论文部分内容阅读
由于PM_(2.5)日均浓度值受外界多重复杂因素的影响,其较强的自相关性使得时间序列模型ARIMA构建难以实现,因此,给出高映射能力的非线性神经网络预测模型,并分别建立基于BP神经网络和GRNN神经网络的预测模型,进行PM_(2.5)浓度预测实验.结果表明,BP神经网络回检过程和检测过程存在不稳定性,预测残差波动较大,而GRNN神经网络检测残差呈完全U型,回检过程和检测过程较稳定,并且GRNN神经网络回检数据拟合度、预测数据精度和运算速度均优于BP神经网络,建模过程更为方便,易于实际应用.
Because the daily average PM_ (2.5) concentration is influenced by multiple complicated factors, its strong autocorrelation makes the time series model ARIMA difficult to be constructed. Therefore, a nonlinear neural network prediction model with high mapping capability is given and separately A prediction model based on BP neural network and GRNN neural network was established to predict the PM 2.5 concentration.The results show that the BP neural network has instability in the process of backtesting and testing and the prediction residual fluctuates greatly while the GRNN neural network The detection residuals are completely U-shaped, and the process of backtesting and testing are relatively stable. Moreover, the fitting degree of GRNN neural network test data, the accuracy of prediction data and the computing speed are better than BP neural network, and the modeling process is more convenient and practical application.