A PCA-LSTM neural network-integrated method for phreatic line prediction

In order to prevent dam ̄breaking accidents of tailings pondsꎬ to excavate effective information of online monitoring system and improve prediction accuracy of phreatic linesꎬ a prediction model was set up based on PCA and LSTM neural network. Thenꎬ with Chenkeng tailings pond as an exampleꎬ Pearson correlation coefficient and variable combination method were introduced to determine 18 features of model inputsꎬ including location of phreatic line of measuring point in the first three daysꎬ location of two adjacent surrounding saturation linesꎬ water level of pondsꎬ longitudinal displacement of dam body and rainfall. Finallyꎬ PCA was used to eliminate data redundancy between input variablesꎬ and LSTM neural 第 3 期 戴健非等: 集成 PCA 和 LSTM 神经网络的浸润线预测方法 network was applied to predict location of phreatic line for the next three days. The results show that PCA ̄ LSTM neural network ̄based method presents higher predication accuracy with an average absolute error of 0􀆰 011 and a decision coefficient of 0􀆰 805. And it can achieve stable prediction of phreatic lines for tailings ponds under different rainfall conditions.