Deep reinforcement genetic learning -based model reference adaptive inverse control applied on the floating wind turbine

The ulterior motive of studying floating wind turbine is to decrease the harmful effects of global warming issue. Owing to the crucial changes of the Earth planet, the high potential of active structural control and the technical requirements associated with adaptation of floating wind turbine is applicable. An artificial intelligent controller involving deep reinforcement learning method of both state and action pairs is proposed as useful tool to cope with the variant environmental situations. The floating wind turbine tracks an input reference signal driving by a controller that approximates the inverse of plant model. The proposed adaptive algorithm should minimize the tracking error of the plant output with respect to the reference model output and the controller parameters are updated as well. The model reference adaptive control system is affected by sensor noise and exogenous disturbances. Modeling uncertainties and exogenous inputs are imposed in the overall control system through the proposed intelligent controller. The gathered rewards of both the identification and feed forward terms with control actions in loss function undergoes a Genetic Algorithm (GA) optimization process. Through software implementation results, removing the disturbance and noise effect on the tracking performance of the wind turbine and its stability is vivid.

Keywords:

deep reinforcement learning , artificial intelligence , inverse control , model reference adaptive control , Floating Wind Turbine , genetic algorithm