基于LSTM和时间序列分析法的短期风速预测  

Short-Term Wind Speed Prediction Based on Short and Long Time Memory Network and Time Series Analysis Method

作　　者：李蓉蓉[1] 戴永[2] LI Rong-rong;DAI Yong(College of Computer,Guangdong University of Science and Technology,Dongguan Guangdong 523083,China;College of Information Engineering,Xiangtan University,Xiangtan Hunan 411105,China)

机构地区：[1]广东科技学院计算机学院,广东东莞523083 [2]湘潭大学信息工程学院,湖南湘潭411105

出　　处：《计算机仿真》2020年第3期393-398,共6页Computer Simulation

基　　金：广东省教育厅青年创新人才类项目(自然科学类)(2017KQNCX226)。

摘　　要：短期风速对输电线路影响巨大,由于短期风速的随机性和非线性特性,使得短期风速难以精确预测。提出了一种将长短时记忆网络和时间序列分析法相结合的组合预测算法来实现对短期风速的预测。首先,利用时间序列分析法对短期风速进行预测得到预测结果和预测残差,然后利用长短时记忆网络对预测残差进行预测,最后将两种方法得到的预测结果进行线性组合得到最终的预测结果序列。为验证所提出的算法的实际效果,将提出的算法与时间序列分析法、长短时记忆网络以及BP神经网络等进行对比。实验结果表明,组合算法有效提高了短期风速序列预测精度,是一种可行的分析方法。The short-term wind speed has a great impact on the transmission line.Due to the randomness and nonlinear characteristics of the short-term wind speed,it is difficult to accurately predict the short-term wind speed.This paper proposes a combined prediction algorithm combining long-short-time memory network and time series analysis to predict short-term wind speed.Firstly,the short-term wind speed was predicted by time series analysis to obtain the prediction result and the prediction residual.Then,the prediction residual was predicted by the long-short-time memory network.Finally,the prediction results obtained by the two methods were linearly combined to obtain the final prediction result sequence.In addition,in order to verify the actual effect of the proposed algorithm,the proposed algorithm was compared with time series analysis method,long and short time memory network and BP neu?ral network.The experimental results show that the combination algorithm effectively improves the prediction accuracy of short-term wind speed sequences and is a feasible analysis method.

关 键 词：短期风速预测 时间序列分析法 长短时记忆网络 组合算法 

分 类 号：TP391.9[自动化与计算机技术—计算机应用技术]