为有效提升配电网韧性,提出了一种基于数据-模型混合驱动的多类型移动应急资源优化调度方法。首先,考虑到交通道路状态动态变化对移动储能车(mobile energy storage system,MESS)和应急抢修队(repair crew,RC)策略的影响,构建了以电力-...为有效提升配电网韧性,提出了一种基于数据-模型混合驱动的多类型移动应急资源优化调度方法。首先,考虑到交通道路状态动态变化对移动储能车(mobile energy storage system,MESS)和应急抢修队(repair crew,RC)策略的影响,构建了以电力-交通耦合网总损失成本最小为目标的多类型移动应急资源随机优化调度模型。然后,为了实时准确地求解MESS和RC最优路由和调度策略,提出了一种数据-模型混合驱动方法对所构建的复杂非线性随机优化模型进行求解。在数据驱动部分提出一种图注意力网络多智能体强化学习算法,以求解考虑交通网道路修复时间和移动应急资源邻接关系动态变化等不确定因素的MESS和RC最优路由策略。所提算法有效结合多种改进策略和优先经验回放策略以提高算法的采样效率和训练效果。在模型驱动部分采用二阶锥松弛和大M法将多类型移动应急资源优化调度问题构建为混合整数二阶锥规划模型以求解可再生能源出力和配电网负荷变化影响下MESS和RC最优调度策略。最后,在2个不同规模的电力-交通耦合网中验证所提方法的有效性、泛化能力和可拓展能力。展开更多
针对电网积累的海量工单数据未被深度挖掘、人工分析效率低等问题,本文提出一种基于命名实体识别模型的配网客户诉求知识图谱构建方法。该方法通过融合基于Transformer的双向编码器表征(bidirectional encoder representation from tran...针对电网积累的海量工单数据未被深度挖掘、人工分析效率低等问题,本文提出一种基于命名实体识别模型的配网客户诉求知识图谱构建方法。该方法通过融合基于Transformer的双向编码器表征(bidirectional encoder representation from transformers,BERT)预训练语言模型、双向长短期记忆(bidirectional long short-term memory,BiLSTM)语义提取层和条件随机场(conditional random field,CRF)解码层构建实体识别模型,显著提升了配网工单文本中“客户”“故障设备”“反馈类型”等关键实体的识别准确率(F1值达84.21%);利用Neo4j图数据库对抽取的实体关系进行知识融合与存储,实现了客户诉求、设备信息、处理状态等多维度数据的可视化关联分析。经某供电公司8 764份工单数据验证,该方法能有效推动配网供电服务指挥系统转型升级,提升诉求处理效率和业务管控能力,为供电服务质量优化提供数字化决策支持。展开更多
To address the challenges of ill-defined optimization objectives,difficult constraint coordination,and lack of quantitative basis for interconnection splicing and switch placement in current distribution network topol...To address the challenges of ill-defined optimization objectives,difficult constraint coordination,and lack of quantitative basis for interconnection splicing and switch placement in current distribution network topology optimization,this paper proposes a data-driven intelligent optimization method for panoramic construction of distribution network topology based on the Common Information Model(CIM).This method integrates multi-source heterogeneous data relationships-including equipment,terminals,and connection nodes-through joint analysis of multi-line CIM and hierarchical topology extraction.It automatically identifies feeder trunk paths and branch structures,incorporates inter-connection switch splicing and intelligent path optimization strategies,and performs topology opti-mization and switch placement based on the principle of minimizing outage impact.This constructs a complete,robust main-branch topology graph model.The algorithm employs depth-first search(DFS)for supply path modeling,complemented by semantic analysis of equipment attributes and hierarchical node classification to refine topology simplification.Batch testing on a dataset of 6880 medium-voltage feeders in a Central China city achieved a 98.30%successful modeling rate for complete interconnection information,with an average processing time of approximately 4.57 s per feeder.Further validation using representative overhead,cable,and hybrid lines demonstrated high consistency between the automatically generated topology and the original system diagram in node identification,path con-struction,and information annotation,confirming the algorithm's structural adaptability and engi-neering practicality.These findings provide dynamically interactive topology model support for multiple distribution network scenarios-including planning,operation,and maintenance-offering significant application and promotion value.展开更多
文摘为有效提升配电网韧性,提出了一种基于数据-模型混合驱动的多类型移动应急资源优化调度方法。首先,考虑到交通道路状态动态变化对移动储能车(mobile energy storage system,MESS)和应急抢修队(repair crew,RC)策略的影响,构建了以电力-交通耦合网总损失成本最小为目标的多类型移动应急资源随机优化调度模型。然后,为了实时准确地求解MESS和RC最优路由和调度策略,提出了一种数据-模型混合驱动方法对所构建的复杂非线性随机优化模型进行求解。在数据驱动部分提出一种图注意力网络多智能体强化学习算法,以求解考虑交通网道路修复时间和移动应急资源邻接关系动态变化等不确定因素的MESS和RC最优路由策略。所提算法有效结合多种改进策略和优先经验回放策略以提高算法的采样效率和训练效果。在模型驱动部分采用二阶锥松弛和大M法将多类型移动应急资源优化调度问题构建为混合整数二阶锥规划模型以求解可再生能源出力和配电网负荷变化影响下MESS和RC最优调度策略。最后,在2个不同规模的电力-交通耦合网中验证所提方法的有效性、泛化能力和可拓展能力。
文摘针对电网积累的海量工单数据未被深度挖掘、人工分析效率低等问题,本文提出一种基于命名实体识别模型的配网客户诉求知识图谱构建方法。该方法通过融合基于Transformer的双向编码器表征(bidirectional encoder representation from transformers,BERT)预训练语言模型、双向长短期记忆(bidirectional long short-term memory,BiLSTM)语义提取层和条件随机场(conditional random field,CRF)解码层构建实体识别模型,显著提升了配网工单文本中“客户”“故障设备”“反馈类型”等关键实体的识别准确率(F1值达84.21%);利用Neo4j图数据库对抽取的实体关系进行知识融合与存储,实现了客户诉求、设备信息、处理状态等多维度数据的可视化关联分析。经某供电公司8 764份工单数据验证,该方法能有效推动配网供电服务指挥系统转型升级,提升诉求处理效率和业务管控能力,为供电服务质量优化提供数字化决策支持。
基金supported by the State Grid Corporation of China science and technology project funding(5400-202322560A-3-2-ZN).
文摘To address the challenges of ill-defined optimization objectives,difficult constraint coordination,and lack of quantitative basis for interconnection splicing and switch placement in current distribution network topology optimization,this paper proposes a data-driven intelligent optimization method for panoramic construction of distribution network topology based on the Common Information Model(CIM).This method integrates multi-source heterogeneous data relationships-including equipment,terminals,and connection nodes-through joint analysis of multi-line CIM and hierarchical topology extraction.It automatically identifies feeder trunk paths and branch structures,incorporates inter-connection switch splicing and intelligent path optimization strategies,and performs topology opti-mization and switch placement based on the principle of minimizing outage impact.This constructs a complete,robust main-branch topology graph model.The algorithm employs depth-first search(DFS)for supply path modeling,complemented by semantic analysis of equipment attributes and hierarchical node classification to refine topology simplification.Batch testing on a dataset of 6880 medium-voltage feeders in a Central China city achieved a 98.30%successful modeling rate for complete interconnection information,with an average processing time of approximately 4.57 s per feeder.Further validation using representative overhead,cable,and hybrid lines demonstrated high consistency between the automatically generated topology and the original system diagram in node identification,path con-struction,and information annotation,confirming the algorithm's structural adaptability and engi-neering practicality.These findings provide dynamically interactive topology model support for multiple distribution network scenarios-including planning,operation,and maintenance-offering significant application and promotion value.
