Abstract
Distributed generation resources have become significantly more prevalent in the electric power system over the past few years. This warrants a reconsideration on how the coordination of generation resources is achieved. In this paper, we particularly focus on secondary frequency control. It is assumed that the system operator first selects the resources which are expected to provide secondary frequency control by carrying out a co-optimization between energy and ancillary service markets. Two kinds of payments are taken into account for the frequency control participants: the capacity payment and the performance payment. We first restructure the participants’ cost functions to decouple these two payments. Then, we introduce a fully distributed real-time control framework based on a consensus-plus-innovations approach, providing an alternative to the centralized automatic generation control approach. A key aspect is that the objective of the real-time dispatch is to choose the most cost-effective real-time deployment of the secondary frequency resources. We prove that our distributed control approach performs c-close to the centralized optimal economic dispatch, where c is a positive constant depending only on the cost parameters and the communication topology. Simulation results confirm the fast convergence and the cost effectiveness of the proposed distributed control approach.
Introduction
The modern power grid is undergoing a paradigm shift. Traditional centralized generation is being supplemented, and in some cases replaced, by thousands of small-scale distributed energy resources (DERs) like solar arrays, wind turbines, and battery storage. This decentralization poses a significant challenge for secondary frequency control, traditionally managed by a central operator via Automatic Generation Control (AGC).
Centralized control faces two primary bottlenecks in this new landscape:
- Scalability: Managing a massive fleet of DERs from a single point creates computational strain and increases vulnerability to single-point failures.
- Economic Efficiency: Traditional AGC typically relies on static participation factors, failing to optimize resource dispatch based on real-time marginal costs.
While distributed economic dispatch has been proposed as a solution, many existing algorithms require full convergence before physical action can be taken—a delay the grid cannot afford during rapid frequency fluctuations.
The Consensus + Innovations Framework
To address these challenges, this paper proposes a distributed peer-to-peer (P2P) control law based on a "consensus-plus-innovations" approach. This hybrid mechanism allows resources to act immediately while moving toward a global optimum.
The control law operates through two simultaneous components:
- The Consensus Term (Local): Resources communicate with immediate neighbors to synchronize their marginal costs. This peer-to-peer exchange ensures the collective output is generated as cheaply as possible without needing a central coordinator.
- The Innovation Term (Global): Every resource monitors a global "innovation" signal—the frequency deviation ($\Delta f$). This ensures that while optimizing for cost, the primary objective of balancing supply and demand is never compromised.
A significant theoretical contribution of this work is proving that this distributed law is mathematically equivalent to a Proportional-Integral (PI) controller. This equivalence provides a rigorous guarantee of system stability, as PI controllers are the industry standard for robust industrial processes.
Performance and Economic Dispatch
The effectiveness of the proposed framework was validated through numerical simulations of single-area and multi-area power systems. The results demonstrate that the distributed approach significantly outperforms traditional AGC in both speed and cost-effectiveness.
In simulations with a 4-second control interval, the distributed method stabilized grid frequency in approximately 8 seconds. In contrast, traditional AGC required over 12 seconds to reach stability, even with faster update rates. Furthermore, the algorithm successfully drove the system toward the optimal "Economic Dispatch" solution, reducing the relative error from optimal allocation from 25% to just 7% over a 5-minute window.
The system also proved robust in multi-area configurations, maintaining zero net tie-line flow between regions while restoring frequency. This confirms that sparse communication—where devices only talk to a few neighbors—is sufficient to manage complex, interconnected grids.
Conclusion and Future Directions
The consensus-plus-innovations approach offers a scalable, privacy-preserving, and cost-effective alternative to centralized frequency control. By allowing resources to exchange only marginal cost data, it achieves economic optimality without exposing private cost functions to a central authority.
Future research will focus on the impact of communication latencies and packet loss in P2P networks. Additionally, the team aims to explore "fully" distributed versions that utilize local frequency measurements, further reducing reliance on global signals and enhancing the grid's resilience in the face of localized disturbances.
Summary updated