基于完全集成经验模态分解和深度学习的PM2.5浓度预测模型
PM2.5 Concentration Prediction Model Based on Complete Ensemble Empirical Mode Decomposition and Deep Learning
陈学斌 1陈春晖1
作者信息
- 1. 福建理工大学 互联网经贸学院,福建 福州 350014
- 折叠
摘要
为应对PM2.5浓度预测精度不高的问题,提出了一种结合完全集成经验模态分解与Informer预测模型的方法.该方法以历史污染物数据为输入,通过完全集成经验模态分解技术将其分解为不同频率的本征模态函数(IMFs),并重构序列.重构后的序列输入Informer模型,以捕捉长期依赖性并建模影响因子间的复杂非线性关系,从而提升预测准确性.对空气污染物数据进行训练、验证和测试后,结果表明该方法预测指标最优,预测精度显著提升.
Abstract
This research proposes a novel approach based on Complete Ensemble Empirical Mode Decomposi-tion and the Informer forecasting model to address the issue of low prediction accuracy in existing PM2.5 concentra-tion forecasting models.Historical pollutant data are used as input,which are decomposed into Intrinsic Mode Functions(IMFs)of different frequencies using Complete Ensemble Empirical Mode Decomposition,and the se-quences are reconstructed.These reconstructed sequences are then fed into the Informer model to capture long-term dependencies within the input sequence and model the complex nonlinear relationships among influencing factors,thereby improving prediction accuracy.The results show that the reconstructed model achieves the best performance metrics and significantly improves the prediction accuracy after training,validating,and testing the air pollutant da-ta.
关键词
PM2.5浓度预测/Informer/时间序列Key words
PM2.5 concentration prediction/Informer/time series引用本文复制引用
出版年
2024
NETL
NSTL
万方数据