Variable Frequency AC Network

A group of wind turbines can share a common converter as shown in Figure 3.7. This concept network topology benefits from centralising the converters as discussed in Section 3.1 and improving cable interconnection for reduced cable length and cost as discussed in the beginning of Section 3.2. This topology eliminates the need of a frequency converter directly at the WT. A generator will be connected directly to the network via a transformer and switch gear as a Type A or Type B configuration WT would. The total number of converter stages that the power has to go through in the line of power flow before reaching the onshore grid is reduced, which contributes to reduced converter losses as well as saving cost, weight and space [123].

Wind turbine generators Variable frequency Medium voltage AC cluster network Transformers VSC Power collection in an AC or DC network before HVAC or HVDC transmission SCIG/ WRIG

The variable frequency concept topology has been investigated in literature. Jov- cic presented a system where the VSC-HVDC offshore converter is divided into 25 modules connecting to the same HVDC link as the multi-terminal configura- tion. Each module is connected to an independent cluster network consisting of 4 WTs [124]. The operation of this network topology is similar to a WT with a fully rated frequency converter, except that multiple WTs are sharing a single VSC. The AC frequency of a network cluster is controlled using MPPT, which is determined for the average wind speed in that group. For generators connected directly to the network like the Type A configuration, the rotor speeds of the WTs are assumed to be synchronised and proportional to the network frequency. The SCIG would be the simplest and least expensive for this application. The WT rotor frequency of the generator can be varied by varying the frequency of the cluster network. How- ever, the actual wind speed varies between WTs due to the wake effect caused by neighbouring WTs as mentioned in Section 2.2.9. The tip speed ratio λ would not be optimal and there will be a loss in the coefficient of performance Cp compared to

individual MPPT control in Type C and Type D wind turbines where the maximum power could be extract from wind [125]. This loss increases with the number of WT per cluster, but the percentage loss begins saturating at about 16% after having more than 6 WTs per cluster [126].

Improving the Overall MPPT

In the variable frequency network topology, it is more advantageous to use an asyn- chronous induction generator with a wide slip speed range than a synchronous gen- erator. WTs that experience a higher than average wind speed will produce more power and increase the induction generator slip speed with torque. As a result, more power can be further extracted as the tip speed ratio λ is improved towards the maximum power point. When there is high variability in the wind speed across the WTs in the cluster, the constant rotor resistance in the induction generator could be tuned to optimise the slip speed and the tip speed ratio λ for a better over- all power extraction. Furthermore, a wide slip speed variation with torque helps to reduce the rotor shaft stress as the asynchronous speed of the rotor Na increases.

3.2. Wind Turbine Interconnection 73 Variable slip control in Type B configuration WTs can be used to vary the rotor speed in the individual WTs, which is achieved by varying the rotor resistance and therefore the slip speed as described in Section 2.2.5 and as shown in the torque-slip characteristic curve for a variable rotor resistance in Figure 2.13. The individual rotor speed control with Type B WTs further optimises the tip speed ratio λ of the turbines to improve the overall MPPT in the variable frequency network to minimise the Cp losses and extract the maximum power [127]. A Type B configurations can

be achieved without slip ring and brush connections, by using an OptiSlipr _type

technology, as mentioned in Section 2.2.5. This can more robust compared with frequency converters in Type C and D systems where a potential high failure rate in the power electronics can render the WT system to be unstable or useless during operation. A failure in the variable resistor mechanism only affects the MPPT and does not have to affect the operability of the WT to generate power, assuming the failure mode is a short circuit or a constant stuck resistance state. The variable rotor speed range of the induction generator Na can only be above the synchronous

speed of the stator field Ns and the network frequency. Therefore, the set variable

network frequency will be skewed for WTs experiencing the lower spectrum of the wind speeds instead of the average wind speed in the cluster, while the rotor speed of Type B WTs can be adjusted to cater for higher wind speeds.

When combining the variable frequency network and Type B limited variable speed WT, the variable speed range of the turbines is maximised on a per WT basis, which are then able to individually fine tune the tip speed ratio for maximum power extraction. The WF does not need to be split into smaller clusters and controlled in groups to minimise the MPPT loss as Jovcic and Strachan did without flexible rotor speeds [126]. The variable frequency can be applied for entire WF, which offers simplicity and greater degree of freedom to connect in any WF layout and cable path, without being restricted into small clusters.

The variable frequency AC network topology already uses the frequency as a guide for the synchronous speed of the WT generators. Therefore it cannot use the frequency droop communication method to communicate a signal to the wind turbines to indicate how much power it should generate in the same way as described

in Section 2.4.6. Alternatively, it would be possible to manipulate the frequency away from the maximum power point to stall the WTs to curtail and control the power output of the wind farm.

Concern of Low Frequencies in Transformers

A concern for a variable frequency network is transformer saturation at low frequen- cies. When the frequency is reduced while voltage level remains constant, the flux density B in transformer core increases, which may exceeding the saturation level. When the transformer core saturates, the magnetising inductance becomes a hun- dred to a thousand times smaller, and the magnetising current becomes larger as a result, until it effectively becomes like shorting the transformer windings. Increasing the core cross-sectional area to accommodate more flux (or volt-seconds) at lower frequencies is not practical as it will be larger, heavier and more costly. Transform- ers are an essential and cost effective method for isolation and voltage stepping for more efficient network transmission and thinner copper in the cables.

According to the transformer EMF equation (3.3), the flux density B in the transformer is proportional to the ratio of the EMF voltage and the frequency (E/f ), assuming the turns ratio N and the transformer core cross-sectional area A are constant.

E = 4.44f N AB (3.3)

The EMF voltage level should be reduced linearly with frequency in order to keep the flux in the transformer constant and prevent transformer saturation. As the voltage and frequency is reduced, the current rating of the transformer is also reduced linearly [128]. The WT power has a cubic characteristic with respect to frequency as described in Section 2.2.7, assuming constant Cp with MPPT and that

the WT rotor speed and frequency are proportional to the wind speed. When the WT is operating below the rated wind speed, the power output level is significantly reduced at lower frequencies and is well below the dynamic rating of the transformer. Therefore, there will be no danger of current overloading at the transformer [126]. Low frequency AC is also more efficient due to lower reactive losses, eddy currents and hysteresis losses in both cables and transformers.

3.2. Wind Turbine Interconnection 75 The network voltage is controlled by the VSC-HVDC and the start up current is minimised by starting from low voltage and low frequency. Therefore, the soft- starter device shown in Figure 2.12 for Type A and B configurations would not be necessary.

Low frequency AC circuit breaking would usually take longer to extinguish be- cause the current is cutting through the zero crossing less often. However, the system only operates at lower frequencies when the wind speed is low. Therefore the lower maximum power point, together with a lower voltage level at low frequencies should aid quicker circuit breaking.

The transformer design needed must take into account of the increased mag- netising currents at lower frequencies. This is due to the magnetising reactance decreasing as the frequency is reduced (X = ωL). Therefore more turns per unit voltage are needed.

Constant Current

The network voltage characteristic against frequency may follow a constant current regime instead of a linear characteristic for constant transformer flux. Although the overall I2R conduction losses will be higher when operating at constant current when operating below rated wind speed, the loss will be constant and therefore the device temperature will be constant. This would help reduce the frequency of deep thermal cycling in transformers, and therefore reduce fatigue failure and increase the reliability performance. It may then not be required to overrate the devices or even use heating elements to minimise the thermal cycling. However, the thermal cycling in the individual junction of the switching devices in the VSC will be worse with low frequency and high peak currents [105]. Furthermore, it will be difficult to implement constant current control in an interconnected system with many WTs because the margin for variability in the load sharing is low and the odd WT may overload.