Flow Rate Estimation of Downhole Leaks Using Acoustic Sensor Array

Abstract Well integrity is becoming a key area of focus for most operators during the entire well life cycle. Acoustic conformance-monitoring systems have been developed to detect undesired leaks in terms of depth and radial location and to estimate leak flow rates, enabling operators to assess the severity of the leak and to plan remedial action. If the leak flow rate were substantial based on a given threshold, the remedial action would be scheduled with high priority; otherwise, repair work would be deferred to a suitable maintenance time. This paper presents a systematic statistical methodology to estimate leak flow rates using data acquired by an acoustic sensor array. The first stage of this methodology is data-level fusion through the beamforming technique, in which all signals acquired by an array of acoustic sensors are fused into a new signal. The second stage consists of feature-level fusion, in which representative and distinct features are extracted from the fused signal to construct the feature space. The third stage is decision-making, in which a statistical model is fitted in the feature space and then used to estimate flow rates in the sense of probability. A series of experiments was conducted in which a variety of leak flow rates were simulated with various combinations of orifice sizes and differential pressures in various leak scenarios. A low-dimension, third-order logistic regression model was demonstrated that can be fitted to the data and used to effectively separate small/large flow rates (90.0% specificity and 91.7% sensitivity).