摘要
人工智能和机器学习的发展为入侵电网数据采集与监视控制(supervisory control and data acquisition,SCADA)系统的虚假数据检测,提供了新的高效解决方案.目前,针对运用机器学习中的单分类器对电网中虚假数据的检测,出现的准确率低、误检率高、模型区分能力差等问题,提出了一种基于集成学习的检测方法对电网数据进行二分类,以GBDT(gradient boosting decision tree)、XGBoost、LightGBM、RF-LightGBM和Bagging分类器为基分类器,经过贝叶斯调参后,最后通过投票策略进行集成.集成学习在融合各分类器优点的同时,不仅降低了检测的误检率还提高了检测准确率及模型区分能力的稳定性.经实验对比分析,该算法在数据检测领域具有一定的应用和借鉴价值.
The development of artificial intelligence and machine learning provides a new and efficient solution for the false data detection of supervisory control and data acquisition(SCADA)system.At present,using single classifier in machine learning to detect the false data in power grid has some problems,such as low accuracy,high false detection rate,poor model differentiation ability and so on.This paper proposes a detection method based on ensemble learning to binary classify the power grid data,such as gradient boosting decision tree,XGBoost,LightGBM,RF-LightGBM and so on.Bagging classifier is used as the base classifier.After Bayesian parameter adjustment,the voting strategy is used to integrate.Ensemble learning not only integrates the advantages of each classifier,but also reduces the false detection rate,and improves the detection accuracy and the stability of model distinguishing ability.The experimental results show that the algorithm has certain application and reference value in the field of data detection.
作者
戚元星
崔双喜
QI Yuanxing;CUI Shuangxi(School of Electrical Engineering, Xinjiang University, Urumqi 830047, China)
出处
《河北大学学报（自然科学版）》
CAS
北大核心
2022年第1期105-112,共8页
Journal of Hebei University(Natural Science Edition)
基金
国家自然科学基金资助项目(51667020)
新疆大学自然科学基金资助项目(BS160246)。
关键词
SCADA系统
集成学习
贝叶斯调参
入侵检测
SCADA system
integrated learning
Bayesian parameter adjustment
intrusion detection
