Economic Dispatch in Unbalanced Distribution Networks via Semidefinite Relaxation

Emiliano Dall’Anese, Georgios B. Giannakis, B.f. WollenbergarXiv 2012 · DOI · arXiv

Abstract: The economic dispatch problem is considered for unbalanced three-phase power distribution networks entailing both non-deferrable and elastic loads, and distributed generation (DG) units. The objective is to minimize the costs of power drawn from the main grid and supplied by the DG units over a given time horizon, while meeting the overall load demand and effecting voltage regulation. Similar to optimal power flow counterparts for balanced systems, the resultant optimization problem is nonconvex. Nevertheless, a semidefinite programming (SDP) relaxation technique is advocated to obtain a (relaxed) convex problem solvable in polynomial-time complexity. To promote a reliable yet efficient feeder operation, SDP-compliant constraints on line and neutral current magnitudes are accommodated in the optimization formulated, along with constraints on the power factor at the substation and at nodes equipped with capacitor banks. Tests on the IEEE 13-node radial feeder demonstrate the ability of the proposed method to attain the globally optimal solution of the original nonconvex problem.

Introduction

Modern power distribution systems are evolving with the integration of distributed generation (DG) and flexible, "elastic" loads like electric vehicles. Optimizing these systems, however, is complicated by their inherently unbalanced nature. Unlike high-voltage transmission lines, distribution feeders often serve unequal single-phase loads and have asymmetrical line configurations. Traditional "balanced" models fail to capture these nuances, leading to sub-optimal or even physically damaging operational decisions. This paper addresses the economic dispatch problem—minimizing costs while maintaining network stability—specifically for these complex, unbalanced three-phase networks.

The Challenge of Unbalanced Networks

The core difficulty lies in the non-convexity of the power flow equations. The relationship between voltage and power is non-linear, making the search for a global minimum mathematically challenging. Standard solvers often converge to local minima or fail entirely when faced with the high dimensionality of three-phase systems. Furthermore, unbalanced systems introduce neutral currents, which can cause overheating in transformers and cables if not explicitly managed.

Proposed Methodology: Semidefinite Relaxation (SDR)

To overcome these hurdles, the authors propose a convex optimization framework based on Semidefinite Relaxation (SDR):

  1. System Modeling: The network is modeled using detailed $\pi$-equivalent circuits for three-phase lines. The formulation accounts for non-deferrable loads, elastic loads (which can shift in time to meet a total energy quota), and DG units.
  2. Convexification via SDP: The non-convexity typically arises from rank constraints on voltage-related matrices. By redefining variables in terms of an outer-product matrix of voltages and relaxing the "rank-1" constraint, the problem is transformed into a Semidefinite Program (SDP). SDPs are convex and can be solved efficiently in polynomial time.
  3. Operational Constraints: The model incorporates "SDP-compliant" constraints to ensure grid health, including thermal limits for both line and neutral currents, voltage regulation within strict bounds (e.g., $\pm 5\%$), and power factor maintenance at the substation.

Experimental Validation

The framework was validated using the IEEE 13-node radial test feeder, a standard benchmark for distribution analysis that captures the complexities of unbalanced phases.

The IEEE 13-node radial test feeder used for validation.

To simulate realistic economic conditions, the authors utilized locational marginal prices (LMPs) from the MISO database alongside standard residential and commercial load profiles to simulate daily demand fluctuations.

Key Results and Insights

The numerical tests yielded several significant findings:

  • Global Optimality: Remarkably, for the tested scenarios, the optimal matrix solution was always rank-1. This indicates that the SDR was "exact," providing the globally optimal solution to the original non-convex problem.
  • Economic Load Shifting: The model successfully demonstrated "peak shaving." Elastic loads, such as electric vehicle charging, were automatically scheduled during early morning hours (2 AM – 5 AM) when grid prices and base loads were at their lowest.
Aggregate real load profile and locational marginal prices (LMP) over a 24-hour period.
  • DG and Power Factor Management: While DG units were heavily utilized when grid prices spiked, their operation can negatively impact the power factor at the Point of Common Coupling (PCC). The proposed constraints successfully forced the system to maintain a minimum power factor of 0.8, ensuring power quality.
  • Neutral Current Regulation: By explicitly constraining neutral currents, the model prevents the potential hardware damage that simplified balanced models might overlook.

Conclusion and Future Work

This research provides a robust mathematical foundation for managing modern distribution grids. While the SDR approach proved highly effective for the IEEE 13-node system, future work will focus on scaling these techniques to larger networks with thousands of nodes using techniques like chordal sparsity. Additionally, the authors aim to integrate discrete control variables, such as capacitor bank switching, more seamlessly into the convex framework.

Summary updated

Abstract

The economic dispatch problem is considered for unbalanced three-phase power distribution networks entailing both non-deferrable and elastic loads, and distributed generation (DG) units. The objective is to minimize the costs of power drawn from the main grid and supplied by the DG units over a given time horizon, while meeting the overall load demand and effecting voltage regulation. Similar to optimal power flow counterparts for balanced systems, the resultant optimization problem is nonconvex. Nevertheless, a semidefinite programming (SDP) relaxation technique is advocated to obtain a (relaxed) convex problem solvable in polynomial-time complexity. To promote a reliable yet efficient feeder operation, SDP-compliant constraints on line and neutral current magnitudes are accommodated in the optimization formulated, along with constraints on the power factor at the substation and at nodes equipped with capacitor banks. Tests on the IEEE 13-node radial feeder demonstrate the ability of the proposed method to attain the globally optimal solution of the original nonconvex problem.
BibTeX
@misc{dallanese2012economic,
  author = {Emiliano Dall’Anese and Georgios B. Giannakis and B.F. Wollenberg},
  title = {{Economic Dispatch in Unbalanced Distribution Networks via Semidefinite Relaxation}},
  year = {2012},
  eprint = {1207.0048},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/1207.0048},
}
Figure