小波神经网络模型在河道流量水位预测中的应用

Application of RBF and GRNN neural network model in forecast of water runoff and head

鉴于BP神经网络学习收敛速度慢、参数选择困难、易陷入局部极值等缺点,提出小波神经网络河道流量水位预测模型,以盘龙河天保站流量水位预测为例进行分析。采用循环算法确定最佳BP神经网络结构,并在相同网络结构及期望误差等条件下,运用GA优化BP神经网络初始权值和阈值,构建传统BP、GA-BP神经网络河道流量水位预测模型作为对比预测模型。结果表明:小波神经网络结合了神经网络与小波分解在函数逼近上的优点,其预测精度高于传统BP和GA-BP网络模型,表明小波神经网络用于河道流量水位预测是合理可行和有效的,可为水文预测预报提供新的途径和方法。且小波神经网络模型具有计算简便、逼近能力强、收敛速度快,能有效避免局部极值等特点,有着广阔的应用前景。

Based on RBF and GRNN neural network algorithm, the paper constructed RBF and GRNN neural network water demand prediction model, the model was applied to river water demand prediction, and with the basic BP neural network model and the gray GM （ 1,1 ） water demand prediction model fitting, prediction results were compared and analyzed. The results showed that RBF and GRNN neural network model has higher fitting, prediction accuracy, the average relative errors are within 5%. The RBF and GRNN neural network model witch is applied to water demand prediction is reasonable and feasible, the model generalization capability, high precision, stable algorithm, and BP algorithm compared with GRNN, RBF network model also has the advantages of fast convergence speed, a few tuning parameters and is easy to avoid failing into local minimum and other advantages, has a good application prospects.