基于神经网络集中供热二次回水温度预测研究  

Research on Secondary Network Backwater Temperature Forecast for Centralized Heat-Supply System Based on Neural Network

作　　者：冯敬芳[1] 谢慕君[1] 步伟明[1] 姜长泓[1] 

机构地区：[1]长春工业大学电气与电子工程学院,吉林长春130012

出　　处：《计算机仿真》2014年第3期351-354,共4页Computer Simulation

基　　金：吉林省科技支撑重大项目(20126040);吉林省自然科学基金(201115151)

摘　　要：研究供热质量优化控制问题,针对集中供热二次网回水温度控制系统的实时性要求及供热系统时变性、非线性和时滞性的特点,为了准确获取回水温度控制系统的设定值,分别采用LM、Elman和RBF三种神经网络算法进行回水温度预测研究。选取一次网供水温度、一次网供水流量、室外温度、二次网供水温度、二次网供水流量作为预测模型的输入,二次网回水温度作为输出,分别构建了LM、Elman和RBF三种预测模型,仿真结果表明,三种预测模型均能实现回水温度的预测,RBF预测模型具有更快的收敛速度和更高的预测精度,可以更好地满足回水温度预测的实时性要求。The problem of heating quality optimization control was researched. According to the real-time require- ment of central heating secondary network backwater temperature control system and the features of central heating system such as time-varying, nonlinear and time lag, in order to obtain the accurate given value of the system, we re- searched backwater temperature prediction based on LM, Elman and RBF neural network algorithm respectively. The inputs of prediction model include supply water temperature and supply water flow of primary network, outdoor tem- perature, supply water temperature and supply water flow of secondary network. The output of prediction model is the secondary network backwater temperature. The simulation results based on LM, Elman and RBF prediction model re- spectively show that all the three prediction models can realize the prediction of backwater temperature. But the RBF prediction model has faster convergence speed and higher prediction precision, which can meet the real-time require- ment of backwater temperature prediction.

关 键 词：神经网络 集中供热 回水温度 预测 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]