基于人工智能算法的上海市生活垃圾产量与组分预测研究

打开文本图片集

关键词：人工智能；生活垃圾；产量预测；组分拟合

中图分类号：X705 文献标识码：A 文章编号：1008-9500（2026）02-0092-05

DOI：10.3969/j.issn.1008-9500.2026.02.027

Research on Prediction of Urban Waste Yield and Composition in Shanghai Based on Artificial Intelligence Algorithms

HE Chao

（Shanghai Environment Group Co.，Ltd.，Shanghai2OO12O，China）

Abstract：The yield and compositionof urban domestic solid waste arecore fundamental dataforthe planing of environmentalsanitationfacilities，transportationscheduling，andresourceutilization.Taking Shanghaiastheresearch object，thisstudyconstructsamulti-algorithmintegratedartificialinteligencepredictionmodelbyintegratingdatafrom the 2024 Shanghai Statistical Yearbook，measureddata fromthe greening and urban appearance management department， and socio-economic panel data.The model conducts time-series predictionand composition fiting forthedaily average transportationvolumeofdomesticwaste，theproportionofclasifiedcomponents，andphysicalandchemicalproperties.Four algorithms—LongShort-TermMemory（LSTM），SupportVectorRegresion （SVR），eXtremeGradientBoosting （GBoost）， and BackPropagation Neural Network（BPNN）—areselected forcomparativeexperiments.Multiplecharacteristic factors such as populationsize，residents’consumption，temperatureand precipitation，andtheimplementationdegreof waste clasificationpoliciesareintroduced，andthegeneralizationabilityofthemodelisimproved throughfeatureengineering andhyperparameteroptimization.TheresultsshowthattheintegratedLSTM-XGBoosthybridmodelachievesacoefcient of determination R2 of 0.978 in yield prediction，and the mean absolute error of component proportion prediction is less than （20 2.8% ，which issuperior to traditional time-seriesmodelsand single machinelearning models.This modelcanrealize highprecisionshort-term（weekly/monthly）and medium-term （annual） predictions，providingdata supportanddecision-making basisfortheexpansionofShanghai'sdomestic wasteterminaltreatmentfacilities，optimizationof transportationroutes， improvement of resource utilization efficiency，and realization of carbon emission reduction goals. Keywords： artificial intelligence; domestic waste; yield prediction; composition fitting

随着超大城市常住人口集聚、消费结构升级与垃圾分类政策全面落地，生活垃圾产生呈现总量稳中有调、组分动态分化、时空分布不均的特征。（剩余5351字）