鲸鱼算法优化变分模态分解和支持向量机的日蒸发量预测模型  

Forecasting the Daily Evaporation by Using the Whale Algorithm Optimizes the Variational Modal Decomposition and Support Vector Machine

作　　者：付桐林[1] 杨明霞[1] FU Tong-lin;YANG Ming-xia(College of Mathematics and Information Engineering,Longdong University,Qingyang 745000,Gansu)

机构地区：[1]陇东学院数学与信息工程学院,甘肃庆阳745000

出　　处：《陇东学院学报》2025年第2期1-9,共9页Journal of Longdong University

基　　金：甘肃省自然科学基金项目“机器学习耦合启发式算法和数据预处理的不完备气象参数下干旱区蒸散量估算研究”(23JRRM734);甘肃省高等学校创新基金项目“基于机器学习的黄土高原人工植被区蒸散量估算研究”(2023B-211);庆阳市联合科研基金专项一般项目“机器学习耦合启发式算法和数据预处理的不完备气象参数下蒸散量估算研究”(QY-STK-2024A-068)。

摘　　要：准确预测蒸发量能够为干旱半干旱沙区水资源的有效利用提供依据。已有的基于时间序列预测建模的研究忽略了变分模态分解(VMD)参数的优化,导致预测精度有待提高。采用鲸鱼优化算法(WOA)优化变分模态分解(VMD)的两个参数,借助于VMD提取出蒸发量时间序列的主要变化特征,选用支持向量机SVM作为预测主模块,并采用WOA优化支持向量机(SVM)的超参数,构建了不完全数据下一种新的蒸发量预测模型VMD-WOA-SVM,对甘肃省环县北部沙区日蒸发量进行预测。结果表明,与混合模型DWT-WOA-SVM相比,VMD-WOA-SVM具有更高的计算精度,在预测阶段,Nash–Sutcliffe效率系数(NSCE)的平均值由0.9143增加到0.9154,平均绝对百分比误差(MAPE)的平均值由17.37%下降到16.92%。Accurate prediction of evaporation is the basis for effectively utilizing the limited water resources in arid and semiarid sandy areas.The previous forecasting model based on time series analysis neglects the optimization parameters of variational mode decomposition(VMD),and the prediction accuracy needs to be improved.In this study,the whale optimization algorithm(WOA)was adopted to optimize the two parameters of VMD,and the VMD with optimal parameters was employed to extract the main characteristics of evaporation time series,then,support vector machine(SVM)was selected as the main forecasting module,and the hyperparameters of SVM were also optimized by using WOA.The hybrid evaporation prediction model VMD-WOA-SVM with incomplete data was proposed to realize the accurate prediction of daily pan evaporation in the northern sandy area of Huan County,Gansu Province.The results show that VMD-WOA-SVM has higher computational accuracy than that of the hybrid model DWT-WOA-SVM,the average Nash-Sutcliffe coefficient of efficiency(NSCE)increases from 0.9143 to 0.9154,and the average of the mean absolute percentage error(MAPE)decreases from 17.37%to 16.92%in the prediction stage.

关 键 词：变分模态分解 支持向量机 鲸鱼算法 蒸发量 

分 类 号：O212.3[理学—概率论与数理统计] TP391.9[理学—数学]