自动化学报2025，Vol.51Issue(12)：2648-2663,16.DOI:10.16383/j.aas.c250395
基于大模型量化突发事件影响的工业能源价格预测研究
Research on Quantifying Unexpected Event Impacts in Industrial Energy Price Forecasting Based on Large Language Model
摘要
Abstract
Energy is the core element of modern industrial operations,with its price volatility significantly impact-ing the cost structure and competitiveness of industrial enterprises.However,industrial energy prices exhibit high-frequency fluctuation characteristics affected by unexpected events such as policy changes and geopolitical conflicts,making accurate forecasting challenging.Furthermore,existing deep learning approaches exhibit limitations in event quantification,dynamic responsiveness,and interpretability,leading to a significant increase in the uncertainty of prediction results.To address these issues,this paper proposes an industrial energy price forecasting method based on large language model for quantifying unexpected event impacts.First,we establish a two-stage collaborative mechanism integrating domain knowledge fine-tuning and structured data fine-tuning to achieve quantitative char-acterization of unexpected events and extract explicit reasoning paths.An event shock function-driven multi-scale event parser is designed,combined with an attention mechanism to adaptively adjust attenuation rates,decompos-ing event impacts into short-term shocks,intermediate transmission,and long-term trends.Second,we unify event quantification and price forecasting process by incorporating quantification results as external inputs to enhance model responsiveness,subsequently feeding them into a gated recurrent neural network for price prediction.Finally,systematic experiments were conducted on two real-world industrial datasets:The monthly coal prices at Qinhuang-dao and the monthly Brent crude oil prices.These experiments validated the effectiveness of the proposed method and further analyzed the contribution mechanism of large language model's reasoning text to prediction performance.关键词
事件驱动/工业能源/价格预测/大模型/量化分析框架Key words
Event-driven/industrial energy/price forecasting/large language model/quantitative analysis frame-work引用本文复制引用
LUO Jiang,WANG Ya-Lin,LIU Chen-Liang,YUAN Xiao-Feng..基于大模型量化突发事件影响的工业能源价格预测研究[J].自动化学报,2025,51(12):2648-2663,16.基金项目
国家自然科学基金(U25A20466,92267205,62503507),湖南省自然科学基金(2025JJ60423,2025JJ10007),中南大学中央高校基本科研业务费(2023ZZTS0163)资助 Supported by National Natural Science Foundation of China(U25A20466,92267205,62503507),Natural Science Foundation of Hunan Province(2025JJ60423,2025JJ10007),and Fundament-al Research Funds for the Central Universities of Central South University(2023ZZTS0163) （U25A20466,92267205,62503507）
