基于遗传算法优化的小波神经网络在地下水位预测中的应用  

Application of Optimized Wavelet Neural Network Based on Genetic Algorithm in Groundwater Level Prediction

作　　者：徐强[1] 束龙仓[1] 杨桂莲[2] 刘晋[2] 杨丹[2] 

机构地区：[1]河海大学水文水资源学院,江苏南京210098 [2]水利部地下水监测中心,北京100053

出　　处：《水文》2010年第1期27-30,共4页Journal of China Hydrology

基　　金：水利部公益性行业科研专项"京津冀地下水严重超采区地下水预测预报研究"(200801020);江苏省"青蓝工程"

摘　　要：地下水位是衡量生态环境优劣和地下水资源的一个重要指标。地下水位下降,将引发地面沉降、地面塌陷和降落漏斗等。因此,地下水位预测对保护地质生态环境和实现地下水资源严格管理至关重要。由于BP算法存在极易收敛于局部极小点与过拟合等缺点,导致网络泛化能力不足,本文在构建小波神经网络基础上并引入遗传算法加以优化,以解决上述不足,并与BP和WNN对比预测了天津市深层承压水水位。预测结果表明,GA-WNN模型拟合精度较高,模型的预测能力有较大幅度提高。Groundwater level is an important indicator in measuring ecological environment and groundwater resources. Decline of groundwater level will lead to land subsidence, surface collapse and depression cone. Therefore, groundwater level forecasting is essential for protection and improvement of geo-ecological environment and strict management of groundwater resources. BP algorithm has the shortcomings of easily converging to local minimum point and over-fitting, which will result in insufficient capacity of network generalization. In order to overcome the shortcomings, this paper built the wavelet neural network and introduced genetic algorithms to be optimized. And the deep confined groundwater level was predicted in comparison of BP with the wavelet neural network in Tianjin. The predicted results show that, GA-WNN model fitting with high precision, the model＇s predictive power has increased substantially.

关 键 词：遗传算法 小波 神经网络 地下水位 预测 

分 类 号：TV124[水利工程—水文学及水资源]