An open-source framework for centralized and distributed Optimal Power Flow.

Mandala Lirion

What is Lirion

Optimal Power Flow is the central decision problem of operating an electric power system: given a network, a set of generators with cost functions, and a forecast demand, what is the cheapest dispatch that respects every physical and operational limit?

Formally it is a constrained optimization problem whose constraints are the nonlinear equations of AC power flow — equations rooted in Maxwell’s laws as they manifest in lumped-element circuit models of transmission networks. Because the AC formulation is non-convex, decades of research have produced a rich landscape of approximations, relaxations, and decompositions, each with its own assumptions, guarantees, and computational profile.

Lirion brings nine of these formulations together under a single API, alongside distributed solution methods (ADMM today, ALADIN soon) — so that researchers and practitioners can compare, combine, and extend them without rewriting infrastructure.

Centralized Models

// centralized.models
Family 0202 Models
Convex Relaxations
loosen constraints to convexify; give lower bounds
Family 0303 Models
Linear & Polynomial Approximations
simplify physics for tractability
Family 0401 Model
Topology Baseline
no electrical physics — pure network flow

Distributed Algorithms

[
ADMM
]
ADMM / Ipopt · Clarabel

Alternating Direction Method of Multipliers

Decomposes the network into zones exchanging dual prices on tie-lines. Implemented for DC, AC, and SOCWR.

[
ALADIN
]
ALADIN / Ipopt · QP

Augmented Lagrangian Alternating Direction Inexact Newton

Coordinates local solutions with a centralized QP for faster non-convex convergence.

Foundations