[1]袁溪,张邵欣*,张超,等.基于Transformer的电动汽车充电站能耗预测研究[J].计算机技术与发展,2025,(02):213-220.[doi:10.20165/j.cnki.ISSN1673-629X.2024.0326]
 YUAN Xi,ZHANG Shao-xin*,ZHANG Chao,et al.Research on Electric Vehicle Charging Station Energy Consumption Prediction Based on Transformer[J].,2025,(02):213-220.[doi:10.20165/j.cnki.ISSN1673-629X.2024.0326]
点击复制

基于Transformer的电动汽车充电站能耗预测研究()

《计算机技术与发展》[ISSN:1006-6977/CN:61-1281/TN]

卷:
期数:
2025年02期
页码:
213-220
栏目:
新型计算应用系统
出版日期:
2025-02-10

文章信息/Info

Title:
Research on Electric Vehicle Charging Station Energy Consumption Prediction Based on Transformer
文章编号:
1673-629X(2025)02-0213-08
作者:
袁溪1张邵欣2*张超3王宁宁4张萌萌5
1. 西安文理学院 商学院,陕西 西安 710065;
2. 绿盟科技,陕西 西安 710076;
3. 西安索思信息技术有限公司,陕西 西安 710075;
4. 平凉信息工程学校,甘肃 平凉 744000;
5. 西安邮电大学 计算机学院,陕西 西安 710121
Author(s):
YUAN Xi1ZHANG Shao-xin2*ZHANG Chao3WANG Ning-ning4ZHANG Meng-meng5
1. School of Business,Xi’an University of Arts and Sciences,Xi’an 710065,China;
2. Green Alliance Technology,Xi’an 710076,China;
3. Xi’an Sosi Information Technology Co. ,Ltd. ,Xi’an 710075,China;
4. Pingliang Information Engineering School,Pingliang 744000,China;
5. School of Computer Science,Xi’an University of Posts and Telecommunications,Xi’an 710121,China
关键词:
电动汽车能耗预测Transformer模型深度学习自注意力机制
Keywords:
electric vehiclesenergy consumption predictionTransformer modeldeep learningself-attention mechanism
分类号:
TP391
DOI:
10.20165/j.cnki.ISSN1673-629X.2024.0326
摘要:
针对传统模型在处理高维非线性数据时预测精度不足的问题,该文提出了一种基于 Transformer 的电动汽车充电站能耗预测方法。 该方法通过引入自注意力机制,有效捕捉时间序列中的长短期依赖关系,显著提升了模型在复杂时序数据中的特征提取与预测能力。 为了验证该模型的有效性,将其与几种主流深度学习模型(包括 RNN、TCN、LSTM、GRU及其结合注意力机制的变体模型)进行了对比分析。 通过对宝鸡市某微电网电动汽车的真实能耗数据进行实验验证,结果表明,基于 Transformer 的模型在预测精度上显著优于其他模型,且在平均绝对误差(MAE)、均方根误差(RMSE)、以及判定系数(R2)等评价指标上表现出明显优势,证明了其在电动汽车能耗预测任务中的可靠性与先进性。 该方法为电动汽车能源管理、充电站规划和电网优化提供了有力支持,对推动电动汽车行业的可持续发展、优化充电设施布局以及提升电网运营效率具有重要的实际应用价值。
Abstract:
In addressing the issue that traditional models exhibit insufficient prediction accuracy when handling high - dimensional nonlinear data,we put forward a Transformer-based energy consumption prediction approach for electric vehicle charging stations. This method incorporates the self - attention mechanism to effectively capture long - term and short - term dependencies within time series,markedly enhancing the model’s feature extraction and prediction capabilities in complex time-series data. To validate the efficacy of the proposed model,it was compared with several mainstream deep learning models (including RNN,TCN,LSTM,GRU,and their variants based on the attention mechanism). Through experimental verification using the real energy consumption data of a microgrid electric vehicle in Baoji City,it is indicated that the Transformer-based model demonstrates significantly superior prediction accuracy compared to other models and presents distinct advantages in terms of mean absolute error (MAE),root mean square error (RMSE),and coefficient of determination (R2),substantiating its reliability and advanced nature in the task of electric vehicle energy consumption prediction.The proposed approach offers substantial support for electric vehicle energy management, charging station planning, and grid optimization,and holds significant practical application value for promoting the sustainable development of the electric vehicle industry,optimizing the layout of charging facilities,and enhancing the operational efficiency of the grid.
更新日期/Last Update: 2025-02-10