Positioning & ecosystem

BMOPFTools is infrastructure for distribution-system optimization research: it constructs, validates, converts, and benchmarks conductor-level network models. It is not positioned as another distribution OPF solver — although it ships a reference four-wire rectangular current–voltage OPF (solve_opf) used to validate cases and profile solutions, the product is the model and the tooling around it, not the optimizer.

The reference engine is nevertheless a reusable research substrate: downstream packages can stage the JuMP build, replace selected device formulations, bind semantic parameters and coefficients, and place the model behind DiffOpt. This does not reposition BMOPFTools as a differentiable-optimization service; outer-level semantics, sensitivity execution, and bespoke study logic remain in the downstream package. See Parameterized and differentiable extensions.

The transmission/distribution maturity gap

Power-system software is often discussed as one ecosystem, but transmission and distribution have matured very differently.

Transmission optimization has decades of shared conventions and benchmarks — MATPOWER, PowerModels.jl, and PGLib-OPF among them — which let researchers compare algorithms on common test systems with well-understood problem definitions, and reproduce each other's results.

Distribution optimization remains fragmented. Tools such as PowerModelsDistribution.jl, ppOPF, Open-DSOPF, and various OpenDSS- and GridLAB-D-based workflows solve overlapping problems — Volt/VAr optimization, DER coordination, network reconfiguration, hosting-capacity analysis, distribution OPF — but rely on different data models, feeder representations, assumptions, formulations, and conversion pipelines. As a result, reproducing results and fairly comparing methods is hard.

What BMOPFTools is

Rather than competing with optimization frameworks, BMOPFTools strengthens the ecosystem around them. It focuses on:

• Network representation — a faithful, conductor-level (up-to-four-wire) data model (conventions);
• Data interoperability — OpenDSS ingestion via from_dss and export to PowerModelsDistribution via to_pmd, treated as a semantic projection onto a canonical model rather than a byte-faithful transcode (conversion guide);
• Analytical model construction & validation — provenance, data-quality, and spec-conformance analysis (analysis & reports);
• Benchmark generation — structural repair and standards-grounded gap-filling that turn raw utility feeders into well-posed cases (case augmentation);
• Reproducible workflows — a stable, code-addressable diagnostic vocabulary (finding-code reference).

How it relates to existing tools

BMOPFTools sits between the simulation tools that produce network data and the optimization frameworks that consume it — supplying the shared representation and benchmark cases that let those frameworks be compared. It is complementary to all of the below, not a replacement for any.

The benchmarking gap, and the opportunity

The throughline above is a missing layer of shared benchmark infrastructure for distribution optimization: feeder conversion is lossy, assumptions are inconsistent, data models are incompatible, and reproducibility is limited, so published methods are difficult to compare fairly.

Transmission optimization matured rapidly once the community converged on common benchmark systems and data formats. Distribution likely needs the same shift. BMOPFTools is intended to help create the conditions for it — reusable benchmark feeders, common network representations, transparent model transformations, and cross-framework interoperability — so the broader distribution-optimization ecosystem can mature, without replacing the optimization frameworks at its core.

For the full argument — why the missing benchmarks mislead algorithmic research, why utilities are right to reject methods that work for "80% of the networks, 80% of the time", and the literature supporting that view — see Why benchmarks matter: the 80% problem.