摘要
建立了一个能够预测电动汽车充电站用户未来是否违约的模型,为优化充电站调度提供支撑。首先,分析了影响用户行为的特征因素,并据此建立决策树模型进行预测,验证特征工程是否可行;然后,在确定特征工程的基础上,尝试了2种主流集成学习算法,即基于Bagging算法的随机森林模型和基于Boosting算法的提升树模型;最后,为对比2种算法的准确率,对其进行算例仿真分析,结果表明后者准确率更高。
A model that can predict whether electric vehicle charging station users will default in the future is established to provide support for optimizing charging station scheduling.Firstly,the characteristic factors affecting user behavior are analyzed,and a decision tree model is established to predict and verify whether feature engineering is feasible.Then,on the basis of determining the feature engineering,two mainstream ensemble learning algorithms are tried:the random forest model based on bagging algorithm and the lifting tree model based on boosting algorithm.Finally,in order to compare the accuracy of the two algorithms,the above integrated learning algorithm is simulated and analyzed.The results show that the lifting tree model performs better.
作者
袁天笑
贾鹏
顾云长
冯浩
姜雨馨
吴宇辰
伍兴达
YUAN Tianxiao;JIA Peng;GU Yunchang;FENG Hao;JIANG Yuxin;WU Yuchen;WU Xingda(Nanjing Institute of Technology,Nanjing 211167,China;Xinghua Power Supply Branch,State Grid Jiangsu Electric Power Co.,Ltd.,Xinghua 225700,China)
出处
《电工技术》
2023年第10期69-72,共4页
Electric Engineering
基金
南京工程学院基金项目(编号TB202204028)。
关键词
预测模型
用户行为
充电站
电动汽车
集成学习
prediction model
user behavior
charging station
electric vehicle
integrated learning
