区域工业用水量非线性预测模型的优选

Preferred Nonlinear Model for Predicting Industrial Water Demand of District

针对BP神经网络在确定输入因子时的任意性,将相关性分析引入BP神经网络输入因子的选取中,通过计算输入因子和输出因子之间的相关系数,并根据相关程度来确定输入因子,同时利用遗传算法优化BP神经网络的权值和阈值的方法,建立了改进BP神经网络模型.将该模型应用于永定河山区工业用水量预测中,通过和传统非线性回归法进行比较,结果表明改进BP神经网络拟合和预测精度均较高.改进BP神经网络法的平均相对误差达到1.4100/,预测的2010年和2030年工业用水量将分别达到3.63×108m3和4.46×108m3.预测结果可为水资源规划和管理提供依据.

The improved BP neural network model was built, in which correlation analysis was adopted to choose input gene. Based on the correlation coefficient of input and output gene, the input gene was chosen. The genetic algorithm was used to optimize the weights and thresholds of BP neural network. The model was applied to predicting the industrial water demand of the mountainous area along Yongding River, and the predicted values were compared with the values by the traditional non-linear regress methods. The results indicated that both the fitting and prediction precision of the improved BP neural network method are better than that of the traditional non-linear regress methods, and the improved BP neural network method is feasible and effective for application. The mean relative error of the improved BP neural network method is 1.41%, the industrial water consumption is 3.63 × 10^8 m^3 in year 2010 and 4.46 × 10^8 m^3 in year 2030. The results are available in water resource planning and management.