基于EEMD-WOA-SVM的土石坝渗流量预测

EEMD-WOA-SVM Based Seepage Prediction for Earth and Rock Dams

为准确预报土石坝渗流量的变化趋势,针对传统的时间序列模型存在非线性处理能力较差、捕捉序列依赖关系能力不足等问题,建立了基于EEMD-WOA-SVM的土石坝渗流量预测模型.该模型采用集合经验模态(EEMD)有效分解土石坝渗流时间序列,引入鲸鱼优化算法(WOA)寻找支持向量机模型(SVM)的最优超参数组合,再将各模态分解分量代入组合模型预测并重构预测结果.案例分析结果表明,所建EEMD-WOA-SVM模型与传统的SVM模型相比,其拟合优度R2提升了19.8%,均方误差EMS、均方根误差ERMS、平均绝对误差EMA和平均绝对百分比误差EMAP分别降低了76%、50.3%、45.2%和43.2%.另外,与GA-SVM和WOA-SVM模型相比,R2值达0.9486,EMS、ERMS、EMA和EMAP分别降低至0.0012、0.0352、0.0289和0.0176,进一步说明了该组合模型具有较高的预测精度,为土石坝渗流量的精确预测提供了新途径.

In order to accurately forecast the trend of seepage flow from earth and rock dams,an earth and rock dam seepage flow prediction model based on EEMD-WOA-SVM was established to address the problems of poor nonlinear processing ability and insufficient ability to capture sequence dependencies in the traditional time series model.The model uses the ensemble empirical modal(EEMD)to effectively decompose the seepage time series of earth and rock dams,introduces the Whale Optimization Algorithm(WOA)to find the optimal hyper-parameter combination of the Support Vector Machine(SVM)model,and then substitutes the decomposed components of each modal into the combined model for prediction and reconstruction of the prediction results.The case study results show that the proposed EEMD-WOA-SVM model improves its R2 by 19.8%,and reduces the mean square error,root mean square error,mean absolute error,and mean absolute percentage error by 76%,50.3%,45.2%,and 43.2%,respectively,when compared with the traditional SVM model.In addition,compared with the GA-SVM and WOA-SVM models,the goodness-offit reached 0.9486,and the EMS,ERMS,EMA,and EMAP are reduced to 0.0012,0.0352,0.0289,and 0.0176,respectively.The combined model has high prediction accuracy and provides a new way for the accurate prediction of seepage flow from earth and rock dams.