摘要
利用4个基准函数对HBBO进行仿真测试;采用WT分解处理年径流时序数据;通过突变点检测方法Mann-Kendal(M-K)划分训练、预测样本,构建RBF适应度函数,利用HBBO优化RBF神经网络输出层权值、基函数中心和隐含层节点宽度,建立WT-HBBO-RBF模型,并构建WT-HBBO-SVM、WT-HBBO-BP、WT-RBF、WT-SVM、WT-BP、HBBO-RBF、HBBO-SVM、HBBO-BP作为对比分析模型。以云南省龙潭站、落却站年径流时间序列预测实例对模型进行验证的结果表明,HBBO具有较好的寻优精度及全局搜索能力;WT-HBBO-RBF模型对龙潭站、落却站年径流时间序列预测误差小于其他对比模型,具有较好的预测精度和泛化能力;HBBO能有效优化RBF神经网络输出层权值、基函数中心和隐含层节点宽度,提高RBF神经网络预测性能;WT能科学降低径流序列的复杂性,提高预测精度。
The human behavior-based optimization(HBBO)algorithm is simulated with four benchmark functions.The wavelet transform(WT)decomposition is used to process annual runoff time series data.By using Mann Kendal(M-K),a mutation point detection method,to divide training and prediction samples,the radial basis function(RBF)fitness function is constructed.The HBBO is used to optimize RBF neural network output layer weight,basis function center and hidden layer node width,and then the WT-HBBO-RBF model is established.The WT-HBBO-SVM,WT-HBBO-BP,WT-RBF,WT-SVM,WT-BP,HBBO-RBF,HBBO-SVM,HBBO-BP models are also established as comparative analysis.The annual runoff time series of Longtan and Luoque hydrological stations in Yunnan Province are used as prediction examples to verify the model.The results show that,(a)the HBBO has better optimization accuracy and global search capability;(b)the WT-HBBO-RBF model has better prediction accuracy and generalization ability because its prediction error of annual runoff time series of Longtan and Luoque stations is smaller than that of other comparison models;(c)the HBBO can effectively optimize RBF neural network output layer weight,basis function center and hidden layer node width,and improve RBF neural network prediction performance;and(d)the WT can scientifically reduce the complexity of runoff sequence and improve the prediction accuracy.
作者
徐成贵
崔东文
XU Chenggui;CUI Dongwen(Jingdong County Reservoir Management Center,Puer City,Puer 676200,Yunnan,China;Yunnan Province Wenshan Water Bureau,Wenshan 663000,Yunnan,China)
出处
《水力发电》
CAS
2023年第4期17-22,95,共7页
Water Power
基金
云南省创新团队建设专项(2018HC024)
云南重点研发计划(科技入滇专项)
国家澜湄合作基金项目(2018-1177-02)。
