基于VMD-RIME-LSTM算法的天然气负荷预测  

Natural gas load forecasting based on VMD-RIME-LSTM algorithm

作　　者：张凯 高伟 刘晓磊[2] 孙旭 卜跃刚 张宏喜 ZHANG Kai;GAO Wei;LIU Xiaolei;SUN Xu;BU Yuegang;ZHANG Hongxi(Department of Energy Engineering,Hebei University of Architecture,Zhangjiakou 075132,China;Zhangjiakou Cigarette Factory Co.,Ltd.,Zhangjiakou 075000,China;Hebei Energy Storage Heating Technology Innovation Centre,Zhangjiakou 075132,China)

机构地区：[1]河北建筑工程学院能源工程系,河北张家口075132 [2]张家口卷烟厂有限责任公司,河北张家口075000 [3]河北省储能供热技术创新中心,河北张家口075132

出　　处：《区域供热》2025年第2期51-59,106,共10页District Heating

基　　金：河北省高等学校科学技术研究项目资助(QN2023215)。

摘　　要：针对某企业各用能端用能无序、多台燃气锅炉交互使用,天然气日用气负荷波动性大等问题,提出了一种变分模态分解(VMD)和霜冰优化算法(RIME)与长短期记忆神经网络(Long Short-Term Memory,LSTM)相耦合的天然气负荷预测模型。首先使用VMD对经过数据清洗的天然气负荷序列进行分解,将复杂的信号分解为若干个不同频率的模态分量(Intrinsic Mode Function,IMF);然后将各模态分量输入到经霜冰优化算法优化过的长短期记忆神经网络模型中进行预测,最后将各子序列预测结果叠加重构得到最终预测结果。实验结果表明:相比于单一长短期神经网络模型LSTM以及VMD-LSTM模型,这种VMD-RIME-LSTM模型在天然气负荷预测方面具有较好的预测精度,可为企业燃气锅炉系统实现更精确的运行管理和能源利用提供数据支撑。A natural gas load prediction model coupled with the Variational Modal Decomposition(VMD)and the Frost and Ice Optimization Algorithm(RIME)and the Long Short-Term Memory(LSTM)is proposed to address the problems of disordered energy use at each energy end of a certain enterprise,the interaction of multiple gas boilers,and the large fluctuation in the daily natural gas load.Firstly,the natural gas load sequence after data cleaning is decomposed by VMD,and the complex signal is decomposed into several Intrinsic Mode Functions(IMFs)with different frequencies;then each modal component is inputted into the Long Short-Term Memory neural network model optimized by RIME to make the prediction,and then the predictions of each sub-sequence are superimposed and reconstructed to obtain the final prediction result.Then,the modal components are input into the long and short-term memory neural network model optimized by the Rime-ice optimization algorithm algorithm.The experimental results show that the proposed VMD-RIME-LSTM model has better prediction accuracy in natural gas load prediction than the single long and short-term neural network model LSTM and VMD-LSTM model,and it can provide data support for the enterprise gas boiler system to realize more accurate operation management and energy utilization.

关 键 词：变分模态分解 霜冰优化算法 长短期记忆神经网络 天然气负荷 时序预测 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]