Closed Loop Gas-Lift Optimization

Abstract

Significant value can be achieved by optimizing production of a gas-lift network. Operators have traditionally performed this work manually using network models, but maintaining these models is often labor-intensive. To address this challenge, a closed-loop optimization system was developed that leverages both advanced analytics and physics-based techniques, as well as Internet of Things (IoT) Edge technology. The objectives of such system are to control and optimize the gas-lift network automatically, reduce downtime during compressor upsets, and mitigate any potential flare events. The new closed-loop gas-lift optimization algorithm consists of well and surface network models, optimization and regression solvers, and disturbance adaptation, all running in real time. The closed-loop optimizer works as follows: (1) in every cycle, the optimizer receives measurements; (2) disturbance adaptation compares the model's estimates with the measurements and adapts the surface network model to make it more accurate; (3) the adapted surface network model and well models are used to find the optimum lift gas setpoints; and (4) the calculated setpoints are sent to the automation system through IoT Edge technology. Integral to this system is a single-well nodal analysis model that automatically generates updated models daily for all gas-lift wells. This model is used both as a monitoring tool by the engineers and as part of the network model in the closed-loop gas-lift optimizer, which has been deployed in multiple fields and is running continuously (24/7). The optimizer has saved both production engineering time per network and well specialist time per compressor upset event. Field case studies have shown that the closed-loop optimizer has been successful in maintaining compressor station outlet pressure and optimizing the gas-lift networks during compressor upsets or potential flare events. A significant improvement in oil production has been achieved in fields using optimizer due to both optimized lift gas distribution and reduced downtime. This new algorithm can optimize gas-lift networks during normal operating conditions, compressor upsets, or potential flare events, while simultaneously controlling compressor station outlet pressure within an acceptable range in real time. Significantly, disturbance adaptation is used for the first time to improve the surface model accuracy immediately as additional measurements are received.