基于混合算法的径流预报模型研究  

Neural network model based on hybrid algorithms and its application in runoff for ecasting

作　　者：蒋林利[1] 吴建生[2] 曾文华[1] 

机构地区：[1]厦门大学软件学院,福建厦门361005 [2]武汉理工大学信息工程学院,湖北武汉430070

出　　处：《广西大学学报（自然科学版）》2014年第2期351-357,共7页Journal of Guangxi University（Natural Science Edition）

基　　金：国家自然科学基金资助项目(61202143;51065002;11161029)

摘　　要：为了提高径流预报的精度,采用一种基于粒子群和遗传的混合方法同时优化人工神经网络结构、连接权和偏置,在进化过程中采用训练样本和验证样本共享适应度技术,并以此建立径流预报模型。通过对柳州径流实例分析,并与离子群优化的人工神经网络模型、遗传进化的人工神经网络模型和时间序列模型方法对比,研究结果表明,该方法学习能力强、泛化性能高和有效提高系统预测的准确率,为获得更高预测精度的径流预报提供了一种有效的建模方法。In order to improve the accuracy of runoff forecasting, a hybrid algorithm combining PSO and GA algorithm with optimizing artificial neural network structure, connection weights and bias was proposed and used to establish a runoff forecasting model. This hybrid algorithm adopts training samples and validation samples to share fitness in the evolutionary process. The algorithm was com-pared with two forecasting models including PSO-ANN and GA-ANN through the actual examples of Liuzhou runoff forecasting. The results show that the new approach has strong learning ability and high generalization performance and can improve the accuracy of forecasting system effectively. Thus, it is an effective modeling method to get high precision of runoff forecasting.

关 键 词：人工神经网络 粒子群 遗传算法 径流预报 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]