Physics-Augmented Data Driven Method to Detect and Quantify the Impact of Well Interference

Abstract

Abstract In unconventional reservoirs, well interference caused by hydraulic fracturing of a child well can significantly affect productivity of the parent well, leading to prolonged shut-ins and increased operational costs. Therefore, a reliable approach is needed to detect well interference, quantify impact, and forecast the resulting production profile. This study proposes a physics-based data-driven method for identifying well interference events and quantifying their effects on subsequent well production using minimal and easily accessible input data. In this work, we suggest a workflow for detecting well interference and quantifying its impact based on a transient well performance (TWP) methodology. The main steps in the workflow include computing the dynamic drainage volume (DDV) using the diffusive time of flight (DTOF) method, history matching on pressure and rates, calculating transient productivity index (PI), detecting breakpoints in PI evolution, and fitting a multi-segment TWP model to forecast production rates. We also implement physically consistent validation criteria based on directional variations in water-cut and pressure to determine if the anomaly is the result of well interference. The method is scalable and is run on a routine basis (daily) to pick up new events and automatically update the forecast for all wells in the field. The improved forecast is a result of two major effects – first, accounting for pressure variations as part of the TWP method that normalizes operational variations; and second, the ability to account for interference and extraneous influence to recalibrate well performance. We verified and validated the proposed method on multiple US onshore fields with extremely high accuracy. Both before and after well interference, the production forecast results were compared with actual production rates, and the results were highly accurate. Further, a root cause analysis was performed to find correlations between the interference results and potential driving factors. Our results indicate that not all interference events are detrimental to long term well performance. Some of the results generated by this method could be useful in understanding the mechanisms to predict or control future interference events. This study offers a systematic and computationally robust approach to identifying well interference events and quantifying their effects on subsequent well production using minimal and easily accessible input data. The proposed method provides a quick, distinct, and understandable physics-based data-driven approach that can be used as part of closed-loop reservoir management to scale up to the entire field.