摘要
针对城市轨道交通短时进出站客流的强随机性、周期性及非线性的特征,提出了一种基于小波变换与Adam算法优化的长短时记忆网络(LSTM)短时客流组合预测模型(即WT-LSTM组合模型),同时基于非饱和激活函数ReLU函数实现了LSTM的学习与训练。采用LSTM模型与WT-LSTM组合模型对广州地铁广州塔站的客流量进行预测,并对预测结果的误差进行对比分析。结果表明,WT-LSTM组合模型能够较好地预测短时客流,预测结果优于单一LSTM模型。
In view of short-term passenger flow in urban rail transit having characteristics of strong randomness,periodicity and non-linearity,a combined forecasting model(WT-LSTM combined model)of short-term passenger flow based on wavelet transform and LSTM(long short-term memory network)optimized by Adam algorithm is proposed.Meanwhile,the LSTM model is learned and trained by the unsaturated activation function ReLU function.Passenger flow at Guangzhou Metro Canton Tower station is predicted using LSTM single model and WT-LSTM combined model,and the error of the prediction results is compared and analyzed.Result shows that the WT-LSTM model can predict short-term passenger flow well,and the prediction effect of it is better than that of LSTM single model.
作者
蔡昌俊
CAI Changjun(Guangzhou Metro Group Co.,Ltd.,510335,Guangzhou,China)
出处
《城市轨道交通研究》
北大核心
2021年第9期14-19,24,共7页
Urban Mass Transit
关键词
城市轨道交通
短时客流量预测
组合模型
小波变换
长短期记忆网络
urban rail transit
short-term passenger flow prediction
combined prediction model
wavelet transform
LSTM(long short-term memory network)
