基于自回归LSTM神经网络的地铁车站环境参数概率预测  

Probabilistic Forecasting of Metro Station Environmental Parameters Based on Autoregressive LSTM Network

作　　者：曲洪权[1] 李博[1] 庞丽萍[2] 梁思远 Qu Hongquan;Li Bo;Pang Liping;Liang Siyuan(School of Information Science and Technology,North China University of Technology,Beijing,100144;School of Aeronautic Science and Engineering,Beijing,100191)

机构地区：[1]北方工业大学信息学院,北京100144 [2]北京航空航天大学航空科学与工程学院,北京100191

出　　处：《制冷与空调（四川）》2020年第6期676-681,共6页Refrigeration and Air Conditioning

基　　金：国家重点研发计划(2017YFB1201100)资助项目;辽宁省“兴辽英才计划”项目资助(XLYC1802092)。

摘　　要：在地铁车站这种人员密集的场景中,预测未来一段时间内的环境参数变化对于列车正常运行和乘客安全有重要意义。和传统的点预测方法相比,提出了一种基于自回归循环LSTM网络的概率预测方法。不同于点预测,概率预测给出预测变量在下一时刻的概率密度函数,考虑了地铁车站环境参数预测的不确定性,对于站内提前、可靠对紧急情况做出反应有重要的意义。提出的预测模型,将预测变量的历史数据和车站外部环境参数作为输入变量来预测的环境参数的下一时刻的概率密度函数,进而得到下一时刻该环境参数的变化范围和分位数等信息。为了验证所提方法的准确性,收集了3个地铁站的环境参数数据,并使用概率预测方法进行了预测。结果显示,提出的方法不仅可以预测最可能的环境参数预测结果,而且可以预测极端情况的发生概率,对于预防紧急事故和决策有重要意义。In the metro station which is crowded with people,it is very important to predict the changes of environmental parameters in the future for the normal operation of trains and the safety of passengers.Compared with the traditional point prediction method,this paper proposes a probabilistic forecasting method based on autoregressive LSTM network.Different from the point prediction,the probabilistic forecasting model gives the probability density function of the prediction variable at the next moment,and calculates the uncertainty of the environmental parameter’s prediction,which is of great significance for the station to respond to the emergency in advance.The proposed model in this paper uses the historical data of the prediction variables and the external environmental parameters as the input variables to predict the probability density function of the environmental parameters at the next time,so as to obtain the information of the changing range and quantiles.In order to verify the accuracy of the method,we collected the environmental parameter data of 3 subway stations,and used the probabilistic forecasting method to predict.The results show that the method can not only predict the most likely environmental parameters prediction results,but also predict the probability of extreme cases,which is of great significance for the prevention of emergency accidents and decision-making.

关 键 词：地铁车站 环境参数 概率预测 LSTM网络 

分 类 号：X820[环境科学与工程—环境工程]