基于FHO-CatBoost的分布式电源调控异常事件检测  

Anomalous Events Detection for Distributed Generations Control Based on FHO-CatBoost

作　　者：谢国强 卢志学 陈明亮 余滢婷 潘本仁 孙鹤洋 李元诚[3] XIE Guoqiang;LU Zhixue;CHEN Mingliang;YU Yingting;PAN Benren;SUN Heyang;LI Yuancheng(State Grid Jiangxi Electric Power Co.,Ltd.,Research Institute,Nanchang 330096,Jiangxi Province,China;School of Electrical Engineering,Xi’an Jiaotong University,Xi’an 710049,Shaanxi Province,China;School of Control and Computer Engineering,North China Electric Power University,Changping District,Beijing 102206,China)

机构地区：[1]国网江西省电力有限公司电力科学研究院,江西省南昌市330096 [2]西安交通大学电气工程学院,陕西省西安市710049 [3]华北电力大学控制与计算机工程学院,北京市昌平区102206

出　　处：《电网技术》2025年第4期1625-1634,共10页Power System Technology

基　　金：国家电网有限公司总部科技项目“新型电力系统下分布式电源调度控制安全防护关键技术研究与应用”(5108-202325046A-1-1-ZN)。

摘　　要：新型电力系统的全面推进仍然面临多重安全挑战,特别是分布式电源系统容易受极端天气、自然灾害和网络攻击等威胁,从而导致系统波动异常和设备故障,使得分布式电源调度控制面临更加复杂的局面。为应对这些挑战,提高异常事件的检测效率和准确率,以辅助分布式电源系统的调控决策技术,提出了一种基于火鹰优化的CatBoost算法(fire hawk optimizer-CatBoost,FHO-CatBoost)的分布式电源调控异常事件检测模型。该模型充分利用了CatBoost的强大梯度框架和自动处理类别特征的能力,通过FHO算法的调整超参数优化模型,提高了检测效率与识别准确率。实验结果证明,FHO-CatBoost模型在不同类别异常事件准确检测和整体性能上均表现优越,并在多方面性能评估中均优于其他主流梯度提升算法,在准确率上达到了91.59%,较最好的CatBoost方法提升了6.58%,具有更出色的性能表现,在分布式电源调控异常事件检测中具有显著优势,为电力系统安全运行提供了重要支持。The comprehensive promotion of the new power system still faces multiple security challenges,especially the distributed power system is vulnerable to threats such as extreme weather,natural disasters and cyber attacks.These situations will lead to abnormal system fluctuations and equipment failures,making distributed power scheduling and control face a more complex situation.To cope with these challenges,the detection efficiency and accuracy of anomalous events are improved to assist the regulation decision-making techniques of distributed power systems and to improve the chain fault blocking capability.In this paper,we propose a distributed power control anomalous event detection model based on FHO-CatBoost.The model makes full use of CatBoost's powerful gradient framework and automatic processing of category features,and optimizes the model by adjusting the hyperparameters of the Fire Hawk Optimization(FHO)algorithm,which improves the performance of the model,and efficiently and accurately detects and identifies anomalous events.Experimental results demonstrate that the FHO-CatBoost model exhibits superior performance in accurately detecting different categories of abnormal events and overall performance.It outperforms other mainstream gradient boosting algorithms in multi-faceted performance evaluations,achieving an accuracy of 91.59%,which is a 6.58%improvement over the best CatBoost method.It demonstrates better performance and significant advantages in detecting abnormal events in distributed generations power control,providing important support for the safe operation of power systems.

关 键 词：分布式电源 异常事件检测 CatBoost 火鹰优化算法 

分 类 号：TM721[电气工程—电力系统及自动化]