A Dynamic Extremum Seeking Scheme Applied to Gas Lift Optimization

Abstract

This paper presents the application of a data-driven optimization scheme using transient measurements to a gas-lift optimization problem. Optimal operation of a gas-lifted field involves controlling the marginal gas-oil ratio (mGOR), which is the steady-state gradient of the oil rate from the gas lift injection rate. In this paper we apply a dynamic extremum seeking scheme to estimate the marginal GOR online using transient measurements, which is based on identifying a local linear dynamic model around the current operating point instead of a local linear static model. By doing so, we can use the transient measurements and effectively remove the time-scale separation between the plant dynamics and the perturbation signal, that is typically required in the classical extremum seeking scheme. This results in significantly faster convergence to the optimum compared to classical extremum seeking scheme. The effectiveness of the proposed method is demonstrated using simulation results for a single gas lifted well, as well as a network of gas lifted wells.