Efficient Decentralized Economic Dispatch for Microgrids with Wind Power Integration

Yu Zhang, Georgios B. Giannakis

Abstract: Decentralized energy management is of paramount importance in smart microgrids with renewables for various reasons including environmental friendliness, reduced communication overhead, and resilience to failures. In this context, the present work deals with distributed economic dispatch and demand response initiatives for grid-connected microgrids with high-penetration of wind power. To cope with the challenge of the wind's intrinsically stochastic availability, a novel energy planning approach involving the actual wind energy as well as the energy traded with the main grid, is introduced. A stochastic optimization problem is formulated to minimize the microgrid net cost, which includes conventional generation cost as well as the expected transaction cost incurred by wind uncertainty. To bypass the prohibitively high-dimensional integration involved, an efficient sample average approximation method is utilized to obtain a solver with guaranteed convergence. Leveraging the special infrastructure of the microgrid, a decentralized algorithm is further developed via the alternating direction method of multipliers. Case studies are tested to corroborate the merits of the novel approaches.

Introduction: Managing Stochastic Wind Power

Microgrids are localized energy systems capable of operating autonomously or in conjunction with the main power grid. They are pivotal for integrating renewable energy sources like wind power. However, wind energy is inherently stochastic and unpredictable. Scheduling energy in such environments presents a significant challenge: overestimating wind availability leads to shortages, while underestimating it results in wasted clean energy.

Traditional Economic Dispatch (ED) aims to minimize costs while meeting demand, but high wind penetration typically necessitates a centralized controller. Centralized architectures are prone to high communication overhead and represent a single point of failure. This research addresses these vulnerabilities by proposing a decentralized framework for economic dispatch and demand response in grid-connected microgrids.

Typical microgrid architecture featuring diverse generation sources, energy storage, and dispatchable loads connected to the main grid.

Methodology: Stochastic Optimization and Decentralization

The proposed approach integrates three technical innovations to handle uncertainty and eliminate the need for central control.

1. Transaction Mechanism for Wind Uncertainty

Instead of reacting purely to real-time wind fluctuations, the microgrid commits to a scheduled wind power amount ($P_R^t$). Any deviations—surpluses or shortages—are managed through a transaction mechanism with the main grid. These trades occur at predefined purchase ($\alpha^t$) and selling ($\beta^t$) prices, ensuring the microgrid can balance its energy budget.

2. Sample Average Approximation (SAA)

Calculating the expected cost of wind uncertainty involves complex, high-dimensional integrals. To bypass this, the authors employ Sample Average Approximation (SAA). By averaging a large set of Monte Carlo wind scenarios, the stochastic problem is transformed into a solvable convex optimization problem. The authors proved that as long as the selling price back to the grid does not exceed the purchase price ($\beta^t \leq \alpha^t$), the cost function remains convex, guaranteeing convergence to a global optimum.

3. Decentralized ADMM Solver

To achieve decentralization, the Alternating Direction Method of Multipliers (ADMM) is utilized. This algorithm decomposes the global optimization problem into smaller sub-problems handled locally by individual controllers (e.g., for generators, loads, and wind farms). These local controllers iteratively exchange a single vector of "price-like" information—Lagrange multipliers ($\lambda$)—to reach a consensus on the optimal schedule without sharing raw local data.

Key Results and Performance

The framework was validated using a simulated microgrid with three conventional generators, three dispatchable loads, and four wind farms. The decentralized algorithm demonstrated robust performance, converging to the optimal solution in fewer than 50 iterations.

Convergence of the decentralized ADMM algorithm, showing the evolution of Lagrange multipliers ($\lambda$) and scheduled power for generation ($P_G^S$), demand ($P_D^S$), and renewables ($P_R^S$).

As shown in the convergence plots, the scheduled wind power ($P_R^S$) reached its upper limit of 60 kWh when purchase prices from the main grid were high relative to local costs. The system also effectively managed "demand response." During periods of high fixed demand, elastic loads were automatically scaled back to maintain balance. Conversely, when demand was low, elastic consumption increased to maximize consumer utility. Economic analysis revealed that the microgrid's net cost is highly sensitive to the grid's price ratio; as selling prices approach purchase prices, the microgrid significantly reduces costs by exporting surplus renewable energy.

Conclusion and Future Directions

This work provides a resilient, decentralized solution for microgrid energy management under wind uncertainty. By combining SAA with ADMM, the authors achieve a computationally efficient and communication-light dispatch mechanism. Future research will look to incorporate real-time dispatch capabilities to handle second-by-second fluctuations and extend the stochastic modeling to include consumer load uncertainty.

Summary updated

Abstract

Decentralized energy management is of paramount importance in smart microgrids with renewables for various reasons including environmental friendliness, reduced communication overhead, and resilience to failures. In this context, the present work deals with distributed economic dispatch and demand response initiatives for grid-connected microgrids with high-penetration of wind power. To cope with the challenge of the wind's intrinsically stochastic availability, a novel energy planning approach involving the actual wind energy as well as the energy traded with the main grid, is introduced. A stochastic optimization problem is formulated to minimize the microgrid net cost, which includes conventional generation cost as well as the expected transaction cost incurred by wind uncertainty. To bypass the prohibitively high-dimensional integration involved, an efficient sample average approximation method is utilized to obtain a solver with guaranteed convergence. Leveraging the special infrastructure of the microgrid, a decentralized algorithm is further developed via the alternating direction method of multipliers. Case studies are tested to corroborate the merits of the novel approaches.
BibTeX
@misc{zhang2014efficient,
  author = {Yu Zhang and Georgios B. Giannakis},
  title = {{Efficient Decentralized Economic Dispatch for Microgrids with Wind Power Integration}},
  year = {2014},
  eprint = {1401.5956},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/1401.5956},
}
Figure