E-Elman神经网络在冰蓄冷空调系统建模中的应用  

Application of Modeling in Ice Storage Air Conditioning System Based on E-Elman Neural Network

作　　者：赵志达[1] 余南阳[1] 

机构地区：[1]西南交通大学机械工程学院,成都610031

出　　处：《计算机测量与控制》2017年第6期135-138,共4页Computer Measurement &Control

基　　金：四川省省级建筑节能专项资金项目(2013-02-05)

摘　　要：冰蓄冷技术能够为空调系统带来显著的节能效果,已广泛地应用在现代建筑当中,准确地监测数据预测值有利于系统合理地运行;针对实际冰蓄冷空调工程中的能源管理控制系统(energy management and control system,EMCS)数据采样周期较长所导致在系统制冰与融冰阶段数据不足的问题,提出了一种改进的机组运行状况预测模型;模型算法以数据变化趋势为依据,在传统Elman中引入评价层以约束网络输出值,增加计算针对性,从而提高模型输出的准确性;仿真结果表明,此种建模方法解决了系统融冰与制冰阶段的数据突变及网络输出值局部最优解等问题,与传统Elman网络结构相比,其输出值更为接近测量值,有效地提高了模型输出的真实性;通过关联函数,设计的模型对冷水机组的能源消耗也可起到预测作用,进一步说明了其实用性。Ice storage has been widely used in modern buildings, because of the prominent energy saving effect for air conditioning sys- tems, and the accurate predicted value of monitoring data is propitious to reasonable operation of the system. To solve the problem of data de- ficiencies in the EMCS with long sampling period when the ice is made and melted, an improved chiller operation prediction model was pro- posed. This model based on the data variation trend in the actual project, and the evaluation layer was added in Elman neural network to re- strain the network output, it had increased the computational pertinence and accuracy. The result showed that the improved model solved the problems about data mutation and local optimum, and enhanced the facticity of predicted outputs significantly. Compared with Elman net- work, the output value of the improved model was more close to the measured value. Meanwhile, the energy consumption of chiller also could be predicted using this improved model and correlation function, the practicability was explained.

关 键 词：冰蓄冷 空调系统 ELMAN神经网络 数学建模 

分 类 号：TU831[建筑科学—供热、供燃气、通风及空调工程]