I have a program (an external simulation) that receives a JSON file as input called input.json, does the simulation process as a black box (I do not have access) and gives as return a JSON file as output.json. The output file is used to calculate some parameters that compose an objective function. I am trying to use Pyomo to optimize this problem, however, it seems that it is not able to recognize any gradient in the function because of the transfer of information between the input and output files. I was thinking about a gradient-free optimization that surely will take more time but at least can have a good result.
I tried using Mindtpy but it was not able to find an optimum path to decrease the variable or variables.