地铁冰蓄冷空调系统负荷预测研究  

Study on load prediction of ice storage air conditioning system in metro

作　　者：陈哲[1] 张九根[1] 陈浩[1] CHEN Zhe;ZHANG Jiugen;CHEN Hao(Intelligent Building Institute,Nanjing Tech University,Nanjing 211816,China)

机构地区：[1]南京工业大学建筑智能化研究所,江苏南京211816

出　　处：《现代电子技术》2018年第23期169-174,178,共7页Modern Electronics Technique

摘　　要：地铁的运营特点和负荷分布特点使得其应用冰蓄冷技术能够实现"削峰填谷"以及提高空调设备使用率,均衡峰谷电网负荷,节省运行费。以南京地铁2号线某地下车站通风空调系统为研究对象,将实时负荷计算结果作为神经网络的训练数据,并通过遗传算法优化神经网络的初始权值和阈值构成GA-BP算法,采用分段训练的方法,建立地铁工作日和休息日逐时负荷24 h的提前神经网络预测模型。运用BP算法和GA-BP算法对该模型进行预测。结果表明,基于分段预测的神经网络预测模型能够满足地铁冰蓄冷空调系统负荷预测的要求,并且采用GA-BP算法的预测精度更高。The operation characteristics and load distribution characteristics of the metro make ice storage air conditioning technology realize peak cut,improve the utilization ratio of air conditioning equipment,balance the peak valley grid load,and save the operating cost.Taking the ventilation and air conditioning system of an underground station of Nanjing Metro Line 2 as the study object,using the real-time load calculation results as the training data of the neural network,the GA-BP algorithm is constructed by optimizing the initial weights and thresholds of neural network by means of genetic algorithm.The method of segment training is used to establish a 24 hours-ahead neural network prediction model of hourly load for subway working days and rest days.The BP algorithm and GA-BP algorithm are adopted to predict the model.The results show that the neural network prediction model based on segment prediction can meet the requirement of ice storage air conditioning system in metro for load forecasting,and the GA-BP algorithm has higher prediction accuracy.

关 键 词：冰蓄冷空调 节能 BP神经网络 遗传算法 负荷预测 建模 

分 类 号：TN876-34[电子电信—信息与通信工程] TP183[自动化与计算机技术—控制理论与控制工程]