[Paper Review] Phase Transitions for Optimality Gaps in Optimal Power Flows A Study on the French Transmission Network

Terrence W. K. Mak, Lyndon Shi+1 more|arXiv (Cornell University)|Jul 14, 2018

Optimal Power Flow Distribution5 references3 citations

TL;DR

This paper investigates phase transitions in optimality gaps for Optimal Power Flow (OPF) on the French transmission network using real-world data. By uniformly increasing loads, it reveals that large optimality gaps—up to 20%—occur even far from congestion points, driven by voltage magnitude constraints. These gaps emerge via complex bifurcations not explained by current limits, challenging assumptions about relaxation performance in non-congested systems.

ABSTRACT

This paper investigates phase transitions on the optimality gaps in Optimal Power Flow (OPF) problem on real-world power transmission systems operated in France. The experimental results study optimal power flow solutions for more than 6000 scenarios on the networks with various load profiles, voltage feasibility regions, and generation capabilities. The results show that bifurcations between primal solutions and the QC, SOCP, and SDP relaxation techniques frequently occur when approaching congestion points. Moreover, the results demonstrate the existence of multiple bifurcations for certain scenarios when load demands are increased uniformly. Preliminary analysis on these bifurcations were performed.

Motivation & Objective

• To identify and analyze conditions under which large optimality gaps emerge in real-world AC Optimal Power Flow (AC-OPF) problems.
• To investigate whether large optimality gaps occur in non-congested regions of transmission networks, challenging the assumption that gaps are primarily congestion-related.
• To examine the role of voltage magnitude constraints and current limits in inducing phase transitions and bifurcations in relaxation-based OPF solutions.
• To evaluate the numerical behavior and convergence properties of AC-OPF solvers and convex relaxations (QC, SOCP, SDP) under increasing load conditions.
• To determine whether these phenomena are artifacts of a single test case or representative of broader structural behaviors in real transmission networks.

Proposed method

• The study uses a real-world subset of the French transmission network (MSR) from the ARPA-E Grid Data project, with detailed network parameters and real load profiles.
• Load levels were uniformly increased from 10% below to 80% above base case, generating over 6,000 scenarios across varying voltage feasibility regions and generation capabilities.
• AC-OPF was solved using IPOPT, while QC, SOCP, and SDP relaxations were solved via interior-point methods to obtain lower bounds and assess optimality gaps.
• Primal solutions were recovered from relaxations using power flow procedures to validate feasibility and compare with AC solutions.
• The percentage of tight voltage magnitude constraints and current limits was tracked across scenarios to correlate with optimality gap behavior.
• CPU runtimes and convergence behavior were monitored to assess numerical stability and computational challenges.

Experimental results

Research questions

• RQ1 Do large optimality gaps in AC-OPF occur in real transmission networks even when the system is far from congestion?
• RQ2 What causes phase transitions and bifurcations in optimality gaps, particularly when the network is not heavily loaded?
• RQ3 How do voltage magnitude constraints influence the emergence of large optimality gaps compared to current flow limits?
• RQ4 Do convex relaxations (QC, SOCP, SDP) exhibit similar bifurcation patterns as the AC solution, or do they remain stable?
• RQ5 Can the observed phase transitions be reproduced across other large-scale networks, suggesting a general structural property?

Key findings

• Large optimality gaps of up to 20% were observed in real-world AC-OPF instances on the French transmission network even at 22% load increase, well before congestion points.
• These large gaps emerged not at congestion, but via a phase transition around 20% load increase, followed by a recovery to smaller gaps by 30%, indicating complex, non-monotonic behavior.
• The phase transitions in optimality gaps strongly correlated with the percentage of buses operating at tight voltage magnitude limits in the AC solution, suggesting voltage constraints as the root cause.
• The relaxations (QC, SOCP, SDP) did not exhibit the same bifurcation behavior, indicating that the gap dynamics are driven by the primal solution’s sensitivity to voltage bounds.
• Current limits did not correlate with optimality gap transitions, implying they are not the primary driver of the observed phenomena.
• CPU runtimes for load flow recovery from relaxations increased significantly when optimality gaps were large, indicating numerical instability even in non-congested regimes.