I have an MINLP model implemented in BONMIN. The model functions correctly; however, since it is applied to distribution power systems, its complexity increases as the number of nodes grows. For this reason, I have attempted to initialize the model’s binary variables using:
@variable(OPF, Y[l in 1:NL, c in 1:NC], Bin, start=Y_sol[l,c])
where Y_sol[l,c] is a binary matrix that contains the initial solution of the problem.
I have noticed that, whether the variables are initialized or not, the solver always starts from the same point, and the iterations proceed in the same way.
In this regard, I also attempted to initialize the model with a suboptimal solution for further refinement. That is, I ran the model for a set period, saved the solution, and then tried to initialize the model with this new solution, hoping it would start from this point and find the optimal solution. However, I again noticed that the model behaves as if the variables were not initialized.
In the documentation, I found that `
warm_start
` allows initializing the model with a previous solution, but I have not been able to understand how to implement it.
I am currently working with Bonmin 1.8.9, Cbc 2.10.8, and Ipopt 3.14.13.
