Sliding Window-LSTM and VMD-LSTM Model for Yungang Grottoes Weather Predictions

This paper addresses the prediction of external environmental parameters at the Yungang Grottoes by proposing two deep learning models: 1) a sliding window-Long Short Term Memory (LSTM) model for air temperature and humidity forecasting, designed to capture temporal dependencies in meteorological data; and 2) a variational mode decomposition (VMD)-optimized LSTM model for wind speed prediction, which enhances accuracy through signal decomposition. Experimental validation using field monitoring data from July 1 to October 1, 2024, demonstrates that the proposed models achieve high-precision 30-minute ahead weather forecasting, providing a scientific basis for the preservation of grotto relics.