Hierarchical speed control of an infinitely variable transmission in tidal current energy converters using integral reinforcement learning

Abstract

This paper presents a novel data-driven internal reinforcement learning (IRL)-based optimal speed control of an infinitely variable transmission (IVT) for tidal current energy converters (TCECs). The main objective of this study is to develop an effective control strategy that can maintain a stable output speed of the IVT system under varying tidal current conditions without relying solely on a dynamic model. The proposed method integrates a crank length controller with an IRL-based speed controller. The crank length controller adjusts the speed ratio of the IVT in real time, while the IRL-based speed controller optimizes the input speed using only sampled input-output data. An actor-critic neural network architecture is used to approximate the optimal control policy. The performance of the proposed approach is validated through simulations under both constant and time-varying tidal speed scenarios. Control results demonstrate that the data-driven IRL-based speed control approach of the IVT can effectively regulate its output speed with average tracking errors of 2.6 % for time-varying inputs and 6.3 % for constant inputs, outperforming traditional proportional-derivative control. This data-driven method eliminates the need for complex dynamic modeling of the IVT system, offering a more practical and adaptable solution for TCEC applications.