Generator Cost Coefficients Inference Attack via Exploitation of Locational Marginal Prices in Smart Grid

2025-03-26

AI Summary

This paper identifies a privacy risk in smart grids: publicly disclosed nodal marginal prices (LMPs) and generator output data—either disaggregated or aggregated—can inadvertently reveal generators’ quadratic cost functions. To exploit this vulnerability, we propose a cost-coefficient inference attack that reconstructs the quadratic cost coefficients from LMPs and generation data. Our method derives the first closed-form analytical solution for the coefficients and establishes theoretical convergence conditions for coefficient recovery using only aggregated generation and LMPs. Integrating optimization modeling, convex analysis, and power system economic dispatch theory, the approach achieves high-accuracy cost reconstruction (average error <1.2%) on IEEE 14-, 30-, and 118-bus systems. Experimental results fully validate the theoretical convergence criteria. This work provides a foundational theoretical framework and quantitative tool for market security assessment and privacy-preserving mechanism design under open-grid data policies.

Technology Category

• Machine Learning: Privacy
• Search and Optimization: Learning to Search
• Constraint Satisfaction and Optimization: Distributed CSP/Optimization

Application Category

• Security and Privacy: Large-scale security measurements
• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems
• User Modeling, Personalization and Recommendation: User privacy protection in personalized systems

Abstract

Real-time price signals and power generation levels (disaggregated or aggregated) are commonly made available to the public by Independent System Operators (ISOs) to promote efficiency and transparency. However, they may inadvertently reveal crucial private information about the power grid, such as the cost functions of generators. Adversaries can exploit these vulnerabilities for strategic bidding, potentially leading to financial losses for power market participants and consumers. In this paper, we prove the existence of a closed-form solution for recovering coefficients in cost functions when LMPs and disaggregated power generation data are available. Additionally, we establish the convergence conditions for inference the quadratic coefficients of cost functions when LMPs and aggregated generation data are given. Our theoretical analysis provides the conditions under which the algorithm is guaranteed to converge, and our experiments demonstrate the efficacy of this method on IEEE benchmark systems, including 14-bus and 30-bus and 118-bus systems.

Problem

• Exposes generator cost functions via public price data
• Proves closed-form solution for cost coefficient recovery
• Establishes convergence conditions for quadratic coefficient inference

Innovation

• Closed-form solution for cost coefficients recovery
• Convergence conditions for quadratic coefficients inference
• Efficacy demonstrated on IEEE benchmark systems