[Paper Review] Enabling Distributed Optimization in Large-Scale Power Systems

Junyao Guo, Gabriela Hug+1 more|arXiv (Cornell University)|May 31, 2016

Optimal Power Flow Distribution25 references3 citations

TL;DR

This paper demonstrates for the first time that distributed optimization via ADMM can efficiently solve non-convex AC Optimal Power Flow (OPF) in a large-scale real-world transmission network—the 2383-bus Polish system—using a spectral clustering-based partitioning technique. The method enables fast convergence and near-optimal solutions without pre-defined system partitions, proving distributed OPF is viable for real-world large-scale power systems.

ABSTRACT

Distributed optimization for solving non-convex Optimal Power Flow (OPF) problems in power systems has attracted tremendous attention in the last decade. Most studies are based on the geographical decomposition of IEEE test systems for verifying the feasibility of the proposed approaches. However, it is not clear if one can extrapolate from these studies that those approaches can be applied to very large-scale real-world systems. In this paper, we show, for the first time, that distributed optimization can be effectively applied to a large-scale real transmission network, namely, the Polish 2383-bus system for which no pre-defined partitions exist, by using a recently developed partitioning technique. More specifically, the problem solved is the AC OPF problem with geographical decomposition of the network using the Alternating Direction Method of Multipliers (ADMM) method in conjunction with the partitioning technique. Through extensive experimental results and analytical studies, we show that with the presented partitioning technique the convergence performance of ADMM can be improved substantially, which enables the application of distributed approaches on very large-scale systems.

Motivation & Objective

• To investigate whether distributed optimization can be practically applied to large-scale real-world power systems, particularly for non-convex AC OPF problems.
• To evaluate the effectiveness of a spectral clustering-based partitioning technique in enabling efficient distributed optimization for real transmission networks.
• To demonstrate that ADMM, a state-of-the-art distributed method, can achieve fast convergence and near-optimal solutions when combined with this partitioning approach.
• To show that the partitioning technique is generalizable across different distributed optimization methods beyond OCD, such as ADMM.
• To establish that distributed optimization is a viable alternative to centralized approaches in real-world systems with no pre-defined partitions.

Proposed method

• The paper applies the Alternating Direction Method of Multipliers (ADMM) to solve the non-convex AC Optimal Power Flow (OPF) problem in a distributed manner.
• A spectral clustering-based network partitioning technique is used to decompose the Polish 2383-bus system into subregions without relying on pre-defined geographical or operational boundaries.
• The partitioning method defines an affinity metric based on power system coupling to map the network decomposition problem to a graph partitioning problem.
• Each subregion solves a local OPF subproblem using ADMM, with dual variables and consensus constraints enforced via alternating updates of primal and dual variables.
• The method incorporates both equality and inequality constraints, including thermal line limits, and uses Lagrangian relaxation within the ADMM framework.
• The convergence behavior is analyzed under varying numbers of regions, and the impact of partitioning on computational efficiency is evaluated.

Experimental results

Research questions

• RQ1 Can distributed optimization via ADMM be effectively applied to large-scale real-world power systems without pre-defined partitions?
• RQ2 How does the spectral clustering-based partitioning technique affect the convergence speed and solution quality of ADMM in solving non-convex AC OPF?
• RQ3 Is the proposed partitioning method generalizable to other distributed optimization algorithms beyond Optimality Condition Decomposition (OCD)?
• RQ4 What is the trade-off between the number of regions and computational efficiency in distributed ADMM for real power systems?
• RQ5 How robust is the solution quality to variations in ADMM parameters and initial conditions?

Key findings

• The proposed ADMM-based distributed approach successfully solves the non-convex AC OPF problem in the 2383-bus Polish system with a solution close to the local optimum.
• The spectral clustering-based partitioning technique significantly improves ADMM convergence speed, enabling efficient solution of large-scale problems.
• A crossover point exists around 40 regions where the distributed approach becomes faster than the centralized approach in terms of estimated computation time.
• The solution quality is robust to variations in ADMM parameters and initial conditions, and remains stable even when thermal line constraints are included.
• The partitioning method is generalizable and effective not only for OCD but also for ADMM, indicating broad applicability across distributed optimization frameworks.
• The method demonstrates that distributed optimization is practically viable for real-world large-scale power systems, marking a milestone in scalable power system operations.