Abstract
China is one of the countries suffering from the most disasters in the world, so
national emergency warehouse location is critical for the country to reduce loss resulting
from disasters. Considering that emergency management is more concerned
about effectiveness than efficiency, the paper proposes an emergency warehouse location
problem (EWLP) model—an extension of the classic P-center model, for the
Chinese national emergency warehouse location problem. Some features, including
population distribution, economic condition, transportation, and multi-coverage for
some vital areas, are put into the proposed model, which are characterized with data
gathered from the reality. A Variable Neighborhood Search (VNS) based heuristic
algorithm is developed to solve the extensional model. The computational result
gained is cmopared with current emergency warehouse location planning in China.
It is shown that huge saving can be gained with the guarantee that the rescue resources
could be delivered in time. Moreover, the proposed VNS based algorithm
1 Introduction
China is one of the countries that suffer from the most disasters in the world.
Mega-disasters occur quite often, such as the massive flood in the Yangtze
river in 1998, the SARS epidemic situation in 2003, a series of earthquakes in
Xingjiang in 2005, the ice disaster in the Southern China in early 2008 and
the 5.12 Wenchuan earthquake in 2008.
In order to reduce loss from disaster, dispatching rescue resources is very
important. Fortunately, recent years Chinese government pays more and more
attention to emergency rescues management. Since 1998 when the law of
setting up reserve system of emergency supplies was first issued, there have
been 10 cities who set up national emergency resource warehouses. They are
Harbin, Shenyang, Tianjin, Zhengzhou, Hefei, Wuhan, Changsha, Chengtu,
Xi’an and Nanning. After the 5.12 earthquake in 2008, the ministry of Civil
Affairs makes a new plan which aims to increase the number of national emergency
warehouses from 10 to 24. However, building too many warehouses
does not make any sense. First, constructing and administering warehouses
needs a large amount of money. Second, the resource stored in warehouses
will occupy funds but bring no profits. Meanwhile, some supplies have to be
disposed when they exceed the time limit. So reasonable decision should be
made on Chinese national emergency warehouse locations.
In this study, an emergency warehouse location problem (EWLP) model
— an extension of the classic P-center model is set up for the Chinese national
emergency warehouse location problem. With data gathered from the
reality, a modified Variable Neighborhood Search (VNS) heuristic algorithm
is developed to solve the extensional model. It is shown that the result gained
in this study reduce the number of the wareshouse from 24 to 11, with the
guarantee that the rescue resources could be delivered to any disaster areas in
1 The work of this study is supported by National Natural Science Foundation of China
(No. 91224007), by the teacher research fund of China Earthquake Administration (No.
20140106).
2 Email:
time. Moreover, the proposed VNS algorithm shows its good performance in
the computational experiment.
2 Literature review
Researches on Location problem mainly focuses on location models and theories.
Some relative literature is listed as follows. Cooper (1963) first addressed
location problem [3]. Hakimi (1994, 1995) raised P-median and P-center problems
respectively [7,8]. Church and ReVelle (1974) firstly concerned maximal
covering location problem [2]. P-median problem, P-center problem and covering
problem are the basic problems in the field of location science because
some other problems are extensions of these three problems. For example,
P-center problem is useful in plenty of fields. So many scholars have studied
it and made some modifications. The a-neighbor P-center problem, which
is first presented by Krumke (1995) [10], can be seen as a generalization of
the P-center problem. R-all-neighbor P-center problem [12] has the similar
idea with the a-neighbor P-center problem, both of which are developed for
emergency systems.
Meanwhile, some scholars pay attention to algorithms solving location
problems. Most location problems are NP-hard, no deterministic polynomial
algorithms has been found to solve them. Therefore, a lot of heuristic
methods are developed, such as genetic algorithm, ant colony algorithm, simulated
annealing algorithm, etc. For instance, Kratica et al. (2001) presented
a genetic algorithm to solve the simple plant location problem [9]. Alp et al.
(2003) accelerated a genetic algorithm by combining a greedy algorithm for the
P-median problem [1]. Mladenoviดc et al. (2003) presented a variable neighborhood
search algorithm for both P-median and P-center problems [13,14].
Recently, Davidoviดc et al. (2011) also published a bee colony optimization
method for P-center problem [5].
Apart from mathematical models and algorithms, some scholars applied
location methods to real-world emergency location problems. For example,
Davoodi et al. (2011) presented a real test problem. The problem is to determine
the location of some new medical emergency centers for part of a new
city in Iran [6]. Dantrakul et al. (2014) studied a case study about the facility
locations in Chiang Mai city and 5 provinces of Northern Thailand [4]. Lu
(2013) gave a numerical example which demonstrates the application of the
proposed a weighted vertex p-center model to locate urgent relief distribution
centers in a relief supply distribution network responding to the massive
earthquake which hit central Taiwan on September 21, 1999 [11].
F. Ye et al. / Electronic Notes in Discrete Mathematics 47 (2015) 61–68 63
It can be seen from previous studies that researchers who study emergency
facility location problem mainly focus on theoretical methods. There are few
comprehensive researches considering both theoretical studies and their real
application in publications. Therefore, it is necessary to combine facility location
theories with the conditions of nation situation and apply them to the
planning and construction of national emergency facilities.
3 The EWLP model
In this paper, the EWLP model, which is a modification of the P-center model,
is introduced to solve the Chinese national emergency warehouses location
problem. In the original P-center problem, p sites (facilities) are selected from
the set of potential locations, and a set of clients are asssigned to the selected
sites. The objective is to minimize the maximum distance between a client
and the facility to which he or she is assigned. In the P-center model, each
client has only one facility. However, in the EWL problem, the risk of the
destruction of warehouses or the failure of transportation must be taken into
consideration. Therefore, each demand city in the EWL problem may have
one or more warehouses, the number of warehouses that each demand city
should be assigned can be settled after an investigation. Moreover, in the
EWLP model, the objective is to minimize the number of selected facilities so
that the total operation expenses of these warehouses can be decreased. The
above two aspects of modification fits to the condition of the Chinese national
emergency warehouses location problem.
The following notations are introduced to formulate the mathematical
model.
V : set of potential locations;
U: set of demand cities;
ni: number of warehouses which are assigned to demand city i, j ∈ U;
dij : distance between warehouse j to demand city i, i ∈ U, j ∈ V ;
C: distance limit for rescuing;
z: the longest distance from any warehouse to any demand city assigned to it;
p: number of warehouses that will be opened;
xij =
_
1, if warehouse j is assigned to demand city i, and
0, otherwise;
, i ∈ U, j ∈ V ;
yj =
_
1, if warehouse j is opened, and
0, otherwise;
, j ∈ V.
F. 64 Ye et al. / Electronic Notes in Discrete Mathematics 47 (2015) 61–68
The model can be formulated as follows.
min p (1)
s.t.
_
j
xij = ni, ∀i, (2)
xij ≤ yj , ∀i, j, (3) _
j
yj = p, (4)
z ≥ max
j
(dij, xij), ∀i, (5)
z ≤ C, (6)
xij, yj ∈ 0, 1, ∀i, j, (7)
The objective function (1) is to minimize the number of warehouses; Constraint
(2) rules that there must be a specific amount of warehouses to supply
emergency materials to any demand city; Constraint (3) ensures that demand
cities will not be served by an unopened warehouse; Constraint (4) guarantees
that the number of opened warehouses is equal to p; Constraint (5) picks up
the longest distance to z; Constraint (6) ensures that the longest distance
should be less than the distance limit for rescuing.
4 Data collecting
To solve the Chinese national EWLP, 66 potential locations for warehouses and
28 demand cities are carefully selected. Among 28 demand cities, according
to the ranking made by AHP (Analytic Hierarchy Process) method, each of
the top 20% of demand cities is equipped with 3 warehouses. While for the
bottom 20% of demand cities, each needs only one warehouse. The remaining
demand cities are assigned 2 warehouses. If you need detailed data, you can
contact Professor Zhao with the Email: qhzhao@buaa.edu.cn.
5 The VNS based algorithm
VNS algorithm is one kind of efficient heuristic algorithms for NP-hard problems.
Mladenoviดc et al. (1997) and Mladenoviดc et al. (2003) successively
solved P-median problem [13] and P-center problem [14] with VNS algorithms.
Although the EWLP model can be seen as an extension of the P-center problem,
it has some significant difference with P-center model and cannot be
solved by the VNS algorithm in Mladenoviดc et al. (2003). The VNS based
F. Ye et al. / Electronic Notes in Discrete Mathematics 47 (2015) 61–68 65
Step 1 Initialization: (1) Determine values of the parameters; (2) Set a value to p, e.g.
2
the area of the region
p 1
C π
= ⎡ ⎤ − ⎢⎢ ⎥⎥
.
Step 2 p := p + 1.
Step 3 With g