Optimize and Interpret Results#

Instantiate#

To optimize, the pulpo_worker must first instantiate the problem. This step combines the provided inputs and passes them to the abstract optimization model, creating a concrete optimization model.

Simple Unconstrained Instantiation#

For an unconstrained problem, the instantiation can be as simple as:

pulpo_worker.instantiate(demand=demand, choices=choices)

Adding Constraints#

Additional constraints can be specified for different aspects of the system:

upper_limit: Constraints on scaling vectors (e.g., capacity or availability).
upper_elem_limit: Constraints on environmental flows (e.g., emissions or resource usage).
upper_imp_limit: Constraints on impact indicators (e.g., global warming potential).

pulpo_worker.instantiate(
    demand=demand, 
    choices=choices, 
    upper_limit=upper_limit, 
    upper_elem_limit=upper_elem_limit, 
    upper_imp_limit=upper_imp_limit
)

Optimize / Solve#

After instantiating the problem, you can solve it using the solve() method:

pulpo_worker.solve()
# Optionally, specify the GAMS solver path:
# pulpo_worker.solve(GAMS_PATH="path/to/gams")

If no GAMS path is specified, the open-source solver highspy will be used. While highspy is slower than commercial solvers like CPLEX and may occasionally struggle with complex problems, it typically performs well for most scenarios.

Summarize / Interpret Results#

To make the results accessible and interpretable, use the following methods:

summarize_results(): Provides a concise overview of the key decisions, objective values, and constraints.
save_results(): Saves detailed results to an Excel file for further analysis.

Example Usage:#

pulpo_worker.summarize_results(choices=choices, demand=demand, constraints=upper_limit)
pulpo_worker.save_results(choices=choices, demand=demand, name='path/to/save/results.xlsx')

Visualization and Iteration#

The interpretation and visualization of results are left to the user. Depending on the insights gained, re-iterations to refine inputs or constraints may be necessary, just as in a traditional LCA workflow.