Wind energy is the most mature and promising renewable energy source in the world at present. The doubly-fed induction generator (DFIG) based and direct-drive permanent magnet synchronous generator (PMSG) based systems are the most popular wind energy conversion systems (WECS), which are widely used in wind farms [1
]. Though both of the two WECSs have their own advantages and disadvantages, the direct-drive PMSG based WECSs are preferred in high power applications owing to the simple structure, high efficiency, and high reliability due to the absence of gearboxes [1
]. However, due to the stochastic nature of wind, the output power of the wind farm is fluctuating, which brings negative influences to the stability and economy of the grid operation. Specifically, the problems caused by the wind power fluctuations are listed as: frequency deviation, voltage flicker, low power quality, and harm to sensitive loads [3
]. To solve these problems, several methods are proposed to smooth the output power of the wind farm, which can be basically divided into two major categories: direct power control and indirect power control.
The direct power control usually depends on using the wind turbine (WT) itself without energy storage devices, such as through the DC bus voltage control, rotor inertia control, and the pitch angle control [3
]. However, the DC bus voltage control may cause the excessive voltage ripple, thus affecting normal operation of the WECS [5
]. For the rotor inertia control, the rotor speed is always controlled to exceed the rated value to store the kinetic energy. According to the research [5
], the efficiency of the kinetic energy control relying on the generator inertia is close to the maximum power point tracking (MPPT) control. The trade-off between the maximum power tracking and power smoothing is explored through the optimal control of WTs based on the rotating kinetic energy reserves in references [7
]. The pitch angle control has to operate at a fast rate, thus mechanical stress accrues in the wind turbine blade. Besides, the direct power control methods always focus on the output power smoothing of a single WT, which may lead to the poor effect for the entire wind farm. In addition, all the smoothing capability of the direct power control method is constrained by the rated power of the WTs.
The indirect power control is realized by using energy storage systems (ESS), such as the battery, the supercapacitor, the superconductor magnetic energy storage (SMES), the flywheel, and so on [12
]. But all existing ESSs have certain problems in practical application. The charge-discharge cycles of the battery are limited, although many types of batteries are quite mature at present [12
]. The supercapacitor has the much longer life of charge-discharge processing than that of batteries, but it cost too much to be used on a large scale [14
]. SMES is a large superconducting coil that can store electric energy in the magnetic field produced by the flow of a DC current through it. The effectiveness of the SMES is demonstrated by many studies, but it is still too costly [15
]. The flywheel ESS utilizes the kinetic energy of a rotating disc, which depends on the square of the rotational speed. It has the advantages of long cycling life, large energy storage capacity, and high reliability [17
]. However, the drawbacks of immature technology, large size, and high standing losses cannot be ignored in practical use at present. Then hybrid ESSs for smoothing out wind power fluctuations are proposed in literatures to overcome the disadvantages of the single ESS, but the control strategies are more complicated [19
As above, all the existing power smoothing control strategies have both advantages and disadvantages, which result in difficulties for practical application. In order to solve such problems, a power smoothing method based on the allocation of the wind turbines and the rotor inertia is proposed in this paper, which is quite different from the above-mentioned power smoothing methods. In this paper, the direct-drive PMSG based WECS is chosen in the wind farm. The WTs of the wind farm are divided into two classes: control wind turbines (CWT) and power wind turbines (PWT). The PWTs are expected to output as much power as possible and then the MPPT control strategies are adopted by combing the power smoothing strategy using rotor inertia. The increment of wind power can be stored in the generator as a form of kinetic energy and released when the wind power decreases. The CWTs are in charge of the output power smoothing for the entire wind farm by giving the appropriate power according to the output power need of the wind farm. The battery energy storage system (BESS) with small capacity is installed as the back support of the CWTs to satisfy the high power demanding of the grid. The simulation model of the wind farm based on the proposed control strategies is built in Matlab/Simulink and the efficiency of the proposed strategy is verified by the simulation results.
4. Simulation Results
The simulation model of the wind farm was built in the MATLAB/Simulink, which was based on the above-mentioned models and control strategies. The DC link voltage and the output reactive power were controlled through the grid-side converter. The variable pitch angle control was also applied in the simulation. The detailed control strategies of the grid-side converter and the pitch angle control were not presented in this paper and the detailed control strategy can be seen in [30
]. The phase-locked-loop method was used to detect the grid voltage and phase information [32
]. The parameters of the wind turbine and the PMSG are listed in Table 1
. The size of the grid-side DC/AC inverter was chosen as 4.16 kV/20 MW which was based on the whole output power capacity of the wind farm (18 MW) by adding some margin. The reference voltage of the HVDC was set to 3500 V and the reference reactive power of the grid-connected inverter was set to zero. The minimum output power of CWTs was set as 200 kW. The size of the BESS was chosen according to the grid demand and the output power of wind turbines, which can be estimated from the wind speed. The small capacity BESS with the parameter 4 MW/3 MWh was configured in the simulation.
Three working modes of the wind farm were studied in the simulation, which were the MPPT mode, the power smoothing mode, and the specified power output mode, respectively. The first working mode of the wind farm was to output the maximum power then the optimum power given MPPT control was adopted for all the nine WTs, while using the control diagram shown in Figure 7
. The BESS was adopted to smooth the output power of the nine WTs based on the traditional method. The reference power of the BESS was given as the difference of the output power of WTs and the output power through a low-pass filter. The simulation results of the WTs under the MPPT control are shown in Figure 9
and Figure 10
The stochastic nature of the wind speed generated by the software TurbSim was given in the simulation. The wind speed waveforms of the nine turbines are shown in Figure 9
a–c, while those of 1–3 WTs are given in Figure 9
a, 4–6 WTs are given in Figure 9
b and 7–9 WTs are given in Figure 9
c. The wind speed of each wind turbine was given different values by considering the wake effect to close to the real wind farm in the simulation. The wind speeds of three group wind turbines were totally different, while the wind speed of each wind turbine in one group was decreasing from the upstream to downstream with the similar shape. Figure 9
d–f show the rotor speed of the nine WTs respectively. It can be seen that the rotor speeds of nine WTs varied with the wind speeds to realize the MPPT control.
The torque of nine WTs is shown in Figure 10
a,c,e and the output power is shown in Figure 10
b,d,f, respectively. The MPPT control was realized while the torque and output power of the WTs varied with the wind speed. The total output power is shown in Figure 10
g and it is clearly observed that the output power was consistent with the ideal MPPT power by ignoring the system losses in the simulation. The output power of the wind farm was smoothed by the BESS, as shown in Figure 10
g. The power of the BESS was presented in Figure 10
i, where the BESS was charging when the value was greater than 0 and discharging when the value was less than 0. It can be seen that the BESS was charging and discharging all the time under the traditional power smoothing control, which would cause the life of BESS to decrease seriously. Figure 10
h demonstrates the DC link voltage of the wind farm, from which it can be seen that the voltage remained at the reference value 3500 V only, with little fluctuation which was about 5 V in the steady state, as shown in the extended waveform. Then the control strategy of the grid-side converter was verified.
The second working mode of the wind farm was to output the smoothing power and the simulation results are shown in Figure 11
and Figure 12
. At this time, the proposed power smoothing control strategy based on the allocation and rotor inertia of wind turbines was adopted. The MPPT method combing the rotor inertia control was applied to the PWTs 1–6, while the given power control was adopted for the CWTs 7–9. The sampling period of the kinetic energy based power smoothing strategy for PWTs was 0.001. The cut off frequency of the low-pass filter was 10 Hz. The simulation waveforms of the PWTs are shown in Figure 11
including the rotor speed, torque, and output power. The rotor speed of PWTs shown in Figure 11
a,d did not follow the optimum MPPT speed, which are shown in Figure 9
d,e. It can be seen from Figure 11
b,e that the torque of the PWTs was almost straight with very tiny fluctuations. Then, the mechanical stress of WTs was reduced effectively. By comparing the output power Figure 11
c,f with Figure 10
d,e, it can be observed that the output power of PWTs under the proposed rotor inertia based method was smoother than the power under MPPT control. Thus, the effectiveness of power smoothing method for PWTs based on the rotor inertia control was verified.
The output power of the three CWTs is shown in Figure 12
a–c respectively. The reference power calculated according to the block diagram in Figure 4
was given to CWTs after 1s, while the maximum power was given to the PWTs before 1s. The power of each CWT was allocated according to the wind speed relationship which was obtained by calculating the wake effect. As shown in Figure 12
a–c, the output power of CWTs was different from that under the MPPT control mode which is shown in Figure 10
f and the actual output power of the three CWT can match the reference given power well. It can be seen from Figure 12
e that the output power of WTs was smoothed through the proposed control strategy by comparing with the MPPT power. However, the output power of WTs could not be smoothed when the needed smoothing power exceeded the maximum output power of CWTs. In this case, the BESS was joined in to smooth the rest power and the power of BESS is shown in Figure 12
d. Thus, the output power of the wind farm was smoother than the output power of WTs. Figure 12
d shows that the BESS was not working all the time and then the charge and discharge times was reduced significantly compared to that using the traditional power smoothing method (shown in Figure 10
i). Thus, the battery lift would be extended greatly. The DC link voltage can remain constant as shown in Figure 12
f with little fluctuation around 5 V in the steady state.
The maximum energy function of the MPPT power, the output power of WTs and output power of the wind farm is shown in Figure 12
g. It reflects that the maximum energy function of the output power of wind farm was slightly less than the MPPT output power. The energy loss mainly happened in the PWTs by adopting the kinetic energy based power smoothing control. The power smoothing function is shown in Figure 12
h, from which it can be seen that the value of output power by WTs was obviously smaller than that of MPPT control. Then, it can be summarized that the proposed method can realize the good performance of power smoothing. The power smoothing effect was better when the BESS is joined in as shown in Figure 12
In some situations, the grid needs the wind farm to have the ability to output the specified power to meet the balance of the grid and loads. Thus, the third working mode of the wind farm was to output the determined power. The step-changed power demand (6 MW, 4 MW, 8 MW, 6 MW, 7 MW) for the wind farm was given in the simulation. The reference demand power was given to test the effectiveness of the proposed method in both cases of power increasing and power decreasing. Besides, the maximum reference demand power is limited by the sum of the maximum output power of wind turbines and the BESS. The simulation results of this case are shown in Figure 13
. Being the same as that in the second working mode, the CWTs were given the reference power calculated according to block diagram Figure 4
after 1 s. The simulation results of PWTs were the same as that shown in Figure 11
and not given here. The output power of CWTs is demonstrated in Figure 13
a–c, from which it can be seen that the CWTs outputted constant minimum power 200 kW between 2 s to 4 s due to the low grid demand power. The power of the BESS is shown in Figure 13
, and it can be found that the BESS was working in the charge, discharge, and rest working modes according to the different power grid demands. The output power of the wind farm can follow the given step-changed demand power as shown in Figure 13
e. As displayed in Figure 13
f, the DC link voltage basically remained at the reference value, but with acceptable mutation due to the sudden power changes.
The new wind speed condition of CWTs was given in the simulation to increase persuasion of the proposed power smoothing strategy, while the wind speed of the PWTs was the same as that in Figure 9
. The wind speed of CWTs is shown in Figure 14
a, from which it can be seen that the shape of downstream WT was different from the upstream WT. Then, the rotor torque of each CWT had different shapes, as demonstrated in Figure 14
b. The actual output power of the three CWT can match the reference given power well as shown in Figure 14
c–e. Figure 14
f shows the output power of the wind farm and it can be observed that the output power was smoothed by comparison with the reference MPPT power.
Thus, the proposed power smoothing control strategy through the reasonable allocation of the WTs in the wind farm and the rotor inertia was verified by the simulation results. The rotor inertia based power smoothing method for PWTs can reduce the torque ripple and mechanical stress significantly. The CWTs can output the appropriate power according to the reference value. The configured BESS were not working all the time to extend the life. The control for the grid-side converter was also verified through the constant DC link voltage. The proposed control strategy can respond to different needs of the grid, which was verified by the simulation results under different working modes.