Abstract:Predictions of the seepage field of underground water-sealed oil storage caverns (UWOCs) are significant for guiding the work of water curtain systems, ensuring the safety of oil storage operations, and reducing the operational cost of oil storage. Based on the field time-series monitoring data of a UWOC project, a long short-term memory (LSTM) model was used to predict the groundwater level of OH-4 and OH-5 and seepage pressure of the PA2-1 and PA3-1 monitoring points, and an error analysis of the prediction results was performed. The results showed that the LSTM model exhibited high prediction accuracy (relative error rate [ – 0.3%, 0.8%] and goodness of fit (R2) of four monitoring points are above 0.9). Simultaneously, by comparing the prediction results of the LSTM model with the generalized regression neural network and backpropagation neural network models, the LSTM model was found to perform better in terms of goodness of fit and prediction accuracy. The corresponding results provide a reference for the performance optimization of water curtain systems of UWOCs.

Seepage field prediction of underground water-sealed oil storage cavern based on long short-term memory model