基于遗传粒子群优化的热负荷预测方法  

Heat load prediction method based on genetic particle swarm optimization

作　　者：谢文举 薛贵军 史彩娟 李水清 XIE Wenju;XUE Guijun;SHI Caijuan;LI Shuiqing(College of Electrical Engineering,North China University of Technology,Tangshan 063210,China;Instrument Factory,North China University of Technology,Tangshan 063210,China;College of Artificial Intelligence,North China University of Technology,Tangshan 063210,China)

机构地区：[1]华北理工大学电气工程学院,河北唐山063210 [2]华北理工大学智能仪器厂,河北唐山063210 [3]华北理工大学人工智能学院,河北唐山063210

出　　处：《中国测试》2024年第6期131-138,147,共9页China Measurement & Test

基　　金：国家自然科学基金(61502143)。

摘　　要：集中供热系统一次网热负荷预测对于换热站能源合理分配具有重大意义,针对换热站之间存在较强的耦合性,如何保证换热站实现节能减排的同时并保证热用户的舒适性是供热行业的首要任务,提出一种基于遗传粒子群的混合神经网络(GAPSO-CNN-BiLSTM)预测模型。首先,利用热负荷历史值、供水流量、供水温度以及回水压力构建模型输入;然后,利用完全噪声辅助聚合经验模态分解(complete ensemble empirical mode decomposition with adaptive noise,CEEMDAN)将供热负荷分解为不同子序列,以弱化供热负荷数据复杂程度,挖掘数据内部潜在特征;其次,为进一步减少模型计算时间,根据样本熵(sample entropy,SE)对子序列进行合并;最后,利用所提模型对不同子序列进行预测重构。实验表明所提模型相比LSTM、CNN-LSTM以及粒子群优化的混合神经网络(PSO-CNN-LSTM)在供热负荷预测中精度分别提高42%,32%,30%拥有更出色的特征提取能力和精度。The primary network heat load prediction of centralized heating system is of great significance for the reasonable energy distribution of heat exchange stations.In view of the strong coupling between heat exchange stations,how to ensure the energy saving and emission reduction of heat exchange stations while ensuring the comfort of heat users is the top priority of the heating industry,a hybrid neural network(GAPSO�CNN-BiLSTM)prediction model based on genetic particle swarm is proposed.First,the model input is constructed using the heat load history,water supply flow,water supply temperature,and return water pressure;then,the heat load is decomposed into different subseries decomposition with adaptive noise(CEEMDAN)to weaken the complexity of the heat load data and to explore the potential features within the data;second,to further reduce the model computation Secondly,in order to further reduce the model computation time,the subsequences are combined according to Sample Entropy(SE);finally,the proposed model is used to reconstruct the prediction of different subsequences.The experiments show that the proposed model has 42%,32%,and 30%better feature extraction ability and accuracy than LSTM,CNN-LSTM,and hybrid neural network with particle swarm optimization(PSO-CNN-LSTM)in heating load prediction,respectively.

关 键 词：完全噪声辅助聚合经验模态分解 样本熵 遗传粒子群 一次网 热负荷预测 

分 类 号：TB9[一般工业技术—计量学] TH16[机械工程—测试计量技术及仪器] TP391.9[自动化与计算机技术—计算机应用技术]