Optimizing MimiDICE

21

Glad that worked! You should be able to apply to apply the constraints over just a time period, by selecting only some of the values out of the array, like model -> sum(model[:emissions, :MIU][10:20]) - 0.2. It’s not pretty, but it will depend on what exactly you want to do.

22

Perfect, that works too!

23

Hello everyone,

I was facing a similar issue when I tried optimising MimiDICE2016R2 this summer. OptiMimi did not work because installing the package in my Julia environment forcefully downgraded my Mimi version:

image
image703×43 7.47 KB

Hence, I let myself be inspired by the code for Budolfson et al. (2021) and created my own fork of MimiDICE2016R2 (OptMimiDICE2016R2) which has an in-built optimise_model function. Maybe that’s helpful to some people landing in this forum!

Cheers!

1 Like
25

Hello jrising, I want to know how to reduce it to 2 parameters? Is it convenient for you to post the total code? It could offer great help.

26

Here’s a simple example that uses paramtrans to optimize DICE with 3 parameters (just an increase in carbon price that takes 3 periods):

using Mimi
using OptiMimi
using MimiDICE2016
using Distributions

model = MimiDICE2016.get_model()
run(model)

cprice_desired = 100
function objective(model::Model)
    -(abs(model[:neteconomy, :CPRICE][1] - 0) + sum(abs.(model[:neteconomy, :CPRICE][2:4] .- cprice_desired)))
end

optprob = problem(model, [:emissions], [:MIU], zeros(3), 0.99 * ones(3), objective,
                  paramtrans=convert(Vector{Function}, [(params) -> [0; params; params[3] * ones(96)]]))

soln = solution(optprob, () -> rand(Uniform(0.01, 0.99)) * ones(3))
println(soln)

update_param!(model, :MIU, [0; soln[2]; soln[2][3] * ones(96)])
run(model)