A Multi-Objective Model for Fire Station Location under Uncertainty

Jian Zhou, Zhen Li, Ke Wang

A Multi-Objective Model for Fire Station Location under Uncertainty

1
Jian Zhou, 2Zhen Li, 3Ke Wang*
1, First Author
School of Management, Shanghai University, Shanghai 200444, China,
[email protected]
2
School of Management, Shanghai University, Shanghai 200444, China,
[email protected]
*3,Corresponding Author
School of Management, Shanghai University, Shanghai 200444, China,
[email protected]

Abstract
The efficient location of fire stations may involve multiple objectives, some of which are possibly
conflicting. A multi-objective goal programming model is proposed in this paper taking both service
quality, setup costs, and operating costs into consideration in the uncertain environment. Based on
uncertainty theory, an equivalent deterministic programming model is developed at a given confidence
level of safety requirement. A numerical example is presented as well for illustration.

Keywords: Fire Station Location, Uncertain Goal Programming, Chance-Constrained

Programming

1. Introduction
The problem of locating fire stations in an efficient manner is a type of emergency facility location
problems. It has attracted much attention since public safety becomes more and more important during
last several decades. Locating fire stations usually involves some different and even conflicting
objectives. For example, the fire engines in Beijing are assumed to arrive at the accident spot within
about five minutes. This requires much more fire stations to cover the urban areas. However, it
certainly goes beyond the annual budget approved by the government and a compromise should be
made. Generally speaking, as suggested by Tzeng and Chen [10], an efficient location of fire stations
should well address the following issues: (a) it can shorten the distance between fire stations and
accident spots so as to improve emergency response efficiency; (b) it can reduce the overlap of service
region so as to improve the resource utility; and (c) it can estimate the reasonable number of fire
stations in a given area considering a trade-off between the loss of accidents and the setup and
operating costs of fire stations.
Recently, quite some methods and techniques have been developed for the fire station location
problem as well as the emergency facility location problems. For example, Badri et al. [2] presented a
multi-objective model to determine the locations of fire stations considering 11 technical and political
objectives. Araz et al. [1] took into account as well the population served by a single fire station and
the population with backup coverage. Tzeng and Chen [10] presented a multi-objective model to
determine the optimal number of fire stations at an international airport and their sites. Yang et al. [12]
suggested a fuzzy decision making model where the fire stations were classified into several categories.
Chevalier et al. [5] developed a decision support system based on a multi-scale geography information
system, where the users may set locations, equipment allocations, staffing, response times,
cost/efficiency trade-off, etc. Besides, some methods for emergency facility location problems have
also been proposed for different situations in the literature (see, e.g., [3,4,11,13]).
In this paper, a goal programming model is presented for the fire station location problem in the
uncertain environment considering service quality, setup costs, and operating costs. The service quality
is measured by the coverage area of fire stations within a predetermined safe distance and the average
traveling time. The setup and operating costs are mainly constrained by the limited budget. According
to the decision criteria, an uncertain chance-constrained programming model is developed with
predetermined confidence levels as the appropriate safety margin set by the decision-maker.
The rest of this paper is organized as follows. The problem description and notations are given in
Section 2. Four objectives concerned are introduced in Section 3. In order to model this type of
problems with uncertain variables, an uncertain goal programming is formulated and transformed into

Advances in information Sciences and Service Sciences(AISS) 1184

Volume5, Number7, April 2013
doi:10.4156/AISS.vol5.issue7.140
 A Multi-Objective Model for Fire Station Location under Uncertainty
Jian Zhou, Zhen Li, Ke Wang

its deterministic counterpart in Section 4. A numerical example is presented in Section 5 for

illustration.

2. Problem description and notations

Usually, the most important decision in the fire station location problem is how many fire stations to
have, and where to build them. On the one hand, to improve the public safety, a sufficient amount of
fire stations should be built. On the other hand, we also cannot ignore the cost of building and
operating these fire stations. Therefore, they should be located in the most effective way that makes an
appropriate trade-off between the service quality and the budget.
In this paper, four objectives are concerned to develop a multi-objective goal programming model in
the uncertain environment. These objectives are
(1) guaranteeing the safe distance (to meet the safety requirement);
(2) minimizing the annual operating costs (to reduce the economical burden);
(3) minimizing the deviation of fixed costs (to make full use of the budget to improve public safety
as possible);
(4) minimizing the average traveling time (to improve the overall service quality level).
Some notations used to formulate the goal programming are
: the number of candidate sites for fire stations;
: the number of demand sites;
: the index of candidate sites for fire stations;
: the index of demand sites;
: the total fixed cost targeted;
: the total annual operating cost targeted;
: the safe distance to ensure the service level;
: the fixed cost associated with locating a fire station in site ;
: the distance between fire station and demand site ;
: the uncertain annual operating cost of fire station ;
: the uncertain traveling time from fire station to demand site ;
: the binary decision variable, and = 1 if fire station is open, otherwise = 0.

3. Objectives
In general, when an accident occurs, the nearest fire station is assumed to response as soon as
possible. To guarantee that each demand site can be served properly, it requires that the distance
between each demand site and its nearest fire station is roughly within a safe distance, say, , which is
usually suggested by various related regulations. Mathematically, it can be formulated as
max min + = (1)
{| }
where max min is the largest distance among all pairs of demand sites and their nearest fire
{| }
stations, and and are negative and positive deviations from , respectively. Note that
= max min 0,
{| }
(2)
= max min 0.
{| }
To guarantee the emergent response performance, is to be minimized.
However, the real world situations are much more complicated. The emergent response performance
depends on not only the distance but also the traffic condition, the traveling time, the capacity of fire
stations, the severity of the accident, and some other unexpected situations. In this paper, we simply
consider the average traveling time over all pairs of fire stations and demand sites as an index of the
overall service quality by which it means to minimize


= . (3)

It should be noted that in practice the traveling time can hardly be explicitly determined or even
statistically estimated. Hence, the traveling time is assumed to be an uncertain variable in the

1185
 A Multi-Objective Model for Fire Station Location under Uncertainty
Jian Zhou, Zhen Li, Ke Wang

sense of Liu [6]. Consequently, the average traveling time is also an uncertain variable. Thus
minimizing may not be understood in the usual sense. In this case, we can handle it in the
framework of uncertainty theory by minimizing the -optimistic value of , i.e.,

= min | (4)

where is the predetermined confidence level. Note that the -optimistic value 0 due
to the nonnegativity of and , = 1, 2, , , = 1, 2, , .
The cost related objectives are formulated analogously. The annual operating cost should be roughly
below a prespecified amount, say, . That is,

+ = (5)

where is the total annual operating cost, and are negative and positive deviations
from , respectively. Furthermore, since part of the operating cost depends on the number of accidents
occurred and the severity of the accidents, which can not be precisely predicted or estimated, the
annual operating cost of fire station is characterized by an uncertain variable . Therefore, the total
annual operating cost is also an uncertain variable. Similarly, we may consider the positive
and negative deviations from at a given confidence level by introducing
= min{ 0|{ } },
(6)
= min{ 0|{ } },

where and are the -optimistic and -negative deviations from , respectively. It is clear
that is to be minimized in this situation.
Another issue concerned in the fire station location problem is the fixed cost associated with
locating a new station. To fully utilize an approved budget, say , it is wise to build fire stations as
many as possible with this budget. Mathematically, it can be formulated as

+ = (7)

where and are negative and positive deviations from , respectively. It is clear that

+ = (8)

is to be minimized in order to make the full use of the budget.
By now, four objectives have been considered, two of which involve uncertain variables. In this
paper, they are assigned different priorities according to the relative importance of various goals. To be
precise, we are going to minimize these objectives lexicographically, i.e.,
lexmin { , , + , }. (9)

4. Uncertain chance-constrained goal programming

In order to handle the multi-objective optimization problem involving uncertain variables, a goal
programming model with uncertain chance constraints is proposed as suggested in Liu and Chen [9].
For the uncertain fire station location problem, the decision-maker may desire the optimization
goals involving uncertain variables to hold at least at some confidence levels. In such a case, these
uncertain objectives may be expressed as deterministic objectives with some uncertain
chance-constrained requirements. With given confidence levels and , the uncertain fire station
location problem can be formulated as

1186
 A Multi-Objective Model for Fire Station Location under Uncertainty
Jian Zhou, Zhen Li, Ke Wang

lexmin { , , + , }

subject to
max min + =
{| }






(10)
+ =







, , , , 0
= 0 or 1, = 1, 2, , .
Furthermore, according to the arithmetic operational laws of independent uncertain variables in [7,8],
we can obtain the following result.

Proposition 1 Assume that and are all independent uncertain variables with regular
uncertainty distributions , , = 1, 2, , , = 1, 2, , , respectively. Then
(1) { } holds if and only if

() ; (11)

(2)
holds if and only if

() . (12)

By Proposition 1, the proposed uncertain chance-constrained programming model (10) can be
further transformed into a deterministic programming model as
lexmin { , , + , }

subject to
max min + =
{| }


()


(13)
+ =



()


, , , , 0
= 0 or 1, = 1, 2, , .
Note that the objective function can also be expressed by
min + + ( + ) + (14)
where , = 1, 2, 3,4, are the preemptive priorities assigned to the four objectives, respectively, with
, which express the relative importance of various goals.
The model (13) is a mixed integer nonlinear goal programming model. Generally, solving such a
model requires much computational effort. However, using some well developed software packages,
for example, CPLEX, and MATLAB, it can be solved to a moderate size or even large size.

1187
 A Multi-Objective Model for Fire Station Location under Uncertainty
Jian Zhou, Zhen Li, Ke Wang

5. Numerical example
In this section, a numerical example of small size is presented for illustration. Assume that there are
10 candidate sites for fire stations and 10 demand sites to be served. The locations of fire stations and
demand sites are shown in Figure 1. The distances of all pairs of fire stations and demand sites are
given in Table 1. In order to meet the safety requirements, the safe distance sets to be 5.

Figure 1. Locations of the candidate fire stations and demand sites

Table 1. The distance from the th fire station to the th demand site
i
1 2 3 4 5 6 7 8 9 10
j
1 8.1 5.8 11.2 19.7 17 14.9 32 30 25.5 35.1
2 2.2 8 5 13 12.6 13.5 25.7 25.5 22.4 30.1
3 10 5.4 6 13 7.3 5.7 23.3 20 15.5 25.3
4 13.6 20.6 11.2 5.4 14.1 20 15.5 21 22.4 23.3
5 18.6 11.2 13.9 18.1 9.2 3.2 24.2 17 10.4 22.8
6 18.9 15 12.8 13.3 5 6.3 17.2 10.4 5.4 16.2
7 18.1 20.1 12.6 5.1 9.4 16.1 8.2 12 14.3 14.3
8 27 27.9 21.1 14.3 15.2 20.8 4.1 7.1 13.4 5.7
9 31.8 30.1 25.5 20.6 18 21.6 12.1 5 11.2 3
10 33.1 34.1 27.3 20 21.5 26.9 7.3 11.4 18.4 6.3

Moreover, the traveling times from fire station to demand site , = 1, 2, , , =

1, 2, , , are assumed to be independent linear uncertain variables with distributions ~( , ),
= 1, 2, , , = 1, 2, , , respectively, where
0, if
() = ( )( ) , if < (15)
1, if > .
The information of the parameters is given in Table 2.

1188
 A Multi-Objective Model for Fire Station Location under Uncertainty
Jian Zhou, Zhen Li, Ke Wang

Table 2. The uncertain travelling time from fire station to demand site
i
1 2 3 4 5
j
1 (6, 9) (4, 8) (8, 13) (15, 20) (16, 19)
2 (3, 5) (6, 10) (5, 7) (12, 15) (11, 15)
3 (10, 15) (5, 8) (5, 7) (10, 15) (6, 8)
4 (11, 18) (15, 25) (9, 12) (6, 7) (12, 15)
5 (15, 25) (10, 15) (11, 13) (15, 20) (8, 10)
6 (20, 28) (12, 18) (10, 13) (11, 14) (5, 7)
7 (16, 25) (18, 25) (10, 14) (5, 5) (8, 10)
8 (25, 32) (23, 30) (20, 24) (12, 15) (13, 16)
9 (27, 35) (30, 38) (20, 28) (20, 26) (15, 20)
10 (30, 40) (35, 45) (25, 30) (20, 25) (22, 26)
i
j 6 7 8 9 10
1 (12, 15) (27, 36) (28, 32) (20, 25) (30, 40)
2 (12, 16) (21, 30) (22, 26) (20, 25) (28, 36)
3 (5, 7) (20, 26) (20, 25) (15, 20) (23, 28)
4 (18, 22) (12, 18) (18, 25) (20, 25) (20, 26)
5 (3, 5) (20, 28) (15, 19) (10, 12) (20, 25)
6 (6, 8) (15, 20) (10, 12) (5, 7) (15, 20)
7 (15, 18) (5, 11) (10, 13) (12, 16) (13, 16)
8 (20, 25) (4, 8) (6, 8) (10, 14) (5, 6)
9 (20, 24) (10, 15) (5, 8) (10, 4) (3, 5)
10 (25, 30) (5, 10) (10, 12) (15, 20) (6, 8)

Besides, assume that the annual operating costs , = 1, 2, , , are independent linear uncertain
variables with distributions ~( , ) , = 1, 2, , , respectively. The information of the
parameters is given in Table 3. The total annual operating cost targeted for all the fire stations is
850,000. The fixed cost associated with locating a fire station is 1,000,000 for each fire station, and the
total approved budget of fixed cost is 5,800,000.

Table 3. The uncertain annual operating costs

candidate site i
1 (100,000, 115,000)
2 (105,000, 122,000)
3 (118,000, 125,000)
4 (100,000, 120,000)
5 (120,000, 130,000)
6 (115,000, 128,000)
7 (100,000, 125,000)
8 (110,000, 119,000)
9 (112,000, 125,000)
10 (105,000, 118,000)

Therefore, for given confidence levels = 0.8 and = 0.8 , the multi-objective goal
programming model can be further transformed into

1189
 A Multi-Objective Model for Fire Station Location under Uncertainty
Jian Zhou, Zhen Li, Ke Wang

min + + ( + ) +

subject to
max min + = 5
{| }


(0.8) 850,000


(16)
1,000,000 + = 5,800,000




(0.8) 10

, , , , 0
= 0 or 1, = 1, 2, , 10
where , , , are the preemptive priorities assigned to the objectives, respectively, with
. Note that for a linear uncertain variable with distribution ~(, ), and its
inverse uncertainty distribution is () = (1 ) + . Hence, the model (16) can be further
simplified.
For this simple example, there are 2 = 1024 possible solutions for the binary variables ,
= 1, 2, , 10, some of which can be easily eliminated. According to the hierarchy of the priorities,
the goal programming can be solved in 4 successive phases. The results are shown in Table 4. The final
optimal locations of the fire stations are 2, 3, 4, 6, 7, 10 as illustrated in Figure 2.

Table 4. The results of the numerical example

phase objectives number of optimal solutions optimal value of deviation
48 ( ) = 1.3
27 ( ) = 0
+
9 ( + ) = 200,000

1 ( ) = 17.1533

Figure 2. Optimal locations of the selected fire stations

1190
 A Multi-Objective Model for Fire Station Location under Uncertainty
Jian Zhou, Zhen Li, Ke Wang

6. Conclusion
As a type of emergency facility location problems, it is concerned in this paper the problem of
locating fire stations effectively in the uncertain environment to guarantee the service quality under
limited budgets. An uncertain multi-objective goal programming model is proposed to deal with such
situations. Based on the uncertainty theory founded by Liu [6], it may be transformed into a mixed
integer nonlinear programming model and solved with the aid of some well developed optimization
software packages. The method presented in this paper illustrates an alternative modeling technique
when there are uncertain factors that cannot be precisely determined or statistically estimated.

7. Acknowledgments
This work was supported in part by grants from the Shanghai Philosophy and Social Science
Planning Projects (No. 2012BGL006 and No. 2012XAL022), and Innovation Program of Shanghai
Municipal Education Commission (No. 13ZS065).

8. References
[1] Ceyhun Araz, Hasan Selim, Irem Ozkarahan, A fuzzy multi-objective covering-based vehicle
location model for emergency services, Computers & Operations Research, vol. 34, no. 3,
pp.705-726, 2007.
[2] Masood A. Badri, Amr K. Mortagy, Colonel Ali Allsayed, A multi-objective model for locating
fire stations, European Journal of Operational Research, vol. 110, no. 2, pp.243-260, 1998.
[3] Mustafa S. Canbolat, Michael von Massow, Locating emergency facilities with random demand
for risk minimization, Expert Systems with Applications, vol. 38, no. 8, pp.10099-10106, 2011.
[4] Mei-Shiang Chang, Ya-Ling Tseng, Jing-Wen Chen, A scenario planning approach for the flood
emergency logistics preparation problem under uncertainty, Transportation Research Part E:
Logistics and Transportation Review, vol. 43, no. 6, pp.737-754, 2007.
[5] Philippe Chevalier, Isabelle Thomas, David Geraets, Els Goetghebeur, Olivier Janssens,
Dominique Peeters, Frank Plastria, Locating fire stations: An integrated approach for Belgium,
Socio-Economic Planning Sciences, vol. 46, no. 2, pp.173-182, 2012.
[6] Baoding Liu, Uncertainty Theory, 2nd ed., Springer-Verlag, Berlin, 2007.
[7] Baoding Liu, Some research problems in uncertainty theory, Journal of Uncertain Systems, vol.
3, no. 1, pp.3-10, 2009.
[8] Baoding Liu, Uncertainty Theory: A Branch of Mathematics for Modeling Human Uncertainty,
Springer-Verlag, Berlin, 2010.
[9] Baoding Liu, Xiaowei Chen, Uncertain multiobjective programming and uncertain goal
programming, Technical Report, 2013.
[10] Gwo-Hshiung Tzeng, Yuh-Wen Chen, The optimal location of airport fire stations: A fuzzy
multi-objective programming and revised genetic algorithm approach, Transportation Planning
and Technology, vol. 23, no. 1, pp.37-55, 1999.
[11] Keming Wang, Zujun Ma, Emergency blood transportation problem based on two-stage
decision-making method, International Journal of Advancements in Computing Technology, vol.
4, no. 19, pp.617-625, 2012.
[12] Lili Yang, Bryan F. Jones, Shuang-Hua Yang, A fuzzy multi-objective programming for
optimization of fire station locations through genetic algorithms, European Journal of Operational
Research, vol. 181, no. 2, pp.903-915, 2007.
[13] Qiansheng Zhang, Fuchun Liu, Yirong Huang, Early warning index weight assignment and
decision-making mechanism for city significant emergency in uncertain environment,
International Journal of Digital Content Technology and its Applications, vol. 7, no. 1, pp.327-335,
2013.

1191