数学规划模型习题-线性规划

上述问题本质为一个线性规划问题。
而线性规划有以下性质。

1. 比例性:每个决策变量对目标函数的贡献，与该决策变量的取值成正比；每个决策变量对每个约束条件右端项的贡献，与该决策变量的取值成正比
2. 可加性:各个决策变量对目标函数的贡献，与其他决策变量的取值无关；各个决策变量对每个约束条件右端项的贡献，与其他决策变量的取值无关
3. 连续性:每个决策变量的取值都是连续的

最终我们可作出如下假设
模型假设

1. 各种蔬菜产品的单价均是与购买量无关的常数，各种蔬菜产品提供的营养价值也是与当前摄入量无关的常数
2. 各个蔬菜的购买单价均是与他们相互间购买量无关的常数（一种蔬菜的购买不会影响另一种蔬菜的价格）；各种蔬菜餐品提供的营养价值也是与相互的摄入量无关的常数（一种蔬菜的摄入并不会影响到另一种蔬菜的营养价值）
3. 各种蔬菜的购买件数（杯数）可以是任意常数

在满足约束满足学生对基本营养的需求情况下，目标函数为投入最小费用。那么我们可以建立数学模型如下
假设苹果、香蕉、胡萝卜、枣汁、鸡蛋的购买量分别为$x_1,x_2,x_3,x_4,x_5$假设每天食材支出为$y$元。那么我们可以建立目标函数为$$Miny=10x_1+15x_2+5x_3+60x_4+8x_5$$
同样我们也可以建立约束条件
$$\begin{gathered} S.T. \\ 0.3\ast x_1+1.2\ast x_2+0.7\ast x_3+3.5\ast x_4+5.5\ast x_5>=50 \\ 73\ast x_1+96\ast x_2+20253\ast x_3+890\ast x_4+279\ast x_5>=4000 \\ 9.6\ast x_1+7\ast x_2+19\ast x_3+57\ast x_4+22\ast x_5>=1000 \\ {x}_i>=0,i=1,2,3,4,5 \end{gathered}$$
我们利用python求解得到结果。
代码

#程序文件Pex5_4.py
import numpy as np
from cvxopt import matrix, solvers
c=matrix([10.,15,5,60,8]); #系数向量
A=matrix([[-0.3,-1.2,-0.7,-3.5,-5.5],
          [-73,-96,-20253,-890,-279],
          [-9.6,-7,-19,-57,-22],
          [-1,0,0,0,0],
          [0,-1,0,0,0],
          [0,0,-1,0,0],
          [0,0,0,-1,0],
          [0,0,0,0,-1],
         ]).T#约束条件左端系数
b=matrix([-50.,-4000,-1000,0,0,0,0,0]); 
sol=solvers.lp(c,A,b)
print("最优解为：\n",sol['x'])
print("最优值为：",sol['primal objective'])
'''
    运行结果
    pcost       dcost       gap    pres   dres   k/t
 0:  1.0965e+03  2.1410e+04  2e+02  2e-03  1e+03  1e+00
 1:  4.1684e+02  1.1224e+05  1e+04  1e-02  7e+03  7e+01
 2:  5.4657e+02  7.8504e+03  4e+02  8e-04  5e+02  1e+02
 3:  4.5291e+02  3.3434e+03  2e+02  3e-04  2e+02  5e+01
 4:  1.2653e+03  1.3518e+04  6e+03  1e-03  8e+02  4e+02
 5:  2.8744e+03  2.1860e+04  1e+05  1e-03  9e+02  6e+03
 6:  2.0199e+03  7.8726e+03  3e+04  6e-04  4e+02  3e+01
 7:  1.3657e+03  2.6481e+03  4e+03  1e-04  8e+01  7e+01
 8:  8.6825e+02  1.6685e+03  3e+03  8e-05  5e+01  8e+01
 9:  4.3787e+02  5.4587e+02  3e+02  1e-05  6e+00  1e+01
10:  3.6601e+02  3.6929e+02  8e+00  3e-07  2e-01  3e-01
11:  3.6390e+02  3.6470e+02  2e+00  7e-08  4e-02  1e-01
12:  3.5529e+02  3.5833e+02  6e+01  1e-07  7e-02  2e+00
13:  3.1848e+02  3.2539e+02  1e+02  7e-08  4e-02  6e+00
14:  3.0232e+02  3.0603e+02  6e+01  4e-08  2e-02  3e+00
15:  2.7112e+02  2.7126e+02  3e+00  2e-09  1e-03  1e-01
16:  2.6938e+02  2.6938e+02  3e-02  2e-11  1e-05  1e-03
17:  2.6936e+02  2.6936e+02  3e-04  2e-13  1e-07  1e-05
18:  2.6936e+02  2.6936e+02  3e-06  1e-13  1e-09  1e-07
Optimal solution found.
最优解为：
 [ 7.15e-09]
[ 8.94e-09]
[ 4.94e+01]
[ 1.67e-09]
[ 2.81e+00]

最优值为： 269.3602711693341
'''

或者:

import cvxpy as cp
from numpy import array
c=array([10.,15,5,60,8])  #定义目标向量
a=array([[-0.3,-1.2,-0.7,-3.5,-5.5],
          [-73,-96,-20253,-890,-279],
          [-9.6,-7,-19,-57,-22],
          [-1,0,0,0,0],
          [0,-1,0,0,0],
          [0,0,-1,0,0],
          [0,0,0,-1,0],
          [0,0,0,0,-1]])  #定义约束矩阵
b=array([-50.,-4000,-1000,0,0,0,0,0])  #定义约束条件的右边向量
x=cp.Variable(5,integer=False)  #定义5个决策变量，True为整数规划，False为实数
obj=cp.Minimize(c*x)  #构造目标函数
cons=[a*x<=b, x>=0]    #构造约束条件
prob=cp.Problem(obj, cons)  #构建问题模型
prob.solve(solver='GLPK_MI',verbose =True)  #求解问题
print("最优值为:",prob.value)
print("最优解为：\n",x.value)

'''
运行结果:
===============================================================================
                                     CVXPY                                     
                                    v1.1.12                                    
===============================================================================
(CVXPY) Aug 11 05:13:02 PM: Your problem has 5 variables, 2 constraints, and 0 parameters.
(CVXPY) Aug 11 05:13:02 PM: It is compliant with the following grammars: DCP, DQCP
(CVXPY) Aug 11 05:13:02 PM: (If you need to solve this problem multiple times, but with different data, consider using parameters.)
(CVXPY) Aug 11 05:13:02 PM: CVXPY will first compile your problem; then, it will invoke a numerical solver to obtain a solution.
-------------------------------------------------------------------------------
                                  Compilation                                  
-------------------------------------------------------------------------------
(CVXPY) Aug 11 05:13:02 PM: Compiling problem (target solver=GLPK_MI).
(CVXPY) Aug 11 05:13:02 PM: Reduction chain: Dcp2Cone -> CvxAttr2Constr -> ConeMatrixStuffing -> GLPK_MI
(CVXPY) Aug 11 05:13:02 PM: Applying reduction Dcp2Cone
(CVXPY) Aug 11 05:13:02 PM: Applying reduction CvxAttr2Constr
(CVXPY) Aug 11 05:13:02 PM: Applying reduction ConeMatrixStuffing
(CVXPY) Aug 11 05:13:02 PM: Applying reduction GLPK_MI
(CVXPY) Aug 11 05:13:02 PM: Finished problem compilation (took 2.800e-02 seconds).
-------------------------------------------------------------------------------
                                Numerical solver                               
-------------------------------------------------------------------------------
(CVXPY) Aug 11 05:13:02 PM: Invoking solver GLPK_MI  to obtain a solution.
-------------------------------------------------------------------------------
                                    Summary                                    
-------------------------------------------------------------------------------
(CVXPY) Aug 11 05:13:02 PM: Problem status: optimal
(CVXPY) Aug 11 05:13:02 PM: Optimal value: 2.694e+02
(CVXPY) Aug 11 05:13:02 PM: Compilation took 2.800e-02 seconds
(CVXPY) Aug 11 05:13:02 PM: Solver (including time spent in interface) took 1.800e-02 seconds
最优值为: 269.36026936026735
最优解为：
 [-0.         -0.         49.38271605 -0.          2.80583614]

'''

最终得到结果上述蔬菜分别购买:

0.        -0.         49.38271605 		0.          2.80583614

最终支出为$269.36026936026735$（角）
灵敏性分析
接下来我们介绍几个概念:

1. 影子价格:影子价格通俗理解

接下来我们求解下面两问，最终我们根据上述求解结果分别购买49.38的胡萝卜和2.80的鸡蛋。我们可以算出若按照这个方案安排，营养含量如下图

由于python灵敏度性不太完善，影子价格的计算不太方便，因此对于后两问需重新调整模型进行计算（后续可考虑Scipy Linprog中的影子价格）