基于灵敏度分析的数据最优筛选与不良数据辨识

Optimal Data Screening and Bad Data Identification Based on Sensitive Analysis

以多时间断面的量测信息为基础,将图论和蚁群优化算法(ant colony optimization,ACO)结合起来,对数据进行筛选,对不良数据进行检测与辨识。首先将量测支路化,把全网数据等效为一张无向图,再利用ACO结合图论中最小支撑树的搜索方法对全部数据进行筛选,并在迭代过程中使用灵敏度分析法得到电网状态和量测的标准化残差,在迭代中实现数据标准化残差和系统状态的对比,以寻优形式找到系统最优量测组合及计算出当时系统状态,最后依据灵敏度分析法进行不良数据检测与辨识。整体算法在保持计算精度的同时,还避免了辨识中重复估计带来的时间损耗,提高了计算速度。最后,以IEEE14节点系统对所提算法进行了仿真验证。

To detect and identify bad data, the measured data of multi-time scale is screened by integrating graph theory with ant colony optimization （ACO）. Firstly, the measurement is changed into paths and the data of the whole network is equivalent to a undirected graph; then combining ACO with the search method of minimum spanning tree （MST） in graph theory, all measured data is screened and during the iteration the power network status and the standardized residuals of measurement are achieved by sensitivity analysis, and the comparison of standardized residuals of the data and that of system status are implemented during the iteration system status; in the form of searching, the optimal measuring combination can be found and the system status at that moment caqn be calculated; finally, utilizing sensitivity analysis the bad data is detected and identified. The proposed algorithm can avoid the time loss brought by repetitive state estimation during the identification while the computational accuracy is kept. The proposed algorithm is verified by the simulation of IEEE/4-bus system.