Reliable and Efficient Algorithms to Optimize Gas Well Intermittent Production Through Automatically Intelligent Control

Abstract

Abstract Liquid loading poses a prevalent challenge in gas well production. To maximize the gas production rate and reduce the cost of deliquification, intermittent production is often applied. However, optimization of operation parameters remains challenging due to the transient behavior of the reservoir-wellbore system. In this study, efficient algorithms are proposed to automatically optimize the parameters in real-time to remove the liquid efficiently and maintain gas production, through which the recovery factor can be extended. Firstly, we build a comprehensive model that couples the reservoir transient behavior with wellbore dynamic characteristics. The coupling model lays out the fundamentals for reservoir-production system analysis. Then, two intelligent optimization algorithms, namely the Genetic Algorithm (GA) and Proximal Policy Optimization algorithm (PPO) are developed to obtain the optimal operation parameters, which are the key elements in intermittent production management. Finally, the optimal production system is determined by jointly controlling the gas production rate and bottom hole pressure in the coupled reservoir-wellbore model to adjust choke and open-shut intervals for obtaining the maximum cumulative gas production rate and drainage rate. It is observed that the proposed reservoir-wellbore coupling model can characterize reservoir transient performance and capture the dynamic behavior of multiphase flow in the wellbore. The model facilitates the real-time analysis of the inflow-outflow system automatically, which can output optimal parameters to manage the operation. The global optimal solution of gas production rate and bottom hole pressure can be discerned from a set of combined parameters by using the GA algorithm. PPO algorithm can optimize the system by applying variable step size search and multivariable synchronous optimization depending on the environmental state, which can be a closed-loop optimization that allows automatic control/adjustment in the system. The simulation results indicate that intermittent production designed by two algorithms has higher cumulative gas production compared with continuous production scenarios, showing the significance of proper management in intermittent production. Additionally, it is noteworthy that GA and PPO exhibit no significant difference in gas cumulative production. Each algorithm has its special adaptability: the GA is suitable for the coupling model characterized by ideal homogeneity, whereas the PPO is more appropriate for the coupling model with complex heterogeneity, interference, transient response, and randomness. This presented work demonstrates the ability of the GA and PPO optimization algorithms to optimize the coupled reservoir-wellbore model for solving the optimal operation parameters in intermittent production, which has great significance in avoiding the liquid loading phenomenon and improving gas recovery.