Application of TTSSA-CNN Algorithm in Fault Diagnosis of Pumping Unit

In the current oil extraction process, frequent failures of beam pumping units seriously affect the economic benefits of the oilfield. This paper proposes a Convolutional Neural Network condition diagnosis algorithm based on an improved Sparrow Search Algorithm (TTSSA-CNN). Firstly, considering the good distribution, randomness, and regularity of Tent chaotic sequences, the initialization population diversity of the algorithm is improved, and the number of sparrow swarm optimization iterations is used as a degree-of-freedom parameter of the t-distribution to enhance the ability of SSA to jump out of the local search space and to improve the local optimization capability and accuracy of SSA. Finally, TTSSA is employed to optimize various hyperparameters of the CNN pumping unit fault diagnosis model, thereby improving training speed and accuracy. Experimental results indicate that the working condition diagnosis rate of the TTSSA-CNN algorithm is 98.7%, which represents a significant increase compared to three other diagnosis algorithms (BP, CNN, LSTM) by 13.7%, 4%, and 5.2%, respectively. It exhibits higher adaptability and diagnostic accuracy, holding practical and theoretical significance for pumping unit fault diagnosis.