Precipitation Prediction Based on ICEEMDAN-VMD and CNN-BiLSTM-AT Models: A Case Study in Lanzhou, Gansu

• Published in: 2023 IEEE International Conference on Sensors, Electronics and Computer Engineering (ICSECE) pp. 710 - 717
• Main Authors: Wang, Suichan, Meng, Panlong, Kong, Lingwang, Qin, Sanjie, Gui, Xiangquan
• Format: Conference Proceeding
• Language: English
• Published: IEEE 18-08-2023
• Subjects: Attention mechanism; Bidirectional long short-term memory; Biological system modeling; Computational modeling; Convolutional neural network; Precipitation; Precipitation forecasting; Predictive models; Spatio-temporal information; Time series analysis; Uncertainty; Urban areas
• DOI: 10.1109/ICSECE58870.2023.10263358

Accurate precipitation prediction is of vital importance to give healthy and sustainable development of agricultural production, ecological environment, economy, and society. The randomness and uncertainty of precipitation events make precipitation prediction challenging. Therefore, a novel hybrid deep learning architecture is proposed in this paper. The model first uses the improved complete ensemble empirical mode decomposition with adaptive noise (ICEEMDAN) and the Variational mode decomposition (VMD) decomposition method to denoise the input sequence to improve the accuracy and stability of the prediction. Then, a convolutional neural network (CNN) is used to perform local feature extraction for each intrinsic mode function (IMF) after decomposition, and the features extracted by the CNN are fed into a bidirectional long short-term memory network (BiLSTM) for sequence modeling. In addition, to further improve the performance of the model, an attention mechanism (AT) is used to dynamically adjust the weights of different time steps to better capture the important information in the sequences. Finally, it is applied to the precipitation prediction in Lanzhou City, and the results show that the proposed model has good prediction performance in precipitation prediction and outperforms other models.