Intelligent Recognition of Sand Plugging Accidents in Fracturing Based on Time Series Feature Enhancement

Fracturing technology, one of the key technologies to increase oil and gas production, is now widely used in the oil and gas recovery industry. However, during large-scale fracturing operations in shale gas horizontal wells, various downhole accidents often occur, especially sand plugging accidents. These accidents not only seriously affect the safety of the operation but also lead to a significant increase in fracturing costs. Due to the randomness of the abnormal changes in the fracturing curve and the substantial hysteresis when the sand-carrying fluid enters the formation, the accident characteristics of the curve are not prominent, which often leads to false or late reports of the accident by manual judgment. Firstly, based on the curve characteristics of fracturing construction data, the relationship between the long time series affected by the discharge fluctuation and the type of fracturing stage is investigated, and a stage identification model based on the LSTM-FCN-Attention Network is established, which is integrated with the standard deviation-based discharge fluctuation monitoring algorithm to realize the delineation of the sensitive stage of sand plugging in the fracturing of a single-stage well. Then, based on the temporal characteristics of oil pressure and discharge, the time series to matrix mapping method and Gramian Angular Field (GAF) were used to construct the 3D data and fused by Convolutional Block Attention Module (CBAM) DenseNet121 to construct a CBAM-DenseNet121-based oil pressure anomaly state recognition model to recognize the accidents in the sensitive phase of sand plugging. The experimental results verified that the accuracy of the sand plug recognition method proposed in this paper reached 93.65%, and the misclassification rate was lower than 3.50%. In practical application scenarios, this method can effectively recognize accidents and remind construction personnel to make further corresponding decisions to avoid serious losses caused by accidents.