How to obtain dual variables (shadow price) for a set of constraints

Specific Domains Optimization (Mathematical)
,
1

Hello,

I have time-related constraints and I modeled it as

for time in 1:nTime
    @constraint(model, a[time] + b[time] == c[time], base_name = "cons" 
end

The model is solved and I want to obtain the dual variables, thus, how to obtain them by the above constraint formulation method?

BTW, I know there are two other ways to obtain the dual:

@constraint(model, cons1, a[1] + b[1] == c[1])
@constraint(model, cons2, a[2] + b[2] == c[2])
...
shadow_price(cons1)
...

and

cons = @constraints(model, [time in 1:nTime], a[time] + b[time] == c[time])
shodow_price(cons)

(My model contains other parameters, so that the top modeling method is preferred for me. )

1 Like
2

Sorry, my method is

for time in 1:nTime
    @constraint(model, a[time] + b[time] == c[time], 
    base_name = "cons$(time)")
end
3

The dual variables can be obtained by:

for time in 1:nTime
    shadow_price(constraint_by_name(model, "cons$(time)")
end
1 Like
4

You should always keep in mind that JuMP variables and constraints are just regular Julia objects.

That means you can use them with any other data structure in Julia. Iā€™d write your model as:

cons = []
for time in 1:nTime
    c = @constraint(model, a[time] + b[time] == c[time], base_name = "cons") 
    push!(cons, c)
end
shadow_price.(cons)

or perhaps

cons = map(1:nTime) do time
    # Stuff here
    return @constraint(model, a[time] + b[time] == c[time], base_name = "cons")
end
shadow_price.(cons)

Your constraint_by_name approach also works, but it involves a couple of lookups and relies on base_name being unique. So having an explicit vector of constraints might be better.

1 Like
5

Indeed, this method could be better, as it returns a vector, which is better for using dual variables.

1 Like