Robust Reinforcement Learning-Based Multiple Inputs and Multiple Outputs Controller for Wind Turbines

The control of variable-speed wind turbines that generate electricity from the kinetic energy involves subsystems need to be controlled simultaneously, namely, blade pitch angle controllers and generator torque controllers. presented study solves problem with multiple inputs outputs (MIMO), using method reinforcement learning–based Trust Region Policy Optimization, through which parameters both are simultaneously optimized. In this case, robust is transformed into a constrained optimal an appropriate choice value functions for nominal system. aims synthesize controller, aim maximizing generated (power) minimizing unwanted forces (thrust). innovative architecture uses extended input space, allows fine-tuning each operating state. Test calculations carried out in simulation experiments models 5 MW NREL turbine 4 Enercon E-126 EP3 illustrate performance practicality proposed approach.