Artificial Intelligence-Based Adaptive Control of a Liquid Resistance Starter for a 5 MW Slip-Ring Induction Motor in Oil and Gas Installations

Abstract

High-power slip-ring induction motors remain important in oil and gas installations where high breakaway torque, restricted network capacity and severe operating conditions complicate starting. Conventional liquid resistance starters (LRSs) generally follow a predetermined electrode-travel schedule and therefore cannot compensate explicitly for changing load torque, electrolyte conductivity or actuator constraints. This study develops an adaptive starting framework combining a quadratic B-spline neural disturbance estimator, a bounded sliding-mode-inspired correction, constrained rotor-resistance selection and a physically consistent electrode/thermal model. The framework is assessed using an original, deterministic reduced-order numerical simulation of a 5 MW, 6.6 kV, four-pole wound-rotor motor; it is not presented as a laboratory experiment, plant trial or MATLAB/Simulink electromagnetic-transient model. For a combined scenario with 30% conductivity degradation, elevated ambient temperature, load changes and oscillatory disturbance, the adaptive controller decreases integral absolute error from 6.398 to 5.553 pu·s (13.2%), integral squared error by 20.7%, and the time to 95% synchronous speed from 33.46 to 30.70 s. Peak current remains essentially unchanged at 1.501 pu, while predicted LRS energy increases from 56.98 to 57.98 MJ. The findings therefore support improved speed tracking and faster acceleration, while identifying a modest thermal-energy trade-off that requires explicit design attention.