Combining Physics and Machine Learning for Multimodal Virtual Flow Metering with Confidence

Abstract

Abstract This paper introduces a multimodal virtual flow meter (VFM) that merges physics-driven multiphase flow simulations with machine learning models to accurately estimate flow rates in oil and gas wells. The combining algorithm takes advantage of the confidence decay and historical performance factors to assign confidence and contribution weights to the base estimators and then aggregates their estimates to arrive at more accurate flow rate estimates. Furthermore, the proposed multimodal VFM provides an indication of the confidence level for each estimate based on the underlying agreement of the base estimates and the historical performance. The proposed VFM was tested in a 6 months online pilot in two oil wells. The proposed multimodal algorithm resulted in almost 50% improvements in performance compared to individual VFMs. The proposed robust multimodal approach can provide a complimentary benefit as an optimal VFM and reduce the overall system uncertainty. The developed VFM can be used for real-time production monitoring, verification and backup of physical meters, and well-test validation.