Process monitoring for tower pumping units under variable operational conditions: From an integrated multitasking perspective

Abstract

Accurately monitoring the safe operation and dynamometer diagram inference of tower pumping units is crucial for process monitoring on drilling platforms. This paper proposes an integrated multitasking intelligent tower pumping unit process monitoring scheme facing variable operational conditions. The scheme proposes an unsupervised fault detection approach utilizing a multi-head self-attention mechanism neural network with a modified denoising autoencoder for tower pumping units without faulty data. The network robustness and reconstruction ability are enhanced through a multi-head attention mechanism layer added to the bottleneck layer, thereby effectively accomplishing the fault detection task. Furthermore, the scheme establishes the mapping relationship between electrical parameters and corresponding operational conditions of tower pumping units through a learning-based algorithm, which enables operational condition identification under variable conditions. Moreover, the scheme proposes a dynamometer diagram inference approach for tower pumping units under variable conditions, which accurately estimates the suspended load and displacement, to achieve an efficient inference process. The effectiveness of the proposed integrated multitasking intelligent tower pumping unit process monitoring scheme is validated through the real-world data provided by the Daqing Petroleum Institute.