电力系统自动化2026，Vol.50Issue(18)：97-107,11.DOI:10.7500/AEPS20250710002
基于知识嵌入超图深度学习的城市快充站运行状态估计
Operation State Estimation for Urban Fast-charging Stations Based on Knowledge-embedded Hypergraph Deep Learning
摘要
Abstract
Accurate operation state estimation for urban fast-charging stations is a crucial prerequisite for the efficient utilization of electric vehicle charging networks.Existing studies lack the consideration of the actual operation mechanisms of fast-charging stations,failing to explore the spatio-temporal interaction characteristics of multi-station operation states.To address this,by transforming state estimation into a sample generation process that integrates operation knowledge and historical data of fast-charging stations,this paper proposes an operation state estimation method for urban fast-charging stations based on knowledge-embedded hypergraph deep learning architecture.First,the dynamic operation states of fast-charging stations and their multi-spatial distribution are represented as a hypergraph model,and the temporal convolution attention network and hypergraph convolution network are used to capture temporal similarity and spatial correlation in graph-structured feature information.Second,operation strategies of fast-charging stations,physical rules,and constraints are decoupled into a knowledge loss function,which is integrated into the sample learning process of generative adversarial networks to construct a knowledge-data dual-driven loss function.Through the"encoding-decoding"paradigm,the conditional probability distribution of samples is implicitly learned to estimate the operation state of individual charging piles across different fast-charging stations.Finally,using the actual road network and operating fast-charging stations in Nanjing,China,as a case study,simulation results demonstrate that the proposed method can accurately characterize the spatio-temporal distribution characteristics of operation states for fast-charging stations,providing theoretical support for subsequent electric vehicle charging scheduling.关键词
电动汽车/充电网络/快充站/状态估计/运行知识/超图深度学习/生成对抗网络Key words
electric vehicle(EV)/charging network/fast-charging station/state estimation/operation knowledge/hypergraph deep learning/generative adversarial network引用本文复制引用
杨璐彤,高辉,邢强,黄宏旭,陈良亮..基于知识嵌入超图深度学习的城市快充站运行状态估计[J].电力系统自动化,2026,50(18):97-107,11.基金项目
国家自然科学基金资助项目(52077107) （52077107）
江苏省基础研究计划自然科学基金资助项目(BK20240652). This work is supported by National Natural Science Foundation of China(No.52077107)and Jiangsu Provincial Natural Science Foundation for Basic Research of China(No.BK20240652). （BK20240652）
