基于模糊C聚类-云模型-LSTM网络的分接开关气象易损性评估预警  

Meteorological Vulnerability Assessment and Warning of On-load Tap Changer Based on Fuzzy-C Means-cloud Model-LSTM Network

作　　者：龚禹璐 崔龙飞 陈静 王典浪 须雷 李东 尹启 GONG Yulu;CUI Longfei;CHEN Jing;WANG Dianlang;XU Lei;LI Dong;YIN Qi(Qujing Bureau of EHV Power Transmission Company of China Southern Power Grid Co.,Ltd.,Qujing 655000,Yunnan Province,China;NR Engineering Co.,Ltd.,Nanjing 210000,Jiangsu Province,China)

机构地区：[1]中国南方电网有限责任公司超高压输电公司曲靖局,云南省曲靖市655000 [2]南京南瑞继保工程技术有限公司,江苏省南京市210000

出　　处：《现代电力》2025年第2期385-392,共8页Modern Electric Power

摘　　要：为解决新型电力系统中极端气象灾害对变压器有载分接开关(on load tap changer,OLTC)破坏严重、易造成故障的问题,提出一种基于模糊C均值聚类(fuzzy c-means,FCM)-云模型-长短期记忆网络(long short term memory network,LSTM)的有载分接开关气象灾害易损性评估预警方法。该方法首先基于变压器气象监测数据,建立致灾因子评估体系;根据FCM聚类算法,将传统云模型的阈值划分进行改进,得到客观云模型,将主客观云模型与其进行结合构建组合云模型。基于自然灾害理论,考虑地理孕灾环境、设备本身的抗灾能力、灾害风险累积程度以及面对灾害的风险综合处理能力等因素,对待评估灾害指标进行动态修正,将修正后的指标计算在组合云模型的隶属度中,从而得到OLTC各灾害易损性等级。最后,应用LSTM神经网络提取各致灾因子与各灾害易损性之间的关联规则,并进行气象灾害预警,形成最佳应对策略。算例结果表明:该文提出的OLTC气象灾害易损性评估预警方法准确率较高,达到了防灾减灾的效果。To address the issue of severe damage caused by meteorological disasters and fault potential,a vulnerability assessment and early warning method for on load tap changer(OLTC)of transformer in the new power system was proposed based on improved cloud model and long short term memory network(LSTM).Firstly,the assessment system of disastercausing factors was established based on the meteorological monitoring data of OLTC.According to the FCM clustering algorithm,the threshold division of the traditional cloud model was improved to obtain the objective cloud model,and consequently a combination cloud model was constructed by combining the subjective and objective cloud models.Based on the natural disaster theory,the disaster indicators to be evaluated are dynamically adjusted by considering the factors such as the geographical disaster pregnant environment,the disaster resistance ability of the equipment itself,the accumulation degree of disaster risk and the comprehensive risk processing ability facing the disaster.These modified indicators were calculated and utilized in the membership degree of the combined cloud model,aiming to acquire the disaster vulnerability level for OLTC.Finally,the LSTM neural network was applied to extract the association rules between each disaster causing factor and each disaster vulnerability,thereby facilitating meteorological disaster early warning and forming the optimal response strategy.The example results indicate that the OLTC meteorological disaster vulnerability assessment and early warning method proposed in this paper exhibits high accuracy and effectively achieves the objective of disaster prevention and reduction.

关 键 词：有载分接开关 极端气象灾害 改进云模型 LSTM神经网络 易损性评估预警 

分 类 号：TM407[电气工程—电器]