Fault Detection Based on Complete Information Principal Component Analysis for Electric Submersible Pump

Abstract

Electric submersible pump (ESP) is the key production equipment for the offshore oil industry. As a result of poor working condition, fault and failures happen frequently and affect the production effectiveness. It is significant to detect the fault and failure in time. The fault detection for ESP mainly depends on the expert experience for decades. The data-driven fault detection methods are widely utilized in recent years with the rapid development of sensor technology, where the PCA is the most widely used methods. However, the data collected from the front line faces the imprecision problem, which reduces the accuracy of data-driven fault detection methods. The interval arithmetic is an effective way to handle imprecision data. However, traditional PCA and interval PCA algorithms cannot handle nonlinear problem. Aiming at this problem, the complete information principal component analysis methods are adopted for the fault detection of electric submersible pump in this paper. Compared with PCA and CPCA methods, the CIPCA method has better detection robustness when more inaccurate point value data.