Cooperative day-ahead bidding strategies for demand-side expected cost minimization

Italo Atzeni, Luis G. Ordóñez, Gesualdo Scutari, Daniel P. Palomar, J.r. Fonollosa

Abstract

The transition to a smart grid requires improved interaction between the supply and demand sides of the electricity network. This paper proposes a holistic, distributed day-ahead demand-side management (DSM) method designed for energy markets under external regulation. Active subscribers solve a nonconvex optimization problem to derive bidding strategies that minimize their expected monetary expenses while optimizing energy generation and storage. The study demonstrates that cooperative user strategies yield significantly greater savings compared to selfish, user-oriented optimization. A cooperative, distributed, and iterative algorithm is presented, along with proofs of its convergence properties.

Introduction: Bridging the Gap Between Planning and Uncertainty

In traditional power grids, matching supply with fluctuating demand is a persistent challenge. Demand-Side Management (DSM) aims to mitigate this by encouraging users to schedule consumption in advance. However, existing day-ahead models often assume deterministic behavior, failing to account for the inherent randomness of real-world energy needs—such as varying weather affecting solar output or unpredictable household activities. Deviations from pre-set bids incur real-time penalties for utility providers, which are passed on to consumers. This research bridges the gap between static planning and real-time uncertainty by introducing a pricing model that penalizes deviations and provides optimal bidding strategies under stochastic conditions.

Methodology: Distributed Algorithms and Stochastic Modeling

The authors employ a game-theoretic framework to model interactions between "active" users equipped with smart meters. The approach utilizes a two-stage pricing mechanism: a day-ahead price based on aggregate bids and a real-time penalty for deviations. Instead of bidding fixed values, users utilize the probability distribution function (PDF) of their expected load to inform their bids.

Two primary distributed algorithms are evaluated:

  1. Proximal Decomposition Algorithm (PDA): A robust method for non-cooperative scenarios, particularly when users integrate local generation (e.g., solar) and storage (batteries).
  2. Distributed Dual Proximal Algorithm (DDPA): A cooperative approach where users collaborate to minimize collective expenses.

The system was tested using a simulated environment consisting of 100 active and 900 passive users over a 24-hour period, with consumption profiles modeled on typical residential patterns.

Comparison of expected aggregate load profiles across different management strategies.

Key Insights: Strategic Displacement and Stability

A significant finding is the concept of "strategic displacement": optimal bids rarely equal the average expected consumption. If penalties for over-consumption exceed those for under-consumption, rational users "over-bid" to create a safety buffer. Furthermore, the study identifies a "flatness" requirement for the probability distribution of energy needs; excessive unpredictability can destabilize the Nash Equilibrium. Crucially, the distributed nature of these algorithms preserves user privacy, as participants only share aggregate bids rather than detailed consumption patterns.

Results: The Economic Advantage of Cooperation

The simulations confirm that strategic bidding significantly reduces costs. Standard active users saw expense reductions of 1.9% to 5%. However, the most substantial gains occurred through cooperation and the integration of hardware. When users collaborated using the DDPA and utilized Distributed Generation (DG) and Storage (DS), cumulative expenses dropped by 25.3% compared to baseline average-load bidding. As shown in the expense analysis, cooperative strategies effectively manage costs even during peak hours.

Average expected expense per time-slot, highlighting the cost-saving benefits of cooperative strategies.

Conclusion

The proposed distributed algorithms successfully converge to stable states in realistic scenarios, proving their feasibility for real-world smart grid implementation. While challenges remain—such as the need for households to accurately characterize their consumption statistics and the potential impact of unpredictable passive users—this framework provides a clear path toward more efficient, cost-effective, and cooperative energy markets.

Summary updated

Abstract

The envisioned smart grid aims to improve the interaction between the supply- and the demand-side of the electricity network, resulting in a great optimization potential. In this paper, we propose a holistic-based, distributed day-ahead demand-side management method that is suitable for energy markets subject to an external regulation. Here, active subscribers solve the nonconvex problem of deriving the bidding strategies that minimize their overall expected monetary expense and simultaneously optimize eventual dispatchable energy generation and storage strategies. We show that, when such users collaborate, they achieve greater saving with respect to the corresponding user-oriented, selfish optimization. In this setting, we propose a cooperative, distributed, and iterative algorithm providing the optimal bidding, production, and storage strategies of the users, along with its convergence properties.
BibTeX
@inproceedings{atzeni2013cooperative,
  author = {Italo Atzeni and Luis G. Ordóñez and Gesualdo Scutari and Daniel P. Palomar and J.R. Fonollosa},
  title = {{Cooperative day-ahead bidding strategies for demand-side expected cost minimization}},
  booktitle = {IEEE International Conference on Acoustics, Speech and Signal Processing},
  year = {2013},
  url = {https://openreview.net/forum?id=0J7FaJTxd_E},
}
Figure