基于修正组合模型的青海省城市需水量预测

根据城市用水量系统具有非线性和随机波动性的特点,为了充分发挥组合灰色神经网络预测模型能够综合单变量预测及非线性处理的优势,同时降低组合权系数计算方法的不确定性对模型预测效果的影响,论文提出了基于马尔科夫链修正的组合灰色神经网络预测模型。将其应用于1980—2009年青海省城市用水量序列的拟合分析,并预测其2010、2015以及2020年的城市需水量。结果表明:基于马尔科夫链修正的组合灰色神经网络预测模型预测结果的误差更小,精度更高。

Forecast of Qinghai Urban Water Demand Based on Amending Combination Model

Urban water demand system is always nonlinear and stochastic,although combination model based on Grey Model and BP Neural Network has strength both on single variable forecasting and nonlinear problem,its weight coefficient calculating is uncertain and would affect forecasting result.Therefore,a combination model based on Grey Model and BP Neural Network and amended by Markov chain is set up to forecast Qinghai eight cities’water demand.The water consumption of Qinghai from 1980 to 2009 is used to verify this amended combination model and forecast eight cities’water demand in 2010,2015 and 2020.The forecasting result shows that the combination model based on Grey Model and BP Neural Network and amended by Markov chain has smaller deviation and higher forecasting precision accuracy.