Accelerated decentralized dispatch of energy storage via constrained gradient tracking

Abstract

Distributed renewable generation introduces spatially dispersed power fluctuations that must be balanced under local operating and communication constraints. Carbon dioxide energy storage systems (CO 2 -ESS) offer a viable option for distributed storage applications, yet their economic dispatch is complicated by thermo-mechanical operating constraints, heterogeneous unit characteristics, and the need for global power balance. This study develops a decentralized energy management framework for distributed CO 2 -ESS based on a constrained distributed Nesterov-accelerated primal-dual gradient tracking (C-DNGT) algorithm. The method embeds the power-balance constraint into the distributed optimization process and employs a dynamic gradient tracker to exchange local information and compensate for power mismatch without a centralized coordinator. Nesterov acceleration is introduced to improve convergence under heterogeneous storage parameters and operating limits. Theoretical analysis shows that, under standard convexity and connectivity assumptions, the CDNGT iterates converge linearly to the Karush-Kuhn-Tucker saddle point of the economic dispatch problem. Numerical studies on PJM Interconnection load data under heterogeneous CO 2 -ESS configurations indicate that C-DNGT reduces the number of communication iterations required for convergence compared with conventional distributed primal-dual and gradient-tracking methods, while preserving power-balance feasibility. The proposed C-DNGT framework enables fully decentralized, coordinator-free economic dispatch of distributed CO 2 -ESS with accelerated convergence and guaranteed power-balance feasibility, demonstrating its potential for heterogeneous distributed storage networks.