中国电力 ›› 2020, Vol. 53 ›› Issue (6): 56-63.DOI: 10.11930/j.issn.1004-9649.201911078

• 人工智能在电力系统的应用 • 上一篇    下一篇

基于FCE和SVM融合的线路典型冰风灾害算法分析

谷凯凯1,2, 陈凯1,2, 顾然1,2, 彭仲晗1,2, 吴启瑞1,2, 宋友1,2   

  1. 1. 南瑞集团有限公司,江苏 南京 211106;
    2. 国网电力科学研究院武汉南瑞有限责任公司,湖北 武汉 430074
  • 收稿日期:2019-11-14 修回日期:2020-03-03 发布日期:2020-06-05
  • 作者简介:谷凯凯(1987-),男,博士,高级工程师,从事电网状态评估与检修,E-mail:gukaikai@sgepri.sgcc.com.cn;彭仲晗(1991-),男,通信作者,硕士,工程师,从事电力设备状态监测及故障诊断,E-mail:pengzhonghan@sgepri.sgcc.com.cn
  • 基金资助:
    国家电网有限公司总部科技项目(基于小样本机器学习方法的输电线路典型冰风灾害特征识别及预测技术研究,524625180051)

An Algorithm for Analyzing Typical Transmission Line Icing and Wind Disasters Based on Integration of Fuzzy Comprehensive Evaluation and Support Vector Machine

GU Kaikai1,2, CHEN Kai1,2, GU Ran1,2, PENG Zhonghan1,2, WU Qirui1,2, SONG You1,2   

  1. 1. NARI Group Corporation, Nanjing 211106, China;
    2. Wuhan NARI Limited Company of State Grid Electric Power Research Institute, Wuhan 430074, China
  • Received:2019-11-14 Revised:2020-03-03 Published:2020-06-05
  • Supported by:
    This work is supported by Science and Technology Project of State Grid Corporation of China(SGCC)(Research on Characteristics Recognition and Prediction of Ice and Wind Disasters in Transmission Lines Based on Small Sample Machine Learning, No.524625180051)

摘要: 目前,国内外对于线路覆冰和风耦合作用的灾害分析较少,因此,提出了一种基于模糊综合评价法(fuzzy comprehensive evaluation, FCE)和支持向量机(support vector machine,SVM)融合的线路典型冰风灾害分析算法。通过分析典型冰风灾害影响因子及类型,借助模糊综合评价法提取了关键的灾害影响指标,并对风速和风向关键指标进行修正。在提取的温度、相对湿度、风速、风向和地貌5类致灾相关程度高指标的基础上,提出了采用径向基RBF核函数的非线性SVM小样本灾害分析模型。通过历史的冰风故障和非冰风故障数据建立训练样本和测试样本,仿真结果表明,建立的模糊综合评价和支持向量机融合的冰风灾害模型可有效分析判断冰风灾害发生的概率,实现了对冰风灾害小样本数据的可靠分析。

关键词: 输电线路, 模糊综合评价, 支持向量机, 冰风灾害, 故障分析

Abstract: At present, researches on transmission line faults caused by the coupling effect of icing and wind are few, an innovative algorithm is therefore proposed for analyzing the icing-wind disasters of transmission lines based on integrated fuzzy comprehensive evaluation (FCE) and support vector machine (SVM) method. Firstly, by analyzing the influencing factors and their types of typical ice-wind disasters, the key influencing indicators are extracted using FCE method, and the key indicators of wind speed and direction are corrected. And then, based on extraction of five indicators that are highly correlative to disasters, including temperature, relative humidity, wind speed, wind direction and landforms, a nonlinear SVM model with RBF kernel function is proposed for disaster analysis of small samples. Finally, the training samples and test samples are established from historical icing-wind caused faults and non icing-wind caused fault data. Simulation results show that the ice-wind disaster model established by integrated FCE and SVM can effectively judge the probability of ice-wind disaster, and realize the reliable ice-wind disaster analysis with small samples of data.

Key words: transmission line, fuzzy comprehensive evaluation (FCE), support vector machine (SVM), icing and wind disaster, fault analysis