Analytic Solutions Help Predict Sucker-Rod-Pump Failure

Abstract

This article, written by JPT Technology Editor Chris Carpenter, contains highlights of paper SPE 200825, “Failure Prediction for SRP Using Analytic Solutions,” by Viorica Sirghii, SPE, Rudolf Konrad Fruhwirth, and Herbert Hofstaetter, University of Leoben, et al. The paper has not been peer reviewed. A method for prediction of sucker-rod-pump (SRP) failure based on improved, completely connected perceptron (ICCP) artificial neural networks (ANNs) is presented. Results are compared with historical wellbore data. The validity of the ANN model is proved to meet SRP failure prediction with a ±5% error, helping identify and prevent potential damage, reduce costs and risks, and optimize production. SRP Diagnosis To evaluate the performance of SRPs, dynamometer cards are used. The load on the top rod is measured and plotted in relation to the polished rod position as the pumping unit moves through each stroke cycle. The result is a surface card that illustrates the polished rod load vs. position (downhole diagnostics). To determine above-pump liquid accumulation, pump-filling efficiency, and displacement, the well’s producing and lifting efficiencies can be analyzed and quantified. Depending on the results, operational and equipment changes can be made to improve efficiency and production optimization. How the plunger picks up, holds, and releases the fluid load at each stroke can be seen in a dynamometer card’s shape. The most-common and most-harmful pump states are fluid pound, gas interference, and tagging. When the pump is not completely full, the rod string is damaged because of an excessive number of strokes.