基于LSTM的地铁车站设备间设备发热量预测  

LSTM-based equipment's heat release prediction method for equipment rooms in subway station

作　　者：孙心明[1] 李国栋[1] 刘军 夏三县 篮杰[1] 余伟之 王海涛[3] SUN Xinming;LI Guodong;LIU Jun;XIA Sanxian;LAN Jie;YU Weizhi;WANG Haitao

机构地区：[1]中铁第四勘察设计院集团有限公司,湖北武汉430063 [2]郑州地铁集团有限公司,河南郑州410104 [3]河南工业大学土木工程学院,河南郑州450001

出　　处：《节能》2024年第10期98-100,共3页Energy Conservation

基　　金：河南省科技攻关项目(项目编号:232102320225)。

摘　　要：设计地铁车站设备间空调系统时,缺少可靠的发热量数据会影响其精细化设计和节能减排效果。结合实测地铁车站设备间的设备发热量数据,通过定量分析特征变量对地铁车站设备间设备发热量变化的影响,确定地铁车站设备间设备发热量预测的5个稳定特征变量和9个时变特征变量。考虑地铁车站设备间设备发热量数据的非线性和时序相关性特点,给出一种基于长短期记忆神经网络(LSTM)的地铁车站设备间设备发热量预测方法。该方法可以明显提高地铁车站设备间设备发热量预测的准确性,为未来解决地铁车站空调系统设计、节能控制和智能控制等问题提供参考。In the design of air conditioning systems for subway station equipment rooms,the absence of reliable heat generation data can impact the precision of the design and the effectiveness of energy conservation and emission reduction.By integrating actual measured heat generation data from equipment in subway station equipment rooms and conducting a quantitative analysis of the influence of characteristic variables on the heat generation changes in these rooms,five stable characteristic variables and nine time-varying characteristic variables for predicting equipment heat generation in subway station equipment rooms are identified.Considering the nonlinear and time-series correlation characteristics of the equipment heat generation data in subway station equipment rooms,a prediction method for equipment heat generation based on Long Short-Term Memory(LSTM) neural networks is proposed.This method significantly enhances the accuracy of predicting equipment heat generation in subway station equipment rooms,providing a reference for future solutions to issues such as the design of subway station air conditioning systems,energy-saving control,and intelligent control.

关 键 词：地铁车站 设备发热量 特征变量 LSTM神经网络 

分 类 号：TU831.3[建筑科学—供热、供燃气、通风及空调工程]