摘要
针对用水量预测组合模型的优选问题,根据海河流域1988-2003年生活用水量时间序列的具体变化特征和拟合精度进行组合模型初选,并引入最优组合原理,确定组合权系数．所用的预测模型主要有回归模型、灰色模型和BP神经网络模型．以滦河分区和子牙河分区为例,计算的平均相对误差(MRE)分别为1．91％和1．54％, 2010年生活用水量将分别达到6．15×108m3和8．72×108 m3,2030年将分别达到7．37×108 m3和1.27×109 m3．结果表明,该方法提高了预测精度,可为水资源规划和管理提供依据．
In order to choose the optimal model for water demand forecasting, a combined forecasting method based on the optimal principle was put forward. According to the characteristics of domestic water consumption time series data from 1988 to 2003 and the fitting accuracy of the model, appropriate models were chosen, then based on the optimal principle the optimal weight coefficients were obtained, and finally a combined forecasting model was set up, which mainly included regression analysis model, grey model and BP neural network model. Two of the subareas were used as case studies. The mean relative error （MRE） in the Luanhe subarea is just 1.91%, and that in the Ziyahe subarea is only 1.54%. Domestic water consumption in the Luanhe subarea is 6.15 ×10^8m^3 in 2010 and 7.37 ×10^8m^3 in 2030, and that in the Ziyahe subarea is 8.72 ×10^8m^3 in 2010 and 1.27×10^9m^3 in 2030. The results show that the methodology improves prediction accuracy.
出处
《天津大学学报》
EI
CAS
CSCD
北大核心
2006年第6期745-750,共6页
Journal of Tianjin University(Science and Technology)
基金
天津市自然科学基金资助项目(043605611).
关键词
流域生活用水量
预测
最优组合原理
组合预测模型
海河流域
basin domestic water consumption
forecast
the optimal principle
combined forecasting model
the Haihe river basin
