Chapter 5
Test Bench for the Doubly-Fed Induction Generator based Wind Turbine Generator System
This chapter describes in detail the realized test bench. The dimensioning of major components is also discussed. It is organized such that Section-5.1 provides the test bench layout while Section-5.2 and Section-5.3 describe the power handling and the control system components respectively.
The developed setup is for a 10 kW DFIG based WTGS, controlled by Digital Signal Processing (DSP) cards from the manufacturer dSPACE [dSPACE 2011].
The Semiteach converter is from the manufacturer Semikron [SEMIKRON 2005].
The necessary transducers and hardware protection against short-circuit, over- voltage at the DC-link and over-temperature have been included inside chariots especially made for the test bench. The user’s manual attached as Appendix-D provides detailed information on the chariots. Special effort has been placed to ensure safe operation of the equipment and safety of the user. Apart from the functions provided by the IGBT drivers, the additional fault protection features designed are
• The measurement signals (low power) are accessible at the front while the high power currents and voltages are connected at the back of the chariot.
• Phase indicator lamps: to show the availability of each of the grid phases.
• Status LEDs: to indicate which of the contactors are currently operating.
• DC-link voltage panel meter: to indicate capacitor voltage at all times.
• DC-link over-voltage circuit: to avoid damage to the capacitors.
• Thermal over-load circuit: to avoid damage to the semiconductors.
• Audible alarm generator and latch function: to prevent further operation unless the cause of the fault has been removed.
• Desktop emergency push-button: to isolate the system quickly and from a safe distance, to avoid fault propagation.
5.1 Test Bench Layout
The layout of the test bench is presented in Figure-5.1. The components, marked with numbers, are listed in Table-5.1. The arrow, to the left of the figure, signifies the fact that the wind turbine drive train i.e. the blades, the low-speed shaft and the gear-box are to be emulated by a motor controlled by a drive. This drive takes its reference from a DSP card that runs the model of the drive-train. The other DSP cards control the electrical part of the generator. They send proper switching signals to the IGBT drivers and the measured currents and voltages are fed-back to the cards, after signal-conditioning and filtering. Several of the components have been recuperated from existing material in the laboratory due to large time delays, up to 4 months, in delivery.
Figure 5.1 – The DFIG based WTGS test bench.
Table 5.1 – List of components for the DFIG based WTGS test bench.
Component
No. Description Component
No. Description
1
Drive-Train Emulator-Motor
Drive (30 kVA)
2
Drive-Train Emulator-Motor
(15 kW)
3 Speed Encoder
(360 pulses/rev) 4 Torque Meter
(100 Nm) 5
Doubly-Fed Induction Generator
(15 kW)
6 & 7
IGBT based Converter
(20 kVA) 8 DC-link Capacitor
(1100 μF, 750 V) 9 Inductance Filter (6 mH, 30 A) 10
Machine Contactors, Circuit Breakers, Over-load Relays
(for a 15 kW machine)
11
Digital Signal Processing Card
(DS1104) 12
Digital Signal Processing Card
(DS1104)
13
Current and Voltage Transducers (50 A, 860 V) 14
Signal Conditioning + Filtering (±10 V, ≤ 500 Hz)
15 DC Power Supply (± 15V, 5 V, 24 V) 16 Desktop Computer (PC) 17 Charging Resistors
(330 Ω, 300 W) Miscellaneous Cables, Connectors, Panel Meters, Indicator Lamps etc.
(5 % added to the total component costs)
5.2 Power Handling Components
Proper dimensioning is required especially for components which handle power, such that they do not fail for routine operation.
5.2.1 Back-to-Back Converter
The 20 kVA back-to-back converter is formed using two commercially available Semiteach converters [SEMIKRON 2005]. Each converter has a diode bridge rectifier, a three-phase IGBT inverter, an IGBT chopper leg and DC-link capacitors. The drivers for the IGBTs, the heat-sink temperature measurement sensor, a protective thermal switch and a cooling fan are integrated in the converter module. The converter schematic is given in Figure-5.2.
Figure 5.2 – The schematic of the Semiteach converter [SEMIKRON 2005].
5.2.1.1 Component Description
5.2.1.1.1 IGBT Modules
The individual IGBT modules SKM 50 GB 123 D are rated at 1200 V, 50 A but the maximum current recommended by the manufacturer for the Semiteach unit is 30 A [Sargos 2008] [SEMIKRON 2005]. The input voltage signal level required to turn the IGBT on and off is +15 V and -15 V respectively. The typical turn-on threshold voltage is 5.5 V. The maximum saturation collector to emitter voltage is 3 V (3.7 V), for junction temperature of 25 ˚C (125 ˚C), at the rated current of 50 A [SEMIKRON 2006]. The peak collector current can be 80 A but with a maximum pulse duration of 1 ms. The IGBTs have a high short-circuit current capability of 500 A, at full DC voltage of 1200 V, but this short-circuit must be detected and removed within 10 μs otherwise there is a risk of thermal breakdown [Sargos 2008]. A maximum of 1000 short-circuit events are allowed, which must be at least 1 s apart. When the short-circuit current reaches 10 times the nominal, the IGBT starts to de-saturate, the voltage between collector and emitter increases and the current is limited [Sargos 2008].
5.2.1.1.2 IGBT Drivers
The SKHI 22A drivers can provide an output peak current of 8 A with the +15 V and -15 V required to switch the IGBT. The gate resistor value is 30 Ω. The drivers require a supply voltage of 15 V and consume 16 mA per driver. The maximum switching frequency for the drivers is 50 kHz. There is interlocking to prevent simultaneous switch-on for the IGBTs in the same inverter leg. The driver inserts an interlock dead time of 3.3 – 4.3 µs by default.
The drivers also provide an error output for protection at faults like short-circuit of the IGBT, by monitoring the collector to emitter voltage VCE, and the driver supply under-voltage 13 V. A short-circuit is detected within 5 μs [Sargos 2008].
The error input to output propagation time is 0.6 μs. In fault case, the input firing pulses are ignored and the output error latch is set. In order to reset the latch, both the signal inputs should be put to zero for at least 9 μs. The absence of a fault condition is indicated by a +15 V level at the error outputs. This error output has to be included when designing an application. The driver also has an integrated short pulse suppression function, which suppresses switching pulses less than 500 ns, caused by high frequency interference at the driver input [SEMIKRON 2008].
5.2.1.1.3 Snubber Capacitors
The MKP type snubber capacitors, rated at 22 μF, 1600 V, are mounted directly between the collector and emitter terminals of the IGBT modules. They limit over-voltage, due to the parasitic inductances, and the high rate of change of IGBT currents. A snubber also reduces the switching losses but this function is less important for an IGBT, since it can be switched at full current with full rated voltage e.g. 1200 V [Sargos 2008].
5.2.1.2 Component Dimensioning
For a DFIG, the power flows through the stator ps and the rotor pr are given as [Santos-Martin et al. 2006]
s 1 ps pe
≈ − (5.1)
s r s.p
p ≈− (5.2)
The selected rated slip s = – 0.12, for a rated power pe = 10 kW of the DFIG system, gives a rotor power pr equal to 1.1 kW. Now, the rotor voltage vr at a specific slip s is given as
s . v
vr s = r s=1 (5.3)
t s 1 r s
r v = v
= (5.4)
rt is the turns ratio between the stator and the rotor while vs is the stator voltage.
Turns ratio is an important factor for converter current dimensioning since a large ratio requires larger currents in the rotor, for the same slip power. It is therefore recommended to select a small turns ratio, if possible [Rabelo &
Hofmann 2005] [Petersson, Lundberg & Thiringer 2005] [Serban et al. 2006]
[Lindholm 2003]. The turns ratio for the generator used in the test bench is 1.91, which is rather large. With a rated line (phase-phase) voltage of 380 V, this gives a rotor line voltage around 24 V at the rated slip and the resulting rotor phase current is 26.6 A, which is within the recommended limit.
5.2.2 DC-link Capacitor
Electrolytic capacitors are mostly used in automotive and industrial applications from 1 kW to 1 MW and DC voltages from 200 V to 1000 V [Sargos 2008]. The advantage, when compared to other capacitor technologies, is greater capacitance per unit volume while the disadvantages are the higher ESR and the inability to withstand reverse voltages of a significant value. The heating losses, incurred due to current flow through the ESR, put a limit on the RMS value of the ripple current that the capacitor can withstand [Anwar & Teimor 2002].
5.2.2.1 Component Description
Two electrolytic DC-link capacitors, each rated at 2200 μF, 400 V, are connected in series to provide an equivalent 1100 μF, 800 V rating. The capacitor voltage balancing resistors are 22 kΩ each. The DC-link capacitor will discharge through the voltage balancing resistor across it with a time constant of about 48 s.
5.2.2.2 Component Dimensioning
The maximum DC-link voltage specified in the datasheet of the Semiteach converter is 750 V. The peak DC-link voltage ripple ∆vDC for the converter at rated conditions, Sconv = 20 kVA, CDC = 1100 μF, vDC = 750 V and ω1 = 314.16 rad/s, is calculated to be 38.6 V or 5 % of the average DC-link voltage using Equation (5.5) [Lindholm 2003] [Bojrup 1999].
sync DC
DC conv
DC 2.
. 1 C . v v S
∆ = ω (5.5)
For the rated slip power Sconv = 1.1 kVA with CDC = 1100 μF, vDC = 380 V and ωsync = 314.16 rad/s, it gives a 2.0 % peak-peak voltage ripple on the DC-link.
Apart from ensuring equal voltage drop on the series connected capacitors, the voltage sharing resistors provide a discharge path to the capacitors. This is important from a safety point of view. This Equation (5.6) gives the expression to calculate the voltage sharing resistance Rvsr1 for a capacitance value C1 [Evox Rifa n.d.]. For C1 = 2200 μF capacitance value this gives 30 kΩ which is close to the available 22 kΩ resistances on the Semiteach.
[ ] [
Ω]
µF k C
015 . 0 R 1000
1 1
vsr = × (5.6)
For high reliability, the power loss in the resistor should be less than 50 % of its rated value while the tolerance should be better than 5 %. The consequence of omitting the voltage sharing resistor is that the DC-link voltage is limited, due to tolerance in capacitance value. This is found using Equation (5.7) [Evox Rifa n.d.].
For an electrolytic capacitor of the type used in this converter, a tolerance of 20
% normally exists. For a vDC = 700 V, maximum tolerance Tmax = 1.2, minimum tolerance Tmin = 0.8 and number of capacitors in series nc = 2, the voltage vcap on the minimum value capacitor would be 420 V which exceeds its rating.
min c
max
max DC
cap T (n 1).T
T . v v
−
= + (5.7)
The determination of the capacitor combination, series or parallel, is dependent on voltage rating and the calculation of power losses. The power losses depend on the ESR value and the RMS ripple current through the capacitor, which are application dependent themselves. The current through the capacitor consists of different frequencies, which depend on the converter switching frequency and the duty cycle of the output [Anwar & Teimor 2002]. The power loss should be determined at each frequency, with the RMS value of current and the corresponding ESR, and added up [Evox Rifa n.d.]. The hot spot temperature, and thus the operational life time, can then be determined using the thermal resistance table usually available on request from the manufacturer. The ripple current limit, for the capacitor used in this converter, is 8.2 Arms at 100 Hz, 85 ˚C [Sargos 2008].
5.2.3 Inductance Filter
The idea of the inductance filter or the L filter is to put adequate impedance at the output of the converter to limit the flow of currents around the switching frequency. Although additional inductance is contributed by the grid interface transformer, but if the transformer leakage inductance is significant i.e. the grid is weak, the harmonic currents can cause voltage harmonic distortion at the low- voltage side of the transformer [Chen 2005]. This can disturb the PLL on which the complete control loop relies. Therefore, the major impedance contribution to the harmonic current flow should come from the inductance of the filter.
The advantage of using an L filter as opposed to an LCL filter is the simple design concept. The disadvantage is that, for a specific amount of attenuation required at a switching frequency, a larger overall value of inductance is needed. The larger the inductance, the greater is the voltage drop across it thereby requiring a larger converter output voltage and the required minimum DC-link voltage [Lindholm 2003]. For good dynamic current control performance an additional increase in DC-link voltage is often required [Lindgren & Svensson 1995]. The larger DC-link voltage leads to a higher ripple in the current component associated with active power and therefore a higher ripple in the active power [Lindgren & Svensson 1995]. Furthermore, the converter output has a higher dv/dt, when the voltage switches alternately between 0 and vDC, which can cause EMI problems for nearby equipment and insulation stress. Also the amplitude of the harmonics in the converter output voltage depends on the ratio of the DC- link voltage and fundamental output voltage. It is therefore recommended to use the PWM strategy that gives a higher converter output voltage thereby limiting the DC-link voltage required. The DC-link voltage is usually not set more than two times the phase-phase voltage [Lindgren & Svensson 1995].
To reduce the inductor size, a higher switching frequency can be used but the overall switching losses increase since there will be more switching events. It might also lead to sub-harmonic output currents due to non-ideal switching, if the time to turn an IGBT on and off becomes larger than a critical percentage of
the overall switching time period [Lindgren & Svensson 1995]. One solution to reduce EMI is also to turn an IGBT on and off more slowly but the sub-harmonic problem may appear, as explained above. To realize an inductance value for higher power ratings, the cost of the material and weight of the product also become a concern.
5.2.3.1 Component Description
The three L filter units are three-phase single-core 2 mH, 30 A, 50 – 60 Hz each, provided by the manufacturer Gestufir. These filter inductances are shown in Figure-5.3.
Figure 5.3 – The L filter units.
5.2.3.2 Component Dimensioning
The converter can be modelled as a superposition of two voltage sources, one with a frequency equal to that of the grid and the other with the switching frequency. On the other hand, the grid itself can ideally be modelled as a single voltage source at the fundamental frequency, while being a short circuit for other frequencies [Liserre, Blaabjerg & Hansen 2005]. Thus, the harmonic frequency circuit is just a voltage source with frequency dependent impedance at its output.
The transfer characteristics from converter voltage to phase current are given by Equation (5.8). ih, vh are the h order harmonic current and voltage of the fundamental frequency fsync respectively.
f sync h
h
L ).
f . h .(
. 2
1 v
i
= π (5.8)
The frequency response is shown in Figure-5.4 for different values of Lf. As a 1st order filter it has an attenuation of -20 dB/decade [Serban et al. 2006] [Lindgren
& Svensson 1995] [Peltoniemi et al. 2006]. The standards related to harmonic quality of currents exchanged with the grid are IEEE 519-1992 and IEC 1000-3-4 [Liserre, Blaabjerg & Hansen 2005] [Roufi & Lamchich 2004]. A criterion of a maximum THD of 5 % is usually set for the current exchanged with the grid [Liserre, Blaabjerg & Hansen 2005] [Peltoniemi et al. 2006]. This is met at a minimum fundamental-frequency RMS current of 10 A, with simulations carried out using an inductance of 6 mH and switching frequency of 1.95 kHz. The voltage drop over the filter, in this case, is 2.0 % of the grid line voltage 220 V.
From simulations, for a fundamental-frequency RMS current of 2.9 A equivalent to 1.1 kW slip power, the THD of the currents is 11 % but this current is only a fraction of the total current of the generator. The THD is calculated using Equation (4.24).
Figure 5.4 – Frequency response of the inductance filter transfer characteristics.
5.2.4 Emulator Motor
As mentioned in Section-5.1, the drive train of the WTGS is emulated through an induction motor coupled to the DFIG, as shown in Figure-5.5.
Figure 5.5 – The emulator motor (left) and the DFIG (right).
5.2.4.1 Component Description
The squirrel-cage induction motor selected for the emulator function is an IEC 34 15 kW, 400 V, 50 Hz, 1460 rpm machine from the manufacturer 2EC. The name plate is shown in Figure-5.6.
Figure 5.6 – Name plate of the induction motor for the wind turbine emulator.
5.2.4.2 Component Dimensioning
The catalogue data of the 15 kW slip ring induction machine, used as a DFIG, is given in Table-5.2 [VEM n.d.]. It is necessary to determine the rating of the emulator motor, which allows one to generate 10 kW from this machine. With a rated efficiency of 85 %, the input power required at the shaft is 10 kW/0.85 = 11.76 kW.
Table 5.2 – Catalogue data of the 15 kW, 400 V, 50 Hz, 4-pole slip ring induction machine SPH 160 M4 from VEM [VEM n.d.].
Stator Voltage
[ V ]
Stator Current
[ A ]
Speed [rpm]
Rotor Voltage
[V]
Rotor Current
[A]
Eff.
[%]
Power Factor
Inertia [kg.m2]
Y 380
∆ 220
Y 31.5
∆ 54.5 1440 205 45 85 0.85 0.128
The rated slip s for the DFIG is -0.12. This is equivalent to 176 rad/s or 1680 rpm.
The torque required at the shaft is therefore 11760/176 = 66.84 Nm.
To run the emulator motor at nm = 1680 rpm the output frequency fd of the drive, for a machine with poles p = 4, is 56 Hz determined using
120 . p n
fd = m (5.9)
The load capacity curves of IEC 34 motors are given in Figure-5.7 [ABB n.d.].
Curve 1 is the typical continuous load capacity curve of a self-ventilated motor.
The motor rated frequency and the flux weakening point are at 50 Hz.
The shape of the curve is defined by the required cooling and flux weakening.
Below 50 Hz, the cooling capacity of the self-ventilated motor is reduced while at frequencies above, the flux weakening region, the voltage cannot be increased further. Thus in both cases, the load capacity is reduced.
At 56 Hz, the specified load capacity is 90 %. Thus the nominal torque of the emulator motor is 66.84/0.90 = 74.27 Nm. Finally, the rated power of the required emulator motor is 74.27*1500*(2*π/60) = 11.67 kW.
The higher standard offering after 11 kW is 15 kW. Thus a 15 kW squirrel-cage induction machine is selected for the emulator function.
Figure 5.7 – Load Capacity Curves for IEC 34 motors [ABB n.d.].
5.2.5 Generator
The name plate data of the DFIG is given in Table-5.2. The stator of the generator has been rewound to study winding short-circuit faults. The construction details of the machine are given in Table-5.3. All dimensions are in mm. Figure-5.8 shows the rotor.
Figure 5.8 – The DFIG rotor.
Table 5.3 – Construction details of the DFIG with dimensions in mm.
Stator
No. Property Value
1 Stack length 170
2 Stack diameter 176
3 End turns length (terminals side) 60 4 End turns length (opposite side) 50 5 Winding diameter (internal) 185 6 Winding diameter (external) 230
7 Slots 48
8 Slot opening 2.65
9 Slot width 7
10 Slot depth 21
11 Conductors in parallel 2
12 Conductor diameter 1.6
13 Poles 4
14 Cable terminals 6
15 Coil pitch 11th slot
16 Turn length 700 approx.
17 Sensor PT 100
Rotor
No. Property Value
1 Stack length 169
2 Stack diameter 175
3 Slots 36
4 Slot opening 2.65
5 Slot width 8
6 Slot depth 34
7 Conductors in parallel 4
8 Conductor diameter 0.95
9 Coil pitch (concentric) 8th and 6th slot
10 Turns 15 approx.
11 Turn length 770 approx.
12 Skew angle 5˚
The stator winding scheme is presented in Figure-5.9. Figure-5.10 shows the rewound stator.
Figure 5.9 – Scheme of the rewound DFIG stator.
(a)
(b)
Figure 5.10 – The rewound DFIG stator. (a) Stator core. (b) Winding terminals.
5.2.6 The Encoder
The speed measurement is provided by an encoder integrated within an inline torque meter DR-2112 from Lorenz Messtechnik [Lorenz n.d.]. The torque output is ±5 V for 100 Nm with a sample rate 10 kSamples/s. Two TTL channels, A and B, provide 360 pulses per revolution, which are 90˚ phase shifted to allow the determination of the direction of rotation. The torque meter is shown in Figure- 5.11.
Figure 5.11 – The torque meter.
5.3 Control Hardware
The control hardware consists of data acquisition and data processing devices.
5.3.1 Current Transducers
Five current transducers are provided for each converter. The instantaneous value transducers, LT-100S from LEM, have a bandwidth DC…150 kHz and an output of ±9.8 V for a ±50 A current input (DC/AC/pulsed). The response time at 90 % of the nominal current is less than 1 μs while the accuracy and linearity are
±0.5 % and less than 0.1 % respectively [LEM n.d.].
5.3.2 Voltage Transducers
Five voltage transducers are provided for each converter. The instantaneous value transducers, LV100 from LEM, provide ±7.5 V output for an input of ±860 V (DC/AC/pulsed). The response time at 90 % of the nominal voltage is 20 μs to 100 μs while the accuracy and linearity are ±0.7 % and less than 0.1 % respectively [LEM n.d.].
5.3.3 DS1104 Rapid Prototyping System
The dSPACE DS1104 R&D Controller Board is based on MPC8240 processor with PPC603e core and on-chip peripherals. It is a 64 bit floating-point processor with a 250 MHz CPU. The slave DSP responsible for PWM generation is based on TMS320F240 from Texas Instruments. It is a 16 bit fixed-point processor with a 20 MHz clock frequency. The hardware interface panel is shown in Figure-5.12.
Figure 5.12 – Connector panel for the DS1104 R&D Controller Board.
Other features of the rapid prototyping system, related to the control system of the DFIG, are described below.
5.3.3.1 Analogue to Digital Converters (ADC)
The ADC inputs are present to interface with the transducers. They consist of 8 channels. ADCH1 to ADCH4 are multiplexed, 16 bit resolution, to a single ADC while ADCH5 to ADCH8 are four independent ADC, each with 12 bit resolution.
The input voltage limit is ±10 V. The conversion time for the multiplexed channels is 2 μs while for the parallel channels it is 800 ns.
5.3.3.2 Digital to Analogue Converters (DAC)
The DAC outputs are present to interface with actuators. There are 8 independent DAC channels. DACH1 through DACH8 have a 16 bit resolution and voltage output limits of ±10 V. The output current limit is ±5 mA.
5.3.3.3 Master PPC Digital Input/Output
The 20 parallel Input/Output (I/O) provided by the master processor are indicated as Digital I/O on the connector panel. They have a TTL level and a current limit of ±5 mA.
5.3.3.4 Slave PWM Digital Input/Output
The digital I/O controlled by the slave DSP and marked as Slave I/O PWM on the connector panel are reserved for functions related to the generation of PWM signals. They also have a TTL level but a higher current limit of ±13 mA.
5.3.3.5 Incremental Encoder Interface
The 24 bit digital incremental encoder interfaces, 2 in number, are present for speed and position measurement. They are single-ended TTL or differential RS422, input selectable, with a maximum input frequency of 1.65 MHz.
5.3.3.6 Serial Interface
Two serial Universal Asynchronous Receiver and Transmitter (UART), one with RS232 and the other, selectable, RS422/RS485 transceiver mode are also present.
5.3.4 Measurement Filter
The prototype of the 2nd order low-pass Butterworth measurement filter, discussed in Section-3.4.8, is shown in Figure-5.13. The electronic circuit equivalent to the transfer function Hf of the 2nd order filter, given in Equation (5.10), is shown in Figure-5.14. OP07 is the operational amplifier used.
1 G s
1 G
C 1 G s
G C C ) 1 s ( H
1 f 2 f 2 f 2 2 f 1 f
2 f 1 f f
+
+
+
= (5.10)
Gf1 and Gf2 are the respective conductance of resistors Rf1 and Rf2 while Cf1 and Cf2
are the capacitance values. This transfer function form implies that as long as the
ratio between the reactance and the resistance is maintained, the frequency response will remain the same [Maxim 2003]. This will help in adjusting the values of components.
Figure 5.13 – The measurement filter.
Figure 5.14 – Circuit of the 2nd order low-pass filter.
Equation (3.11) can be rearranged to the form of Equation (5.10). A comparison of the corresponding coefficients of s, in the denominator of the two functions, yields the components values. With the resistor values chosen to be Rf1 = Rf2 = 1 kΩ, the capacitance values are found to be Cf1 = 450 nF and Cf2 = 225 nF. Using standard values of capacitors Cf1 = 470 nF and Cf2 = 220 nF, the cut-off frequency is recalculated to be 494.95 Hz which is close enough to the design value of 500 Hz.
The delay and the attenuation introduced by the realized filter are given in Table-5.4 measured using the HAMEG 8030-5 function generator. For rotor electrical frequencies, maximum 6 Hz at s = -0.12, the delay is not noticeable even when observed with an oscilloscope.
Table 5.4 – Performance of the realized measurement filter with a sinusoidal input signal of 10 V peak value using HAMEG 8030-5 function generator.
Test No. Input Frequency [ Hz ]
Time Delay [ ms ]
Output Voltage [ V ]
1 5 0 10
2 50 0.4 10
3 500 0.5 7
5.4 Conclusions
This chapter has introduced the components related to the developed DFIG based WTGS test bench. Component dimensioning has also been discussed.
Complete details of the fabricated chariots such as the protection circuits and wiring schematics can be found in the user’s manual attached as Appendix-D.
Chapter 6
Experimental Results
This chapter describes the tests to determine parameter values of components and the experiments carried out on the test bench. It is organized such that Section-6.1 discusses the estimation of grid impedance, DFIG electrical circuit parameters and the delay of the converter. The experiments reported in Section- 6.2 and Section-6.3 are in the context of DFIG start-up, discussed in Section-3.7.
Section-6.2 demonstrates the operation of the GSC, with the charging of the DC- link capacitor and controller performance for reference tracking, disturbance rejection and reactive power generation. Section-6.3 follows the start-up steps of stator voltage generation, phase difference minimization and grid synchronization. A grid line voltage of 220 V is used. The simplified phase difference minimization technique, developed in Section-3.7.2.1, and the harmonic suppression property of the control system, highlighted in Section- 4.2.3, are shown to work in reality. The stand-alone operation of the generator is also tested to determine the performance of the voltage controller. An investigation into the erroneous oscillations observed in the RSC operation is carried out in Section-6.4. Section-6.5 is a brief treatment of the stator current frequency harmonics, for the rewound machine in the motor mode, to demonstrate the ones most affected by the winding asymmetry. Finally, some conclusions are drawn in Section-6.6.
6.1 Parameter Estimation Tests
The parameter values of interest need to be determined for simulations, to decide the controller gains and to account for non-ideal factors such as the converter delay.
6.1.1 Grid Impedance Estimation
In order to make simulations representative of the test bench, measurements to ascertain the grid impedance have been carried out. The three-phase Dy11 distribution transformer in the laboratory is rated at 100 kVA, 380 V/500 V and 50 Hz. The test circuit layout is presented in Figure-6.1. The voltage at the secondary side of the transformer can be changed in steps of 20 V. The measurements have been carried out at a line voltage of 100 V and maximum 86 A, to avoid exceeding the rated voltage limits of the devices used to load the transformer and to respect the transformer fuse limit of 100 A. Vo, ZT and ZL
represent the open-circuit voltage, the transformer impedance and the impedance of the load respectively.
Figure-6.2 plots the measurements carried out on the secondary side of the transformer. Two types of load were used for the test. RL represents the lighting load, which is predominantly resistive. It consists of 110 V, 115 W lamps, a maximum of 60 in each phase. XL is predominantly inductive and consists of three banks. Each bank is made up of three 25.68 mH inductances, which are delta-connected. Each bank is rated at 4500 VA, 50 Hz, Y 190 V / ∆ 110 V, Y 13.6 A
/ ∆ 23.6 A, cosΦ = 0.05. Therefore, the calculated impedance per phase of a bank is 0.401 + i8.056 Ω.
Figure 6.1 – Circuit layout to measure the grid impedance.
Figure 6.2 – Voltage regulation of the grid transformer.
An almost linear voltage reduction with a similar slope is observed for the two cases. The transformer per-phase impedance is calculated to be 0.018 Ω and the total three-phase short-circuit power Skg at 100 V is therefore 586.8 kVA.
6.1.2 Machine Parameter Estimation
The equivalent circuit parameters of the rewound DFIG have been determined by the classical no-load, blocked-rotor and DC resistance tests [Chapman 1999].
The name plate data of the DFIG are given in Table-5.2. For the no-load test, the emulator motor could not be decoupled from the DFIG since the speed measurement is given by the inline torque meter. For this reason the slip is larger. The rotor is short-circuited via a liquid rheostat and the stator is fed from the grid. The no-load measurements have been recorded in plots of Figure-6.3, where the measurement corresponding to the rated phase-neutral voltage of 220 V is marked. Figure-6.3(a) to Figure-6.3(f) present the stator current, the slip, the rotor current, the per-phase active power, the per-phase reactive power and the
estimated rotational losses respectively. Below 100 V, the slip is already quite large to assume the rotor as an open-circuit.
(a) (b)
(c) (d)
(e) (f)
Figure 6.3 – No-load test on the rewound DFIG to determine equivalent circuit parameters. (a) Stator current. (b) Slip. (c) Rotor current. (d) Active power per-
phase. (e) Reactive power per-phase. (f) Estimated rotational losses.
Table 6.1 – Blocked-rotor test on the rewound DFIG to determine equivalent circuit parameters.
Stator Voltage
[ V ]
Stator Current
[ A ]
Active Power [ W ]
Reactive Power [ VAr ]
Rotor Current
[ A ]
Rotor Speed [ rpm ]
22.1 16.35 185.8 315 28.85 0.0
Figure 6.4 – The calculated magnetizing inductance of the rewound DFIG.
Table 6.2 – The estimated electrical parameter values of the rewound DFIG determined with classical tests (referred to the stator).
Stator Resistance
Rs [Ω]
Stator Leakage Inductance
Lls [mH]
Magnetizing Inductance
Lm [mH]
Rotor Leakage Inductance
Llr [mH]
Rotor Resistance
Rr [Ω]
0.37 1.90 62.03 1.90 0.28
The blocked-rotor test, Table-6.1, has been performed for a stator current lower than rated, since the applied line voltage could only be increased in steps of 20 V.
The magnetizing inductance curve, determined from the tests, is shown in Figure-6.4. Finally, the determined parameter values, referred to the stator, of the electrical equivalent circuit are given in Table-6.2.
6.1.3 Converter Delay
The consequences of converter and control system delay are readily apparent when synchronization has to be carried out with the grid. This delay causes a current flow, as soon as the converter function is enabled, which activates either the DC-link over-voltage protection or the over-current protection. Figure-6.5 shows the typical delays of components in the test bench obtained from the datasheets. It includes the delay to generate the voltage, the delay of the measurement transducer and the delay of the ADC. The components are: the buffer GD74LS07, the driver SKHI 22 A, the IGBT module SKM50GB123D, the voltage transducer LEM LV-100 and the ADC of DS1104 [Datasheet4U n.d.]
[SEMIKRON 2008] [SEMIKRON 2006] [LEM n.d.] [dSPACE 2011]. tPLH, tPHL, tr, tf
and Tdt represent the propagation time for low to high logic, high to low logic, rise time, fall time and the dead time respectively.
Two types of tests have been carried out to determine the source and the dependence of the system delay. In one type, the sampling and the switching frequency are changed while a 50 Hz reference signal is given to the converter.
For the other type, the frequency of the reference signal is changed while the sampling and the switching frequency are kept the same. These tests determine the total system delay, starting from reference generation by the control system.
Figure 6.5 – System delay for generation of voltage through the converter.
The total system delay Tsd is determined experimentally in the form of phase difference θdiff, by recording both the reference voltage and the measured converter voltage in steady-state and doing an FFT on both. The equivalent time to the phase difference, for the reference voltage frequency fref, is found using Equation (6.1a). The results of the tests are given in Table-6.3 and Table-6.4. It can be seen that the major contributor to the total system delay is the selected switching frequency. When the switching frequency is doubled, keeping the same sampling frequency, the phase difference reduces almost by half. The net system delay, apart from the converter, is approximated by subtracting the switching time period Tsw from Tsd. It can also be seen that the net system delay is affected by the sampling frequency.
Table 6.3 – System delay with different sampling and switching frequencies at the reference voltage frequency fref = 50 Hz.
Test No.
Sampling Frequency
[ Hz ]
Switching Frequency
[ Hz ]
Switching Time Period
Tsw
[ ms ]
Phase Difference
θdiff [ ° ]
Equivalent Time Delay
Tsd
[ ms ]
Net System
Delay [ ms ]
1 7500 750 1.33 26.3 1.46 0.13
2 7500 1500 0.67 14.3 0.79 0.13
3 15000 1500 0.67 13.6 0.76 0.09
4 15000 750 1.33 25.6 1.42 0.09
Table 6.4 – System delay at different frequencies fref for reference voltage with 1.25 kHz switching and 10 kHz sampling frequency.
Test No.
Frequency of the Reference
Signal fref [ Hz ]
Phase Difference
θdiff [ ° ]
Equivalent Time Delay Tsd
[ ms ]
Net System Delay
[ ms ]
1 6 1.98 0.92 0.12
2 12 4.01 0.93 0.13
3 24 7.88 0.91 0.11
4 50 16.33 0.91 0.11
ref diff
sd f
. 1 T 360
θ
= (6.1a)
2 T 2 Tsd Ts + sw
= (6.1b)
The average total system delay can also be approximated by Equation (6.1b), provided the delay introduced by the components apart from the converter is small. Consider Figure-6.6, which has been adapted from [le Roux & van Wyk 2000]. The reference voltage amplitude and the step-wise output, fundamental- frequency voltage amplitude, from the converter can be seen. The instants when the reference voltage is sampled are marked, on the time axis, with solid vertical lines while the instants when the converter generates the corresponding output are marked with dashed vertical lines.
Figure 6.6 – Approximation of total system delay Tsd [le Roux & van Wyk 2000].
The converter delay is compensated by adding an equivalent angle to the angle used for the dq-abc conversion of the control voltage output [Park & Kwon 2004]
[Kuperman, Rabinovici & Zhong 2004]. This way the initial large current transient is avoided. Thus the converter can be connected to the grid.
Afterwards, the current control implicitly nullifies this angle by modifying the reference voltages in response to the current observed.
The switching waveforms are presented in Figure-6.7(a). The first two plots are the switching signals, for the top and bottom IGBTs of a branch, sent to the drivers. The last plot is the corresponding phase voltage at the output of the converter, measured w.r.t. the negative terminal of the DC-link. This figure is redrawn from the observation made using an oscilloscope, at a switching frequency of 1.25 kHz with 325 V at the DC-link. Figure-6.7(b) shows a typical
charging of the gate-emitter circuit of an IGBT [Sargos 2008]. This is the output of the driver. The gate resistor along with the, IGBT characteristic, gate capacitance can be approximated as an RC circuit. The gate resistor is used to improve the Electro-Magnetic Compatibility (EMC) performance and limit the rate of gate current and consequently the over-voltage due to parasitic inductance but causes switching losses.
(a)
(b)
Figure 6.7 – Converter switching.
(a) Input switching signal and the converter output.
(b) A typical driver output waveform [Sargos 2008].
The CMOS compatible IGBT drivers have the following characteristics, related to the switching performance [SEMIKRON 2008]:
• The maximum and minimum input thresholds are 12.5 V and 4.7 V for the high and low logic respectively.
• The maximum input to output turn-on and turn-off propagation time delay is 1.15 μs.
• An IGBT interlock dead-time of 3.3 – 4.3 μs is inserted by default and cannot be changed.
• The maximum frequency for the input switching signal is 50 kHz.
6.2 Grid-Side Converter Operation
This section discusses the performance of the GSC control. Tests for reference tracking, disturbance rejection and reactive power generation are carried out. A comparison between simulation and experimental results is also provided, where some difference is observed between the two despite efforts to model the system as accurately as possible. These differences can be due to the values assumed for components in simulations. For example, the actual value of ESR and the impedance of cables used to connect different components are not known. Nevertheless, the system model is close enough to the real setup to allow the use of same controller gains. The schematic of the GSC control part, from the overall DFIG system, is shown in Figure-6.8. Figure-6.8(a) shows the layout of the test setup components while Figure-6.8(b) shows the control system layout.
The switching frequency is 1950 Hz or 39 p.u., while the sampling frequency is 3900 Hz or 78 p.u. The compensation angle for converter delay is 13˚, determined experimentally for the sampling and switching frequency used. The gains of the PI controllers are found in Table-3.3.
The component values used in simulations are repeated here: Lf = 6 mH, Rf = 0.1 Ω, C1 = C2 = 2CDC = 2200 μF, RDC = 0.03 Ω, Rvsr1 = Rvsr2 = 22 kΩ, Skg = 586.8 kVA and the DC-link charging resistors Rch = 330 Ω. The actual X/R ratio of the grid is not known, thus for simulations a value of 2, 5, 7 and 9 was taken. The measurement noise has not been considered to observe the differences clearly. In general, the X/R ratio of transformers increases with their power rating. An X/R value of 7 is the default value suggested by SimPowerSystems while an X/R value of 2 was taken from the ANSI Standard C37.010 for a 100 kVA power transformer [Arcadvisor n.d.].
+ -i +v -
Switch
Step
Rvsr 2 Rvsr 1
Idc RSC2 Idc
RSC1
Van Vbn Vcn Va
Vb Vc n
Grid Voltage Measurement
N A
B C
Grid 220 V, 50 Hz Skg = 586.8 kVA IdcRSC
Vdc Vgrid
If g
A B C +
- GSC pulses
A B C A B C
Filter Inductance Lf = 6 mH
s -+
Currrent
Source Vabc
Iabc A B C
a b c Current Measurement
A B C
a b c Contactor
A B C A B C
Charging Resistance Rch = 330 Ohm C2
C1
(a)
Ifd*
Vdc*
Ifd
Ifq
Vd,grid
Vq,grid
thetaPLL
Vdc
Vfa
Vfb
Vfc
Vector Control Vdc*
2
Vdc*
1
Switch 2 Switch
1
Step 2 Step
1
Vfa
Vfb
Vfc
Vdc DC_a
DC_b
DC_c SVM_PWM
PWM On/Off PLL
Enable
Vac
Vbc
Vd,grid Vq,grid thetaPLL fPLL PLL
Ifd*
2
Ifd*
1
Vgrid_ac Vgrid_bc Grid Voltage Measurement
pulses
Vdc
DC_a
DC_b
DC_c
PWM Stop Pulses
DS1104SL_DSP_PWM Ifa
Ifb Ifc Current Measurement
Control On/Off
ifa ifb ifc thetaPLL
Ifd Ifq (a,b,c)--> (d,q) Power Invariant
(b)
Figure 6.8 – Schematic of the GSC system. (a) Test setup. (b) Control system.
Figure 6.9 – GUI for the GSC control in dSPACE Control Desk software.
Figure-6.9 shows the Graphic User Interface (GUI) developed in dSPACE Control Desk software. The experiment can be controlled in steps using the On/Off buttons. The status of I/O can be seen and the captured variables displayed. The
layout is divided into different sections, like the ‘start-up’ to charge the DC-link capacitors via the anti-parallel diodes of the GSC, the ‘vector control’ to boost the DC-link voltage to the operational level and the ‘acquisition’ section to control how the data is acquired.
6.2.1 Reference Tracking
The reference tracking performance is tested through charging of the DC-link capacitors up to the operational DC voltage of 380 V. Figure-6.10 presents the experimental results of the charging process and the performance of the PLL. At time t = 1 s, capacitor charging begins via the charging resistors. The resistors are shorted out at t = 12 s and the DC-link voltage achieves the nominal value of about 320 V for the grid voltage of 225 V, see Figure-6.10(a). The PLL is enabled at t = 13 s and its recorded output is shown in Figure-6.10(c). At t = 15 s, the control system is enabled with a reference voltage of 320 V for the DC-link. A transient is observed in the DC-link voltage when this is done. This is due to small error between the reference and the measured value resulting from the measurement noise, for example. After t = 18 s, the reference voltage is stepped up to 380 V and at around t = 23 s it is stepped down to 320 V again. The currents measured on the grid side are shown in Figure-6.10(b) and Figure-6.10(d). It is seen in the close-up provided that although the currents are periodic they are dominated by harmonics. The current harmonic content is presented in Figure- 6.10(e), where the harmonic magnitudes Ih are reported with respect to the magnitude I1, of the 50 Hz fundamental frequency component, and the numbers indicate the harmonic order.
(a) (b)
(c) (d)
(e)
Figure 6.10 – DC-link capacitor charging. (a) Capacitor voltage. (b) Filter currents. (c) PLL output. (d) Filter currents – close-up. (e) FFT ifb.
A comparison between experiments and simulations is presented in Figure-6.11.
The experimental results are in the left-hand column while the simulations are in the right-hand column. Figure-6.11(a) is a close-up of Figure-6.10(a) and focuses on the time during which the control is ON. In Figure-6.11(b), two cases are presented. For X/R = 2, the controller is already ON when a step change, at t = 1 s from 320 V to 380 V, in voltage reference is made. For all others, the controller is turned ON with 380 V as reference. It can be seen that the overshoot is larger in the latter case. It is due to the fact that it takes some time for the controller to establish the right voltages as reference to the converter since the initial condition for the integrators of the PIs is zero. A similar observation, for the same application, has been reported in literature with 1.5 p.u. current flow due to the control system delay [Teodorescu et al. 2003]. This is easily understood, since the grid can readily source current while the response time of the control system to achieve the proper converter voltage, to limit this current, is larger.
The reference is achieved in about 200 ms.
Figure-6.11(c), (d) and Figure-6.11(e), (f) provide the comparison for ifq and ifd
current components respectively. For the experiments, the increased oscillations in the dq component currents in steady state are due to offset errors of the transducers and measurement noise. For the simulations, there is only a slight difference in the reference tracking performance for different X/R ratios where a larger overshoot is seen for a higher value of the ratio, see Figure-6.11(b).
(a) (b)
(c) (d)
(e) (f)
Figure 6.11 – Comparison between experiment (left-hand column) and simulation (right-hand column) for reference tracking performance of the GSC
control.
(a) and (b) DC-link voltage. (c) and (d) ifq. (e) and (f) ifd. 6.2.2 Disturbance Rejection
The disturbance rejection performance is tested by connecting a load at the DC- link. For the experiments, a resistive load consisting of lamps while for the simulations a controlled current source, shown in Figure-6.8(a), is used. A current of 3.15 A is drawn at the operational DC-link voltage of 380 V. The disconnection of this load is equivalent to the injection of current in the DC-link, from the RSC. It corresponds to a power injection of 1197 W. Figure-6.12 presents the disturbance rejection test. The experimental results are in the left- hand column while the simulations are in the right-hand column.
(a)
(b) (c)
(d) (e)
(f) (g)
(h)
(i)
Figure 6.12 – Comparison between experiment (left-hand column) and simulation (right-hand column) for disturbance rejection performance of the GSC
control.
(a) Lamp load current. (b) and (c) DC-link voltage. (d) and (e) ifq. (f) and (g) ifd. (h) Filter currents. (i) FFT ifb.
Figure-6.12(a) shows the lamp load current. A circuit breaker is used as a switch to manually connect and disconnect the load. Therefore, the connection is not instantaneous due to the spring loaded mechanism. The simulations are started with a DC-link voltage of 380 V. At t = 0.1 s, the control is enabled and a transient can be seen. The load is connected at t = 1.3 s and disconnected at t = 4 s. For the experiments, the load was connected and disconnected at t = 3 s and t = 8.5 s respectively. A comparison of Figure-6.12(b) and Figure-6.12(c) indicates a notable difference in the DC-link voltage waveforms. The X/R ratio does not seem to make much difference in these simulations. The filter currents are shown in Figure-6.12(h) and the harmonic content is presented in Figure-6.12(i).
6.2.3 Reactive Power Generation
The current control loop performance has been tested by injecting reactive power into the grid. A step change in ifd from 0 A to 9 A is made at the DC-link voltage of 380 V. This is equivalent to an injection of about 2 kVAr. Figure-6.13 presents the comparison for reactive power injection into the grid. Here a significant difference is observed for the DC-link voltage comparison shown in Figure-6.13(c) and Figure-6.13(d). Figure-6.13(g) shows the filter currents while Figure-6.13(h) shows the current harmonic content.
(a) (b)
(c) (d)
(e) (f)
(g)
(h)
Figure 6.13 – Comparison between experiment (left-hand column) and simulation (right-hand column) for reactive power generation performance of
the GSC control. (a) and (b) ifd. (c) and (d) DC-link voltage. (e) and (f) ifq. (g) Filter currents. (h) FFT ifb.
6.3 Rotor-Side Converter Operation
The schematic of the RSC control part, from the overall DFIG system, is shown in Figure-6.14. Figure-6.14(a) shows the layout of the test setup components while Figure-6.14(b) shows the control system layout. The voltage source shown on the DC-link in Figure-6.14(a) signifies that the voltage is maintained constant either through the GSC or the diode bridge rectifier supplied from the grid.
Figure-6.15 shows the GUI developed in dSPACE Control Desk software. The experiment can be controlled in steps using the On/Off buttons. The status of I/O can be seen and the captured variables displayed. The layout is divided into different sections, like the ‘start-up’ to measure the grid voltages and activate the PLL, the ‘vector control’ section for stator voltage generation and synchronization with the grid etc.
v +- Vabc
Iabc A B C
a b c Stator Output Measurement
Speed
Rvsr 2 Rvsr 1 Vabc
Iabc A B C
a b c Rotor Current Measurement
g A B C
+
- RSC A Vabc
B C
a b c Grid Voltage Measurement
N A B C Grid 220 V, 50 Hz
Vs
Speed Vdc
Ir Is
Vgrid pulses
w m
A B C
a b c DFIG A
B C
a b c Contactor
C2 C1
(a)
Switch 2 Switch
1
Step 2 Step
1
Var
Vbr
Vcr
Vdc DC_a
DC_b
DC_c SVM_PWM Qs*
2
Qs*
1 Ps*
2
Ps*
1
PWM On/Off PLL
Enable
Vac
Vbc
Vd,grid Vq,grid thetaPLL fPLL PLL
Ir Vs Is Measurements
Vac,grid Vbc,grid Grid Voltage Measurement
pulses
Vdc Speed
DC_a
DC_b
DC_c
PWM Stop Pulses
DS1104SL_DSP_PWM Ps*
Qs*
Ir Vs Is
Current Control Enable poz_error_enable Voltage Control Enable Power Control enable Vd,grid
Vq,grid thetaPLL fPLL Speed Vdc
Var
Vbr
Vcr
Control System Control
On/Off3 Control On/Off2 Control On/Off1 Control On/Off
(b)
Figure 6.14 – Schematic of the RSC system. (a) Test setup. (b) Control system.
Figure 6.15 – GUI for the RSC control in dSPACE Control Desk software.
As mentioned at the beginning of this chapter, the RSC operation has some problems. Erroneous periodic oscillations are observed as shown in the next section. Nevertheless, the tests done show that the steps defined in Section-3.7.2 for DFIG start-up are indeed valid. Furthermore, same parameter and gain values are applicable here. The results of the investigation carried out to isolate the cause or determine the conditions under which this oscillation problem occurs are reported in Section-6.4. The sample-and-hold technique presented in Section-3.7.2.1, to minimize the phase difference between the stator and the grid voltage due to the unknown position of the rotor when the control is activated, is tested here experimentally. It has not been found in the consulted literature and is a contribution of this thesis.
6.3.1 Generation of Stator Voltage and Synchronization with the Grid This section corresponds to the discussion in Section-3.7.2 regarding the generation of stator voltages, correction of the initial rotor position error and the synchronization of stator voltage to the grid before connection. The results of the same experiment have been divided into Figure-6.16 and Figure-6.17 to make the commentary easier. The sampling and switching frequency are both 1950 Hz while the bandwidth of the current and voltage control loop is 0.07 p.u. and 0.007 p.u. respectively on a 50 Hz basis, as used for simulations in Section-3.7.
(a) (b)
(c) (d)
(e) (f)
(g)
Figure 6.16 – Stator voltage generation and rotor position error correction.
(a) Rotor dq currents close-up. (b) Stator and grid dq voltages close-up.
(c) Speed. (d) Stator and grid dq voltages. (e) Rotor dq currents.
(f) Correction angle. (g) Synchronized stator and grid voltages.