一种基于Spark模型的电力异常数据检测方法

A Method for Identifying Bad Data of Power System Based on Spark

电网的信息化与智能化程度不断的提升使得电力数据量越来越大,给数据的处理和分析带来很大的困难。在智能电网大数据应用处理的过程中,数据的实时性存储、高效处理、多源异构数据的融合以及数据的可视化方面面临着严峻的挑战,需要深入对这些方面开展研究,切实发挥大数据在保障电网安全稳定运行的作用。这些异常数据的存在对现代电力系统状态的估计结果的影响是不容忽视的。现有的电力数据异常检测方法未能充分挖掘数据特征,存在计算复杂、灵活性差、精度较低等缺点。目前已有的预测算法无法满足预测速度和精度的要求,因此基于大数据计算平台,提出一种基于Spark的改进ISODATA聚类算法对异常数据进行检测与修正。实验结果表明,该方法对异常数据的检测和修正有很好的效果,降低了检测时间,有效提高了状态估计结果的准确性。

The continuous improvement of the informatization and intellectualization of the power grid makes the power data more and more large,which brings great difficulties to the data processing and analysis.In the application and processing of smart grid big data,there exists severe challenges in real-time data storage,efficient processing,multi-source heterogeneous data integration and data visualization.It is necessary to conduct in-depth research on these aspects and give full play to the role of big data in ensuring the safe and stable operation of the power grid.The influence of these abnormal data on the state estimation of modern power system cannot be ignored.The existing abnormal detection methods of power data fail to fully exploit the data features,and have the disadvantages of complex computation,poor flexibility and low accuracy.At present,the existing prediction algorithms cannot meet the requirements of prediction speed and accuracy.Therefore,based on big data computing platform,we propose an improved ISODATA clustering algorithm based on Spark to detect and correct abnormal data.Experiment shows that the proposed method has a better effect on the detection and correction of abnormal data,reduces the detection time and effectively improves the accuracy of state estimation.