| 引用本文: | 刘文霞,徐晓波,周 樨.基于支持向量机的纯电动公交车充／换电站日负荷预测[J].电力自动化设备,2014,34(11): |
| LIU Wenxia,XU Xiaobo,ZHOU Xi.Daily load forecasting based on SVM for electric bus charging station[J].Electric Power Automation Equipment,2014,34(11): |
|
| 摘要: |
| 讨论了基于相似日选取的支持向量机电动汽车日负荷预测方法。通过对北京现有纯电动公交车充／换电站充电负荷的大量调研，分析了公交车充电站充电负荷的数据特征，采用关联分析方法提取了影响电动公交站充电负荷的因素，基于相关因素应用灰色关联理论构建相似日的小样本集合，而后建立多输入单输出的支持向量机预测模型。针对支持向量机预测模型，提出了两阶段确定模型参数的方法，首先直接确定不敏感损失参数ε，再通过遗传算法寻找最优核参数p和正则化参数C，以提高参数ε选取范围设置较大时的预测精度。实例测试结果表明，日负荷预测的均方根误差为10.85 %，能基本满足有序控制的要求；与其他预测方法相比，改进方法具有较高的预测精度和稳定性。 |
| 关键词: 电动汽车 负荷预测 支持向量机 参数选择 充电 关联理论 相似日 |
| DOI： |
| 分类号: |
| 基金项目:国家高技术研究发展计划（863计划）资助项目（2011AA05A109） |
|
| Daily load forecasting based on SVM for electric bus charging station |
|
LIU Wenxia, XU Xiaobo, ZHOU Xi
|
|
North China Electric Power University，Beijing 102206，China
|
| Abstract: |
| The method of daily electric vehicles load forecasting based on the similar days by the SVM（Support Vector Machine） is discussed. The charging load data of electric bus charging stations in Beijing are researched and their characteristics are analyzed. The correlation analysis is applied to extract their influencing factors and the gray relational analysis is applied to establish the small sample of similar days，based on which a multi-input single-output SVM forecasting model is built. Two steps are proposed to determine the model parameters：the insensitive loss parameter ε is directly set in the first step，while the optimal kernel parameter p and the regularization parameter C are determined by the genetic algorithm to improve the forecast accuracy when the selection range of ε is larger. Results of case test show that，the RMSE（Root Mean Square Error） of daily load forecast is 10.85 %，basically meeting the requirement of coordinated control. Compared with other forecasting methods，the proposed method has better accuracy and stability. |
| Key words: electric vehicles electric load forecasting support vector machines parameter selection charging correlation theory similar days |
