Wind Turbine Individual Pitch PI and MPC Controller Designs Considering Actuator Rate Limits

Abstract

Wind turbine blade and tower sizes have increased over the last decade, leading to an increase in mechanical damage equivalent loads (DEL) acting on the tower and blades. This has sparked research interest in control algorithm solving the inherent constrained optimization problem: Maximize the energy capture, such that the operation and maintenance (O&M) costs is kept as low as possible. Keeping the DELs small while capturing maximal power is the obvious solution to this problem. Wind turbine control is often implemented by two independent SISO control loops: Torque or rotor speed control for below rated wind condition to maximize energy output and collective pitch control (CPC) for above rated condition, keeping energy output constant and limiting loads. Individual pitch control (IPC) can be added for further DEL reduction. IPC was shown to be an efficient approach and does not require any additional mechanical components, differentiating it from other active load control techniques e.g. trailing edge flaps. However, often the additional strain on the pitch actuators is neglected in simulations with the pitch actuator rate set to 10 deg/s per default. While modern actuators might allow even larger pitch rates, the pitch rate might also be reduced to prevent imminent actuator failure. The NREL 5 MW [1] research turbine is used as a test bench to investigate changes on pitch actuator constraints. A non-linear Simulink model, based on the physics of the wind turbine rotor, generator, tower and blade movement, is set up to investigate the effect of IPC with different actuator rate constraints and allow a time-efficient offline PI parameter optimization. The open loop frequency response as well as the closed-loop dynamics are compared to NREL’s FAST. An existing torque and CPC control algorithm is applied. The IPC controller is implemented as two PI controller. The controller parameters are optimized for different actuator constraints according to the cost function, a weighted sum of power reference tracking, power deviation, actuator, blade and tower damage criteria [2]. This simple, highly structured, non-predictive control scheme is compared to an MPC design, which is inherently suited for control with input constraints.