Abstract
Waterflood management in large mature fields is often time intensive due challenges in integrating and analyzing large volumes of data. Maintaining an updated dynamic model may not be practical for day to day decisions and as such data driven analysis becomes the preferred approach. The conventional workflows usually rely on geometry-based allocation factors of injected water and is not readily integrated with other data sources such as injection-production trend correlations, cased hole logs, pressure and production chemistry data. This paper presents a case study of advanced analytics application to a mature waterflood field where available data was rapidly integrated into an integrated visualization dashboard and machine learning was used to identify injector producer connectivity and allocation factors.
This study was carried out for a mature waterflood field with over 15 years of water injection history and over 100 active producers. Python programming was used to clean up and integrate various data sources into an integrated visualization dashboard. Following attributes were identified as indications of producer injector connectivity: (a) Correlation in injection water salinity and produced water salinity (b) Correlation in injection rate trends and produced liquid rate trends, (c) Clear jump in producer ESP intake pressure trends as a sign of response from injection. Machine learning was used to cluster producers based on their produced water salinity trends which enabled shortlisting of injectors potentially connected to them. Another machine learning algorithm was then used to estimate connectivity factors between producers and injectors based on their distance and correlation in injection-production trends. The integrated dashboard was used to quality check results against other data sources e.g. trends in ESP intake pressure, PLT and RFT. Also, for each injector, water utilization factor was calculated based on correlation in cumulative water injection vs cumulative oil production in neighboring producers over last 2 years.
The study was completed within a tight timeframe of 5 weeks. The results comprised of an injector-producer connectivity map and each injector’s water utilization factor, which formed the basis for re-evaluation of the existing injection water allocation strategy. The injection targets of individual injectors were revised within their operating constraints by prioritizing them based on their connectivity and water utilization factors, aiming at an estimated gain of 5% in reservoir oil rate.
This paper demonstrates the potential of advanced analytics in unlocking valuable insights from various overlooked data sources. For example, in waterflood fields with multiple injection water sources, the contrast in their chemical compositions and aquifer water may partially serve as tracers providing useful information on reservoir connectivity. This paper also serves as a practical example of how digitalization can be adopted in subsurface community’s ways of working and hence be part of the ongoing journey of digital transformation in oil and gas industry.
