基于智能报警和故障定位的电网智能化调度研究

Research on intelligent dispatching of power grid based on intelligent alarm and fault location

  • 摘要: 电网规模与网架结构日趋复杂,给调度系统故障诊断带来挑战,为此构建了一套贯穿“一次设备故障-保护动作- 开关跳闸”全链路的时序约束网络诊断模型。在时标处理环节,将多源数据的时标不一致性转化为包含误差变量的 线性不等式组,借助消去法判定不等式组的相容性,剔除存在逻辑冲突的故障假设,保留与实际告警序列吻合的诊 断结论。此外,将故障录波器捕获的暂态波形特征与备自投装置的动作简报纳入诊断流程,前者用于锁定故障相别 与测距信息,后者用于追踪负荷转供路径并识别转供后的设备过载风险,二者共同支撑故障的精细化研判。在此基 础上,依托时序约束网络的子图结构,将分散的告警、录波及简报信息重组为可追溯的事件链。经由实际工程场景 的反复验证,该方法在区段级故障定位的正确率与完备性上均达到了实际调度运行的要求,且所形成的结构化事件 序列也为未来深度融合数据挖掘与智能推理技术奠定了实践基础。

     

    Abstract: The expanding scale and increasingly complex topology of power grids bring severe challenges to fault diagnosis of dispatching systems. To address these bottlenecks, this paper develops a diagnostic model based on a time-series constraint network that spans the entire chain from primary equipment fault through protection response to breaker execution. For timestamp processing, the inconsistency of multi-source timestamps is formulated as a set of linear inequalities containing error variables, and an elimination method is applied to determine the compatibility of the inequality set, thereby filtering out logically contradictory fault hypotheses and retaining diagnostic conclusions consistent with the observed alarm sequences. Additionally, transient waveform features captured by fault recorders and action briefs from automatic bus transfer devices are incorporated into the diagnostic workflow—the former to identify the faulted phase and estimate fault distance, and the latter to trace load transfer paths and detect equipment overload risks after transfer, jointly supporting refined fault analysis. On this basis, the subgraph structure of the time-series constraint network is utilized to reorganize scattered alarms, fault recordings, and action briefs into a traceable event chain. Repeated validation in practical engineering scenarios confirms that the proposed method meets operational dispatching requirements in both accuracy and completeness of section-level fault location, and the resulting structured event sequences also lay a practical foundation for future deep integration with data mining and intelligent reasoning technologies.

     

/

返回文章
返回