Long-Term Time Series-Based Prediction of Mine Slope Displacement

Slope displacement is an important indicator for assessing slope stability, and predicting its future trend can provide early warning of abnormal conditions and prevent accidents from happening. Currently, slope displacement prediction models perform well in short-term predictions. However, their performance in long-term predictions is limited due to the inadequate exploitation of potential information in time series data. In this paper, we propose an improved slope displacement prediction model based on TimesNet. We design a Periodic Feature Extraction Network (PERBET) to extract features between different periods, and a Periodic Inter-scale Attention Network (PERIN) to enhance the model's feature extraction capability by using multiple scales and frequencies, thereby improving the ability to mine information from the data. We conduct experiments on multiple long-term public datasets and slope radar datasets, and the results show that our method outperforms existing models.