电力系统自动化2026，Vol.50Issue(8)：196-205,10.DOI:10.7500/AEPS20250624013
面向省内电网企业代理购电的液体神经网络推理模型
Liquid Neural Network Based Inference Model for Agency Power Procurement of Intra-provincial Power Grid Enterprises
摘要
Abstract
Intra-provincial power grid enterprises face dual uncertainties in agency power procurement due to fluctuations of renewable energy output and electricity spot market prices.Traditional price-volume prediction models are poorly suited for the highly volatile scenarios of agency power procurement.To address this issue,an inference model for trading trends based on a liquid neural network(LNN)is developed.This model considers the multi-factor coupling characteristics of agency power procurement and quantifies the spot trading trends using a vertical-horizontal filter indicator from financial domains.By incorporating feature data such as day-ahead settlement power volume,settlement power price,cumulative temperature,and supply-demand ratio,the LNN is adopted to dynamically adjust neuron states and liquid time constants,achieving the adaptive learning of complex temporal patterns in the electricity market.Finally,the verification is conducted by using the price-volume data from the electricity spot market in a region of Hebei Province,China.The results demonstrate that the proposed model can effectively identify both the trend and oscillation intervals in agency power procurement data,outperforming conventional price-volume models in terms of trend segmentation and inflection point prediction.关键词
量价预测/十字过滤指标/代理购电/电力现货市场/液体神经网络/不确定性/趋势划分/拐点预测Key words
price-volume prediction/vertical-horizontal filter indicator/agency power procurement/electricity spot market/liquid neural network(LNN)/uncertainty/trend segmentation/inflection point prediction引用本文复制引用
李彬,李若松,孟子轩,张雨蒙,陈宋宋,周颖..面向省内电网企业代理购电的液体神经网络推理模型[J].电力系统自动化,2026,50(8):196-205,10.基金项目
国家电网公司科技项目:"省级电网企业代理购电交易策略与风险防控技术研究及应用"(5400-202313231A-1-1-ZN). This work is supported by State Grid Corporation of China(No.5400-202313231A-1-1-ZN). （5400-202313231A-1-1-ZN）
