Collaborative Optimization Method of County-Level Distributed Energy and People-Enriching Industrial Chain Based on Graph Neural Network

Abstract

This article proposes a collaborative optimization framework based on graph neural networks to address the problems of multi-source heterogeneity, high-frequency fluctuations, low energy efficiency, and high emissions in county-level distributed energy systems. By constructing heterogeneous graphs including photovoltaics, energy storage, loads, and industrial terminals, the dynamic coupling relationship of nodes is modeled using graph attention networks (GAT), and green potential node recognition is achieved by combining GraphSAGE. A cascaded optimizer is designed to integrate graph convolution and reinforcement learning to improve scheduling response capability and structural coupling level. Introduce multi-scale graph partitioning and boundary node synchronization mechanism to achieve regional autonomy and global coordination. The experimental results show that this method outperforms the baseline strategy in terms of carbon efficiency, response delay, and comprehensive benefits, effectively supporting the green and intelligent transformation of county-level energy systems.