摘要
对热网进行负荷预测是保障供热系统按需供热和精准调控的前提和依据。针对供热系统中由于高延迟、大惯性特点造成的调控不灵和能源浪费问题,构建了基于EEMD-LSMD方法的区域热网负荷预测模型,先通过EEMD(集合经验模态分解)对某热网换热站历史运行数据进行分解,并对分解的分量进行选择和辨别,再结合LSTM(长短期记忆神经网络)算法对选择的各个变量分别进行预测,最后叠加预测值得出预测结果。算例结果表明,采用EEMD与神经网络相结合的方法进行区域热网负荷预测大大减少了能源浪费,部分站点节能率甚至达到了30%以上。
Load forecasting of heating network is the premise and basis to satisfy heating demand and precisely regulate heating system.In this paper,aiming at avoiding ineffective regulation and energy waste caused by high delay and large inertia in heating system,a load forecasting model of regional heating network based on EEMD-LSMD method was proposed.Firstly,EEMD(Set Empirical Mode Decomposition)was used to decompose the historical operation data of a heating network heat exchange station,and the decomposed components were selected and discriminated.Then LSTM(Long-term and Short-term Memory Neural Network)algorithm was used to forecast the selected variables respectively.Finally,the predicted results were obtained by superimposing predicted values.The results of an example show that the method of combining EEMD with neural network for load forecasting of regional heating network greatly reduces energy waste,the energy saving rate of some stations is even above 30%.
作者
张月宇
李德成
方大俊
王凯
ZHANG Yue-yu;LI De-cheng;FANG Da-jun;WANG Kai(Linyi blue sky heat power Co.,Ltd,Linyi 276017,China;Changzhou Engipower Technology Co.,Ltd,Changzhou 213022,China)
出处
《能源工程》
2022年第1期1-6,共6页
Energy Engineering
关键词
区域热网
负荷预测
集合经验模态分解
长短期记忆神经网络
regional heat network
load forecasting
set empirical mode decomposition
long and short-term memory neural network