基于HHO-QRNN模型的大坝变形预测方法

Dam Deformation Prediction Method Based on HHO-QRNN Model

为有效利用大坝位移数据集中的真实信息,提高预测模型精准度,缩减建模分析训练时间,提出基于卡尔曼滤波算法、完全噪声辅助聚合经验模态分解和准循环神经网络的大坝位移预测方法。首先,模型采用卡尔曼滤波算法对原始输入数据进行处理,提取行有效信息,消除观测噪声影响;其次,设计一种信号分解算法,从累计位移值提取出趋势项、周期项和随机项数据集,以分离不同诱发因素对于大坝位移量的影响;最后,提出一种基于改进哈里斯鹰算法优化准循环神经网络的位移预测算法,对不同数据集分别采用此算法建模预测,将预测结果对应叠加得到最终预测结果。以某水库大坝的历史位移观测数据集为例,将所提模型与其他传统预测模型进行对比分析,结果表明,该模型预测精度和训练速度等方面均有显著提升,验证了其可行性和先进性。

In order to effectively utilize the real information in the dam displacement dataset,improve the accuracy of the prediction model,and reduce the training time required for modeling and analysis,this paper proposes a dam displace-ment prediction method based on Kalman filter algorithm,complete ensemble empirical mode decomposition with adap-tive noise and quasi-recursive neural network.Firstly,the model uses Kalman filtering algorithm to process the original input data for effective information extraction to eliminate the influence of observation noise.Secondly,a signal decompo-sition algorithm is designed to extract trend term,periodic term and random term data sets from cumulative displacement values to separate the influence of different inducing factors on the amount of dam displacement.Finally,a displacement prediction algorithm based on the improved Harris Hawk algorithm optimized quasi-recursive neural network is proposed,and the prediction results are superimposed on different data sets to obtain the final prediction results.Taking the histori-cal displacement observation data set of a reservoir dam as an example,the model of this paper is compared with other traditional prediction models.The results show that the prediction accuracy and training speed of this model are signifi-cantly improved,which verifies its feasibility and advancement.