摘要
应用线性回归预测模型、人工神经网络预测模型、灰色系统预测模型等对天然气消费需求量进行预测后,又运用优化组合预测模型和变权重的自适应递推优化组合预测模型对天然气消费需求量进行了动态预测。分析预测效果发现变权重的自适应递推优化组合预测模型计算出的权系数,假定了“移动样本数据”的权重向量为常值,权重没有充分体现出数据的时效性。为此依据信息论“远小近大”的观点,首次引入时效函数对自适应优化组合预测模型进行改进。结果表明融入时效函数的自适应递推优化组合预测模型比采用的其它预测模型预测的结果更好。
The paper forecasts natural gas demand by applying linear regression forecast model,artificial neural network forecast model, grey system forecast model. To integrated utilize information offered by forecast models,avoid dropping useful information by a model, decrease randomness and increase accuracy, which the paper applies optimization combination forecast model and self-adapt optimization combination forecast model to forecast natural gas demand and obtains better results. But the paper finds weight coefficients of self-adapt optimization combination forecast model which is presumed a constant of moved sample data. The method does not embody forth time effect of data. Based far-smallness near-bigness theory of informa-tionism,the paper firstly import time effect function to improve model for natural gas demand selfadaption optimization combination forecast. As a result,it is suggested that the model is better than others.
出处
《油气储运》
CAS
北大核心
2005年第10期17-20,共4页
Oil & Gas Storage and Transportation
基金
四川省教育厅自然科学基金项目资助(编号:2004A150)。
