Advance ESP Failure Detection by Fuzzy Logic

Abstract

Electric Submersible Pump (ESP) is the most widely used artificial lift (AL) equipment across Sumatera Light Operation (SLO). At a time, ESP experiences failures due to different causes which impact its efficiency. ESP early pump-failure detection is a crucial step to minimize LPO. In the current fluctuating or even low oil price environment, executing the wells with pump efficiency issues becomes one of the economic jobs to optimize the production level. Challenges for this option are: how fast petroleum engineer can detect early pump failure symptoms? How many resources will be allocated for the daily review of a total 2,480 producers in SLO? The previous ESP failure detection tool has been established based on manual data feeding and hard cut-off criteria to generate an “exception signal”. Unfortunately, there were many ESPs which have longer ESP failure detection time that continuously sliding. In several cases, the wells are already down due to late being captured as pump failure. There is an opportunity to develop a more realtime/online data processing and analysis using artificial intelligence (AI) approach. This project has been developed with a more robust and more parameter integration, such as fluid test data, ampere data, fluid level, wellhead pressure and other supporting data, like job history. Fuzzy logic is implemented to assess the degree of potential pump failure. Online surveillance from SCADA is utilized to improve the data quantity and data quality on the fuzzy logic tool to be more like petroleum engineers for decision-making in ESP failure detection. The result of this project is presented as Fuzzy Confidence Index (FCI), in which the highest value of FCI correspondents to a higher potential of pump failure. ESP early failure detection output is broadcasted daily through “Integrated Exception Management Signals” (IEMS), providing daily potential pump failure wells to petroleum engineers for further process.