Wind farm Simulink modelling and control through dynamic adjustment of wind turbines power set-point Saman Poushpas, Prof. W.E Leithead Department of Electronics and Electrical Engineering, University of Strathclyde Address saman.poushpas@strath.ac.uk w.leithead@strath.ac.uk University of Strathclyde R222e, Royal College Building, 204 George Street Glasgow G XW Scotland, UK Tel: +44(0)4 548 2077 Introduction A Simulink wind farm model suitable for fast simulation and control design has been proposed. Structural loads on a wind turbine stem from the rotor/wind-field interaction, wind turbine structural dynamics and its controller. Therefore, the wind farm model must contain all of these elements, although only the most significant structural dynamic modes are required. Approach A generic algorithm for developing wind farm models in Simulink is developed based on the Supergen 5MW Exemplar wind turbine model, the Veers three dimensional turbulence model algorithm [], and the Frandsen analytical wake effects modelling method [2]. The wind farm generator algorithm is written in Matlab script and utilizes various functions and libraries to generate a wind farm with a large number of turbines. Wind farm model utilizes Matlab level2 S- Functions for most of the calculations. Therefore, the simulation speed is increased significantly. For instance, simulation time for a wind farm model with 0 turbines for 4000s is about 600s. Main body of abstract Wind farms are increasingly subjected to higher regulations from the transmission system operators. Mainly, wind turbines are required to provide ancillary services to the grid similar to the conventional power plants. Particularly, primary frequency support is of the most interest to the TSOs. In this work a novel approach has been proposed to provide primary frequency response by a wind farm. The primary frequency support contains synthetic inertia response to compensate for high rate of change of frequency events in the system and droop control for low frequency events. Wind farm modelling The wind farm model must include a decent representation of the wakes including their propagation through the wind farm. A wind-field model is developed based on the Veers algorithm that
generates low frequency turbulences for the wind farm simulation. High frequency turbulence is represented locally at each wind turbine. The wind-field model includes wake effects according to the Frandsen analytical wake modelling method and considers delays between the turbines for wake propagation. Wake deficits are calculated and added to the mean wind speed at each turbine and updated at each sampling time. Effective wind speeds at each wind turbine are calculated by adding the mean wind speed and wake deficits to the spatially filtered turbulence time series. A Supergen 5MW Exemplar wind turbine model is utilized in the model of the wind farm. The wind turbine model contains representations of the drive-train, blade and tower dynamics and includes a full envelope controller augmented with a Power Adjusting Controller (PAC)[3]. The full envelope controller and the PAC are implemented in C code to reduce simulation run-times. The PAC operates as a feed-forward controller and dynamically adjusts the power-set point of the wind turbines in all operational modes of the full envelop controller. The PAC includes a set of flags that indicate the operating status of the wind turbine. These indicators can be used by the wind farm controller to operate the wind farm in a more intelligent and optimal manner. A wind farm control algorithm is developed with objectives of power reference tracking and power maximization of the wind farm. The wind farm controller utilizes a Matlab S-Function and can operate at different sampling times. The primary frequency support of the wind farm controller runs at 00ms and includes synthetic inertia response and droop control. The synthetic inertia response of the wind farm controller utilizes status flags from each wind turbine s PAC controller to dispatch the required power between the turbines. The wind farm droop controller utilizes a droop curve with specific settings according to the National Grid requirements for droop slope and dead-band to response to the low frequency event in the system. The wind farm controller can be modified to optimize the wind farm power delivery at normal operating conditions. The wind farm controller utilizes operating status flags generated by each wind turbine PAC, and the power demand signal from the network operator to adjust the power set-point of the wind turbines as required. PAC generates status flags based on the PAC supervisory rules which are defined to ensure the safe operation of the wind turbines when PAC is operational. Figure shows the regions defined by the PAC supervisory rules for safe operation of the wind turbine. There are two sets of rules that define the operating boundary regions. Namely, black supervisory rules which define the black boundaries and should not be crossed at any circumstances and traffic light rules which are defined to define sensible maximum power delivery at each region. Wind farm controller Synthetic inertia response The synthetic inertia response of the wind farm controller to a sudden rate of change of frequency is developed based on the well-known swing equation Jω grid dω grid dt = P mec P elc. The required level of power adjustment to compensate for a ROCOF event is calculated as 2S n H dω ω dt = P mec P elc Where, inertia constant H = 8s, S n = 5e6, ω = 50Hz K Simulations are run for a wind farm model with 0 5MW turbines operating at 8ms constant wind speed and also with 0% turbulence intensity. Wind turbines are placed in two rows of 5 turbines parallel to the wind direction. P
Figure 2 shows the inertia response of the wind farm to a ROCOF event in Figure at 8ms constant wind speed. Figure 3 shows the inertia response to the same ROCOF event at 0% turbulent wind. As can be seen from Figure 4 each wind turbine contribute to the inertia response based on their operating status.
0-0.0-0.02-0.03 ROCOF (Hz/s) -0.04-0.05-0.06-0.07-0.08 400 420 440 460 480 500 520 540 560 580 600 Figure :ROCOF x 0 7.65.6.55 Total Power (MW).5.45.4.35.3 400 420 440 460 480 500 520 540 560 580 600 Figure 2:Constant wind speed inertia response x 0 7.5 normal operation Synthetic inertia response.4.3 Total Power.2. 0.9 400 420 440 460 480 500 520 540 560 580 600 Figure 3: turbulent wind speed total inertia response x 0 6 4.5 4 3.5 inertia response of each wind turbine 3 Power 2.5 2.5 0.5 0 400 420 440 460 480 500 520 540 560 580 600 Figure 4:turbulent wind speed individual inertia response
Figure 5 shows the turbulent wind-field wake deficits and wake propagation through the wind farm. 0.98 0.96 initial deficits (def,) (def,2) (def,3) (def,4) (def,5) (def,6) (def,7) (def,8) (def,9) (def,0) Deficit 0.94... 0.92 0.9 500 000 500 2000 2500 3000 3500 Figure 5:Wake deficits Wind farm droop control Droop control of the wind farm controller is developed based on the droop curve shown in Figure 6. Figure 8 shows the droop response of the wind farm controller to a frequency event shown in Figure 7. x 0 6 5 max power: 5e6 4.5 4 3.5 Dead band 3 Power 2.5 2.5 0.5 min power:0.2e6 0 48.5 49 49.5 50 50.5 5 5.5 Frequency(Hz) Figure 6:Droop curve
50 49.9 49.8 49.7 system frequency(hz) 49.6 49.5 49.4 49.3 49.2 49. 400 450 500 550 600 650 700 Figure 7:Under-frequency event x 0 7.4 normal operation Droop control.35.3.25 Total Power.2.5..05 450 500 550 600 Figure 8:Droop control response Conclusion A wind farm model with sufficient dynamics for fast simulation and controller design is developed in the MATLAB/Simulink. Abbreviations:. PAC: Power Adjusting Controller 2. ROCOF: rate of change of frequency References:
[] Paul S. Veers, Three-Dimensional wind Simulation, 988. [2] S. Frandsen, R. Barthelmie, S. Pryor, O. Rathmann, and S. Larsen, analytical modelling of wind speed deficit in large offshore wind farms, no. January, pp. 39 53, 2006. [3] A. Stock, Augmented Control for Flexible Operation of Wind Turbines, PhD Thesis, University of strathclyde, 204.