Rise to ridge with it = 0.392, 1,164 16 40 90 57,600
Lp = 1.62 km Lp = 1.32 km
27 Montezic France Down to valley with it = Extend gently 416 25 40 145 145,000
0.110, Lp = 2.52 km
28 Goldisthal Germany Down to valley with it = Down to valley with it = / 26 49 137 174,538
0.410, Lp = 1.67 km 0.206, Lp = 1.85 km
29 Hornbergbecken Germany Down to valley with it = Down to valley with it = 1,000 20 33 219 144,540
(Wehr) 0.218, Lp = 1.93 km 0.285, Lp = 1.05 km
30 Markersbach Germany Down to valley with it = Down to valley with it = 288 24 44 149 157,344
0.147, Lp = 1.44 km 0.230, Lp = 2.74 km
31 Waldeck Germany Down to valley with it = Down to valley with it = 336 34 50 105 178,500
0.315, Lp = 2.05 km 0.203, Lp = 4.12 km
32 Siah Bishe Iran Rise to ridge with it = 0.444, Rise to ridge with it = 0.328, 505 27 46 126 156,492
Lp = 0.98 km Lp = 1.10 km
33 Lago Delio Italy Down to valley with it = Down to valley with it = 720 21 60 172 216,720
0.366, Lp = 2.83 km 0.312, Lp = 3.65 km
34 Entracque Chiotas Italy Rise to ridge with it = 0.503, Rise to ridge with it = 0.801, 1,048 / / / /
Lp = 1.28 km Lp = 1.78 km
35 Anapo (Solarino) Italy Extend gently Down to valley with it = 335 20 40 155 124,000
0.124, Lp = 3.58 km
36 Edolo Italy Down to valley with it = Rise to ridge with it = 0.301, / 16 47 175 131,600
0.199, Lp = 4.62 km Lp = 2.01 km
37 Provvidenza Italy / 19 25 129 61,275
(continued on next page)

3
C. Cao et al. Journal of Hydrology 595 (2021) 126018

Table 1 (continued )
Number Pumped storage Country it of right bank it of left bank Hymax Size of cavern
(m)
Width Height Length V (m3)
(m) (m) (m)

Rise to ridge with it = 0.218, Rise to ridge with it = 0.273,


Lp = 0.55 km Lp = 1.31 km
38 Imaichi Japan Down to valley with it = Down to valley with it = 540 34 51 160 277,440
0.518, Lp = 1.05 km 0.215, Lp = 2.64 km
39 Numappara Japan Down to valley with it = Down to flat with it = 0.167 500 22 46 131 132,572
0.315, Lp = 1.91 km
40 Okumino Japan Down to valley with it = Down to valley with it = 633 20 44 125 110,000
0.299, Lp = 2.54 km 0.223, Lp = 3.62 km
41 Okutataragi Japan Rise to ridge with it = 0.353, Rise to ridge with it = 0.386, 406 25 49 134 164,150
Lp = 0.68 km Lp = 1.27 km
42 Ohkawachi Japan Down to valley with it = Down to valley with it = 395 24 46 134 147,936
0.187, Lp = 3.74 km 0.552, Lp = 1.17 km
43 Sabigawa Japan Rise to ridge with it = 0.143, Down to valley with it = 362 29 52 165 248,820
Lp = 0.28 km 0.463, Lp = 2.66 km
44 Tamahaya Japan Rise to ridge with it = 0.154, Rise to ridge with it = 0.192, 518 26 50 116 150,800
Lp = 2.23 km Lp = 0.52 km
45 Vianden Luxembourg Down to valley with it = Extend gently 275 17 30 330 168,300
0.426, Lp = 1.63 km
46 Aurland Norway Extend gently Down to valley with it = 400 / / / /
0.442, Lp = 6.72 km
47 Porabka-zar Poland Down to valley with it = Down to valley with it = 432 27 40 124 133,920
0.267, Lp = 1.22 km 0.243, Lp = 1.83 km
48 Cierny Vach Slovakia Down to valley with it = Down to valley with it = 434 21 48 90 90,720
0.301, Lp = 1.92 km 0.483, Lp = 1.55 km
49 La Muela I Spain Down to valley with it = Down to valley with it = 522 24 49 111 130,536
0.469, Lp = 1.77 km 0.058, Lp = 0.66 km
50 Linth–Limmern Switzerland Rise to ridge with it = 0.437, Rise to ridge with it = 0.450, / / / / /
Lp = 1.09 km Lp = 1.98 km
51 Cruachan UK Rise to ridge with it = 0.513, Rise to ridge with it = 0.568, 348 24 38 92 83,904
Lp = 1.11 km Lp = 1.08 km
52 Dinorwic UK Down to flat with it = 0.240 Down to valley with it = 544 24 52 180 224,640
0.435, Lp = 0.66 km
53 Racoon US Extend gently Down to valley with it = 310 22 24 150 79,200
0.441, Lp = 0.72 km
54 Bad Creek US Extend gently Down to valley with it = 324 27 40 137 147,960
0.373, Lp = 0.73 km
55 Helms US Extend gently Rise to ridge with it = 0.513, 541 22 47 100 103,400
Lp = 2.04 km
56 Northfield US Extend gently Down to valley with it = 226 25 38 100 95,000
0.273, Lp = 1.12 km

well could be confirmed using the ratio between the drawdown further works were introduced.
threshold and groundwater drawdown at the well, with a suggested ratio
range of 0.001–0.1 (Johnson, 1988; Bresciani et al., 2020). In this study, 2. Methodology
numerical simulation experiments were conducted to establish a rela­
tionship between the cavern and drainage system flow rate (Q) and R, 2.1. Statistics of pumped storage characteristics
and to examine three influence factors. A flow rate threshold (△Qt) was
proposed by referring to the groundwater drawdown threshold. Influ­ Table 1 presents the statistical results for engineering and geological
ence zone of the PS cavern and drainage system could be confirmed with characteristics of 56 operational PSs, including the topography slope
receivable △Qt and the relationship between Q and R, then the optimal ratio (it) of the right and left banks, maximum hydroelectric head
value of R which is equal to the influence zone can be ensured. (Hymax), and size of the cavern. In Table 1, Lp is the planform distance
The paper was organized as follows: In Section 2.1, it is demonstrated from the cavern and drainage system to the ridge or valley. As shown in
that that unclear influence zone or influence zone with large values exist Table 1, the right and left bank topographies of most PSs are uneven. The
in numerous PSs based on a statistical analysis of 56 operational PSs. In topography extension form is illustrated in Fig. 2 and can be divided into
Section 2.2, the influences of different R on Q values and groundwater four types: (1) rise to the ridge with 26.7% topography of the right bank
levels were analysed, and a basic equation was established between Q and 26.7% topography of the left bank; (2) decline to the valley with
and R. In Section 2.3, three major influence factors were confirmed after 53.6% topography of the right bank and 62.5% topography of the left
comparing the PS cavern and drainage system with a pumping latent- bank; (3) decline to the flat with 7.1% topography of the right bank and
water well based on the Dupuit solution. The value distribution of the 5.4% topography of the left bank; (4) extend gently with 12.5% topog­
influence factors was summarised. In Section 2.4, the numerical simu­ raphy of the right bank and 5.4% topography of the left bank.
lation experiments were introduced, and the influence factors were As the watershed usually exits in the ridge, the computation
changed for factor analysis. In Sections 3.1 and 3.2, single factor and boundary should be located at the watershed as a confining boundary
multivariate regression analysis were conducted on the regression co­ (green dashed lines in Fig. 2(a)). The groundwater level in the valley of
efficients of the basic equation for Q and R and influence factors. In Fig. 2(b) was typically a constant result, owing to the steady water
Section 4.1, we proposed a determination method for the influence zone supply from high groundwater. Hence, a constant water head boundary
of PS cavern and drainage system. In Section 4.2, the accuracy of the (red dashed lines in Fig. 2(b)) could be installed. The influence zone of
proposed method was verified. In Section 4.3, limitations and valuable the PS cavern and drainage system was specific in Fig. 2(a) and (b), but

4
C. Cao et al. Journal of Hydrology 595 (2021) 126018

(a) (b)

0.25–2.50 km
0.66–6.72 km
Cavern and drainage system Cavern and drainage system

(c) (d)

Cavern and drainage system Cavern and drainage system

Fig. 2. Topography sketch of PS right or left banks based on Table 1: (a) rise to ridge, (b) decline to valley, (c) decline to flat, (d) extend gently.

k
ban
ht
Rig Up
str Drainage hole arrays
eam
res
erv
oir
1.2 km

0.28 km
m Cavern
1.41 km

65k
1.9 20
10 m
×0 30 °
.2 k
m
Do 5
w 15 m
50.27 m

nst 7.
rea 1.4 k 01
m
res m km
m
erv
oi 65
ro
r ri 3.7 bank
ver 10 × Le
ft 20
0.2 k m .4 m 20 m
m 20 m 25

Fig. 3. Simulation model of the simplified PS, cavern and drainage system to investigate the influence of R on Q and groundwater level.

the influence zone in Fig. 2(b) was large, as the valley was far away from difficult to confirm.
the cavern and drainage system, in some cases, the distance approached
6.72 km. In Fig. 2(c) and (d), the influence zone was unclear. Thus, when
the right and left bank topographies of the PS declined to the valley 2.2. Relationship between model range and seepage analysis results
(which was far from the cavern and drainage system) or declined to the
flat or extended gently, the values of the influence zone and R were To further analyse the influences of R on the seepage analysis results,
a numerical simulation model for a simplified PS with cavern and

5
C. Cao et al. Journal of Hydrology 595 (2021) 126018

1.4 km 1.8 km 2.2 km


2.6 km 3.0 km 3.4 km
3.8 km 4.2 km 4.6 km
5.0 km 5.4 km

Cavern and drainage system

10×0.2 km 1.4 km 10×0.2 km

Fig. 4. Results of the groundwater level with increased R in the simulation model of Fig. 3.

large R (Fig. 4), as did the changes in hydrostatic pressure above the
90 cavern and drainage system. Q decreased from 87.56 L/s to 61.89 L/s
when R increased from 1.4 km to 5.4 km (Fig. 5). The nonlinear decrease
of the groundwater level and Q with increased R demonstrated an
85 evolution law similar to the change in Qw and sw with various Rw ac­
cording to Eq. (1). The relationship between Q and R after the regression
analysis could be expressed as follows:
80
Q=96.2213R-0.2691
Q = aRb (2)
R2=0.9973
Q (L/s)

75 where a and b are regression coefficients.

2.3. Influence factors


70
According to the analysis of the difference between the pumping
latent-water well and PS cavern and drainage system based on the
65 Dupuit solution, the influence factors affecting the influence zone of the
PS cavern and drainage system were summarised as follows: (1) The PS
study area was generally very large with varied topography, and the
60
2 3 4 5 groundwater surface in most situations was not horizontal, which
deviated from the assumption for the Dupuit solution (Huang et al.,
R (km) 2014). Hence, the groundwater surface slope ratio (ig) should be
Fig. 5. Regression analysis of Q and R in the simulation model of Fig. 3. considered. (2) The boundary conditions of the cavern and drainage
system were potential seepage boundaries and not flow rate boundaries.
Thus, the influence zone of the PS cavern and drainage system was
drainage system was established using ABAQUS (Dassault Systèmes,
theoretically unaffected by the hydraulic conductivity of the homoge­
2017). The cavern had a width, height, length and buried depth of
neous stratum. (3) The potential seepage boundary of the cavern and
25.40, 50.27, 157.01 and 1200 m, respectively (Fig. 3). The drainage
drainage system did not intersect the groundwater surface or aquifer,
hole arrays around the cavern had a diameter, space and distance from
and the water head depth above the cavern had a similar effect on H0 in
the cavern of 5 cm, 5 m and 20 m, respectively, and the top drainage
Eq. (1) to a certain extent. Therefore, the maximum water head depth
hole arrays were obliquely distributed upward by 30◦ . The drainage
above the cavern (Hmax) must be considered. (4) According to Eq. (1), Rw
holes in the simulation model were simplified to line elements (Xu et al.,
was affected by rw when other parameters were constant; hence, the size
2020). The groundwater level was assumed to be 50 m below the earth’s
of the PS cavern was another influence factor that must be considered.
surface. The hydraulic conductivity of the homogeneous stratum was set
In this study, similar to the simplified simulation model in Fig. 3, we
as 10− 7 m/s. The walls of the cavern and drainage hole were potential
assumed that the groundwater level was 50 m below the earth’s surface.
seepage boundaries, and the boundary conditions of the upstream,
Thus, the slope ratio of the groundwater surface and topography were
downstream and right and left banks were constant water head
equal, and Hmax was equal to the maximum buried depth of cavern
boundary. The influence zone of the cavern and drainage system was
minus 50 m. Based on Table 1, Fig. 6 shows the weight distribution of it,
constant in upstream and downstream, owing to the upper and lower
Hymax and the size of the cavern with different value ranges. As indicated
reservoirs. Hence, we symmetrically extended R in both the right and
in Fig. 6(a), the downhill it in the right bank changed from 0.040 to
left banks by 200 m each time (all simulation models in this study abided
0.526 and that in the left bank changed from 0.035 to 0.719; the uphill it
by this principle), to investigate the changes in Q and the groundwater
in the right bank changed from − 0.633 to − 0.143, and that in the left
level. Thus, a group of numerical simulation experiments (11 numerical
bank changed from − 0.810 to − 0.273. the downhill it in the right bank
simulation experiments in a group) was conducted with steady seepage
was mostly distributed in 0–0.2 (38.9%) and 0.2–0.4 (36.1%), and that
analysis.
in the left bank was predominantly distributed in 0.2–0.4 (38.9%). Ac­
The results of the groundwater level and Q with different R values are
cording to Fig. 6(b), the Hymax of PSs in Table 1 significantly changed
shown in Figs. 4 and 5, respectively. R strongly influenced the results of
from 210 m to 1164 m. In addition, most PSs were operational with a
the groundwater level and Q, the value of groundwater level and Q was
low hydraulic head in which 40.4% of the PS hydraulic heads were in
substantially overstated with small values of R. The groundwater level
the range of 400–600 m, while 40.4% were lower than 400 m. Fig. 6(c)
significantly decreased with a small R but only slightly decreased with a
shows that the size of the PS cavern was considerably different. It also

6
C. Cao et al. Journal of Hydrology 595 (2021) 126018

16.1%

4
0.
16.1%

i≥
5.8%
39.3%

ax
2

ym
0.

H
i≥
23.2%
4>
0.

25.0% 3.8%

ax
0

ym

>H
i

37.5%
2>

00
Range of Hymax (m)
0.
Range of it

10
0.0%
.2
-0

3.6%
i≥

ax
0>

ym
H
0>
7.1%
4

80
0.
≥-

5.3%
>i
.2

40.4%
-0

ax
8.9%

ym
6
0.

>H
≥-

10.7%
>i
.4
-0

0
3.6% it of right bank

40
40.4%
6

ax <
0.

3.6% it of left bank


i<

ym
H
0 5 10 15 20 25 0 5 10 15 20 25
Statistical magnitude Statistical magnitude
(a) (b)
0
≥3
0 ≥6
th 5.9% Width g ht 5.9% Height
i d ei
W H 0
≥2
5 ≥5
th g ht
id 29.4%
Range of Height (m) ei 31.4%
Range of wideth (m)

>W 0>H
3 0 6 0
≥2
0 ≥4
th g ht
i d 51.0% ei 49.0%
>W 0 >H
2 5 5 5 30
≥1 h t≥
d th ei
g
i 11.7% 9.8%
>W 0 >H
2 0 4
5 30
<1 ht<
d th 2.0% ei
g 3.9%
i H
W
00 25
≥3 2.0% Length 7.8% V
t h V≥
ng
Le 0 0
23 ≥2
h≥ V
Range of length (m)

t
Range of V ×104 (m3)

2.0% > 9.8%


ng 25
> Le
0 60 5
30 h ≥1 ≥1
gt 21.5% > V 33.3%
L en 20
0> 0
23 ≥9 0
g th ≥1
en 72.5% > V 33.3%
0>L 15
16 0
<9 10
th 15.8%
n g 2.0% V<
Le
0 10 20 30 40 0 5 10 15 20 25 30
Statistical magnitude (c) Statistical magnitude

Fig. 6. Weight distribution of the engineering and geological characteristics of 56 operational PSs in Table 1: (a) it (the minus sign signifies topography rises to the
ridge), (b) Hymax, (c) cavern size.

shows that the width changed from 12 m to 34 m, the height changed 2.4. Simulation models
from 24 m to 67 m, length changed from 64 m to 330 m, and V changed
from 23,040 to 378,000 m3. The width values were mainly distributed in Based on Table 1 and the assumptions of this study, the value ranges
20–25 (51.0%) and 25–30 (29.4%) m, and 49.0% and 31.4% of the of ig, Hmax and V are shown in Fig. 7. Corresponding numerical simu­
height values were in the range of 40–50 and 50–60 m, respectively. lation experiments were conducted by changing the model in Fig. 3 to
Moreover, 72.5% of the length values were in the range of 90–160 m, examine the influences of ig, Hmax and V on the relationship between Q
and 33.3% of the V values were in the range of 105–1.5 × 105 and 1.5 × and R. We divided ig, Hmax and V into five values (Table 2). The width,
105–2 × 105 m3, respectively. height and length values in Table 2 were confirmed by the average value
ratio of the width, height and length of the cavern in Table 1 at a scale of

7
C. Cao et al. Journal of Hydrology 595 (2021) 126018

Earth surface of a increased non-linearly with the increases in ig, Hmax and V. The
increased ig, and V values led to decreased b when b increased with the
increase in Hmax. The nonlinear effects of ig, Hmax, and V on a and b could
ig=0.000 be summarised as follows:

ig=0.553 a = eα1 ig , a = Hmax


α2
, a = V α3 (3)

b = eβ1 ig , b = Hmax
β2
, b = V β3 (4)

where αi and βi (i = 1, 2, 3) are regression coefficients.


Hmax=210 m
3.2. Multivariate regression analysis

Fifteen groups of numerical simulation experiments were conducted


Hmax=1,164 m to further scrutinise the comprehensive influences of ig, Hmax and V on
the regression coefficients a and b of Eq. (2). Fig. 11 displays the results
of Q and R when fitted using Eq. (2). Table 3 shows the relationships
V=23,040 m3
between i, Hmax and V and the regression coefficients a and b. In addi­
V=378,000 m3 tion, a multivariate regression analysis was performed using the results
in Table 3.
The results show that regression coefficients a and b changed non­
linearly with various ig, Hmax and V values. The typical multiplicative
Fig. 7. Sketch of the range of ig, Hmax and V in Table 1. form was used to establish the nonlinear empirical relationships, and
functional forms among a and b and ig, Hmax and V were expressed as
follows (Peng and Zhang, 2011; Wen et al., 2017):
Table 2
Value classification of ig, Hmax and V. a = α1 eα2 ig ⋅Hmax
α3
⋅V α4 (5)
ig Hmax (m) Size of cavern
b = β1 eβ2 ig ⋅Hmax
β3
⋅V β4 (6)
V (m3) Width (m) Height (m) length (m)

0 150 23,000 12.34 24.43 76.29 where αi and βi (i = 1, 2, 3, 4) are regression coefficients.
0.14 400 111,750 20.90 41.37 129.21 Eqs. (5) and (6) could be transformed into the linear functions by
0.28 650 200,500 25.40 50.27 157.01 taking the natural logarithm on both sides before conducting the mul­
0.42 900 289,250 28.70 56.80 177.41 tiplicative regression analysis, as follows:
0.56 1150 378,000 31.38 62.10 193.96
lna = lnα1 + α2 ig + α3 lnHmax + α4 lnV (7)

1:1.979:6.181. Based on simulation models in Fig. 3 and Table 2, five lnb = lnβ1 + β2 ig + β3 lnHmax + β4 lnV (8)
groups of numerical simulation experiments were conducted, based on
changing the ig values of the right and left banks simultaneously (Fig. 8 Eqs. (7) and (8) could be expressed after the multivariate regression
(a)). In addition, five groups of numerical simulation experiments were analysis as follows:
conducted by changing Hmax (Fig. 8(b)). Another, five groups of nu­
a = 0.028e0.366i ⋅Hmax
0.995
⋅V 0.095 R2 = 0.9290 (9)
merical simulation experiments were conducted by changing V (Fig. 8
(c)).
b = − 8.356e2.237i ⋅Hmax
− 0.595
⋅V 0.056 R2 = 0.8610 (10)

3. Results
4. Discussion
3.1. Single factor analysis
4.1. Determination of the influence zone of the PS cavern and drainage
system
Fig. 9 shows the relationship between Q and R with the influence of
ig, Hmax and V. The relationship between Q and R still followed Eq. (2)
An infinite R was required in the PS cavern and drainage system in
when ig, Hmax and V varied within the ranges of the statistical results. As
this study when the reduction of Q tended to be infinitesimal or zero.
shown in Fig. 9(a) and (b), Q decreased with increased ig and decreased
According to the relationship between Q and R with the nonlinear in­
Hmax (when R was constant) because of the reductions in the hydrostatic
fluences of ig, Hmax and V, we established arbitrary Ri (i = 1, 2, 3, …, n)
pressure and hydraulic gradient. Fig. 9(c) illustrates that the area of the
values, which led to corresponding Qi (i = 1, 2, 3, …, n) values. Referring
potential seepage boundary increased immediately with increasing V.
to the determination of the influence radius, the flow rate threshold
Hence, additional groundwater flowed into the cavern and drainage
(△Qt) could be expressed as follows:
system with a corresponding increase in Q. The impact of V on Q was
consistent with Eq. (1), as Qw increased with increased rw when other ΔQt = Qn − Qn− 1 = aRbn − aRbn− 1 (11)
parameters were set to a certain value. Furthermore, Fig. 9 explicitly
shows that ig, Hmax and V had strong influences on the slope of the where a and b can be confirmed using Eqs. (9) and (10).
regression. That is, with a small R and increased ig, Hmax and V values From another perspective, △Qt can be treated as the receivable
overestimated the absolute value of the slope and delayed the stable calculation error of Q. For different PS engineering, the value of △Qt or
tendency of the evolution of Q and the groundwater level. Thus, ig, Hmax the calculation error is influenced by the corresponding characteristics
and V strongly affect the influence zone of the PS cavern and drainage and requirements of such techniques. If △Qt is confirmed in the specific
system. PS cavern and drainage system, then the value of R representing the
Fig. 10 illustrates the influence mechanism of ig, Hmax and V on the influence zone of the PS cavern and drainage system, can be calculated
regression coefficients a and b of Eq. (2). As shown in Fig. 10, the value using the following steps: (1) Calculate the regression coefficients a and
b according to Eqs. (9) and (10), respectively, with specific values of ig,

8
C. Cao et al. Journal of Hydrology 595 (2021) 126018

10×200 m 1400 m 10×200 m


i =0
i =0.14
i =0.28
i =0.42
i =0.56

Cavern and draiange system

Model range extension Model range extension


(a)

10×200 m 1400 m 10×200 m


H max=1150 m

H max=900 m

H max=650 m

H max=400 m

H max=150 m

Cavern and draiange system

Model range extension (b) Model range extension

10×200 m 1400 m 10×200 m

V =23000 m3
V =111750 m3
V =200500 m3
V =289250 m3
V =378000 m3

Draiange system Cavern

Model range extension (c) Model range extension

Fig. 8. Sketch of the extension of the simulation model range in PS right and left banks with the changes of influencing factors: (a) ig, (b) Hmax and (c) V.

Hmax and V. The value range of ig, Hmax and V in Eqs. (9) and (10) should 4.2. Verification of the influence zone determination method
be 0–0.553, 150–1,150 m, and 23,000–378,000 m3, respectively. (2)
Substitute the calculated a and b values and accredited △Qt into Eq. A group of simulation models for a simplified PS with cavern and
(11), in which Rn− 1 = Rn − 0.4 km. Then, the influence zone of the drainage system (Fig. 12) was used to verify the determination method
cavern and drainage system can be inversely computed, and is equal to of influence zone, and a contrastive analysis was conducted between the
the optimal R. results of the numerical simulation and those from the proposed
methods. The values of ig, Hmax and V, which accounted for a large
proportion of the statistical results, were confirmed in Fig. 6 at ig = 0.2,

9
C. Cao et al. Journal of Hydrology 595 (2021) 126018

ig=0 ig=0.14 ig=0.28 ig=0.42 ig=0.56 Hmax=150 m Hmax=400 m Hmax=650 m Hmax=900 m Hmax=1150 m
Q R2=0.9973 R2=0.9990 R2=0.9981 R2=0.9978 R2=0.9974 Q R2=0.9975 R2=0.9984 R2=0.9983 R2=0.9987 R2=0.9973
Slop Slop
90 0 0

80
80
-10

-5

Slop of regression curve

Slope of regression curve


70 60
-20
Q (L/s)

Q (L/s)
60
-30 40
-10
50

-40
20
40
-15
-50
30
2 3 4 5 0
2 3 4 5
(a) R (km) (b) R (km)

V=23000 m3 V=111750 m3 V=200500 m3 V=289250 m3 V=378000 m3


Q R2=0.9989 R2=0.9994 R2=0.9974 R2=0.9996 R2=0.9993
Slop
100

90
-5

Slope of regression curve


80

-10
Q (L/s)

70

60
-15

50

40 -20
2 3 4 5
(c) R (km)

Fig. 9. Regression analysis of Q and R with the changes in the influence factors: (a) ig, (b) Hmax and (c) V.

Hmax = 450 m and V = 126,562 m3 (the width, height and length were results. Thus, the method proposed in this study could effectively
22.5, 45 and 125 m, respectively). confirm the influence zone of the PS cavern and drainage system with
The regression coefficients a and b could be confirmed with the credible accuracy.
values of ig, Hmax and V using Eqs. (9) and (10), with a = 40.1434 and b
= − 0.6658. The relationship between Q and R in the simplified PS could 4.3. Analysis of limitations and valuable further works
be subsequently calculated using Eq. (2). Accordingly, Fig. 13 shows the
relationship between Q and R as computed with the simulation models Despite the meticulous care that we considered, the following limi­
and Eqs. (2), (9) and (10). In Fig. 13, the results of Q as confirmed by tations were unavoidable and further works are suggested to address
numerical simulation and those from the proposed methods were close, them:
such that the maximum absolute value of the difference between them
was only 0.39 L/s. Additionally, the values of regression coefficients a (1) Owing to the lack of field-measured data, we assumed the
and b as computed with the simulation models were 40.7141 and − groundwater level to be 50 m below the earth’s surface and to be
0.6764, respectively, which were close to the results when Eqs. (9) and parallel to the topography. Thus, ig = it and Hmax was equal to the
(10) was applied. Furthermore, assuming that △Qt = 1 L/s, according to maximum buried depth of the cavern minus 50 m. For a specific
the numerical simulation results the decrement of Q was 1.02 L/s when PS, ig and Hmax should be confirmed based on explicit ground­
R increased from 3.8 km to 4.2 km and 0.99 L/s when R increased from water data. In addition, we neglected the undulation of the
4.2 km to 4.6 km. Hence, the value of the influence zone was in the range topography when we extended R. So in the Dupuit solution, the
of 4.2–4.6 km. By substituting the calculated a and b values in Eq. (11), seepage medium was considered as homogeneous in this study,
the value of the influence radius was determined as approximately and the influences of flow regime, fault zone and fracture group
4.352 km, a precise value within the range of the numerical simulation

10
C. Cao et al. Journal of Hydrology 595 (2021) 126018

Fig. 10. Variation of regression coefficients a and b of Eq. (2) with the changes in influence factors: (a) ig, (b) Hmax and (c) V.

on the influence zone were not considered. Thus, additional


studies are encouraged to investigate more influence factors.
100 1 R2=0.9959 (2) We symmetrically extended R in the right and left banks. Table 1
2 R2=0.9995
3 R2=0.9984
shows that the right and left banks of most PSs were asymmet­
4 R2=0.9992 rical. We suggest mirroring the right and left banks to another
80
5 R2=0.9993 side and then confirming the optimal right and left banks R with
6 R2=0.9947 the accredited △Qt for these cases. Furthermore, the application
7 R2=0.9966
60 8 R2=0.9988
of the correct units for the parameters should be emphasised in
Q (L/s)

9 R2=0.9974 this study owing to the conflicting dimensions of the regression


10 R2=0.9959 equations. The units of L/s, km, m, and m3 should be used for Q,
40 11 R2=0.9972 R, Hmax, and V, respectively.
12 R2=0.9983
13 R2=0.9931
(3) Fig. 5 indicates that when the constant water head boundary was
14 R2=0.9979 close to the cavern and drainage system, the groundwater
20 15 R2=0.9904 drawdown caused by the cavern and drainage system was
neglected, and that the results for Q and the groundwater level
were overestimated. With the increased R, the influence of
0
2 3 4 5 groundwater drawdown on the computation boundary gradually
R (m) became feeble, and the reductions of Q and the groundwater level
tended to zero. Thus, from another perspective, we can coun­
Fig. 11. Regression analysis of Q and R of 15 groups of numerical simulation teract the influence by directly reducing the groundwater level of
experiments. the constant water head boundary. However, the mechanisms of
the groundwater level of the computation boundary, influence

11
C. Cao et al. Journal of Hydrology 595 (2021) 126018

Table 3
Synthetic relations of ig, Hmax and V and regression coefficients a and b.
ig Hmax (m) V (m3) a b ig Hmax (m) V (m3) a b

0 150 23,000 10.624 − 0.775 0.14 1,150 200,500 98.960 − 0.350


0 400 111,750 26.517 − 0.343 0.28 650 289,250 74.161 − 0.823
0 650 200,500 53.280 − 0.307 0.28 900 378,000 101.839 − 0.607
0 900 289,250 87.782 − 0.283 0.28 1,150 23,000 71.046 − 0.428
0 1,150 378,000 108.501 − 0.282 0.42 900 200,500 87.893 − 0.735
0.14 400 378,000 38.025 − 0.733 0.42 1,150 289,250 130.020 − 0.641
0.14 650 23,000 73.780 − 0.494 0.56 1,150 111,750 107.030 − 0.690
0.14 900 111,750 76.147 − 0.405

Fig. 12. Simulation model of the verification of the proposed influence zone determination method.

and changes in the seepage analysis results with varied values of R were
investigated. Then, the relationship between Q and R with three influ­
ence factors was established based on numerical simulation experi­
ments. Ultimately, the determination method of influence zone for PS
cavern and drainage system was proposed based on the regression
analysis results and receivable △Qt. The following conclusions were
drawn:

(1) The value of R was hard to confirm with the blurry influence zone
of the cavern and drainage system, these commonly exist in PSs.
The seepage analysis accuracy of the PS underground station and
the prediction of ecological environmental pressure caused by the
drawing groundwater level were suspicious with inappropriate
values of R.
(2) The results for Q and the groundwater level were substantially
overstated with small R values, and the reduction of Q and the
groundwater level tended to be infinitesimal when R infinitely
increased. The influence factors i, Hmax and V had non-linear ef­
fects on the values of coefficients a and b. Furthermore, increased
i, Hmax and V delayed the stable tendency of the evolution of Q
and groundwater level within the statistical value range.
(3) The optimal value of R could be the same as the value of the in­
fluence zone calculated by the proposed method. The influence of
the groundwater drawdown in the constant water head boundary
Fig. 13. Comparison of the relationship between Q and R computed with
condition could be counteracted, and the seepage analysis error
simulation models and Eqs. (2), (9) and (10).
caused by inappropriate R values could be eliminated. Addi­
tionally, the range of the groundwater depression cone could be
zone of the PS cavern and drainage system are unclear. Hence, directly predicted using the proposed method.
additional related studies are suggested.
CRediT authorship contribution statement
5. Conclusion
Cheng Cao: Conceptualization, Methodology, Software, Formal
In the present study, the statistical results regarding the engineering analysis, Writing - original draft, Investigation, Validation. Zengguang
and geological characteristics of 56 operational PSs were summarised, Xu: Validation, Resources, Funding acquisition. Junrui Chai:

12
C. Cao et al. Journal of Hydrology 595 (2021) 126018

Supervision. Yuan Qin: Data curation. Jing Cao: Visualization. Goh, A.T.C., Zhang, R.H., Wang, W., Wang, L., Liu, H.L., Zhang, W.G., 2020. Numerical
study of the effects of groundwater drawdown on ground settlement for excavation
in residual soils. Acta. Geotech. 15 (5), 1259–1272. https://doi.org/10.1007/
Declaration of Competing Interest s11440-019-00843-5.
Goh, A.T.C., Zhang, W.G., 2012. Reliability assessment of underground rock cavern limit
The authors declare that they have no known competing financial states. Int. J. Rock Mech. Min. 55, 157–163. https://doi.org/10.1016/j.
ijrmms.2012.07.012.
interests or personal relationships that could have appeared to influence Himi, M., Casado, I., Sendros, A., Lovera, R., Rivero, L., Casas, A., 2018. Assessing
the work reported in this paper. preferential seepage and monitoring mortar injection through an earthen dam
settled over a gypsiferous substrate using combined geophysical methods. Eng. Geol.
246, 212–221. https://doi.org/10.1016/j.enggeo.2018.10.002.
Acknowledgement Huang, C.-S., Yang, S.-Y., Yeh, H.-D., 2014. Groundwater flow to a pumping well in a
sloping fault zone unconfined aquifer. Water Resour. Res. 50 (5), 4079–4094.
This study received financial support from the National Natural https://doi.org/10.1002/2013WR014212.
Javadi, M., Sharifzadeh, M., Shahriar, K., 2016. Uncertainty analysis of groundwater
Science Foundation for Excellent Young Scientists of China (51922088). inflow into underground excavations by stochastic discontinuum method: Case study
of Siah Bisheh pumped storage project. Iran. Tunn. Undergr. Sp. Tech. 51, 424–438.
References https://doi.org/10.1016/j.tust.2015.09.003.
Johnson, P.W., 1988. The relationship between radius of drainage and cumulative
production. SPE-16035-PA 3 (01), 267–270. https://doi.org/10.2118/16035-PA.
Bodeux, S., Pujades, E., Orban, P., Brouyère, S., Dassargues, A., 2017. Interactions
Li, Y., Chen, Y.F., Jiang, Q.H., Hu, R., Zhou, C.B., 2014. Performance assessment and
betwmeen groundwater and the cavity of an old slate mine used as lower reservoir of
optimization of seepage control system: A numerical case study for Kala
an UPSH (Underground Pumped Storage Hydroelectricity): A modelling approach.
underground powerhouse. Comput. Geotech. 55, 306–315. https://doi.org/
Eng. Geol. 217, 71–80. https://doi.org/10.1016/j.enggeo.2016.12.007.
10.1016/j.compgeo.2013.09.013.
Bonalumi, M., Anselmetti, F.S., Wüest, A., Schmid, M., 2012. Modeling of temperature
Peng, M., Zhang, L.M., 2011. Breaching parameters of landslide dams. Landslides 9 (1),
and turbidity in a natural lake and a reservoir connected by pumped-storage
13–31. https://doi.org/10.1007/s10346-011-0271-y.
operations. Water Resour. Res. 8 (48), W08508. https://doi.org/10.1029/
Trautz, R.C., Wang, J.S.Y., 2002. Seepage into an underground opening constructed in
2012WR011844.
unsaturated fractured rock under evaporative conditions. Water Resour. Res. 38
Bresciani, E., Shandilya, R.N., Kang, P.K., Lee, S., 2020. Well radius of influence and
(10), 6-1–6-14. https://doi.org/10.1029/2001WR000690.
radius of investigation: What exactly are they and how to estimate them? J. Hydrol.
Wen, L.F., Chai, J.R., Xu, Z.G., Yuan, Q., Li, Y.L., 2017. A statistical review of the
583, 124646. https://doi.org/10.1016/j.jhydrol.2020.124646.
behaviour of concrete face rockfill dams based on case histories. Géotechnique 68
Bresciani, E., Davy, P., de Dreuzy, J.-R., 2014. Is the Dupuit assumption suitable for
(9), 749–771. https://doi.org/10.1680/jgeot.17.p.095.
predicting the groundwater seepage area in hillslopes? Water Resour. Res. 50 (3),
Xu, Q., Chen, J.-T., Xiao, M., 2020. Analysis of unsteady seepage field and surrounding
2394–2406. https://doi.org/10.1002/2013WR014284.
rock stability of underground cavern excavation. Tunn. Undergr. Sp. Tech. 97,
Cao, C., Xu, Z., Chai, J., Li, Y., 2019. Radial fluid flow regime in a single fracture under
103239. https://doi.org/10.1016/j.tust.2019.103239.
high hydraulic pressure during shear process. J. Hydrol. 579, 124142. https://doi.
Xu, Z.G., Cao, C., Li, K.H., Chai, J.R., Xiong, W., Zhao, J.Y., Qin, R.G., 2019. Simulation
org/10.1016/j.jhydrol.2019.124142.
of drainage hole arrays and seepage control analysis of the Qingyuan Pumped
Chen, Y.F., Hu, R., Zhou, C.B., Li, D.Q., Rong, G., Jiang, Q.H., 2010. A new classification
Storage Power Station in China: A case study. B. Eeg. Geol. Environ. 78 (8),
of seepage control mechanisms in geotechnical engineering. J. Rock. Mech. Geotech.
6335–6346. https://doi.org/10.1007/s10064-019-01527-w.
Eng. 2 (3), 209–222. https://doi.org/10.3724/SP.J.1235.2010.00209.
Zhang, W.G., Goh, A.T.C., 2012. Reliability assessment on ultimate and serviceability
Dang, M., Chai, J., Xu, Z., Qin, Y., Cao, J., Liu, F., 2020. Soil water characteristic curve
limit states and determination of critical factor of safety for underground rock
test and saturated-unsaturated seepage analysis in Jiangcungou municipal solid
caverns. Tunn. Undergr. Sp. Tech. 32, 221–230. https://doi.org/10.1016/j.
waste landfill, China. Eng. Geol. 264, 105374. https://doi.org/10.1016/j.
tust.2012.07.002.
enggeo.2019.105374.
Zhang, W.G., Goh, A.T.C., 2015. Regression models for estimating ultimate and
Dassault Systèmes. 2017. Abaqus Analysis User’s Manual.
serviceability limit states of underground rock caverns. Eng. Geol. 188, 68–76.
Dupuit, J., 1863. Études théoriques et pratiques sur le mouvement des eaux dans les
https://doi.org/10.1016/j.enggeo.2015.01.021.
canaux découverts et à travers les terrains perméables. Dunod, Paris, p. 352.

13

You might also like