基于降阶核范数的PMU数据快速恢复方法

Fast PMU Measurements Recovery Algorithm Based on Reduced Nuclear Norm

相量量测装置(phasor measurement unit,PMU)数据由于受通信问题、网络攻击与电磁干扰等因素影响可能会出现数据缺失、虚假数据与噪声等量测污染。现有PMU数据恢复方法无法同时满足高精度、快速与适应多场景数据恢复的要求,可能影响电力系统的安全稳定运行。提出一种基于降阶核范数的数据恢复方法,利用数据的高维低秩性在每次算法迭代中进行低阶奇异值分解,同时将数据主成分与量测污染进行特征分离,能以较高精度对多种数据污染场景与系统不同状态下的量测数据进行恢复。为进一步减少大规模电力量测数据恢复的时间消耗并提高算法的迭代效率,提出自适应惩罚因子与并行分布式交替乘子算法数据恢复框架,能有效将数据恢复时间控制在秒级,可为基于PMU数据的各电力系统应用提供有效保障。

PMUs suffer from corrupted measurements such as data missing, false data and noise due to different factors such as communication contingency, cyber-attacks and electromagnetic interference. Existing phasor measurement unit(PMU) measurement recovery algorithms cannot simultaneously meet the requirements of high precision, rapid processing and adaptability to multiple scenarios. This paper proposed a measurement recovery algorithm based on the reduced nuclear norm, which used the high-dimensional and low-rank properties of PMU measurements to perform low-order singular value decomposition in the proposed algorithm, and extracted the principal components from the corrupted measurements. It could restore the measurements in multiple modes and different states of the system with high accuracy. In order to further reduce the recovery time consumption of large-scale power system measurements and speed up the iteration efficiency, the adaptive penalty factor and the distributed alternating direction method of multipliers(ADMM),framework were proposed to effectively limit the data recovery time at the level of seconds, which rendered its applicability in practical power systems.