基于季节趋势分解的 PM2.5 浓度混合预测模型
打开文本图片集
关键词： PM2.5 浓度预测；季节趋势分解；自回归移动平均模型；长短期记忆网络；支持向量机□图分类号：P9 文献标志码：A 文章编号：0253-2395（2025）04-0829-10
Abstract： To improve the accuracy of PM2.5 concentration prediction， this paper introduces a hybrid time series forecasting model withseasonal-trendecomposition（ybrid-X12）.Firstlytesasonal-tredecompositioalgorithmperfosthetaskofdecoposing PM2.5 time series intotrend-cycle，sasonaland iregularsub-sries; Then，ARIMA （Autoregressive IntegratedMovingAverage Model），LSTM（Long Short Term Memory），andSVM（Support Vector Machine）are appliedtotheabove sub-series prediction tasksespectively;Finalytefialpredictionresultcomesfromtheintegrationofthepredictedresutsofsubseries.Thesulation experiment selected PM2.5 monthly concentration fromsix major cities in North China and used MAE （MeanAbsolute Error），RMSE （Rot Mean SquareEror），andIA（IndexofAgreement）asmodelevaluationindicators.Theexperimentalresultsemonstratedthat thehybridpredictionsystemcansignificantlyenhancepredictionacuracy.Compared withthetraditionalsingle modelARIMA， LSTMand SVM，the MAE of the proposed model in Beijing is reduced by 18.72% 60.14% and 43.15% ，respectively. This verifies thattheseasonal-trenddecompositionalgorithishelpfulforminngseasonal-trendinformationintimeseries.Itcanbeocluded thatselectingappropriatealgoritmsforsubseriesithdiferentcharacteristicsensuresthefullutlzationoftheadantagesofdif ferent models， providing new ideas for PM2.5 concentration prediction.
Key words： PM2.5 concentration prediction； seasonal trend decomposition;autoregressive integrated moving average model; long short term memory; support vector machine
0 引言
城市化和工业化的快速发展带来了日益严重的大气污染问题，特别是北方地区冬季采暖及气象条件的叠加往往导致污染事件频发，严重影响人民的身体健康和生产的有序进行[1-2]。（剩余12216字）
