Why Design a Pyomo/CVXPY Neutral Library?

Bullets to formulate into paragraphs:

Both libraries are widely used

Pros and cons of each

cvx: native support for matrices, computation time

pyo: nonconvex support, object-oriented (could be considered a con as well - not as lightweight)

etc.

Do not want energy research community siloed

Reproducability and code quality

Using the same functions allows for results between researchers using different methods to be consistent

We know that any discrepancies are due to their model itself, not the cost calculation

Bugs are more likely to be discovered the more widely used the package is, and when they are fixed there will not be two versions out of sync with one another

Distributed optimization with multiple libraries

Maybe this is a niche use case, but we use this package to link together cost calculations from a joint opt. problem with both CVXPY and Pyomo subproblems

It’s nice to have the same dependency for every subproblem