Development of Hybrid Proxy Model for Single Platform Gas Lift Optimization

Production optimizations require production rate data but currently only available from well test or installing costly multiphase flowmeter. Hence this paper present an approach which combines physics knowledge with machine learning for real-time prediction of liquid rates in individual wells Specifically, it leverage on an easy approach to combine data-driven approach and physics-driven approach by developing a proxy hybrid model [ 1 ] for the well. The hybrid proxy model is created within the Integrated Production Modelling software (IPM) to easily leverage physics-based model for generating data for unseen operating conditions. The developed proxy hybrid model is then used for gas lift optimization on single platform where injected gas oil ratio (IGOR) for each well is calculated based on the predicted liquid rate to rank the wells and more gas is allocated to more efficient well based on gas lift performance curve (GLPC)