[1]颜珂 彭星煜 刘小琨 张昆 张瑜春 穆卫巍 李富生.基于CEEMD-LSTM的短期天然气负荷预测[J].油气储运,2024,43(03):1-11.
YAN Ke,PENG Xingyu,LIU Xiaokun,et al.Short term natural gas load prediction based on CEEMD-LSTM[J].Oil & Gas Storage and Transportation,2024,43(03):1-11.
点击复制
《油气储运》[ISSN:1000-8241/CN:13-1093/TE]
卷:
43
期数:
2024年03期
页码:
1-11
栏目:
出版日期:
2024-03-25
- Title:
-
Short term natural gas load prediction based on CEEMD-LSTM
- 作者:
-
颜珂 彭星煜 刘小琨 张昆 张瑜春 穆卫巍 李富生
-
- Author(s):
-
YAN Ke; PENG Xingyu; LIU Xiaokun; ZHANG Kun; ZHANG Yuchun; MU Weiwei; LI Fusheng
-
-
- 关键词:
-
天然气负荷预测; 互补集合经验模态分解; 长短期记忆神经网络; 超参数优化
- Keywords:
-
Natural gas load forecasting; Complementary Ensemble Empirical Mode Decomposition(CEEMD); Long Short-Term Memory Network(LSTM); Hyperparameter optimization
- 分类号:
-
TE832
- 文献标志码:
-
A
- 摘要:
-
【目的】管网公司为保障稳定供气,需对下游用户的短期天然气负荷进行预测,但传统负荷预测方法存在拟合效果差,预测精度低等问题。为提高短期天然气负荷预测的精度,提出了一种基于互补集合经验模态分解(Complementary Ensemble Empirical Mode Decomposition, CEEMD)和长短期记忆神经网络(Long Short-Term Memory Network, LSTM)的短期天然气负荷组合预测模型。【方法】为充分挖掘负荷序列的内部藏特征信息,避免不同分量特征及额外噪声的相互干扰,通过CEEMD分解将原始负荷序列分解为有限个本征模态函数(Intrinsic Mode Function, IMF)分量,随后将不同IMF分量输入LSTM模型进行多步迭代预测,并基于贝叶斯算法对LSTM模型的超参数进行优化以提升学习效果和预测精度,最后将各分量的预测结果叠加重构得到最终预测结果。【结果】将不同组合模型预测结果进行对比,CEEMD-LSTM预测值与真实值误差更小,预测精度更高,贝叶斯调参进一步提升了预测精度,但预测耗费时间更长。【结论】相比其它预测模型,CEEMD-LSTM组合预测模型能够有效提取负荷序列的时序信息并消除非线性因素的影响,在抑制模态混叠的同时减小了重构误差,提升了预测精度,该方法可为天然气管网的调度管理和运行优化提供参考。
- Abstract:
-
【Purpose】 In order to ensure stable gas supply, pipeline companies need to predict the short-term natural gas load of downstream users. However, traditional load forecasting methods have problems such as poor fitting effect and low prediction accuracy. A short-term natural gas load combination prediction model based on Complementary Ensemble Empirical Mode Decomposition (CEEMD) and Long Short Term Memory Network (LSTM) is proposed to improve the accuracy of short-term natural gas load prediction.【Method】In order to fully explore the internal feature information of the load series, avoid the mutual interference of different component features and additional noise, the original load series is decomposed into a finite number of Intrinsic Mode Function (IMF) components through CEEMD decomposition, and then different IMF components are input into the LSTM model for multi-step iterative prediction, And based on the Bayesian algorithm, the hyperparameters of the LSTM model are optimized to improve learning effectiveness and prediction accuracy. Finally, the prediction results of each component are superimposed and reconstructed to obtain the final prediction result.【Result】 Using mean absolute percentage error (MAPE) and root mean square error (RMSE) as evaluation indicators for model prediction accuracy, the prediction results of different combination models were compared. The CEEMD-LSTM prediction value had a smaller error compared to the true value, resulting in higher prediction accuracy. Bayesian parameter tuning further improved prediction accuracy, but the prediction time was longer. 【Conclusion】 Compared with other prediction models, the CEEMD-LSTM combined prediction model can effectively extract the temporal information of load series and eliminate the influence of nonlinear factors. It reduces reconstruction errors while suppressing mode mixing and improves prediction accuracy. This method can provide reference for the scheduling management and operation optimization of natural gas pipeline networks.
相似文献/References:
[1]田文才,傅宗化,周国峰,等.基于WT+改进SSA-LSTM模型的短期天然气负荷预测算法[J].油气储运,2023,42(02):231.[doi:10.6047/j.issn.1000-8241.2023.02.013]
TIAN Wencai,FU Zonghua,ZHOU Guofeng,et al.Short-term natural gas load forecast algorithm based on WT+ improved SSA-LSTM model[J].Oil & Gas Storage and Transportation,2023,42(03):231.[doi:10.6047/j.issn.1000-8241.2023.02.013]
[2]颜珂,彭星煜,刘小琨,等.基于CEEMD-LSTM的短期天然气负荷预测模型[J].油气储运,2024,43(03):351.[doi:10.6047/j.issn.1000-8241.2024.03.012]
YAN Ke,PENG Xingyu,LIU Xiaokun,et al.Short-term natural gas load forecasting model based on CEEMD-LSTM[J].Oil & Gas Storage and Transportation,2024,43(03):351.[doi:10.6047/j.issn.1000-8241.2024.03.012]