Abstract
Curtailed operation -when a wind farm is requested by the grid operator to produce less than what it could- is expected to become more and more common as the share of renewables in the energy market increases. In these conditions, wind farm control enables de-rating and hence offloading specific turbines to slow down degradation of components for reduced maintenance costs and lifetime extension. Drivetrain components are known to be prone to costly repairs, and hence particularly suited to establish this bridge between farm control and maintenance planning.
This study work is a step in this direction, linking dynamic wind farm control based on observed wind speed by individual turbines and predicted fatigue damage of the drivetrain components.
The farm controller sends power setpoints to the individual turbine controllers. It is based on a hierarchical structure, as described in [..]. In a first step, only power tracking is looked at: the controller compensates for the decrease in power for turbines where the wind speed is low by increasing setpoints for turbines where the wind speed is high. The final goal being to include information about turbine degradation to both try to keep the farm output power constant while derating the most damaged turbines. This supervisory control strategy is highly dependent on the total area to average power and load fluctuations over, hence all the farm's turbines are needed for the simulation to be representative.
On the other hand, the drivetrain necessitates a detailed aero-servo-elastic model of the turbine, that feeds into a component-level model. The computationally fast lumped-parameter model of drivetrain as described in [.] is used, consisting of a state-space model able to capture the dynamics of gears and shafts of the gearbox in interaction with the rotor and structural loads. This vibrational response may then be used to estimate load and stress in gears, shafts and bearings, as the input for degradation model to estimate damage.
This paper presents the efforts made to handle the complexity induced by this simultaneous need for farm-wide effects including farm controller on the one side, and detailed component-level effects on the other side. Emphasis is put on keeping the computational efficiency of engineering mid-fidelity models. To this end, add-ons to NREL's farm simulation tool FAST.farm are developed, incorporating farm-wide turbulence and linking to the farm controller and drivetrain model in a flexible, non-intrusive way.
A case study on a utility-scale wind farm is presented. Although not thorough, it shows the potential for flexible development integrating variable degrees of modelling fidelity and complexity for the turbine/flow, controller and drivetrain (and by extension any component). It paves the way for fully coupled integration of drivetrain in farm simulations, for instantaneous fatigue damage estimation and its use in load mitigation control.
References
[.] Moghadam, F. K., Rebouças, G. F. D. S., & Nejad, A. R. (2021). Digital twin modeling for predictive maintenance of gearboxes in floating offshore wind turbine drivetrains. Forschung im Ingenieurwesen, 85(2), 273-286.
[..] Merz, K., Chabaud, V., Garcia-Rosa, P. B., & Kolle, K. (2021, September). A hierarchical supervisory wind power plant controller. In Journal of Physics: Conference Series (Vol. 2018, No. 1, p. 012026). IOP Publishing.
This study work is a step in this direction, linking dynamic wind farm control based on observed wind speed by individual turbines and predicted fatigue damage of the drivetrain components.
The farm controller sends power setpoints to the individual turbine controllers. It is based on a hierarchical structure, as described in [..]. In a first step, only power tracking is looked at: the controller compensates for the decrease in power for turbines where the wind speed is low by increasing setpoints for turbines where the wind speed is high. The final goal being to include information about turbine degradation to both try to keep the farm output power constant while derating the most damaged turbines. This supervisory control strategy is highly dependent on the total area to average power and load fluctuations over, hence all the farm's turbines are needed for the simulation to be representative.
On the other hand, the drivetrain necessitates a detailed aero-servo-elastic model of the turbine, that feeds into a component-level model. The computationally fast lumped-parameter model of drivetrain as described in [.] is used, consisting of a state-space model able to capture the dynamics of gears and shafts of the gearbox in interaction with the rotor and structural loads. This vibrational response may then be used to estimate load and stress in gears, shafts and bearings, as the input for degradation model to estimate damage.
This paper presents the efforts made to handle the complexity induced by this simultaneous need for farm-wide effects including farm controller on the one side, and detailed component-level effects on the other side. Emphasis is put on keeping the computational efficiency of engineering mid-fidelity models. To this end, add-ons to NREL's farm simulation tool FAST.farm are developed, incorporating farm-wide turbulence and linking to the farm controller and drivetrain model in a flexible, non-intrusive way.
A case study on a utility-scale wind farm is presented. Although not thorough, it shows the potential for flexible development integrating variable degrees of modelling fidelity and complexity for the turbine/flow, controller and drivetrain (and by extension any component). It paves the way for fully coupled integration of drivetrain in farm simulations, for instantaneous fatigue damage estimation and its use in load mitigation control.
References
[.] Moghadam, F. K., Rebouças, G. F. D. S., & Nejad, A. R. (2021). Digital twin modeling for predictive maintenance of gearboxes in floating offshore wind turbine drivetrains. Forschung im Ingenieurwesen, 85(2), 273-286.
[..] Merz, K., Chabaud, V., Garcia-Rosa, P. B., & Kolle, K. (2021, September). A hierarchical supervisory wind power plant controller. In Journal of Physics: Conference Series (Vol. 2018, No. 1, p. 012026). IOP Publishing.