Revolutionizing Reservoir Surveillance: A Deep Dive Into Virtual Well Testing Technology

Abstract

Abstract This paper analyses a five-year implementation of an innovative Virtual Well Testing (VWT) technology for reservoir surveillance. VWT enables continuous, real-time insight into reservoir performance and well productivity. The tool utilizes the power of artificial intelligence, making it a live asset management tool with higher accuracy and lower cost. This study showcases the outcomes of this technology and examines its functionality, complexity, limitations, advantages and challenges addressed in comparison to conventional well testing methods. The VWT project was initiated to develop a technology aimed at improving field's reservoir performance, characterization, and management, utilizing artificial-intelligence algorithms. VWT reduces costs and maximizes data availability - something that cannot be achieved practically with most conventional well testing methods. The tool runs on AI engine based on proprietary algorithm and technology. It integrates well testing data, metering, fluid dynamics, physics, mathematics, and AI to determine real time oil, water, and gas flow. Experimental models were designed and hypothesis testing was conducted using the following criteria: Verification against a physical test separator. Target 3 phase prediction accuracy above 80%. VWT has been deployed on 95 wells over a five-year period, achieving an overall accuracy level of over 80% compared to a 3-phase separator. It was observed that more than 50% of the tested wells demonstrated an accuracy exceeding 90% primarily due to their stable production profile. On the other hand, "chaotic" wells - those with frequent and abrupt changes in production rates and composition - exhibited lower accuracy in the beginning (around 75%). However, it was observed that with additional training, most of these wells were able to improve to over 80% accuracy. It was also observed that short-term inaccuracies in predictions were often linked to well interventions, downhole equipment and sensor failures - events that the system can detect and filter in real time. In summary, VWT has exhibited a consistent upward trend in accuracy since it was initially deployed. It also demonstrated a high repeatability as more wells are added in the system. Despite some challenges posed by erratic well behaviour, the VWT has been able to deliver acceptable results. Its continuous machine-learning feature running in the background allowed the AI engine to target a repeatable optimization accuracy of over 95% - surpassing the capabilities of most conventional systems. The VWT represents a transformational innovation in reservoir performance and management, going beyond the conventional methods limitations. It empowers petroleum engineers to make more informed business decisions that positively impact field production and financial performance. In addition to reducing cost and risks, and enabling simultaneous well testing, VWT maximizes the availability of critical data - achievements that conventional methods in the industry today cannot match.