Water Level Prediction Of Pumping Stations Based On The Trpe-Informer Algorithm

Water level data in pump stations can optimize their operation, and precise water level prediction is crucial for the safe operation of pump stations. Existing pump station water level prediction methods often have low training efficiency and prediction accuracy, and they struggle to analyze data based on the correlation between different features. We propose a water level prediction model called Trpe-Informer to address these issues. Introducing a Temporal Convolutional Network (TCN) allows for more precise feature extraction from pump station data by expanding the receptive field. These feature data are fed into the Informer model, which uses relative position encoding. Rotational matrix relative encoding captures the correlations between feature data more accurately, improving the model's prediction accuracy and training speed. Experiments using the Jinhu Pump Station dataset from the South-to-North Water Diversion Eastern Route Project demonstrate that the proposed method achieves high prediction accuracy, providing a reliable basis for pump station operations.