基于深度学习的PM2.5预测模型建立

Establishment of PM2.5 Forecasting Model Based on Deep Learning

本文以京津冀地区为例,选择大气气溶胶光学厚度AOD(Aerosol Optical Depth)和气象参数为影响因子,建立基于深度置信网络DBN(Deep Belief Nets)的PM2.5预测模型,对PM2.5进行有效预测,并与BP神经网络预测结果对比,最后形成整个京津冀地区的PM2.5预测专题图。实验结果表明基于深度学习的置信网络对PM2.5浓度预测效果比BP神经网络更佳,预测精度有较大提高。

In this paper,the aerosol optical depth（AOD）and the meteorological parameters were selected to predict the concentration of PM2.5 base on Deep Belief Nets（DBN）in Beijng-Tianjin-Hebei region,and compared the result with BP networks.Finally the PM2.5 prediction map of whole Beijing-Tianjin-Hebei region is formed.The results showed that the prediction of PM2.5 concentration was more exact than the results of BP networks,the accuracy had greatly improved compared with BP networks.