输电线路覆冰厚度短期预测模型研究

发布时间：2018-05-20 13:08

本文选题：覆冰厚度预测 + 关联分析　；参考：《华北电力大学》2017年硕士论文

【摘要】：输电线路覆冰常常对电网的安全运行构成严重的威胁,因此针对覆冰的在线监测及预测、预警技术的研究,是保障输电线路安全稳定运行的关键,也是建设智能化输电体系的要求。本文运用机器学习智能算法对线路覆冰的智能预测进行了研究,具体工作如下:1)研究了输电线路覆冰机理和覆冰形态,分析了输电线路中雨凇覆冰、混合凇覆冰和雾凇覆冰等多种覆冰形态,并从不同的分类标准描述了相应的线路覆冰类型及其特征。2)分析了在线监测数据库中的覆冰数据,通过对这些数据的大量筛选和预处理,挑选出可用于预测实验的完整覆冰过程数据。运用灰色关联分析方法将对覆冰等效厚度进行了关联分析,得出了影响输电线路覆冰的关键影响因素,并将这些关键因素作为后文中建模的输入量。3)研究了基于极限学习机(ELM)的输电线路覆冰预测模型,利用遗传算法对ELM网络的参数进行了优化,提出了改进型的极限学习机覆冰等效厚度预测模型。最后通过预测实验,证明了此模型的覆冰预测精度高于BP神经网络和GRNN神经网络。4)提出了一种新的线路覆冰厚度预测思路,首先从覆冰历史数据分析出发,通过主成分分析方法对实验数据进行降维后,然后利用基于思维进化算法优化的支持向量机进行覆冰厚度预测,达到了更理想的实验效果。
[Abstract]:The icing of transmission lines often poses a serious threat to the safe operation of the power network. Therefore, the research of on-line monitoring and prediction and early warning technology for icing is the key to ensure the safe and stable operation of transmission lines. It is also the requirement of building intelligent transmission system. In this paper, the intelligent prediction of icing on transmission lines is studied by using machine learning intelligent algorithm. The detailed work is as follows: 1) the mechanism and form of icing on transmission lines are studied, and the icing of rime in transmission lines is analyzed. The ice forms of mixed rime and rime icing are analyzed, and the icing data in the online monitoring database are analyzed according to different classification criteria. Through a large number of screening and preprocessing of these data, the complete ice process data which can be used to predict the experiment are selected. The equivalent thickness of icing is analyzed by using the method of grey correlation analysis, and the key factors affecting the icing of transmission lines are obtained. Taking these key factors as input quantity. 3) the ice prediction model of transmission line based on LLM) is studied. The parameters of ELM network are optimized by genetic algorithm. An improved prediction model of the ice equivalent thickness of the ultimate learning machine is presented. Finally, through prediction experiments, it is proved that the prediction accuracy of this model is higher than that of BP neural network and GRNN neural network. After reducing the dimension of the experimental data by principal component analysis (PCA), the ice thickness prediction is carried out by using the support vector machine (SVM), which is optimized by the thought evolution algorithm, and the experimental results are more satisfactory.
【学位授予单位】：华北电力大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP18;TM75

【参考文献】