Optimization 古罗比-俾莫误差的灵敏度分析

我正在用Pyomo建模一个MIP，我想用gurobi做一个敏感性分析，以得到仍然确保最优组合解的系数范围（决策变量是二进制的）。 据我所知，所有决策变量的值必须固定，模型应重新求解，解算器给出系数的范围

gm

。 我尝试了不同的编码（案例1，案例2）

以下是我的模型的一些部分：

from __future__ import division
from pyomo.environ import *
from pyomo.opt import SolverFactory
#from gurobipy import *



model = AbstractModel ()


model.f = Param(within=PositiveIntegers)
model.F = RangeSet(1, model.f)

model.v = Param(within=PositiveIntegers)
model.V = RangeSet(1, model.v)

model.m = Param(within=PositiveIntegers)
model.M = RangeSet(1, model.m)


...
model.yw = Var(model.M, within=Binary)

model.ys = Var(model.index_fvm, within=Binary)

model.bs = Param(model.index_fvm, within=Binary, default=0)


def Obj_rule(model):
    expr = 0.0
    for (f,v,m) in model.index_fvm:
        expr += model.g[m] * model.ys[f,v,m]    
    for m in model.M:
        expr += model.g[m] * model.yw[m]

    return expr

model.OBJ = Objective(rule=Obj_rule, sense=maximize)

...

def C10_constraint_rule (model, m):
    expr= 0.0
    for (f,v) in model.index_fv:
        expr += model.ys[f,v,m]    

    expr += model.yw[m]

    return expr <= 1

model.C10Constraint = Constraint(model.M, rule=C10_constraint_rule)

...

# Pyomo doc 5.1.1. P82-83

# case 1

opt = SolverFactory("gurobi")

instance = model.create_instance()
results = opt.solve(instance)
instance.display()

instance.load(results)

if instance.ys[f,v,m] == 0:
    instance.ys[f,v,m].setlb(1)
    instance.ys[f,v,m].setub(1)
else:
    instance.ys[f,v,m].setlb(0)
    instance.ys[f,v,m].setub(0)
instance.preprocess()
results = opt.solve(instance)


# case 2

#http://www.gurobi.com/documentation/7.0/examples.pdf
if len(sys.argv) < 2:
    print('Usage: sensitivity.py filename')
    quit()

# Read and solve model

model = read(sys.argv[1])

if model.IsMIP == 0:
    print('Model is not a MIP')
    exit(0)

model.optimize()

if model.status != GRB.Status.OPTIMAL:
    print('Optimization ended with status %d' % model.status)
    exit(0)

# Store the optimal solution

origObjVal = model.ObjVal
for v in model.getVars():
    v._origX = v.X

# Disable solver output for subsequent solves

model.Params.outputFlag = 0

# Iterate through unfixed, binary variables in model

for v in model.getVars():
    if (v.LB == 0 and v.UB == 1 \
        and (v.VType == GRB.BINARY or v.VType == GRB.INTEGER)):

        # Set variable to 1-X, where X is its value in optimal solution

        if v._origX < 0.5:
            v.LB = v.Start = 1
        else:
            v.UB = v.Start = 0

        # Update MIP start for the other variables

        for vv in model.getVars():
            if not vv.sameAs(v):
                vv.Start = vv._origX

        # Solve for new value and capture sensitivity information

        model.optimize()

        if model.status == GRB.Status.OPTIMAL:
            print('Objective sensitivity for variable %s is %g' % \
                  (v.VarName, model.ObjVal - origObjVal))
        else:
            print('Objective sensitivity for variable %s is infinite' % \
                  v.VarName)

        # Restore the original variable bounds

        v.LB = 0
        v.UB = 1

我在案例2中得到了这个错误：

ERROR: Unexpected exception while loading model:
        __init__() takes exactly 3 positional arguments (2 given)

gm.dat如下所示：

param g:=
1 7
7 6
54 7
52 7
53 1
44 7
45 7
56 7
46 7
48 7
70 7
71 7
72 7
73 7
100 7 
112 4
;

有没有人知道如何处理这个问题，并且知道如何进行敏感性分析

先谢谢你

拉拉

也许我遗漏了什么——你似乎声明了

Param

s

f

和

v

，但是你的数据文件中有一个

Param

g

的值，也许我遗漏了什么——你似乎声明了

Param

s

f

和

v

，但是您的数据文件具有

Param

g

param g:=
1 7
7 6
54 7
52 7
53 1
44 7
45 7
56 7
46 7
48 7
70 7
71 7
72 7
73 7
100 7 
112 4
;