Pricing Economic Dispatch With AC Power Flow via Local Multipliers and Conic Relaxation

Mariola Ndrio, Anna Winnicki, Subhonmesh BoseIEEE Trans. Control. Netw. Syst 2024 · DOI

TL;DR

This paper introduces and compares two novel pricing mechanisms for electricity markets that account for the inherent nonconvexity of AC power flow equations. The authors show that these prices, derived from local optimality conditions and a convex relaxation, have distinct properties related to revenue adequacy and market equilibrium, offering a more robust approach to pricing in complex power systems.

What Problem Does This Paper Address?

Electricity markets rely on economic dispatch (ED) to determine how much power generators should produce and at what price. Traditionally, ED problems are simplified to be convex, making pricing straightforward. However, real-world power grids are governed by AC power flow equations, which introduce nonconvexities. This means that finding the absolute optimal dispatch is computationally challenging, and standard pricing methods (like Locational Marginal Prices or LMPs) might not accurately reflect the true costs or incentivize participants correctly. This paper tackles the problem of defining meaningful and economically sound prices in these nonconvex ED settings, specifically focusing on the challenges posed by AC power flow equations.

Methods

The authors propose and analyze two pricing mechanisms:

  • AC-LMPs: These prices are derived from the Lagrange multipliers that satisfy the Karush-Kuhn-Tucker (KKT) conditions for a locally optimal dispatch solution of the nonconvex AC power flow ED problem.
  • SDP-LMPs: These prices are derived from the optimal dual multipliers of a convex semidefinite programming (SDP) relaxation of the AC power flow ED problem. This relaxation approximates the original nonconvex problem with a convex one, making it easier to solve and providing a global perspective.

The paper then rigorously compares these two pricing schemes, analyzing their relationships, revenue adequacy properties (whether the system operator collects enough revenue to cover payments), and market equilibrium properties (whether participants have an incentive to follow the dispatch signals). They also explore the connection between SDP-LMPs and existing Distribution Locational Marginal Prices (DLMPs) derived from Second-Order Cone Programming (SOCP) relaxations for radial networks.

Datasets

Not discussed in the paper. The paper focuses on theoretical analysis and uses a simulated three-bus power network for numerical experiments.

Key Insights and Ideas

  • AC-LMPs vs. SDP-LMPs: While AC-LMPs are tied to local optima of the original problem, SDP-LMPs are derived from a global convex relaxation. This fundamental difference leads to distinct properties, especially when the original problem has a non-zero duality gap (meaning the relaxation's solution is better than the best local optimum of the original problem).
  • Duality Gap Matters: When the duality gap is zero (i.e., the SDP relaxation finds the global optimum of the original problem), AC-LMPs and SDP-LMPs coincide. However, with a non-zero duality gap, these prices can differ significantly, impacting market outcomes.
  • Market Equilibrium Support: AC-LMPs, even for a local optimum, are shown to always support a market equilibrium, meaning participants are incentivized to follow the dispatch. SDP-LMPs, on the other hand, only guarantee this when the duality gap is zero. Otherwise, side-payments might be needed to ensure compliance.
  • Revenue Adequacy: The paper provides a sufficient condition for revenue adequacy (system operator collects enough revenue) based on AC-LMPs, particularly when voltage lower limits are not binding. This condition is shown to be sufficient but not always necessary.
  • SDP-LMPs as Generalization: SDP-LMPs are shown to generalize SOCP-DLMPs for radial networks, providing a unified framework for pricing in distribution systems.

Key Results

  • The paper establishes a dual equivalence between the nonconvex AC ED problem and its SDP relaxation, meaning their dual problems are essentially the same.
  • It's proven that the epigraph of the optimal cost of the SDP relaxation contains the convex hull of the epigraph of the optimal cost of the nonconvex AC ED problem.
  • AC-LMPs are shown to always support a market equilibrium, while SDP-LMPs only do so when the duality gap is zero.
  • A sufficient condition for revenue adequacy with AC-LMPs is derived and illustrated.
  • For radial networks, SDP-LMPs are shown to be equivalent to SOCP-DLMPs.

Limitations and Future Work

  • Nonconvexity from Costs: The current analysis focuses on nonconvexity arising from AC power flow equations. Future work aims to extend the analysis to include nonconvexities from unit commitment decisions (startup/no-load costs) alongside AC power flow.
  • Stochasticity: The paper considers a deterministic setting. Future research will explore how to handle uncertainties in renewable energy supply, making the pricing mechanisms more robust.
  • Side-Payments: While SDP-LMPs can lead to situations requiring side-payments to ensure market participant compliance, the paper shows they aim to minimize a sum of lost opportunity costs and product revenue shortfall, offering some principled behavior even in the presence of duality gaps.

Summary updated

BibTeX
@article{ndrio2024pricing,
  author = {Mariola Ndrio and Anna Winnicki and Subhonmesh Bose},
  title = {{Pricing Economic Dispatch With AC Power Flow via Local Multipliers and Conic Relaxation}},
  journal = {IEEE Trans. Control. Netw. Syst},
  year = {2024},
  doi = {10.1109/tcns.2022.3203470},
}
Figure