基于Log-GMDH模型的我国能源消费中长期预测

Medium and Long-term Forecasting of Energy Consumption Based on Log-GMDH Model

利用Logistic函数作为GMDH两水平自回归算法的传递函数构建了新模型:Log-GMDH模型。运用我国1979～1999年的历史能源消费总量数据,将Log-GMDH模型在检测集(2000～2010年)上的预测结果与自回归移动平均(ARMA)模型和BP神经网络模型进行了比较,表明Log-GMDH模型有更准确和更稳定的预测效果。对我国未来30年(2011～2040年)的能源消费总量进行预测时,发现Log-GMDH模型更适合于反映我国新形势下可持续发展的能源战略。运用Log-GMDH模型的预测结果得到:我国未来能源消费先将有较大幅度的增长,到2030年总量将达62.55亿吨标准煤,之后能源消费将逐步得到较好的控制,预计将于2040年实现"零增长",届时全国能源消费总量约为65.70亿吨标准煤。

This paper provides a new model, Log-GMDH model, by introducing logistic function as the transfer function of autoregressive two-level algorithm of GMDH model. According to the historical data of Chinese total energy consumption from 1979 to 1999, back propagation （BP） neural network model, autoregressive moving average （ARMA） model and Log- GMDH model are compared on the test dataset （2000-2010）. The results show that the Log-GMDH model is more accurate and stable. It forecasts China＇ s energy consumption in the following 30 years （2011-2040） and the results of Log-GMDH fit the energy sustainable development strategy better than ARMA ＇ s. The results show that China＇ s energy consumption will keep a rapid growth. The total energy consumption of China will be 6255 million tons of standard coat in 2030, and then that number will be controlled in 6570 million tons of standard coal in 2040 when energy demand reaches the ＂zero growth＂.