基于GRO优化的VMD-HKELM月蒸发量预测方法研究  

Research on VMD-HKELM Monthly Evaporation Prediction Based on Gold Rush Algorithm Optimization

作　　者：李菊[1] 崔东文 LI Ju;CUI Dongwen(College of Urban Construction,Yunnan Open University,Kunming 650500,China;Yunnan province Wenshan Water Bureau,Wenshan 663000,China)

机构地区：[1]云南开放大学城市建设学院,云南昆明650500 [2]云南省文山州水务局,云南文山663000

出　　处：《水文》2024年第5期25-31,共7页Journal of China Hydrology

基　　金：云南省教育厅教育科学研究基金(2023J0797);云南省教育厅科学研究基金(2024J0756)。

摘　　要：水面蒸发预测对于水库水量预测、区域水量平衡分析和水资源量核算等具有重要意义。水面蒸发量预测影响因素众多,并最终体现在随时间变化的蒸发量监测数据中。为此,基于淘金热(GRO)算法优化变分模态分解(VMD)-混合核极限学习机(HKELM)提出两种方案。方案Ⅰ先对月蒸发量时间序列分解,后划分训练集、测试集;方案Ⅱ先对月蒸发量划分训练集、测试集,再进行时间序列分解。通过一种新型元启发式算法对分解技术VMD、预测器HKELM超参数进行目标寻优并建立多种模型,采用云南省龙潭寨、西洋街水文站月蒸发量预测实例对方案Ⅰ、方案Ⅱ各模型进行检验。结果表明:方案Ⅰ各模型性能优于方案Ⅱ,各模型的拟合精度和预测精度总体上随分解分量数的增加而提高,但方案Ⅰ使用了测试集信息,导致预测精度虚高;方案Ⅱ各模型具有较好的预测精度和稳健性能,其用于月蒸发量时间序列预测是可行的,反映出客观真实的预测效果,具有较好的实用价值和意义。The prediction of water surface evaporation is of great significance for reservoir storage prediction,regional water balance analysis,and water resources accounting.There are numerous factors that affect the prediction of water surface evaporation,which ultimately manifest in evaporation monitoring data that vary over time.To address this,two schemes were proposed based on the Gold Rush Optimization(GRO) algorithm to optimize the VMD Hybrid Kernel Extreme Learning Machine(HKELM).Scheme Ⅰ decomposed the monthly evaporation data using time series analysis and then dividing it into training and testing sets.Scheme Ⅱ divided the monthly evaporation data into training and testing sets before applying time series decomposition.A novel meta-heuristic algorithm was used to optimize the hyper parameters of both the decomposition technique(VMD) and the predictor(HKELM),establishing multiple models.These models were tested using monthly evaporation data from Longtan village and Western Street hydrologic stations in Yunnan Province for both Scheme Ⅰ and Scheme Ⅱ.The results indicate that the models in Scheme Ⅰ outperform those in Scheme Ⅱ.Overall,the fitting and prediction accuracy of each model increase with the number of decomposition components.However,it should be noted that Scheme Ⅰ utilizes information from the testing set,leading to inflated prediction accuracy.In contrast,the models in Scheme Ⅱ demonstrate good prediction accuracy and robustness,making them feasible for predicting monthly evaporation time series.These models reflect objective and realistic prediction results,demonstrating their practical value and significance.

关 键 词：变分模态分解 淘金热优化算法 混合核极限学习机 超参数优化 月蒸发量预测 

分 类 号：P332.2[天文地球—水文科学] TV124[水利工程—水文学及水资源]