基于GA-BP神经网络大气中PM2.5软测量研究

摘要

Because Of the varying cOncentratiOn Of atmOspheric PM2. 5 have strOng nOnIinear characteristics,traditiOnaI sOft sensOr methOds are difficuIt tO make accurate measuring and mOnitOring. AccOrding tO traditiOnaI BP neuraI netwOrk is easy tO faII intO IOcaI minimum,BP neuraI netwOrk is cOmbined with genetic aIgOrithm tO estabIish the GA-BP neuraI netwOrk sOft sensOr mOdeI. The mOdeI is appIied tO the mOnitOring Of the atmOspheric cOncentratiOn Of PM2. 5,and cOmpared with the resuIts Of the mOnitOring Of the traditiOnaI BP neuraI netwOrk mOdeI,the resuIts shOw that the genetic aIgOrithm OptimizatiOn mOdeI has a better nOn-Iinear fitting abiIity and higher mOnitOring accuracy.%大气中PM2.5质量浓度变化具有较强的非线性特性，传统的软测量方法很难对其做出准确的计量监测。针对传统BP神经网络易陷入局部最小值的缺陷，将遗传算法和BP神经网络相结合建立了GA-BP神经网络软测量模型，将该模型应用到大气PM2.5质量浓度的计量监测中，并与传统BP神经网络模型的监测结果进行对比，结果表明经过遗传算法优化后的模型具有更好的非线性拟合能力和更高的监测精度。