低秩矩阵分解在母线坏数据辨识与修复中的应用

Abstract

母线负荷分析与预测对电力系统的安全稳定具有重要意义。目前我国采集到的母线负荷数据中含有较多且类型不同的坏数据,给母线负荷的分析的准确性与预测的精确性带来较大影响。文中提出了一种基于低秩矩阵分解的母线坏数据辨识与修复方法。从母线数据本身出发,首先分析了母线数据的低秩特性,研究不同类型坏数据产生的原因;然后建立了一种基于低秩矩阵分解的母线坏数据辨识与修复的模型,并给出了基于阈值迭代法（iterative thresholding,IT）的模型求解方法;最后,利用广东省母线负荷实际算例进行了分析,并利用修复前后的母线数据进行虚拟预测对比,结果实现了坏数据的有效恢复和预测精度的提高。

Bus load analysis and forecast is of great significance for power system security and stability. At present, there still exist many different types of bad data in bus load, causing great impacts on accuracy of analysis and forecast of bus load. In this paper, a novel method for identification and recovery of bad bus load data based on low-rank matrix theory is proposed for data-drivenperspective. This paper firstly analyzes low-rank characteristics of bus load and studies the causes of different types of bad data. Then a low-rank matrix based model is set up to identify and recover bad data. A method based on iterative thresholding (IT) to solve the model is proposed. Finally, a case study on bus load of Guangdong province is conducted. Impact on forecast with data before and after recovering is also compared. Result shows that bad data is effectively recovered and forecasting accuracy is improved.