Abstract
Electrifying offshore platforms offers significant environmental benefits but introduces challenges in voltage stability, especially in systems with long AC cables and low short-circuit levels. As gas turbines are replaced by converters, the control of reactive power for controlling the voltage becomes critical. This paper evaluates the performance of various tuning methods of converter voltage controllers, focusing on their ability to reduce voltage dips by controlling the reactive power. Conventional methods like Modulus Optimum (MO)/Symmetrical Optimum (SO) criteria, Ziegler-Nichols (Z-N) and Chien-Hrones-Reswick (CHR) PI tuning method are compared with a Reinforcement Learning (RL) approach. Tuning based on RL achieves the lowest voltage deviations by prioritizing the minimization of the voltage deviations, though with a slower response. The RL-tuned voltage controllers enhance damping and stability, even under large disturbances.