基于复合权重的天然气需求组合预测模型的构建

Coweight-based combination forecast model of natural-gas demand and its application

为了提高非线性复杂系统预测的精度,采用多维信息的复合权重组合预测模型,以期达到组合预测进一步降低误差的目的。首先,利用误差平方和倒数法、加权平均法、熵值法将灰色GM(1,N)模型与多元线性回归模型进行组合,构成3种组合模型。然后,采用灰色关联分析法将3种组合模型进行融合,构建复合权重组合预测模型。最后,以中国城市天然气需求量预测为例进行验证分析。研究结果表明:①组合预测比单一预测模型的预测结果更准确,避免了单一预测模型存在的预测风险;②复合权重组合预测模型较组合预测模型的误差更低,提高了预测精度;③到2025年,中国城市天然气需求量将达到2884×10^(8) m^(3),平均增速约为10%。结论认为,复合权重的组合预测模型在非线性复杂系统中具有良好的预测性能,对城市天然气需求量的预测结果可作为制定城市燃气政策的依据。预测结果表明,未来中国城市天然气需求量仍呈高速增长趋势。

In order to improve an accuracy to forecast a non-linear complex system,a new composite weight(so-called co⁃weight)combination forecast model based on multi-dimensional information was proposed to achieve one goal of decreasing the sum of squared error(SSE)of combination forecast further.Firstly,one gray GM(1,N)model and another multiple lin⁃ear regression model were combined to constitute three combination models by means of SSE reciprocal method,weighted average method,and entropy value method.Then,these three combination models were integrated to construct the coweight combination forecast model by means of the gray relation analysis method.Finally,it was verified by taking China's urban natural-gas demand forecast as the example.Results show that(1)the combination models can forecast more accurately than single forecast model and avoid some forecast risks in single model;(2)compared with the combination models,the coweight model has lower SSE and higher forecast accuracy;and(3)China's urban natural-gas demand in 2025 will be up to 288.4×10^(9) m^(3) and the average growth rate is about 10%.In conclusion,this coweight combination forecast model has good forecast performance in the non-linear complex system,and its forecast results can be used as the basis to formulate urban gas policies.What's more,China's urban natural-gas demand in the future will still present the trend of rapid growth.