摘要
针对社区微网中电动汽车负荷预测存在的数据样本不足的问题,提出一种将长短期记忆(Long Short-Term Memory,LSTM)与领域自适应神经网络(Domain Adaptive Neural Network,DANN)结合形成DaNN-LSTM的负荷预测算法,从而实现对社区微网小样本的电动汽车负荷数据的准确预测。利用预处理后的源域数据预先训练好LSTM模型,再将LSTM模型的相关参数迁移到DANN的LSTM层中,最后对社区微网中的负荷数据进行重复训练,得出预测结果。预测结果表明,所提到的方法相比于LSTM模型的准确率有了一定程度的提高,可以满足实际需求。
Aiming at the problem of insufficient data samples for electric vehicle load forecasting in community microgrid,this paper proposed a Domain Adaptive Neural Network-Long Short-Term Memory(DANN-LSTM)load forecasting algorithm combining Long Short—Term Memory(LSTM)and Domain Adaptive Network(DANN),so as to achieve accurate prediction of electric vehicle load data of small samples in community microgrid.In this paper,the preprocessed source domain data is used to train the LSTM model in advance,and then the relevant parameters of the LSTM model are migrated to the LSTM layer of DANN.Finally,the load data in the community microgrid is repeatedly trained to obtain the prediction results.The prediction results show that the accuracy of the proposed method is improved to a certain extent compared with the LSTM model,which can meet the actual needs.
作者
刘子博
LIU Zibo(Jiangnan University,Wuxi 214000,China)
出处
《通信电源技术》
2023年第7期34-38,共5页
Telecom Power Technology
作者简介
刘子博(1997—),男,河北邢台人,硕士研究生,主要研究方向为新能源控制及节能技术。E-mail:735734030@qq.com。
