Breaking the Hierarchy: Distributed Control and Economic Optimality in Microgrids

Florian Dörfler, John W. Simpson-Porco, Francesco BulloIEEE Trans. Control. Netw. Syst 2016 · DOI

TL;DR

This paper proposes a novel, unified control framework for microgrids that seamlessly integrates primary (frequency regulation), secondary (load sharing), and tertiary (economic dispatch) control objectives. By demonstrating a surprising equivalence between droop control and economic optimality, the authors show that a simple, decentralized approach can achieve both stability and cost-efficiency without complex hierarchical structures or time-scale separations.

What Problem Does This Paper Address?

Microgrids, envisioned as self-sufficient energy systems, face the challenge of coordinating numerous distributed energy resources (DERs) like inverters, synchronous generators, and storage systems. Traditional power systems use a hierarchical control structure (primary, secondary, tertiary) with distinct time scales. Adapting this to microgrids is difficult due to their low inertia, rapid dynamics, and the need for plug-and-play operation without detailed system knowledge. Specifically, primary droop control, while essential for stability, introduces steady-state errors that secondary control aims to correct. However, existing secondary control methods often struggle to maintain power sharing or require extensive communication. Furthermore, optimizing economic dispatch (minimizing generation costs) in microgrids has been a separate, complex problem, often solved offline. This paper tackles the problem of unifying these control layers in a decentralized, adaptive, and efficient manner for microgrids.

Methods

The authors develop a comprehensive mathematical model for microgrids, including DERs and loads, using differential-algebraic equations. They build upon the concept of primary droop control, where inverters regulate their power output based on frequency. To address the steady-state frequency errors introduced by droop control, they investigate various decentralized secondary control strategies, including:

  • Centralized Averaging PI (CAPI) Control: This approach uses weighted averaging of frequency measurements from all inverters to compute a secondary control signal, requiring all-to-all communication.
  • Distributed Averaging PI (DAPI) Control: This method uses a local communication graph (not necessarily all-to-all) and Laplacian averaging to achieve frequency regulation.
  • Partial Secondary Control: This strategy allows only a subset of inverters to participate in secondary control, reducing communication overhead and increasing adaptability.

A key theoretical contribution is the demonstration of an equivalence between the steady-state solutions of droop control and the optimal solutions of an economic dispatch problem. This allows them to derive simple guidelines for selecting droop coefficients that simultaneously ensure power sharing and economic optimality. They also reformulate the AC economic dispatch problem into a convex DC equivalent for acyclic networks, simplifying analysis.

Datasets

Not discussed in the paper. The paper uses a simulated IEEE 37 bus distribution network for evaluation.

Key Insights and Ideas

  • Droop Control is Sufficient for Economic Optimality: The most surprising finding is that the steady-state operating points achievable by decentralized droop control are precisely the optimal solutions to the economic dispatch problem. This means that by appropriately tuning droop coefficients, microgrids can achieve cost minimization without a separate tertiary control layer.
  • Unified Control Framework: The paper demonstrates that primary, secondary, and tertiary control objectives can be met simultaneously by simple, distributed PI controllers, eliminating the need for strict hierarchical structures and time-scale separations.
  • Plug-and-Play Capability: The proposed decentralized controllers are designed to operate without explicit knowledge of the network topology, load profile, or DER parameters, enabling true plug-and-play functionality.
  • Power Flow Shaping: The work provides a method to select droop coefficients to achieve desired power injection setpoints, effectively "shaping" the power flow in the microgrid.
  • Partial Secondary Control Benefits: Allowing only a subset of inverters to perform secondary control significantly reduces communication complexity and enhances adaptability while still maintaining stability and frequency regulation.

Key Results

  • The paper rigorously proves the equivalence between the steady-state of droop-controlled microgrids and the optimal solutions of the AC economic dispatch problem for acyclic networks.
  • It shows that DAPI and CAPI controllers can effectively regulate frequency, maintain primary droop control properties, and achieve economic optimality.
  • Simulations on the IEEE 37 bus distribution network demonstrate that DAPI control achieves proportional power sharing and economic optimality, outperforming decentralized integral control which leads to suboptimal power injections.
  • The proposed controllers are shown to be robust to variations in time constants and communication topology.

Limitations and Future Work

  • Local Analysis: The analysis is primarily local, meaning stability guarantees are local. The authors acknowledge that future work needs to consider global stability and regions of attraction.
  • Acyclic Networks: The theoretical results are largely derived for acyclic (radial) networks. Extending these to networks with loops (cyclic topologies) is a key area for future research.
  • Simplified Models: The paper uses simplified models for components like inverters (ignoring some dynamics) and assumes constant resistance-to-reactance ratios for lines. Future work should incorporate more detailed models, including reactive power flows, voltage dynamics, and ramping constraints on inverter injections.
  • Communication Assumptions: While decentralized, the controllers still rely on communication. Future work should explore event-triggered or sampled-data schemes to relax communication assumptions and handle delays.
  • General Objective Functions: The current work focuses on quadratic economic dispatch objectives. Extending the framework to more general objective functions and non-linear droop slopes is also suggested.

Summary updated

BibTeX
@article{dorfler2016breaking,
  author = {Florian Dörfler and John W. Simpson-Porco and Francesco Bullo},
  title = {{Breaking the Hierarchy: Distributed Control and Economic Optimality in Microgrids}},
  journal = {IEEE Trans. Control. Netw. Syst},
  year = {2016},
  doi = {10.1109/tcns.2015.2459391},
}
Figure