西安市空气污染指数的神经网络预测模型

摘要

The BP(Back Propagation) neural network forecast method is introduced in air pollution prediction and the nonlinear time serial neural network prediction model for air pollution index in Xi'an is established. The historical data of API from 2005 to 2010 of Xi'an city was applied to neural network. The research results show: the average relative errors of API between forecating and the monitoring are separately 27.4 %,23.2 %, 12.2 %, 25.7 %. The corelation between results of forecating and the monitoring is very well. This model is applied to air pollution forecast with high precision and good generalization ability.%将BP神经网络模型引入到大气污染预测预报领域,利用西安市2005-2010年逐日空气污染指数,建立了一个空气污染指数的非线性时间序列神经网络模型.实验结果表明:独立样本的日空气污染指数的预测值的平均相对误差分别为27.4%,23.2%,12.2%,25.7%,独立样本的日空气污染指数的预测值与真实值的线性相关系数高.该模型应用于大气污染预报具有较高的预测精度和良好的泛化能力.