supplementary material

1
SUPPLEMENTARY MATERIAL
2
This supplementary material mainly describes the simulation optimization, objective
3
functions, constraints and the reservoir operation rules used for simulation
4
optimization. In addition, the tools used are described.
5
Simulation optimization
6
Simulation optimization composed of simulation and optimization is a process of
7
finding the best input variable values (optimal decision variables) from among all
8
possibilities by searching for best output variable values (maximization or
9
minimization of objective functions) (Kang & Park 2014). The simulation is used to
10
represent the process of the reservoirs’ release and storage based on water balance
11
(Equation S.4 provided in the constraints section of the Supplementary materials).
12
The optimization uses the objective functions of simulation to provide feedback on
13
progress of the search for best input variables, and in turn guides further input to the
14
simulation.
15
In the process of simulation optimization, the basic data used are inflows, water
16
demands, leakage losses and initial reservoir storage volume during operation periods.
17
Historical inflow data used are from year 1958 to year 2013 and 10-day average
18
inflow data are used for simulation. The water demands used are the annual local
19
water demand data and joint water demand data provided in the case study section,
20
which are divided into 36 equal 10-day periods. Leakage loss is calculated by using
21
the leakage loss rate of each reservoir: 13.24 × 104 m3/day for Biliuhe reservoir and
22
3.35 × 104 m3/day for Yingnahe reservoir. The initial water storage volume of Biliuhe
23
reservoir is 291.5 × 106 m3 and Yingnahe reservoir is 110 × 106 m3.
24
Objective functions
25
WSR is defined as
WSR 
26
Maximize
T'
T +1
(S.1)
27
where T is the total number of the operation periods, T' is the number of periods that
28
the water demands of 10-day are supplied as required.
29
SP is defined as
SP 
30
Minimize
1 T J
 SPi j
N i 1 j 1
(S.2)
31
where N is the total number of the simulated year and is equal to 56 , i is the period
32
index, j is a reservoir, J is the total number of reservoirs (here J is 2, including Biliuhe
33
reservoir and Yingnahe reservoir), SPij represents spills from reservoir j at study
34
period i.
35
36
DW is defined as
DW 
37
Minimize
1
N
T
 DW
i 1
i
(S.3)
38
where DWi represents the volume of diverted water from Dahuofang Reservoir at
39
study period i.
40
Constraints
41
For the reservoir optimization problem, it has to satisfy the mass balance equation.
42
The mass balance equation is written as
Si 1  Si  Ii  DWi  WSi  SPi  Ei
43
(S.4)
44
where Si and Si+1 represent the initial storage and end storage; Ii, DWi, WSi, SPi and Ei
45
represent the natural inflow, the amount of diverted water, the water supply amount
46
for the urban area, the amount of water spills and the evaporation and leakage loss at
47
study period i respectively.
48
Si and Si+1 are subject to
Smin  Si  Smax
49
(S.5)
50
where Smin and Smax represent the dead storage and maximum storage. The maximum
51
storage is different for the flood and dry season. In the flood season, the storage is the
52
flood-control storage for flood control safety, while the maximum level is the normal
53
storage in the dry season.
54
DWi is subject to
55
0  DWi  DWmax
56
where DWmax is the maximum diversion, which is limited by the diversion pipeline
57
from Dahuofang reservoir to Biliuhe reservoir.
58
Decision variables
59
The decision variables are shown in Table S1.
60
Table S1| Decision variables
Decision variables
Reservoir storage volume(m3):
water-supply rule curve
Reservoir storage volume(m3):
upper water-diversion rule
curve
Reservoir storage volume(m3):
lower water-diversion rule
curve
Allocation proportions for joint
water demand
(S.6)
Scenario 1
Scenario 2
Scenario 3
xi1 , i  1, 2, …, 36
xi1 , i  1, 2, …, 36
xi1 , i  1, 2, …, 36
xi2 , i  1, 2, …, 36
xi2 , i  1, 2, …, 36
xi2 , i  1, 2, …, 36
xi3 , i  1, 2, …, 36
xi3 , i  1, 2, …, 36
xi3 , i  1, 2, …, 36
--
xi4 , i  1, 2, …, 36
xi4 , i  1, 2, …, 36
61
Remarks: i represents time periods of operation. Each simulation year is divided into 36 time
62
periods (with ten days as a time period). There are 36 decision variables on each of three rule
63
curves and allocation proportions for joint water demand.
64
Reservoir operation rules
65
The reservoir operation rules used in this paper are reservoir water supply operation
66
rule curves and reservoir water diversion rule curves. The reservoir water supply
67
operation rule curves, based on 36 10-day periods, are shown in Figure S1(a). When
68
the water storage lies in zone 1, the amount of water supply is the same as the amount
69
of water demand. When the water storage of reservoir is in zone 2, water supply is
70
equal to the value of water demand multiplied by a water rational coefficient µ. The
71
reservoir water diversion rule curves are shown in Figure S1(b). Water diversion rule
72
curves divide the active water storage into three parts, i.e., zone I, zone II and zone III.
73
When the water storage of reservoir lies in zone I, the amount of water diversion is 0.
74
When in zone II, the amount of water diversion is equal to the conveyance capacity of
75
the water diversion pipelines multiplied by a coefficient ɛ. ɛ is defined as

76
Upperi  Si
Upperi  Loweri
(S.7)
77
where Upperi and Loweri represent the upper bound and the lower bound of diversion
78
at study period i; Si represents water storage of reservoir at study period i. When in
79
zone III, the amount of water diversion is equal to the conveyance capacity of the
80
water diversion pipeline.
81
82
83
Figure S1 | Diagrammatic sketch of reservoir operation rule curves. (a) Water supply
84
Each optimal solution of each scenario corresponds to a similar set of reservoir
85
operation rule curves. For practical operability, the operation rule curves should not
86
have frequent fluctuations, but be as flat as possible. Considering reasonability, the
87
points of diversion rule curves and water supply rule curves should be lower, to
88
restrict diversion from Dahuofang reservoir and increase water supply as much as
89
possible in wet periods while higher in dry periods comparatively. To make reservoir
rule curves; (b) water diversion rule curves.
90
operation rules easy to understand, one is obtained from the numerous optimal
91
solutions in the pipeline connection scenario to show the actual reservoir operation
92
curves. The selected one is shown in Figure S2.
93
94
95
Figure S2 | One of the reservoir operation rule curves in the pipeline connection
scenario.
96
Tools used for optimization and visualization
97
The optimization is implemented using ɛ-NSGAII within the MOEA Framework
98
(http://moeaframework.org/), which is a free, open-source Java framework for
99
experimenting with several popular multi-objective evolutionary algorithms. The
100
water supply system is simulated for computing reservoir storage volumes, water
101
supply, diversion and water spills based on water balance.
102
103
Reference
104
Kang, M. G. & Park, S. W. 2014 Combined simulation-optimization model for assessing irrigation
105
water supply capacities of reservoirs. Journal of Irrigation and Drainage Engineering, 140 (5),
106
04014005.
107