Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Rapid Scalable Distributed Power Flow with Open-Source Implementation

Xinliang Dai, Yichen Cai, Yuning Jiang, Veit Hagenmeyer

TL;DR

The paper tackles scalable distributed AC power flow for large and heterogeneous grids by reformulating the problem as a zero-residual least-squares formulation and solving it with a Gauss-Newton based inexact ALADIN. A reduced modelling method further lowers problem size by separating known and unknown variables across bus-types, enabling faster computations. An open-source implementation, rapidPF+, demonstrates large-scale performance up to 10,224 buses, achieving near-centralized runtimes and locally quadratic convergence in few iterations. The approach offers privacy-preserving, fast, and scalable distributed PF suitable for parallel execution in future power system analysis tools.

Abstract

This paper introduces a new method for solving the distributed AC power flow (PF) problem by further exploiting the problem formulation. We propose a new variant of the ALADIN algorithm devised specifically for this type of problem. This new variant is characterized by using a reduced modelling method of the distributed AC PF problem, which is reformulated as a zero-residual least-squares problem with consensus constraints. This PF is then solved by a Gauss-Newton based inexact ALADIN algorithm presented in the paper. An open-source implementation of this algorithm, called rapidPF+, is provided. Simulation results, for which the power system's dimension varies from 53 to 10224 buses, show great potential of this combination in the aspects of both the computing.

Rapid Scalable Distributed Power Flow with Open-Source Implementation

TL;DR

The paper tackles scalable distributed AC power flow for large and heterogeneous grids by reformulating the problem as a zero-residual least-squares formulation and solving it with a Gauss-Newton based inexact ALADIN. A reduced modelling method further lowers problem size by separating known and unknown variables across bus-types, enabling faster computations. An open-source implementation, rapidPF+, demonstrates large-scale performance up to 10,224 buses, achieving near-centralized runtimes and locally quadratic convergence in few iterations. The approach offers privacy-preserving, fast, and scalable distributed PF suitable for parallel execution in future power system analysis tools.

Abstract

This paper introduces a new method for solving the distributed AC power flow (PF) problem by further exploiting the problem formulation. We propose a new variant of the ALADIN algorithm devised specifically for this type of problem. This new variant is characterized by using a reduced modelling method of the distributed AC PF problem, which is reformulated as a zero-residual least-squares problem with consensus constraints. This PF is then solved by a Gauss-Newton based inexact ALADIN algorithm presented in the paper. An open-source implementation of this algorithm, called rapidPF+, is provided. Simulation results, for which the power system's dimension varies from 53 to 10224 buses, show great potential of this combination in the aspects of both the computing.
Paper Structure (15 sections, 2 theorems, 18 equations, 4 figures, 3 tables, 2 algorithms)

This paper contains 15 sections, 2 theorems, 18 equations, 4 figures, 3 tables, 2 algorithms.

Table of Contents

  1. Introduction
  2. Problem Formulation
  3. Distributed Power Flow
  4. Least-Squares Formulation
  5. Sensitivities
  6. Algorithm
  7. Standard ALADIN
  8. Gauss-Newton based inexact ALADIN
  9. Open-source Implementation
  10. Reduced modelling method
  11. rapidPF vs. rapidPF+
  12. Simulation Results
  13. Comparison of different combinations
  14. Convergence behavior of 10224-Bus system
  15. Conclusions

Key Result

Proposition 1

Let the power flow problem eq:power flow equation be feasible, i.e., a primal solution $\chi^{\ast}$ to the problem eq::dist ls problem exists such that the power flow residual $r_\ell(\chi^{\ast}_\ell)=0$ for all $\ell\in\mathcal{R}$ bounded by consensus constraint eq::ls::consensus, and let licq h

Figures (4)

  • Figure 1: Decomposition by sharing components for a two-region system
  • Figure 2: Flow charts for solving distributed acpf by the rapidpf and the rapidpf+ toolbox
  • Figure 3: Connection graph of 10224-bus test case
  • Figure 4: Convergence behavior of 10224-bus system by applying reduced modeling method with the Gauss-Newton based inexact aladin algorithm

Theorems & Definitions (2)

  • Proposition 1
  • Theorem 1