Session 7: Advances in AI, Graph Technologies, GNNs, and LLMs for New-Type Power System Applications

Advances in AI, Graph Technologies, GNNs, and LLMs for New-Type Power System Applications
“面向新型电力系统应用的人工智能、图技术、图神经网络与大语言模型前沿进展”

Driven by carbon peaking and carbon neutrality goals, new-type power systems are facing growing operational complexity and high renewable penetration, which exposes the limitations of traditional methods for planning, operation, control, and asset management. Artificial Intelligence (AI) excels at adaptive learning and fast decision-making, yet it struggles to capture the structural dependencies in power system data. Graph technologies, including Graph Computing, Graph Neural Networks (GNNs), and Knowledge Graphs (KGs), naturally represent grid topology and constraints, and can model carbon emission transfers through power networks and supply chains, thereby supporting granular carbon accounting and trustworthy traceability. Combined with AI, they enhance situational awareness, prediction, and optimization, enabling decisions that balance security, economy, and carbon emissions while promoting renewable accommodation, energy conservation, and loss reduction. Large Language Models (LLMs) further improve interpretability and natural language interaction.
This special session focuses on the research and practice of these technologies in new-type power system analysis and applications, as well as in computing–power coordination, electricity–carbon coordination, the low-carbon transition, and energy conservation and environmental protection.

“双碳”目标下，新型电力系统运行日趋复杂、可再生能源高比例接入，传统规划、运行、控制与资产管理方法面临局限。人工智能（AI）擅长自适应学习与快速决策，但难以揭示电力数据中的结构依赖关系。图计算、图神经网络（GNNs）、知识图谱（KGs）等图技术可自然表达电网拓扑与约束，并刻画碳排放在电网及产业链中的传递，支撑精细化碳核算与可信溯源。二者融合可提升态势感知、预测与优化能力，支持兼顾安全、经济与低碳的规划运行决策，促进新能源消纳与节能降损。大语言模型（LLMs）进一步增强可解释性与自然语言交互。
本专题聚焦上述技术在新型电力系统分析应用及算电协同、电碳协同、低碳转型与节能环保中的研究与实践。

Topics (Including but not limited to)

• 1. Development of AI, graph computing, GNN, and LLM methodologies and tools for improving performance, scalability, usability, and confidence in power system applications, including energy-efficient and low-carbon AI computing;
  面向电力系统性能、可扩展性、易用性、可信度等关键问题的AI、图计算、GNNs、LLMs方法与工具研究，包括高能效、低碳的AI计算方法
• 2. Applications in power system planning, including spatiotemporal load forecasting, capacity expansion, site selection, simulation enhancement, and low-carbon planning under carbon constraints and renewable accommodation targets;
  支撑电力系统规划的应用，包括时空负荷预测、容量优化、站址选择、仿真增强，以及考虑碳约束与新能源消纳的低碳规划
• 3. Applications in system operation, such as real-time dynamic security assessment, OPF/SCUC/SCED approximation, accelerated optimization, and carbon-aware dispatch;
  支撑电力系统运行的应用，包括实时动态安全稳定评估、OPF/SCUC/SCED近似与求解加速，以及计及碳排放的低碳调度
• 4. Applications in control and protection involving preventive/corrective control, special protection schemes, fault detection, and alarm processing;
  面向控制与保护的应用，包括预防/校正控制、特防系统、故障检测、告警处理
• 5. Applications in asset management, including outage management, equipment health assessment, cybersecurity, and life-cycle energy efficiency and carbon footprint management;
  面向资产管理的应用，包括停电管理、设备健康评估、网络安全，以及设备全生命周期能效与碳足迹管理
• 6. Applications in electricity–carbon coordination, including electricity emission factor applications, carbon emissions traceability, certification methodologies, integrated engineering implementations, and electricity–carbon market coordination;
  面向电碳协同的应用，包括电碳因子应用、碳排放溯源技术应用、认证方法学与体系化工程应用，以及电—碳市场协同
• 7. Applications in energy conservation and environmental protection, including loss analysis and reduction, energy efficiency assessment and improvement, demand response and energy saving, and coordinated monitoring and reduction of pollutant and carbon emissions;
  面向节能环保的应用，包括网损分析与降损、能效评估与提升、需求侧响应与节能、污染物与碳排放协同监测及减排
• 8. Demonstration projects and practical experience involving graph databases, knowledge graphs, and integrated graph-AI-GNN-LLM solutions, including low-carbon parks, zero-carbon plants, and other low-carbon transition demonstrations.
  图数据库、知识图谱、AI/GNN/LLM一体化解决方案的工程示范与实践经验，包括低碳园区、零碳电厂等低碳转型示范