计及时序信息的电网故障在线诊断与连锁故障传播脆度机理

With the scale of power systems interconnection increasing, the structure and operation modes being more complex, the large blackout caused by cascading failures occurred many times, and the uncertainty of misoperation and rejection of protection and switch and information missing accelerates the process of cascading failures. It becomes an important problem to study the accurate on-line power system fault diagnosis methoed and reveal the propagation mechanism of cascading failures. The object of the project is to detect the original fault element rapidly and accurately, to on-line diagnose the power system fault and its evolution, to analyze the reason of misoperation and rejection of protection and switch, and to assess the brittleness of nodes and possible propagation paths of cascading failures, preventing the spreading of failures. Using the information of protection and switch and their time sequences and wide-area measurement, some novel models are constructed, such as the fault diagnosis model for power system based on temporal Bayesian knowledge bases, the directed network of virtual interaction among protection logical nodes based on IEC 61850, and abductive event-tree of misoperation and rejection of protection and switch. The original fault element can be detected with the fault degrees by the multi-source data. The diagnosis algorithm of temporal Bayesian knowledge bases is studied to diagnose the fault and the check correctness of protection and switch. The root reasons of misoperation and rejection of protection and switch can be analyzed. The method of assessing the brittleness of nodes during the cascading failures will be constructed. Using the weighted network, the propagation mechanism of cascading failures based on percolation theory will be studied. The strengthening effect that the misoperation and rejection of protection and switch with their time sequences in information network accelerate the cascading failures is to be revealed.

近些年电力系统互联规模不断扩大，系统结构与运行方式变得更复杂，国内外由连锁故障发展的大停电多次发生，保护与开关误动与拒动、信息缺失不确定性加速了连锁故障的进程。研究准确在线诊断电网故障方法、揭示连锁故障传播机理成为重要问题。本项目目标是快速准确检测原发故障元件，在线诊断故障及其演变，分析保护与开关误动与拒动具体原因，在线评估故障跳闸后各节点的脆度、连锁故障可能传播路径，以防止故障蔓延。充分利用故障前后保护与开关动作及其时序信息、有限广域量测量，构建新型电网故障时序贝叶斯知识库诊断模型、基于IEC61850的变电站保护逻辑节点之间虚拟交互的有向网络、保护与开关误动与拒动溯因事件树等，研究原发故障检测、电网故障时序贝叶斯知识库诊断、保护与开关误动与拒动溯因、连锁故障中各节点脆度与相变分析等方法，研究基于渗流理论的连锁故障传播机理，探索信息网络中保护与开关误动与拒动及其时序对连锁故障的助增效应。

近些年电力系统互联规模不断扩大，系统结构与运行方式变得更复杂，国内外由连锁故障发展的大停电多次发生，大量告警信号上送到调度中心，给电网故障诊断带来困难。本项目的目标是研究快速准确检测原发故障元件的方法，在线诊断故障元件，在线评估故障跳闸后各级线路的脆度，预测出连锁故障较可能的传播路径，以防止故障蔓延。本项目从电网故障事后诊断、在线原发故障检测、电网连锁故障脆弱线路识别和连锁传播路径机理三个角度开展研究。在电网故障事后诊断方面，研究了基于时序贝叶斯网络、基于可信度的变结构贝叶斯网络、基于灰色关联度、基于动态推理链、基于希尔伯特-黄变换和模糊支持向量机、基于异常因子和模糊聚类等故障诊断方法，在诊断模型中考虑时序信息及其可信度，有助于更准确诊断电网故障元件和多重故障。在原发故障在线检测方面，研究了利用保护与断路器动作开关量、有限量测量，研究了基于相对保护熵与名义过渡电阻、有限相量测量单元下最小二乘估计、基于有限PMU和虚拟节点的电网故障区域与线路在线识别、基于零序差动阻抗的输电线路保护新原理、基于多点PMU的电压推算、有限PMU下基于差动有功功率抗过渡电阻、分区加权故障匹配等多种在线故障检测方法，利用广域信息能够及时准确高容错性地发现原发故障元件，在5位保护信息出错情况下保证故障检测的100%正确率，对于6位保护出错情况，本算法仍达到了99.87%正确率，具有很高的容错性。在电网连锁故障传播分析方面，研究了基于渗流理论的脆弱线路识别方法、基于风险理论的连锁故障发展路径分析方法、考虑通信脆弱度的电网连锁故障风险评估与告警方法，能够准确识别脆弱线路，预测连锁故障路径与已有方法的预测结果相近。研究构造了电力风险、通信风险及综合风险指标，定量分析了分布式拒绝服务攻击DDoS对电力系统连锁故障发展的早期和中期的综合风险影响很大，但当攻击强度达到一定程度时，电网的综合风险趋于稳定值。