摘要
针对长时间序列电力负荷的预测精度低的问题,应用了基于Informer长时间序列模型的电力负荷预测方法.该方法通过Informer模型中的自注意力蒸馏机制,使得每层的解码器都将输入序列的长度缩短一半,从而极大地节约了Encoder内存开销,并在编码器结构中使用生成式结构,使得预测解码时间极大的缩短;以澳大利亚的电力负荷数据作为测试用例,并与长短时记忆神经网络(long-short term memory,LSTM)和卷积神经网络(convolutional neural network,CNN)模型方法进行对比,结果表明,Informer模型的预测精度更高,Pearson相关系数可以达到91.30%,有效提高了负荷预测精度.
Aiming at the problem of low accuracy of long time series load prediction,a long sequence Time-Series Forecasting method based on Informer is applied.By means of self-attention distillation mechanism in Informer model,the decoder of each layer can shorten the length of the input sequence by half,which greatly saves the memory overhead and time of Encoder.In addition,the generative structure is used in the structure of Encoder to greatly shorten the predictive decoding time.Taking the load data of Australia as actual calculation example,and compared with long-short term memory(LSTM),convolutional neural network(CNN)model methods,the results show that,the prediction accuracy of Informer model is higher,and Pearson’s correlation coefficient can reach 91.30%,which improves the accuracy of load prediction effectively.
作者
刘洪笑
向勉
周丙涛
段亚穷
伏德粟
LIU Hongxiao;XIANG Mian;ZHOU Bingtao;DUAN Yaqiong;FU Desu(School of Information Engineering,Hubei Minzu University,Enshi 445000,China)
出处
《湖北民族大学学报（自然科学版）》
CAS
2021年第3期326-331,共6页
Journal of Hubei Minzu University:Natural Science Edition
基金
2020年硒食品营养与健康智能技术湖北省工程研究中心开放课题(PT082005).
