Evaluation of automated pressure transient detection methods for efficient oil well management

One of the main causes of production decline in oil wells is the occurrence of formation damage. A major tool for monitoring such phenomena is Pressure Transient Analysis (PTA), applied when the well is closed (shut-in periods). With the advent of Permanent Down-hole Gauges (PDGs), this analysis can now be conducted at any time during oil production; however, the substantial volume of available data makes this task challenging and time-consuming. Therefore, several studies have focused on automating steps of PTA. This paper addresses the automation of the first step, transient detection, and presents a comparative study of six methods proposed in the literature based on signal processing techniques. The main contribution of this paper is an exploratory study intended to identify relevant aspects and patterns concerning the practical application of the methods on extensive and real-life datasets. The experimental evaluation of the methods was conducted using four real-world datasets provided by Petrobras S.A. for this research. Additionally, two synthetic datasets were introduced to facilitate a better interpretation of the findings. According to the experiments conducted, the method that achieved the best results was the convolution filter, proving to be a promising tool for transient and shut-in detection.