基于EEMD-SSA-LSTM的PM_(2.5)预测模型  

PM_(2.5)Prediction Model Based on EEMD-SSA-LSTM

作　　者：连达军[1] 吴亚松 LIAN Da-jun;WU Ya-song(School of Geoscience and Surveying and Mapping Engineering,Suzhou University of Science and Technology,Suzhou Jiangsu 215009,China;School of Environmental Science and Engineering,Suzhou University of Science and Technology,Suzhou Jiangsu 215009,China)

机构地区：[1]苏州科技大学地理科学与测绘工程学院,江苏苏州215009 [2]苏州科技大学环境科学与工程学院,江苏苏州215009

出　　处：《计算机仿真》2024年第12期369-374,522,共7页Computer Simulation

基　　金：江苏省高等学校自然科学研究面上项目(22KJB420005)。

摘　　要：针对PM_(2.5)对空气污染的影响较大及预测难度较高的问题,提出了一种结合集合经验模态分解(EEMD)和麻雀搜索算法(SSA)对长短期记忆(LSTM)神经网络进行优化的PM_(2.5)浓度变化趋势预测模型。为了验证上述模型的预测效果,获取了苏州市11个空气监测站的逐小时的空气质量数据并利用门控制单元(GRU)算法对缺失的数据进行补充。在以上模型中,EEMD用于对PM_(2.5)的时序变化进行分解,SSA用于对LSTM的关键参数寻找最优组合,最后再利用LSTM对PM_(2.5)的变化趋势进行预测。实验结果表明,EEMD-SSA-LSTM预测模型在8小时和24小时的预测效果均优于单一LSTM、VMD-LSTM、EMD-SSA-LSTM三种模型。同时对比11个监测站的预测结果表明,该模型的预测效果具有普适性。Aiming at the high impact of PM_(2.5) on air pollution and the difficulty of prediction,this paper proposes a PM_(2.5) concentration trend prediction model combining the Ensemble Empirical Mode Decomposition(EEMD)and the Sparrow Search Algorithm(SSA)to optimize the Long-short-term memory(LSTM)neural network.In order to verify the prediction effect of the model,hour-by-hour air quality data from 11 air monitoring stations in Suzhou City were obtained and the missing data were supplemented using the Gate Control Unit(GRU)algorithm.In this model,EEMD is used to decompose the time-series changes of PM_(2.5),SSA is used to find the optimal combination of the key parameters of LSTM,and finally,LSTM is utilized to predict the trend of PM_(2.5).The experimental results show that the EEMD-SSA-LSTM prediction model outperforms the three models of single LSTM,VMD-LSTM,and EMD-SSALSTM at both 8 and 24 hours.Also comparing the prediction results of 11 monitoring stations showed that the prediction effect of the model is universal.

关 键 词：集合经验模态分解 麻雀搜索算法 长短期记忆神经网络 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]