Sie sind auf Seite 1von 14

Applied Mathematics and Computation 140 (2003) 165178 www.elsevier.

com/locate/amc

Goal programming in a planning problem


Fabiane de Oliveira a, Neida Maria Patias Volpi Carlos Roberto Sanquetta c
a

b,*

Department of Mathematics and Statistics, UEPG, Post Graduation in Engineering Numerical Methods, UFPR, Curitiba, Brazil b Department of Mathematics, UFPR, 81531-990 Curitiba, Brazil c Department of Forestry, UFPR, Brazil

Abstract The objective of this paper is to apply one of the techniques of multiobjective programming (goal programming) in a brazilian forest problem, in a case study accomplished in the Santa C^ andida Farm, Paran a, Brazil. The areas of this farm can be managed for timber (pine and native species), harvesting of erva-mate leaves, pasture, and tourism. There is also a concern of the farm managers with increasing the diversity of ora and fauna, increasing environmental protection conditions and maintaining employees in the farm. Goal programming was used to develop a project of land allocation, in which all the goals would be reached as closest as possible of the ideal, in a way to attend all the operational restrictions considered. In goal programming, the concept of optimum solution of LP problems is substituted by a satisfactory solution (nondominated). Several solutions can be obtained, and the best solution will depend on the priority associated to each goal. 2002 Elsevier Science Inc. All rights reserved.
Keywords: Goal programming; Multiple objectives; Planning problem; Weights; Goal

1. Introduction Problems involving multiples objectives can be solved using linear programming, where one of the objectives, the most important, is optimized and

Corresponding author. E-mail address: nmpv@mat.ufpr.br (N.M.P. Volpi).

0096-3003/02/$ - see front matter 2002 Elsevier Science Inc. All rights reserved. PII: S 0 0 9 6 - 3 0 0 3 ( 0 2 ) 0 0 2 2 0 - 5

166

F. de Oliveira et al. / Appl. Math. Comput. 140 (2003) 165178

the others are considered in the restrictions. This procedure generates some disadvantages such as Representing the goals by means of restrictions of linear programming generally lead to infactibles problems. In large problems it is dicult to nd the restriction that causes the infactibility. The choice of which objective should be optimized is sometimes dicult or subjective. The multiple-objective programming presents a form of solving these problems, where the optimum solution of the problem of linear programming is substituted by a set of solutions, not necessarily optimum in the sense of the linear programming, but ecient solutions. This paper intend to apply one of the techniques of multiple-objective programming denominated goal programming. Considering that many were the reasons why this paper was considered on, such as: The multiple programming objective has great applicability in the area of natural resources, due to the situations of multiple use that frequently imply in multiples objectives [6]. The techniques of multiple-objective programming are useful when there are situations where the decisions to be taken are in a environmental of conicting objectives. Many of the decisions taken in forest problems involve multiples objectives (economic, social, environmental, among other), as well as multiples use, timber and nontimber. Some of these applications are found in the works of Field [4], Dane and Meador [3], Arp and Lavigne [1], Balteiro and Romero [2], where forest areas are designed for several issues. The globalization of the economy and the competition demand, mainly in the external market, and the need of obtaining certication of the forest products insist that other objectives besides the economic are reached.

2. Multiple-objective programming 2.1. Multiple-objective programming A multiple-objective problem can be represented in the way: max fz1 x; z2 x; . . . ; zn xg; x2X

where x: variable of decision; X: set of possible alternatives; fz1 ; z2 ; . . . ; zn g: set of the objectives.

F. de Oliveira et al. / Appl. Math. Comput. 140 (2003) 165178

167

The obtained solution can be dominated or nondominated. A solution is nondominated when it does not exist another solution feasible that improves one of the objectives without decreasing at least another objective. Then given a set of feasible solutions X, the set of nondominated solutions S, is dened as it proceeds: S fx: x 2 X, there is not another y 2 X such that zh y > zh x for some h 2 f1; 2; . . . ; ng and zj y P zj x for all j 6 hg. Otherwise, the solution is said dominated and the set of the same ones, set of dominated solutions. Several methods exist to generate a set of nondominated solutions. Some of them are described in [5]: the weights method, the e constraint method, the multiobjetive simplex method of Philips, the multiobjective simplex method of Zeleny. The method that will be used in this work it is denominated goal programming and will be describe to proceed. 2.2. Goal programming According to Field [4], the characteristic that distinguishes the formulation of goal programming is that one or more goals are directly incorporate in the function objective, thorough deviation variables, that is, the objectives are written in the form of goals restrictions, where each goal represents the value that intends to be reached. The goals can or not be reached completely and, to allow this exibility, deviations variables are used d and d , indicating how much the objective was surpasses or was lacked by that value respectively. Goal programming searches a form of reaching the goals as closest as possible; the objective of this technique is to minimize the sum of the deviations for all the goals. The general model of goal programming can be written in the following manner: min subject to z w d w d Ax d d M Bx % b x P 0; d P 0; d P 0

where % can be 6 or or P; z w w A B M objective function vector 1 m of weights, associated to the positive deviations of the goals and they express the importance of each goal vector 1 m of weights, associated to he negative deviations of the goals and they express the importance of each goal matrix m n of technological coecients of the goal constraints matrix p n of others constraints vector m 1 that represents the goals that should be reached

168

F. de Oliveira et al. / Appl. Math. Comput. 140 (2003) 165178

b d d x

vector vector vector vector

of resources p 1 for the others restrictions m 1 representing the positive deviations of the m goals m 1 representing the negative deviations of the m goals n 1 of the decisions variables

In the literature there are many suggestions about solving the prior problems. See for example [4] or [2]. When we solve a model of goal programming, the interest is on nondominated solutions, since a unique optimum solution satisfying all the goals in general is impossible to be obtained, due to the conicting character of the objectives.

3. Model of forest planning 3.1. A study case The case study was executed in an area of %2000 ha, in a farm, located in General Carneiro in Paran a, for a ve years period of planning. This farm possesses reforestation areas, native areas, permanent preservation, elds and swamp, that are divided in smaller areas with homogeneous characteristics of age, of location and species, called management units. We considered 14 management units. 3.2. Goals Techniques of goal programming were used seeking to reach the following goals: (1) Wood harvest (pine): This objective was accomplished in the reforestation areas in order to attend the wood demand for the plant of the industry of the company. (2) Wood harvest (araucaria): This objective was accomplished in the native areas and in the elds, with the intention of assisting the wood demand for the plant of the industry. (3) Leaves of erva-mate harvest: There exist a demand of erva-mate in the market, and its harvest can be accomplished in the native areas and in elds. (4) Tourism: The objective of the farms manager is to maximize the number of tourists that can visit the farm, taking into account that there is a hotel with daily capacity of 200 peoples. The areas of existent scenic beauties of the farm are exploited. (5) Pasture: The company has a creation of bualos that is explored commercially.

F. de Oliveira et al. / Appl. Math. Comput. 140 (2003) 165178

169

(6) Maintenance of employment: It is necessary to attend the needs of the farms manager so much, as well as the tourism, in the day-to-day of the hotel, among others. The maintenance of employments generates an economic development for the region. (7) Increase in the diversity of the ora: The objective of the farms manager is to reach larger diversity (number of species) that occurs in the area, contributing to the environmental preservation, motivating the rural tourism, besides attracting more tourists. (8) Increase in the diversity of the fauna: For the same reasons exposed in the diversity of the ora. It were solved problems of linear programming, just considering restrictions of area and optimization only one objective each time, in order to verify the maximum production of that goal, case all the area was used for only that objective. The optimum values for the linear programming problems can be found in the work of Oliveira [7]. Comparing these results with the goals proposed by the industry, it was veried that in some periods the goals of wood of pine could not be reached. The maximum production obtained in the problems of linear programming, were used as the goal of the correspondent period with exception when the maximum production was larger than the goal proposed by the industry. The theoretical goals are present in the second column of the Table 1. Based on the used politics and the description of the management unit, it was dened that the management prescriptions were to be used in each one. 18 management prescriptions and 6 actions such as were proposed 1. 2. 3. 4. 5. 6. wood harvest (pine), wood harvest (araucaria), erva-mate harvest, tourism, creation of bualos, environmental protection.

We dene management prescriptions as the group of actions that can be accomplished simultaneously in a unit management. The management prescriptions are present as follows: M1 M2 M3 M4 M5 M6 pine harvest araucaria harvest araucaria harvest and erva-mate harvest araucaria harvest and erva-mate harvest araucaria harvest, erva-mate harvest, creation of bualos araucaria harvest, erva-mate harvest, creation of bualos

170

F. de Oliveira et al. / Appl. Math. Comput. 140 (2003) 165178

M7 M8 M9 M10 M11 M12 M13 M14 M15 M16 M17 M18

erva-mate harvest erva-mate harvest erva-mate harvest and creation of bualos erva-mate harvest and creation of bualo erva-mate harvest and tourism leaves of erva-mate harvest and tourism leaves of erva-mate harvest, tourism and creation of bualo leaves of erva-mate, tourism and creation of bualo tourism tourism and environmental protection creation of bualos environmental protection

For example, the araucaria harvest and erva-mate harvest held in a same management unit constitute the management prescriptions called M3. The management M1 is exclusive for pine harvest. Two variations of management are considered in relation to the erva-mate harvest, the managements M3, M5, M7, M11 and M13 when the harvest erva-mate is in the rst, third and fth periods and the managements M4, M6, M8, M10, M12 and M14 the ones which the erva-mate harvest is in the second and fourth periods. The management prescription M16 and M18 are managements that represent change of use of the land, that is, they convert the management unit in a preservation area. The model of goal programming proposed as follow search for the best managements prescription to be applied in each management unit, in a way to attend the productions, social and environmental goals of the mentioned company. 3.3. Model of goal programming 3.3.1. Objective function The function objective seeks to minimize the sum of the deviations for all the goals. The objective of the model of goal programming considered has the general form: ! K  X X  k k k k min z wMETA dMETA wMETA dMETA
k 1 METAl 2M
l l l l

where M fPIN; ARAU; EM; TUR; PAST; EMP; DDFLO; DDFAUg is the k set of the goals; k 2 f1; . . . ; K g, represents the period; dMETA , represents the l k lower deviation of the goal l in the period k; dMETA , represents the upper l deviation of the goal l in the period k. The deviation variables must be greater

F. de Oliveira et al. / Appl. Math. Comput. 140 (2003) 165178 Table 1 Solutions for dierent weights for the goals of diversity of ora and fauna k Goal theoretical Goal reached Weight 1 Pine 1 29 300 29 300 2 35 300 35 000 3 6750 6700 4 35 300 35 000 5 30 000 30 000 Araucaria 1 170 170 2 170 170 3 170 170 4 170 170 5 170 170 Erva-mate 1 700 700 2 700 700 3 700 700 4 700 700 5 700 700 Tourism 1 200 200 2 200 200 3 200 200 4 200 200 5 200 200 Pasture 1 2800 2800 2 2800 2800 3 2800 2800 4 2800 2800 5 2800 2800 Employee 1 50 39.66 2 50 43.14 3 50 27.2 4 50 43.14 5 50 40.32 Increase in the diversity of the ora 2 7 1.61 3 14 3.23 4 21 4.84 5 28 6.46 Increase in the diversity of the fauna 2 1 0.16 3 1 0.33 4 2 0.49 5 2 0.66 Weight 10 29 300 35 000 6700 35 000 30 000 170 170 170 170 170 700 700 700 700 700 200 200 200 200 200 2800 2800 2800 2800 2800 39.66 43.14 27.2 43.14 40.32 1.61 3.23 4.84 6.46 0.16 0.33 0.49 0.66 Weight 100 13 650 35 500 0 0 0 170 170 170 170 170 674.52 170 674.52 170 674.52 292.72 281.1 269.49 257.88 200 0 0 0 0 0 39.08 50 29.78 27.97 22.8 4.69 9.38 14.06 18.75 0.5 1 1.5 2 Weight 1000 9192.39 3984.36 0 0 0 170 170 170 170 170 0 165.99 0 165.99 0 429.78 429.78 429.78 429.78 429.78 0 0 0 0 0 50 50 48.36 49.39 50 5.9 11.8 17.7 23.6 0.7 1.3 2 2.6

171

172

F. de Oliveira et al. / Appl. Math. Comput. 140 (2003) 165178

or equal to zero. wk and wk , weights attributed to deviations variMETA META l l ables; z, objective function. 3.3.2. Coecients of the technological matrix The technological matrix of the general model presented in Section 2.2 needs production information in each group of goals restrictions. If i 2 f1; . . . ; I g represents the management units and j 2 f1; . . . ; J g represents the managements prescriptions we considered that:
k apij

ak ij aemk ij
k atij

apsk ij aek ij dflk ij dfak ij xij

coecient of production of wood of pine in the management unit i according the management prescription j, in the period k, measured in stereos/ha coecient of production of araucaria in the management unit i according the management prescription j, in period k measured in m3 /ha coecient of production of erva-mate in the management unit i according the management prescription j, in the period k measured in arrobas/ha coecient that represents the number of tourism-day that visit the management unit i in which is applied the management prescription j, in the period k, measured in number of visitor-day/ha coecient regarding the production of meat from the animals of the management unit i in which is applied the management prescription j, in the period k, measured in kg/ha coecient regarding the necessary employees to execute the tasks of the management unit i in which is applied the management prescription j measured in employees-day/ha increase index in the diversity of the ora in the management unit i according to the management prescription j, in the period k, measured in number of dierent species of ora/ha increase index in the diversity of the fauna in the management unit i according to the management prescription j, in the period k, measured in number of dierent species of fauna/ha represent the number of hectares of the management unit i in which will be applied the management prescription j

The restrictions were written as it proceeded, taking into account that in the model of goal programming were considered two groups of restrictions: goals restrictions and areas restrictions. 3.3.3. Restrictions of goals
I X J X i1 j 1 k k k apij xij dPIN dPIN PINk

8 k 1; 2; . . . ; K

F. de Oliveira et al. / Appl. Math. Comput. 140 (2003) 165178


I X J X i1 j 1 k k ak ij xij dARAU dARAU ARAUk ;

173

8 k 1; 2; . . . ; K

I X J X i1 j 1

k k aemk ij xij dEM dEM EMk ;

8 k 1; 2; . . . ; K

I X J X i1 j 1

k k k atij xij dTUR dTUR TURk ;

8 k 1; 2; . . . ; K

I X J X i1 j 1

k k apsk ij xij dPAST dPAST PASTk ;

8 k 1; 2; . . . ; K

I X J X i1 j 1

k k ek ij xij dEMP dEMP EMPk ;

8 k 1; 2; . . . ; K

I X J X i1 j 1

k k dflk ij xij dDFLO dDFLO DDFLOk ;

8 k 1; 2; . . . ; K

I X J X i1 j 1

k k dfak ij xij dDFAU dDFAU DDFAUk

8 k 1; 2; . . . ; K

PINk goal for the production of pine in the period k in stereos ARAUk goal for the production of araucaria in period k in m3 EMk goal for the production of erva-mate in period k in arrobas TURk goal number of tourism-day in the period k PASTk goal of production of meat in the period k in kg EMPk goal of amount of employees-day for the period k DDFLOk goal of increase in the diversity of the ora in the period k, when the use of the area changes, in the management units, measured in number of dierent species of ora DDFAUk goal of increase in the diversity of the fauna in the period k, when the use of the area changes, in the management units, measured in number of dierent species of fauna In the restrictions (7) and (8) was considered that: For i 1; 2; 3; 4; 5; 6; 7; 8; 9; 10; 11 and 13 ) dflij and  6 0 if j 16 or 18 dfaij 0 if j 6 16 and 18 For i 12 and 14 ) dflij and dfaij 0; 8 j. All the decision variables are greater or equal to zero.

174

F. de Oliveira et al. / Appl. Math. Comput. 140 (2003) 165178

k k Depending of the values of dMETA , three situations can happen and d META l l in the period k:

1. dMETA 0 and dMETA > 0, meaning that the goal was surpassed; l l 0, meaning that the goal was not reached; 2. dMETA > 0 and d METAl l k k 3. dMETA 0 and dMETA 0, meaning that the goal was reached exactly.
l l

3.3.4. Restriction of areas Beside de goal restrictions were considered area restrictions that guarantee that the sum of the areas where the managements are applied does not surpass the area of the entire unit. This restrictions is represented by:
J X j 1

xij 6 Ai ;

i 1; 2; . . . ; I

where Ai is the area of the management unit i measured in ha. This model used the data obtained in the industry, in function of the current conditions of the farm. The ve year of planning was chosen for analysis effectiveness of an analysis of short period. The same model can be applied for larger periods, without generality loss. 3.3.5. The use of weights In general, due to the incompatibility among the objectives, it is practically impossible to reach all the goals simultaneously. For example, considering the goal pine harvest and diversity of the fauna, if we increase the diversity of the fauna then appears the necessity to modify the use of the reforestation areas for preservation areas, what could imply in not reaching the goal pine harvest. In the present project there were attributing dierent weights for a deviation in the objective function. Using dierent weights for the goal several solutions can be met. Attributing the same weights for all goal, they will have the same importance. If the main objective is to increase the diversity of the ora, attribute a weight larger to the deviation correspondent to this goal. This way, all the area will be used rstly to attend the goal increase in the diversity of the ora, being the remaining area used for others goals.

4. Results and analysis of goal programming problem The model of goal programming was solved rstly using unitary weights, meaning that all the goals have the same priority. The results are presented in the Table 1. It was veried that the goals of pine, araucaria, erva-mate, tourism and pasture were exactly reached; on the other hand the number of employees

F. de Oliveira et al. / Appl. Math. Comput. 140 (2003) 165178

175

and diversity of the ora and of the fauna obtained were not reached. To reach to the goal of employees maintenance it is necessary that another goal has a upper deviation, increasing the labor and consequently the value of the function objective, generating a nondominated solution. Supposing that the main priority of the industry is the increase of the biological diversity, than that solution is not the best. A manner of solving this problem is to attribute larger weights for the diversity goals. Three weights were considered: 1000, 100 and 10, each one, representing the importance of the goal diversity in relation to the others ones. The results for three cases are also presented in the Table 1. The prescription that should be applied in each management unit depend on the priority of each goal. When attributing the same weights for whole the goals, the solution met, gives priority to the harvest of pine, being the management chosen in the reforestation management units, prescription that include the pine harvest. Attributing weight 1000 for the diversity, most of the management chosen refer to change of the use the area. The areas of reforestation are converted for environmental protection, stopping producing pine, then increasing the diversity. The relationship of the management chosen are presented in the Tables 2 and 3. The graph exhibited in Fig. 1 shows the inuence of the weights in the goals of diversity of the fauna. In x-axis, the weights were represented: 1, 10, 100 and 1000. For each weight, it has four columns, each column representing a period. The rst of the four columns represents the increase of the diversity of the fauna for unitary weights for all the criterions.

Table 2 Designation of the prescriptions managements using weight 1 for all the goals (ha) Management units of reforestation T1 M1 M18 96.73 Native T9 M3 M7 M8 M10 M11 M18 26.33 T10 163.66 T11 85 85 310.2 27.06 T2 39 T3 10 T4 100 5 T5 85.8 T6 62.89 Field T13 119.87 61.67 37.84 111.52 T7 140 30.69 Swamp T14 126.78 T8 13

Preservation T12 432.82

176

F. de Oliveira et al. / Appl. Math. Comput. 140 (2003) 165178

Table 3 Designation of the prescriptions management using weight 1000 for the goals of diversity of ora and of fauna and weight 1 for the others (ha) Management units of reforestation T1 M1 M18 96.73 Native T9 M4 M8 M16 M18 26.33 T10 56.67 106.99 T11 422.26 T2 26.26 8.88 T3 10 T4 105 T5 85.8 T6 62.89 Field T13 190.23 T7 15.94 154.75 Swamp T14 126.78 T8 13

Preservation T12

Fig. 1. The inuence of the weights in the goals of diversity of the fauna.

In the problems investigated it was veried that some goals are in conict, that is, when one of the goals increases, the other decreases. The goals that are in larger conict are the production of pine and the increase of the diversity of the ora or the fauna. Then, it was made an analysis of these goals through a variation of the production of pine, for unitary weights. Fig. 2 shows the relationship that exists between goals of pine and diversity of the ora in the ve years period of planning. As the goal of pine increases, the diversity of the ora decreases. If no goal of pine is reached, the increase in the diversity of the ora

F. de Oliveira et al. / Appl. Math. Comput. 140 (2003) 165178

177

Fig. 2. Objectives conicting: pine diversity of the ora.

is of %22 species. When the goals of production of pine are larger, observed that the number of dierent species from the ora decreases.

5. Conclusions Solving problems with objectives multiples using goal programming has the advantage of to reach all the objectives as closest as possible of the ideal, since usually is impossible to nd an optimum solution when all the objectives were considered simultaneously. Initially the problem was solved with the use of unitary weights. As some objectives are in conict, it is not possible to reach completely all the goals. Goal programming search a form of to reach all the goals with the minimum deviation possible. Several solutions were obtained, depending on the priority and of the weights considered. It was observed that goal programming possesses the following advantages in relation to the linear programming: All the objectives can be simultaneously optimization, without the necessity to select just one of them to be optimized, consisting the strongest point of the method, since in linear programming only one objective is optimized and the feasible area is dened by the restrictions. The responsible for the decision can choose the best solution from among a collection of solutions, in agreement with its main objective. The problem, with all the objectives, solved through linear programming usually causes instability. The limitation of resources can prevent the possibility to reach simultaneously all the goals. In goal programming the goals do not need to be totally reached.

178

F. de Oliveira et al. / Appl. Math. Comput. 140 (2003) 165178

Nevertheless some bad points were observed based on the runs executed: Instability of the solution when altered the feasible set, through changes in coecients of the matrix A of data and changes of goals. For some situations were veried instability in the tests problems. Generates many restrictions of equality. Requests certain experience of the decisor in relation to the weights.

Acknowledgement To the Industries Pizzatto, for the supply of the data used in this work.

References
[1] A.P. Arp, R.D. Lavigne, Planning with goal programming: a case study for multiple-use of forested land, The Forestry Chronicle 58 (1982) 225232. [2] D.L. Balteiro, C. Romero, Modeling timber harvest scheduling problems with multiple criteria an application in Spain, Forest Science 44 (1) (1998) 4756. [3] C.W. Dane, N.C. Meador, J.B. White, Goal programming in land-use planning, Journal of Forestry June (1977) 325329. [4] D. Field, Goal programming for forest management, Forest Science 19 (2) (1973) 125135. [5] A. Goicoechea, D.R. Hansen, L. Duckstein, in: Multiobjective Decision Analysis with Engineering and Business Applications, John Wiley & Sons, New York, 1882, pp. 17165. [6] G. Mendoza, B.B. Bare, G.E. Campbell, Multiobjetive programming for generating alternatives: a multiple-use planning example, Forest Science 33 (2) (1987) 458491. [7] F. Oliveira, Aplicac ao de goal programming em um problema orestal. Curitiba, Universidade ~ Federal do Paran a, 2000 (Dissertac ao de MestradoM etodos Num ericos em Engenharia/ ~ ao Matem atica). Programac ~

Das könnte Ihnen auch gefallen