Model predictive control for wind turbines with distributed active flaps: incorporating inflow signals and actuator constraints

Abstract

This paper describes the implementation of system identification and controller design techniques using model predictive control (MPC) for wind turbines with distributed active flaps for load control. An aeroservoelastic model of the 5 MW NREL/Upwind reference wind turbine, implemented in the code DU_SWAMP, is used in an industry-based MPC controller design cycle, involving the use of dedicated system identification techniques. The novel multiple-input multiple-output MPC controllers, which incorporate flap actuator constraints and the use of local inflow measurement signals, are designed and implemented for various operating points. The controllers are evaluated in standard power production load cases and fatigue load reductions up to 27.3% are achieved. The distributed flaps controller scheme is also compared with simpler single-flap single-input single-output and individual pitch controller schemes.

A smart rotor configuration with linear quadratic control of adaptive trailing edge flaps for active load alleviation

“…The control configuration based on only measurements feedback (d 00) lowers the lifetime fatigue equivalent damage on the blade root flapwise bending moment by about 10 %, a result in line with previous investigations that considered similar setups [12,17]. Periodic load anticipation, which is based on the blade azimuthal position and handled by the LQ algorithm as a prediction on periodic disturbance signals, allows to reach higher lifetime damage alleviation: 13. pation is comparable to the increase previous investigations have attained using additional in-flow sensors [12,14], with the advantage that the periodic load anticipation approach does not require a sensor setup as complicate and delicate as demanded for in-flow measurements. As an effect of active load alleviation with adaptive trailing edge flaps, a significant increase of the blade torsion fatigue damage equivalent load is reported (nearly 10 %); the increase of the torsional loads, and, to a lesser extent, of loads on other components, should be hence taken into account in the design of smart rotor structures.…”

“…All the investigations confirmed that smart rotors with trailing edge flaps have a potential for reducing the fatigue loads experienced by the turbine; nevertheless, they reported rather widespread results, with load reductions figures ranging from 5 to 47 percent, see the summary compiled by Barlas et al [12]. Differences in the alleviation performances can originate from several sources: the models used in the aeroelastic simulations, the conditions of the wind field and its turbulence levels, the maximum deflection and extension of the flap actuators, and also the choices made in designing the flap control system, as the assumptions on the available sensors and measurements, and the type of control algorithm implemented.…”

“…In fact, the variation of blade root flapwise bending moment obtained by the current flap setup is simply too small to compensate to an higher degree for the load variations observed on the blade during normal operation. Future work should thus consider, first, whether a control algorithm that handles flap deflection constraints, as for instance model predictive control [21,12], could improve the load alleviation performances. Secondly, future investigations might focus on whether fatigue loads could be further reduced by fitting the smart rotor with a more powerful actuator setup, either by augmenting the blades surface covered by adaptive flaps, or by complementing the flap efforts with individual blade pitch actions.…”

“…Most of the studies opted for control algorithms based on classic PID methods, applied either to each blade independently [13,14,15,16,17], or to the whole rotor through multi-blade coordinate transformation [18,19]; other investigations have instead applied model based control algorithms, as Linear Quadratic Regulators (LQR) [20], Model Predictive Control (MPC) [12,21], or H ∞ control [9]. The flap control actions respond to the deformation state of the rotor blades; in some cases, rotor sensors are assumed to provide direct measurements of the blade deflection and deflection rate [15,20,22,17], whereas other controllers use measurements of the blade flapwise bending moment, either at selected locations along the span [14,16], or, more simply, at the blade root [19,12,21,10].…”

“…The flap control actions respond to the deformation state of the rotor blades; in some cases, rotor sensors are assumed to provide direct measurements of the blade deflection and deflection rate [15,20,22,17], whereas other controllers use measurements of the blade flapwise bending moment, either at selected locations along the span [14,16], or, more simply, at the blade root [19,12,21,10]. Some studies have also investigated flap control algorithms where additional information on the in-flow condition along the blade are provided, for instance, by measurements performed with Pitot's tubes mounted on the blade leading edge [14,12,21].…”