The approach we recommend is to divide this problem into multiple smaller sub-problems and solve them using the 80-20 approach. The aim is to reach a feasible solution and then attempt to increase its optimality. A few examples of sub-problems are:
For instance, the first objective of making a forecast for the next 3 months can be accomplished using a time-series forecasting technique. It’s a fairly straight-forward approach. You can think about aggregation and disaggregation approaches as well.
Below, we have mentioned the steps you can follow to solve a sourcing/sequencing problem:
Once you have a solution that works for one month, you can then try to validate for constraints.
If required, some “hand adjustments” can also be made. We don’t necessarily need a fully automated process to make this work (at least not in the first attempt).
We can then move onto building a similar solution for the remaining two months.
Don’t lose hope, keep trying!