基于融合注意力机制改进双向长短时记忆网络在电动汽车充电负荷中的预测研究  

Prediction of Electric Vehicle Charging Load Based on Integrating Attention Mechanism to Improve BiLSTM

作　　者：王华彪[1] 李小勇 WANG Huabiao;LI Xiaoyong(Chongqing Electric Power College,Chongqing 400050,China;School of Environment,South China Normal University,Guangzhou 510000,Guangdong,China)

机构地区：[1]重庆电力高等专科学校,重庆400050 [2]华南师范大学环境学院,广东广州510000

出　　处：《电网与清洁能源》2022年第6期104-112,共9页Power System and Clean Energy

基　　金：重庆市教育委员会2020年度科学技术研究项目(KJQN202002601)。

摘　　要：电动汽车的规模化发展以及充电设施的持续建设将给电网带来重要影响,严重威胁到了电力系统频率稳定性。结合电动汽车充电负荷数据特点,在深度学习方法的基础上提出基于融合注意力机制(attention mechanism,AM)改进的双向长短时记忆网络模型(long short-term memory network,LSTM),实现对电动汽车的优化调度。通过使用实测电动汽车充电负荷数据,比较了所提方法与已有方法的性能。结果表明,在LSTM和(bidirectional long short-term memory network,BiLSTM)分别添加了注意力机制的(long short-term memory attention network,LSTMA)和(bidirectional long shortterm memory attention network,BiLSTMA)模型相对于已有方法,在预测结果评价指标上都有明显的提升,证明了注意力机制在电动汽车充电负荷序列预测上的有效性。The large-scale development of electric vehicles and the continuous construction of charging facilities have an important impact on the power grid and seriously threaten the frequency stability of the power system.Electric vehicle charging load forecasting,which is the basis for analyzing the connection of electric vehicles to the grid,has received more attention.It is of great significance to accurately predict the daily load of electric vehicles based on the daily load curve of electric vehicles in the past few days,which is of great significance for realizing the optimal scheduling of electric vehicles and providing effective auxiliary decision-making.According to the characteristics of electric vehicle charging load data and based on the deep learning method,this paper proposes an improved bidirectional long-short-term memory network model(LSTM)based on the fusion attention mechanism(AM)to realize the optimal scheduling of electric vehicles.Finally,the performance of the new method and the existing methods is compared by using the measured electric vehicle charging load data.The LSTMA and BiLSTM models with attention mechanism added respectively have better prediction results than the LSTM and BiLSTM models alone.The simulation results also confirm the effectiveness of the attention mechanism in the prediction of electric vehicle charging load sequence.

关 键 词：充电负荷 回归预测 递归神经网络 双向长短期记忆网络 注意力机制 

分 类 号：TM715[电气工程—电力系统及自动化]