融合PMV物理方程和Attention-LSTM神经网络的铁路客站旅客舒适度模型研究

Model of Passenger Comfort Level in Railway Passenger Stations Based on Fusion of PMV Physical Equation and Attention-LSTM Neural Network

铁路客站的舒适度直接关系着旅客的出行体验和满意度。选取聊城西站作为研究对象,采用PMV物理方程、Attention-LSTM神经网络模型以及PMV&Attention-LSTM融合模型3种方法,针对旅客舒适度开展综合评估与分析。在模型构建过程中,运用了标准化处理、数据集划分、网格搜索交叉验证等技术寻找最佳超参数,并记录了训练过程中的损失函数和均方误差。在模型预测中,充分考虑了温度、湿度、风速、空气质量、二氧化碳、光照、噪声等环境因素对旅客舒适度的影响。对比3种预测方法,结果显示,融合模型在考虑多维环境数据时可更准确地反映舒适度水平,表明该模型更适应铁路客站的复杂环境条件,可为提高候车厅舒适性提供更为可靠的参考依据。

The comfort level of railway passenger stations plays a crucial role in enhancing the overall travel experience and satisfaction of passengers.Liaochengxi Railway Station is selected as the object of this study,with three models—PMV physical equation,Attention-LSTM neural network model,and PMV&Attention-LSTM fusion model—being adopted to carry out comprehensive evaluation and analysis on passenger comfort level in railway stations.In the process of modelling,techniques such as standardization processing,dataset partitioning,and grid search cross-validation are used to find the optimal hyperparameters,and the loss function and mean square error in the training process are recorded.The model prediction takes into account a variety of environmental factors,including temperature,humidity,wind speed,air quality,carbon dioxide levels,lighting conditions,and noise levels.Comparing the three prediction methods,the results show that the fusion model provides a more precise reflection of passenger comfort levels,particularly when accounting for multi-dimensional environmental data.This suggest that the fusion model is better suited for the complex environmental conditions found in railway passenger stations and can offer more reliable reference for enhancing comfort in passenger waiting areas.