论文部分内容阅读
为提高交通运输碳排放量的预测精度,根据交通运输碳排放量时间曲线具有的非线性饱和增长及随机性波动特点,建立基于Richards模型和BP神经网络的组合预测模型;以1985～2010年中国交通运输碳排放量数据为样本对模型进行了拟合和检验,并将Richards-BP神经网络组合模型预测结果与单项Logistic模型、GM(1,1)模型、Richards模型、BP神经网络及Logistic-BP神经网络组合模型、GM(1,1)-BP神经网络组合模型进行了误差对比分析。研究结果表明:3种组合模型的预测误差明显小于单一模型的预测误差,通过BP神经网络对单一预测模型进行误差修正可显著提高交通运输碳排放量预测精度;Richards-BP神经网络组合模型预测结果的平均绝对误差、平均绝对百分比误差及标准差值分别达到118.439×104 t、0.254%及136.915×104 t,比Lo-gistic-BP神经网络组合模型及GM(1,1)-BP神经网络组合模型精度提高了近5倍;以Richards模型的拟合误差作为BP神经网络输入效果要优于其他模型,Richards-BP神经网络组合模型具有更高的预测精度。
In order to improve the prediction accuracy of transport carbon emissions, a combined forecasting model based on Richards model and BP neural network is established according to the characteristics of non-linear saturation growth and stochastic volatility of transport carbon emissions time curve. From 1985 to 2010, China The carbon emissions data of transport were used to fit and test the model. The results of Richards-BP neural network combined model were compared with single Logistic model, GM (1,1) model, Richards model, BP neural network and Logistic- BP neural network combined model and GM (1,1) -BP neural network combined model were compared. The results show that the prediction error of the three combined models is obviously less than that of the single model, and the error correction of the single prediction model by BP neural network can significantly improve the prediction accuracy of transport carbon emissions. The prediction results of Richards-BP neural network combined model, The mean absolute error, mean absolute percentage error and standard deviation were 118.439 × 104 t, 0.254% and 136.915 × 104 t, respectively, which were higher than those of Lo-gistic-BP neural network and GM (1,1) -BP neural network The precision of the model is improved by nearly 5 times. The fitting error of Richards model is better than that of other models. The Richards-BP neural network model has higher prediction accuracy.