自我调节蚁群-RBF神经网络模型在短期径流预测中的应用  

Short-term runoff prediction based on adaptive regulation ant colony system and radial basis function neural network hybrid algorithm

作　　者：白继中[1,2] 师彪[1] 冯民权[1] 周利坤[1,3] 李小龙[1] 

机构地区：[1]西安理工大学水利水电学院,西安710048 [2]山西水利职业技术学院,运城044004 [3]武警工程学院,西安710086

出　　处：《水力发电学报》2011年第3期50-56,共7页Journal of Hydroelectric Engineering

基　　金：国家火炬计划基金(07C26213711606);山西省水利厅科技计划基金(2009WK110)

摘　　要：为提高短期径流预测精度,提出了自适应调节人工蚁群算法(ARACS),对RBF神经网络参数进行优化,建立了自适应调节蚁群-RBF神经网络组合算法(ARACS-RBF)预测模型,综合考虑气象、天气、季节、降雨等影响因素,对上马水库进行径流预测。仿真表明,该方法克服了RBF神经网络和人工蚁群算法易陷于局部极值、搜索质量差和精度不高的缺点,改善了RBF神经网络的泛化能力,收敛速度快,输出稳定性好,提高了短期径流预测的精度,预测相对误差小于3%。可有效用于短期径流预测。Short-term runoff prediction is a major responsibility of water conservancy departments.To improve the accuracy of reservoir short-term runoff forecast,an adaptive regulation ant colony system（ARACS） algorithm is proposed.A forecast model was developed by using this new algorithm in combination with a radial basis function（RBF） neural network that was trained by using ARACS and an ARACS-RBF hybrid algorithm was obtained.This model can automatically determine the parameters of the neural network from the data sample.Predictions of the short-term runoff of practical reservoirs were made using the model with a comprehensive consideration of impacting factors such as meteorology,weather,rainfall and seasonal change.The result shows a faster convergence and a better forecast accuracy of the hybrid method than those of the traditional ant colony system algorithm-RBF neural network or RBF neural network,and also a significant improvement on the generalization capacity of RBF neural network.An average percentage error of no more than 3% was achieved.Thus,the hybrid algorithm enhances the efficiency of short-term load forecast of the reservoir and river.

关 键 词：水文学 ARACS-RBF神经网络模型 自适应调节蚁群算法 短期径流预测 RBF神经网络 

分 类 号：TM614[电气工程—电力系统及自动化]