摘要
电网覆冰由于湿度、温度、海拔等因素可形成不同的覆冰类型,电网系统故障水平会根据覆冰类型的差异而变化。传统的电网覆冰故障预测多聚焦于覆冰厚度与电网故障的内在联系,忽略了覆冰类型对电网故障的影响作用。为解决这一问题,提出了一种改进的逻辑回归多分类算法,通过将拟合出的回归函数值输入Softmax函数转换为多概率形式实现对覆冰类型的分类,根据数据有无覆冰类型的划分,分别在不同的机器学习算法下对电网覆冰故障进行预测比较。经试验,采用改进逻辑回归算法预测覆冰类型的准确率达88%,电网故障预测的准确率较无覆冰类型下的预测在改进逻辑回归算法、朴素贝叶斯算法(Naïve Bayes,NB)、K近邻、支持向量机算法(Support Vector Machine,SVM)中分别平均提高了5.3%、21.7%、7%、5.3%,研究表明,改进的逻辑回归算法可以准确预测电网的覆冰类型,提升电网故障预测的准确率。
To solve the problem that the traditional prediction of power grid icing fault mainly focuses on the internal relationship between ice thickness and power grid fault,and ignores the influence of ice type on power grid fault,an improved logistic regression multi-classification algorithm is proposed in this paper.In this method,the value of the regression function fitted by geographical and meteorological factors is input into the Softmax function and converted into a multi-probability form to realize the prediction of ice cover type.Firstly,through the ice cover analysis,it is determined that the geographical environment and meteorological environment have an effect on the cause of ice cover,and the data of possible related influencing factors are selected to form the original ice cover type data.Then,75%of them are selected as training data sets to train the model.In the initial network parameters,the learning rate is set as 1,and the iterations are 10000 times,with the same weight of each factor.Finally,the model is verified by using the remaining data as the test set.Experimental results show that the prediction accuracy of the training set and test set is 86%and 88%,respectively.This indicates that there is no over-fitting or under-fitting of the model and the model can fit the training data and test data well.Then,the icing type of power grid fault data is predicted and divided under this model,and the data after dividing the icing type are predicted and compared under different machine learning algorithms.The experimental results show that the accuracy of power grid fault prediction is 5.3%,21.7%,7%,and 5.3%higher than that of the prediction without ice in the improved logistic regression algorithm,GNB,K-nearest Neighbor,and SVM algorithm respectively.The research shows that the improved logistic regression algorithm can accurately predict the type of power grid icing and improve the accuracy of power grid fault prediction.
作者
王国庆
翟悦
朱建明
黄钧
WANG Guoqing;ZHAI Yue;ZHU Jianming;HUANG Jun(School of Engineering Science,University of Chinese Academy of Sciences,Beijing 100190,China)
出处
《安全与环境学报》
CAS
CSCD
北大核心
2023年第6期1762-1770,共9页
Journal of Safety and Environment
基金
国家自然科学基金项目(72074202)。
关键词
安全工程
逻辑回归
覆冰类型
电网故障
预测
safety engineering
logistic regression
ice type
power grid fault
prediction
