Research on Self-Optimization Regulation Method of Electric Submersible Pump Efficiency Based on DDPG

Electric submersible pumps (ESPs) are widely used for artificial lift because of their high liquid-handling capacity. However, changes in reservoir inflow and surface operating conditions can shift an ESP away from its high-efficiency operating range, thereby increasing energy consumption and production costs. This study formulates coordinated ESP regulation as a Markov decision process in which pump efficiency is the optimization objective and liquid production rate is an environment response subject to hydraulic feasibility. Pump frequency and wellhead choke diameter are selected as the control actions, and a self-optimizing controller is developed using the deep deterministic policy gradient (DDPG) algorithm within an actor–critic framework. The method is evaluated using a field-derived well model equipped with a G6200 ESP. In the low-frequency, low-flow scenario, the controller increases efficiency from 60.7% to 67.3%, an improvement of 6.6 percentage points, through four principal variable-step adjustments. In the high-frequency, high-flow scenario, it increases efficiency from 63.9% to 67.3%, an improvement of 3.4 percentage points, through two principal adjustments. The continuous-regulation simulation further shows that the controller compensates for changing operating conditions and maintains the operating point near the best-efficiency line within the evaluated range. These results provide a case-specific proof of concept for continuous ESP efficiency optimization; validation across additional wells, ESP types, and reservoir conditions remains necessary.