基于KPCA降维分析的特高拱坝监测模型

Monitoring model for super high arch dams based on KPCA dimension reduction analysis

为提高大坝变形预测精度,针对变形数据影响因子间的多重共线性问题,构建了基于核主成分分析(KPCA)、全局搜索策略的鲸鱼优化算法(GSWOA)和门控循环单元(GRU)的组合预测模型。首先利用KPCA对高维变形序列进行降维处理,同时使用GSWOA对GRU参数进行优化,进而构建出最优变形预测模型。以小湾特高拱坝变形数据为例,将KPCA-GSWOA-GRU模型与KPCA-WOA-GRU模型、PCA-GSWOA-GRU模型以及传统模型进行预测拟合对比。结果表明:KPCA-GSWOA-GRU模型有效降低了多重共线性问题,且在均方根误差(RMSE)、平均绝对误差(MAE)、平均绝对百分比误差(MAPE)和决定系数(R^(2))等方面均优于对比模型。

In order to improve the prediction accuracy of dam deformation,a prediction model based on kernel principal component analysis(KPCA),global search strategy whale optimization algorithm(GSWOA)and gated recurrent unit(GRU)was constructed to solve the multicollinearity problem among influence factors of deformation data.Firstly,KPCA was used to reduce the dimension of high-dimensional deformation sequence,and then GSWOA was used to optimize the GRU parameters,so the optimal deformation prediction model was constructed.Taking the deformation data of Xiaowan super high arch dam as an example,the prediction effect of KPCA-GSWOA-GRU model was compared with KPCA-WOA-GRU model,PCA-GSWOA-GRU model and traditional models.The results showed that the KPCA-GSWOA-GRU model not only effectively reduced the multicollinearity problem,but also outperformed the compared model in terms of root mean square error(RMSE),mean absolute error(MAE),mean absolute percentage error(MAPE)and coefficient of determination(R^(2)).The research results provide a theoretical basis and technical support for verifying the validity of KPCA-GSWOA-GRU model on a wider data set and its application in other dam deformation prediction in the future.