Graph computing technology for ultra-large-scale discrete optimization: a case study on security constrained unit commitment in energy system

This paper proposes a graph computing-based method to address the ultra-large-scale discrete optimization problem of security constrained unit commitment in power systems. Feasible unit commitment states are modeled as nodes in a graph, with edge weights defined based on operational priorities and physical information. The search is guided via depth-first traversal and cumulative weight maximization, effectively reducing the number of state solutions to be computed. Meanwhile, graph parallel computing techniques are employed on the grid graph model to accelerate the solution process at each node. Case studies demonstrate that the proposed method significantly reduces computation time by up to 92% compared to conventional solvers, while ensuring solution quality, particularly in large-scale systems. This approach offers an interpretable, efficient, and scalable paradigm for addressing complex mixed integer programming problems in energy systems.