摘要
针对超高压变电站PMS工作票数据异构、人工处理效率低、锁控任务生成错误率高的问题,提出基于张量分解和时空依赖图的锁控系统定向开锁方法。首先,构建三阶张量统一表征工作票文本、设备拓扑和操作语义三维特征,采用Tucker分解技术提取核心张量和因子矩阵,获得跨模态低维紧凑表示。其次,构建时空依赖邻接矩阵,将操作时序依赖、设备空间邻接和操作设备关联统一建模为分块矩阵形式,定义时空依赖强度矩阵量化不同依赖关系的权重,应用归一化割算法实现多约束条件下的任务分解优化。在业务逻辑层应用张量分解模型结合注意力机制进行工作票信息抽取,最后,在执行控制层应用时空依赖图模型进行动态任务调度,形成递进式智能化流程。实验表明,张量分解模型实现了多维特征联合学习,时空依赖图模型实现了多约束统一建模,消除了人工转录错误的安全隐患,提升了变电站调试运维的安全性和可靠性。
To address the issues of heterogeneous PMS work order data,low efficiency in manual processing,and high error rates in lock control task generation in ultra-high voltage substations,a directional unlocking method for the lock control system based on tensor decomposition and spatiotemporal dependency graphs is proposed.A third-order tensor is constructed to uniformly represent the three-dimensional features of work order text,equipment topology,and operational semantics.The Tucker decomposition technique is employed to extract the core tensor and factor matrices,achieving cross-modal low-dimensional compact representations.A spatiotemporal dependency adjacency matrix is built to unify the modeling of operational temporal dependencies,equipment spatial adjacency,and operational equipment associations in block matrix form.A spatiotemporal dependency intensity matrix is defined to quantify the weights of different dependencies,and the normalized cut algorithm is applied to optimize task decomposition under multiple constraints.In the business logic layer,a tensor decomposition model combined with an attention mechanism is used for work order information extraction,while in the execution control layer,a spatiotemporal dependency graph model is applied for dynamic task scheduling,forming an incremental intelligent workflow.Experiments demonstrate that the tensor decomposition model enables joint learning of multidimensional features,the spatiotemporal dependency graph model achieves unified modeling of multiple constraints,eliminates the safety hazards of manual transcription errors,and enhances the safety and reliability of substation commissioning and operation.
作者
赖崇杰
王睿
彭淑明
LAI Chongjie;WANG Rui;PENG Shuming(Ultra-High Voltage Branch of State Grid Jiangxi Electric Power Co.,Ltd.,Nanchang 330000,China)
出处
《电气应用》
2026年第3期117-124,共8页
Electrotechnical Application
关键词
超高压变电站
锁控系统
张量分解
时空依赖图
工作票解析
ultra-high voltage substation
lockout control system
tensor decomposition
spatiotemporal dependency graph
work ticket parsing
