Reliable power electronics for windmill generators

In the megawatt range, high-power electronics applications need powerful semiconductors. However, even the largest semiconductors available today are still not strong enough for some applications. It is therefore necessary to connect them in parallel. The parallel connection of semiconductor devices in a traditional power electronics circuit is very common.

One possible solution is discussed in this context: power electronics assembly, IGBT base units containing IGBTs and diodes, heat sinks, DC link capacitors, drivers and protection, auxiliary power supply and a PWM controller (one independent unit), arranged into a three-phase inverter.

Such units can be connected in parallel, in the example a four-quadrant drive windmill with permanent magnet generator and a full-size 4 MW converter.

IGBTs are the working horses of power electronics systems. Today, IGBTs are manufactured in various voltage classes - 1200 or 1700 for different industrial applications as well as for the medium-voltage classes 3.3, 4.5, 6.5 kV.

Which voltage class is best suited to high-power applications?

The answer to this question lies in putting the IGBTs in the largest casing available to obtain inverters. Of course, it is much simpler to simulate available power under optimal working conditions.

To do so, the largest standard casing (IHM, 190 mm wide) is taken. The IGBTs are packed into this casing and the optimal operating regimes defined - V_DC DC operational link voltage, V_AC AC output voltage, a carrier switching frequency of 3.6 kHz and best possible cooling conditions.

Figure 1 shows the different available power levels, calculated on the basis of the given parameters.

The results show that the maximum available power using 3.3 kV, 1200 A individual modules would be half the equivalent power obtained using 1.7 kV, 2400 A IGBTs. The 6.5 kV, 600 A IGBT modules provide a quarter of what would be obtained with a 1700 V IGBT.

The reason behind these results is the losses that occur in IGBT modules.

If we calculate the efficiency of the three converters shown in Figure 2, we can see that the losses have a ratio of 1:2:4.

Figure 1 and Figure 2: Comparison of efficiency in IGBTs with different blocking voltages. Three-phase inverter operation at same cooling conditions and Fsw = 3.6 kHz; cosφ = 0.9 and same module.

For this comparison, we have used the same carrier switching frequency, F_sw = 3.6 kHz. This enables us to design inverters with relatively small filters. A comparison using different carrier switching frequencies would lead to variations in the output sinusoidal filters used.

Given the above, it can be seen that the greatest efficiency is accomplished by using the 1700 V IGBT, a standard industrial product with a very reasonable price per module.

IGBTs for 1700 V are packed in various module casings. For comparison, we can take the largest single-switch module, the IHM 2400 A 1700 V and compare two such modules with a dual module of similar size and length, SKiiP1513GB172.

If the two SKiiPs are put back to back on one heat sink, a half-bridge is obtained for currents 2 x 1500 A = 3000 A (case temperature = 25 °C), or 2250 A for a case temperature of 70 °C.

Two single-switch modules will provide a half-bridge for 2400 A. If we compare the results of the calculations, we can see that the SKiiP solution provides higher output currents throughout the complete range of switching frequencies than a standard module in the largest available case would.

If a more powerful SKiiP module is taken, for example the SKiiP 1800A, 1700 V, which uses an aluminium nitrate (ceramic) substrate, even more power is available from a three-phase inverter, ie, 1800 kVA.

Figure 4: Example with 1800 kVA base unit.

Solutions feasible for the parallel operation of IGBT modules include:

• One three-phase inverter for the entire power; the phase leg is constructed with several IGBT modules connected in parallel and one driver. Each IGBT module must have its own gate resistor and symmetrical DC link and AC output connection;
• Hard paralleling of three-phase IGBT base units.

The whole system is controlled via one controller and its PWM signals. All of the three-phase inverters are connected to a common DC link voltage.

Paralleling is achieved using driver boards for each individual base unit driver. Slight variations in driver propagation times (less than 100 ns) are compensated for with small AC output chokes; (<5 µH inductance).

All the three-phase inverters run simultaneously, with the small time delays that occur being compensated for with additional AC chokes.

To ensure proper load-current sharing, symmetrical layouts and positive temperature coefficients for IGBT saturation voltages are used.

A system as described under point 2 - with additional PWM signal correction for each base unit. Additional PWM corrections are performed to control precise load-current sharing in paralleled base units;

Parallel operation of several units with synchronous PWM and the elimination of circulated current using additional sophisticated PWM control;

Galvanic load isolation for each base unit. Each base unit supplies power to the load through insulated windings. Each base unit has its own controller. PWMs are independent, non-synchronous, free-running signals and each base unit has its own separate DC link.

On the grid side, each base unit has its own sinusoidal LC filter. Circulated currents between different DC links do not exist provided the outputs are galvanically insulated.

Figure 3: Available inverter power vs switching frequency.

This is the easiest parallelisation method for standard independent basic units with standard independent controllers.

A simple design based on galvanic insulation on the generator side is shown in Figure 5.

Figure 5: Three independent 4Q drives in parallel with separate motor windings. The drive can operate with one or two drives in parallel.

Three 1500 kVA 4-quadrant drive units are connected to separate generator windings of a permanent magnet windmill generator. Each drive is standard with its own generator-side and grid-side controllers.

The purpose of the fourth controller is to provide uniform generator torque sharing. Should problems occur in one of the 4Q drives during operation, the remaining drives will continue to operate interrupted.

The system described is used in a 3.6 MW windmill with a PM generator with three separate windings. The system is designed for up to 12 four-quadrant drives in parallel and for the connection of 12 generators or 12 generator windings.

Windmill design engineers have a number of aspects to take into their designs:

• High-power wind turbine;
• Low losses;
• Variable speed;
• High efficiency;
• The use of proved semiconductors;
• Clean, sinusoidal line current using a simple line transformer;
• Good line power factor and low THD;
• Active and reactive power control;
• Modular design to allow use with various currents and voltages, plus quick assembly;
• High degree of reliability;
• Lowest possible costs.

Best solution: the medium-voltage generator.

A medium-voltage generator is a must in high-power windmill designs of the future. Medium-voltage silicon, however, is not suitable for such applications. The right solution is therefore to connect base units in series.

For example:

A 5 MW windmill generator with 6.3 kV rated output voltage has output currents of 3 x 436 A_rms. The rectified variable speed generator voltage is in the range of 1 to 10 kVDC.

How can such variable voltage be connected to the grid?

Each windmill needs to have its own transformer to allow for connection to the grid; grid voltage would be in the range of 20 to 30 kV, which would be the transformer output voltage.

The transformer can be produced with several - in this case 10 - three-phase windings, each for 3 x 690 V, which are used as input voltages.

The new medium-voltage windmill principle is shown in Figure 6.

Figure 6: Cell-based medium-voltage windmill.

One base unit, a 600 kVA three-phase inverter, is attached to each three-phase winding. A fourth IGBT leg can be connected in front of each base unit. This arrangement can be referred to as a medium-voltage cell. All the cells can be connected in series, as shown in Figure 6.

If the IGBT switch of the fourth leg is switched off, the generator DC current will charge the cell DC link voltage.

The three-phase inverter on the cell-grid side discharges, controlling its own DC link voltage. For 3 x 690 VAC, the DC link voltage will be 1050.

Ten base units in series can produce a counter electro motive force (EMF) of up to 10.5 kV. The voltage remains balanced with the rectified generator voltage.

If the generator speed is lower, the generator voltage will be lower, too. For this reason, to control the rectified DC current, which in turn means controlling the generator torque, some of the cells have to be bypassed.

If five are bypassed, the remaining counter EMF is 5.25 kV. Bypassing more cells will increase the DC current and the generator torque. Bypassed cells can deliver full reactive power to the grid.

If one cell is not functioning, it will also be bypassed. The maximum cell DC link voltage is 1200. For this reason, even as few as nine cells in series can carry the rectified generator voltage of up to 9 * 1200 V = 10.8 kV.

Variable-speed wind turbines with medium-voltage synchronous generator have features that include:

• Generator DC voltage range from 0 to V_DCmax;
• DC voltage per cell 1050 V (with 1700 V silicon);
• VDC max per cell = 1200 V;
• Number of cells = V_DCmax/V_cell(+1);
• Cell power: P_genmax/number of cells;
• System redundancy (+1);
• Cell turn-on time varies from 0 to 100%;
• A switched-off cell can produce the full reactive power;
• High degree of efficiency at both low and high power values;
• Line side ripple frequency = N_cell * F_swcell;
• Simple line side transformer.

High-power applications use numerous IGBT modules. It is far better, however, to use more switches with separate controls, eg, several units connected in parallel or in series rather than one large single unit.

The advantages are:

• Good line power factor and low current THD with a lower switching frequency and fewer passive components;
• Modular design that is suitable for various powers and voltages, as well as quick assembly;
• The use of proved semiconductor elements;
• Greater efficiency;
• High reliability;
• Low cost per kW.

Written by Dejan Schreiber, senior applications manager, Semikron