A Carbon-Tracking Resource Trading Mechanism Based on Hierarchical Game in Green Computing Power Networks

Abstract

Recently, computing Power Networks (CPNs) have emerged as a critical infrastructure for supporting intelligent services, particularly Large Language Models (LLMs). While integrating renewable energy is imperative for carbon neutrality, it faces significant challenges due to energy intermittency and the rigid reliability requirements of computing resources. The resulting spatiotemporal dynamics create a complex trilateral interdependence among computing, energy, and carbon, complicating precise emission assignment. In this paper, we propose a carbon-tracking hierarchical trading mechanism to address these challenges in green CPNs. Specifically, we formulate a three-stage Stackelberg game to decouple dynamic interactions among consumers, resource providers, and Energy Supply Points (ESPs), facilitating flexible benefit coordination. Furthermore, we simplify the game into convex optimization problems and theoretically prove the existence and uniqueness of the Nash Equilibrium (NE). Simulation results validate the stability of the proposed scheme and demonstrate its superior efficacy in balancing multi-party benefits and reducing carbon emissions compared to existing baselines.