Partial Discharge Measurements PD are a partial breakdown of gas inclusions in insulation where the electric field intensity exceeds the breakdown field strength PD transforms part of the capacitive stored energy into heat and radiation as well as mechanical and chemical energies which can degrade insulation materials This ageing process progressively reduces the insulation thickness and the breakdown voltage until the failure occurs Insulation erosion is  very fast in organic materials (Type 1)  slow in organic/inorganic insulation systems (Type 2)

Electric signals related to currents, electromagnetic waves, electroluminescence are recorded using proper couplers and processed in order to: provide information on the discharge phenomenon  establish PDIV as a function of the supply square wave parameters (mainly frequency and over-voltages)  check the quality of the insulation and compare different materials (PDIV and time-to-breakdown are currently adopted as parameters to be monitored during life tests) 

Ref.- IEC 61934TS: Electrical insulating materials and systems – Electrical measurement of partial discharges (PD) under short rise time and repetitive voltage impulses

PD Basics The local breakdown generates a voltage drop and a consequent fast impulsive current absorption from the supply A suitable test circuit configuration allows to record the PD pulse signals. It is composed by 

Sinusoidal/square wave generator

Couplers and filters  Synchro Units  Digital recorders  Processing unit 

Waveform Generators

Sinusoidal :

Square Unipolar

PWM-like :

PWM+peaks :

Square Bipolar 

with variable Vrms, V0p, Vpp, RR, RT



with selectable duty and wave-shape



able to supply high capacitive currents

e.g., Vpp= 07.5 kV Unip./ Bip. dV/dt=075 kV/µs Freq.= 020 kHz Duty%=0.0190

Typologies of Samples Under Test Twisted Pair (frame to test T2T insulation according to IEC 60851-5) Motorettes (frame to test T2T, P2G and P2P insulation)

Complete coils & stators

PD Signal Characteristics Partial Discharges generate impulsive signals having up to 1-2 GHz of frequency content Fast voltage transition can reach 200 MHz depending to the voltage RT

0.05

1.5

0.025

0.75

0

0

-0.025

-0.75 PD pulses Generator voltage

-0.05

0

0.5

1

1.5

2 2.5 Time (us)

3

3.5

-1.5 4

Generator voltage (kV)

PD magnitude (V)

Suitable couplers and bandwidth must be selected to avoid the commutation interference during PD measurements

Coupler and high-pass filters having a low cut-off frequency higher than 250-300 MHz are required to avoid commutation interferences

IEC 61934 Frequency Band Prescriptions

Volt. RT: 50ns PD RT: 2 ns

800 800

8th order filter with filter cut-off frequency equal to 500 MHz.

800 800

0,15 0.15

0,04 0.04

600 600

600 600 0.10 0,10

400 400

-200 –200 -0.05 –0,05

AppliedVoltage voltage V[V] Applied

0.00 –0,00

–0,00 0.00

00

-200 –200 –0,02 -0.02

Filtered PD signal V

00

0,02 0.02

200 200

Filtered PD Signal [V]

0,05 0.05

200 200

PDSignal signal [V] V PD

AppliedVoltage voltage V[V] Applied

400 400

-400 –400

-400 –400 Square kHz applied voltage Squarebipolar Bipolar1010 kHz Applied Voltage PDsignal signal PD

-600 –600

-0.10 –0,10

Square kHz applied voltage Square bipolar Bipolar1010 kHz Applied Voltage Filtered FilteredPD PDsignal signal

-600 –600

–0,04 -0.04

–800 -800

-800 –800

–0,15 -0.15

-2 –2

-1 –1

0

1

2

3

4

5

–2 -2

–1 -1

0

1

2

3

4

5

Timee s Tim [s]

Time Time [s] s

IEC 553/06

IEC 552/06

Different Low Cut-Off Frequencies

0.15

Commutation 0.10

PD signal

PD signal [V]

0.05

0.00

-0.05

-0.10

10 MHz 100 200 500 MHz -0.15 -2e-6

-1e-6

0

1e-6

2e-6

3e-6

4e-6

Time [s]

Differential Configuration A differential connection of coupler can be adopted to cancel the interferences due to voltage transitions The capacitance of CC and the SUT must be similar Signal relevant to the voltage transition reaches the +/- input of the LF filter and it is cancelled F.Guastavino et al., “Measuring PD Under Pulsed Voltage Conditions”, IEEE Trans. on Diel.El.Ins. Vol.15, pp.1640-1648, December 2008

PD Detection Circuits and Couplers Capacitive Couplers PD free devices To be connected with a proper detection impedance to obtain high cut-off frequencies

Test object

Supply

Z

Filter

PD signal

IEC 540/06

Inductive Couplers

HFCT Supply

Test object

Filter

With very high cut-off frequency PD signal

IEC 543/06

Test object

Supply

Filter

PD signal

HFCT IEC 544/06

Antennas Antenna Supply

Test Object

Suitable for measurements in applications

Acquisition System

Supply

High cut-off frequencies

Test object

Antenna Acquisition system IEC 545/06

PD ASD

Light Detectors

Description

Advantage

Disadvantage

Electrical

Optical

RF / EMI

Acoustic

Electrical circuit that picks up current pulse produced by charge transfer during partial discharge

Measures light emission from partial discharges

Measures radio frequency interference generated by the discharge

Measures the acoustic emissions produced by a partial discharge.

A good sensitivity and standard Non-contact, applicable for all HV equipment during for all voltage types. manufacture Allows testing of equipment in real conditions

Non-contact, applicable Non-contact, applicable for all voltage types. for all voltage types. Allows testing of Allows testing of equipment in real equipment of real conditions conditions

Sensitive to electrical noise. Insensitive to any form of Depending on Sensitive to other Cannot test circuit in operating internal partial discharge. equipment being tested, acoustic emissions. condition in most cases. Most Sensitive to light and EM emissions can Signals cannot always commercial equipment can highly directional. prevent detection of PD propagate through only test at up to 400Hz insulation / casings

Measurement Systems Very large bandwidth oscilloscopes (up to 2GHz)  Quite difficult to interpret the PD measurement results  Data must be organized to synthesize the information Commercial instruments are available Must of them require suitable input filters to tune their input requirements to record PD and reject voltage commutations (>200-500MHz)

Voltage Pulse

PD using Antenna

Residual Noise

PD using Voltage divider

PD Patterns The amplitude and the phase of the PD signal are evaluated The PD pulse sequence is transformed in a sequence of Dirach functions having the same phase

v

v

A 3D histogram is obtained considering the number of discharges having the same amplitude and the same phase of occurrence

A 2D histogram can be derived from the PRPD pattern projecting n in A, plain and using different colors to evidence the different repetition rate A [V]

φ [deg]

Reference sin wave

Two commutations per period

Square wave Volt. DC=50 % Sync frequency = Commutation frequency

square wave

PDIV

Two sub-patterns close to the zero crossing At least 1 PD per pulse Possible PD even during the “flat zone” of the voltage pulse

A low pass filter is required to obtain the phase reference and to derive the so called “Phase Resolved” PD pattern

PWM EUT

VOLTAGE SUPPLY

Voltage divider LP FILTER SYNC Ch DETECTION UNIT

Modulating wave

Example of PRPD pattern when a PWM voltage is adopted Modulating wave 50 Hz • Period of 20ms

Pulse Rep.Rate 1kHz • 40 commutations per period

40 subpatterns per period

Life-test Results twisted pair specimens were fed till failure by different wave forms at several amplitudes, in the presence of PD Data were processed according to the standard procedure PRPD patterns were recorded during the experiment Weibull Plot:

tF  F (t )  1  exp( ) 

IPL model:

tF  kV  N

Result 1: Unipolar-Bipolar Square Waves Twisted pair of having insulation of different materials were tested in air (with PD) and immersed in oil (no PD) Sinusoidal, unipolar and bipolar square waves, the latter having the same RT, RR (50 Hz, 10 kHz), DT=50% and different amplitudes were used in the test Let V0p and Vpp the 0-to-peak and peak-to-peak values

When V0p = Vpp with or without PD, the life curves relevant to unipolar and bipolar pulses are completely overlapped NO PD

PD

The Jump Voltage is the real voltage stress that affects the insulation ageing VEC can be adopted to compare the performances of different materials even in the presence of PD D.Fabiani et al. “Ageing acceleration of insulating materials for electrical machine windings supplied by PWM in the presence and in the absence of partial discharges”, IEEE ICSD pp.283-286, 2001

Result 2: PWM wit and without Overvoltages The picture is much more complicated when PWM and PWM+over-voltages are adopted to stress the materials in the presence of PD Due to the “equivalent derivative effect”, there are different Jump Voltages V

Vaa

Vpp

V0p = voltage amplitude

x

x t

=

overvoltage

Vpp = 2(V0p +x) Vaa = V0p +2x

F.Guastavino et al. “Life Tests on Twisted Pairs Subjected to PWM-like Voltages”, IEEE ICSD pp.238-241, 2004

Result 3: Phase Resolved PD patterns Sinusoidal

V = 1.83 kV

V = 2.44 kV

V = 3.05 kV

V = 3.80 kV

Square

V = 1.60 kV

V = 1.9 kV

V = 3.96 kV

V = 4.32 kV

PWM

V = 1.63 kV

V = 2.73 kV

V= 3.53 kV

V= 4.32 kV

PWM+Over-voltage

V = 2.90 kV V = 2.08 kV

V = 2.38 kV

V = 2.50 kV

Discussion  PWM voltage waveforms can promote PD activity, reducing the life of magnet wires  The predominant factors explaining this behavior are peak-to-peak voltage and switching frequency.  Even in the absence of PD, PWM voltage waveforms can accelerate the intrinsic aging of winding insulation (overvoltages). Very high slew rate (>5 kV/s), in fact, plays also a non negligible role on acceleration degradation due to increased voltage stress and heating. This effect can be evaluate resorting to space charge measurements both in the presence and in the absence of PD

Space Charge and PD Activity Space Charge = charge trapped inside the insulation or on interfaces (depends on the material, the poling field, the temperature and the supply voltage waveform) Above 10 Hz, SC is not trapped appreciably in insulation unless the waveform contains a DC component IN HF pulse waveforms, SP is accumulated mainly by PD Neglecting the SP accumulated in the bulk, the electric field in air, at the insulation surface, E*0 can be: Where: E0 : electric field without accumulated charge

E

0 : the permittivity of air andinsulation d l0 : the insulation thickness and the half air gap 0: the surface charge density

*

0

 sd  E0   0 d  l 0

Behaviour of the electric field in two enamelled wires in a twisted pair configuration

This configuration helps explain why the main ageing factor is associated to the jump voltage for bipolar and unipolar voltage waveforms

Let E*0(t) the behaviour of the electric field in the air gap before and after a PD event (gray line) E0(t) the applied field (black line) -E0p +E0p the bipolar voltage amplitude 2E0p the unipolar voltage amplitude and let E0>PDIV ESC the drop of the electric field due to the PD charge injection

After PD, a residual value of the electric field is:

Eres  E *0  Esc  0 E

*

0

 sd   0 d  l 0

If the charge injected by PD is not rapidly depleted, E*0(t) due to SC deposited on the insulation surface remains constant

When E0 change its polarity, E*0(t) increases by 2Ep in both cases

E*0 2Ep  Eres  PDIV

The influence of the JV was experimentally supplying 4 different kinds of enamelled insulated wires with sinusoidal, unipolar and bipolar square waves in the range of 50 to 10 kHz and with a RT of 50 ns above the PDIV

SC measurement becomes a significant tool to evaluate the ability of the different materials to deplete the SC injected by PD and to compare different enamels

PEA SP Measurement System A Pulse Electro-Acustic method ha been developed to measure the space charge accumulation on magnet wires

(D.

Fabiani et al. “ Relation Between Space Charge Accumulation and Partial Discharge Activity in Enameled Wires Under PWM-like Voltage Waveforms”, IEEE Trans. on Diel.-Elect.Ins., Vol.11, pp.393-405, June 2004)

The specimen is positioned an aluminium ground plate and a semicon-absorber and fed by HV DC power supply A voltage pulse of A=300 V, 10 ns width, 110Hz of RR, is applied through a 220 pF coupling capacitor

PVC Insulation

Semicon Adsorber 2 M

Enameled Insulation Copper Aluminum Ground Plate

220 pF

Amplifier HVDC

Pulse Oscilloscope

Piezoelectric (PVDF) Adsorber (PMMA)

PC GPIB IEEE-488 Acquisition Board

 Charges present in insulation are forced by the fast electric pulse and a pressure wave is generated from the interaction between charges and the material structure  The pressure wave propagates through the insulation and reaches the ground electrode under which the piezoelectric trasducer (PVDF) is located  PVDF generates a voltage signal (PEA output signal) proportional to the pressure wave propagating through it  A proper calibration procedure allows the PEA output signal to be correlated to the amount of trapped space charge  PEA signal is amplified and sent to a recording system  110 PEA outputs per second, synchronized with the rectangular supply voltage, are processed to obtain the charge profile

Test Procedure SC measurements are performed according to a specific polarization/depolarization procedure Polarization: the electric field is applied (volt-on, VP=1 kV) for a period tp=3600 s to achieve steady state conditions for the accumulated charge Depolarization: depolarization (volt-off) follows polarization. It is obtained removing the supply voltage and grounding the high voltage electrode, and lasted 3600 s as well.

t =20 s:  PEA signal is mainly due to the electrode field-induced charge (peaks indicate the electrode location, i.e., anode and cathode, corresponding to the positive and negative signal peaks, respectively)  the injected charge at the electrode-insulation interface is hidden by the electrode charge

t =3602 s:  SC in the insulation bulk can be observed looking at the PEA profile under volt-off (shaded area)  when the poling field is removed the electrode charge is considerably smaller than under polarization, being only due to the image charge (of the internal charge)

Space Charge Data Processing Space charge profile at the beginning of volt on: no space charge present Space charge profile 2s after voltage removal: space charge = gray area

To quantify the charge accumulation:

TIPO A - PROFILO CARICA: Volt-ON 3600 s (+) Volt-OFF 3600 s 1.2

Depolarization characteristic

Charge [p.u.]

1.0

 Total absolute stored charge density (after grounding), QM x1

1 QVO (t )  Q( x, t ) dx  x1  x0  x0

0.8

0.6

0.4

0.2 1

10

100

Time [s]

1000

10000

 the slope of depolarization characteristic, s, is a measure of space charge dynamic, i.e. the speed of charge recombination / expulsion

Material Improvements PD when active, are the most important ageing factor in Type I insulation Even in the absence of PD, jump voltage and switching frequency can accelerate the intrinsic aging of winding insulation due to increased voltage stress and heating Solutions could come from the use of:  VPI technologies  mica-films for turn and strand insulation  with metal oxide or ceramic fillers (micro-fillers)  nano-scale technologies (nano-fillers)

VPI Technologies The stator can be totally impregnated adopting the VPI technology even for small size random wound machines The PD inception voltage is up to 60% higher compared with non-impregnated ones because air gaps are filled (not totally) by the impregnation The time to breakdown of the inter-turn insulation depends on the:  PDIV and intensity of PD  enamel thickness and its resistance against PD erosion But  due to small imperfections, longer lifetime is not guaranteed  VPI process for small machines is expensive

Mica-Films for Inter-Turn or Strand Insulation Currently adopted in MV and HV machines fed with pulsating voltages

They are also adopted in LV machines Improvements were found but  higher costs  slot efficiency reduction  increased dimensions

Micro Fillers Enamel or polymer film may also contain additives such metal oxides or ceramic materials with a natural resistance to discharges (Al2O3, TiO2, SiO2, Fe2O3..), has been adopted The structure is a multi-coating over the copper enameled wire with the PD shield layer and the surface protection coating

The basic idea is to substitute the large mica flakes with a high density inorganic materials that  form barriers for the tree growth  facilitate the space charge diffusion  facilitate the heat transmission  the endurance life is enhanced  no variations in the ground-wall insulation thickness  increased t2t PD and surge withstand capability But It become brittle and develop cracks when subjected to temperature variation in the presence of mechanical stresses Each micro-composite material must be evaluated carefully

Evaluation of Corona Resistant Materials  Accelerated life tests on twisted pairs with conventional and corona resistant (CR) insulation  Tests: - in the absence of PD (in oil), below PDIV - in the presence of PD (in air), above PDIV  Sinusoidal and distorted waveforms Objective: to draw life lines life models voltage endurance coefficient evaluation comparison among insulation systems

Procedure Specimens: twisted pair insulation

50 Hz sinusoidal

Traditional, #A CR #B CR #C

Above PDIV

#A, #B,#C

Below PDIV

#A, #B,#C

Above PDIV

#A, #B,#C

Below PDIV

#A

10 kHz sinusoidal Life tests

10 kHz square unipolar Above PDIV

#A, #B,#C

Duty cycle = 50% Slew rate = 1 kV/s

10 kHz square bipolar Above PDIV

#A, #C

Test Results Sinusoidal Life Test on #1, AIR and OIL, 50 Hz 10 kHz

• Dramatic PD effect especially at HF on #A

10000

#A, #C

Voltage (rms value) [V]

#B

VEC = 11.7

#C #A

VEC = 6.4

VEC = 10.6

#B

VEC = 8.7

VEC = 7.1

50 Hz SIN AIR #A 50 Hz SIN AIR #B 50 Hz SIN AIR #C 50 Hz SIN OIL #A 50 Hz SIN OIL #B 50 Hz SIN OIL #C 10 kHz SIN AIR #A 10 kHz SIN OIL #A

VEC = 4.5

1000 0.1

1

• #B (CR) is worse than #A at 50 Hz

VEC = 11.9

10

Failure time [h]

100

• #C good VEC and life times in the presence of PD

• Increasing frequency reduces life 1000 considerably, both with and without PD

Life test results and life lines for sinusoidal tests (50 Hz-10 kHz ) in the absence (in oil) and in the presence of PD (in air)

Test Sinusoidal Results Life Test on #1, AIR,

• VEC decrease with frequency for #A

50 Hz 10 kHz

10000

Voltage (rms value) [V]

#C

50 Hz SIN AIR #A 50 Hz SIN AIR #B 50 Hz SIN AIR #C 10 kHz SIN AIR #A 10 kHz SIN AIR #B 10 kHz SIN AIR #C

#A

#B VEC = 7.1

VEC = 10.6

• #B is the worst at 50 Hz but the best at 10 kHz (for high fields)

VEC = 6.4

#B VEC = 8.8

#A VEC = 4.5

VEC = 12.6

#C

1000 0.1

1

10

• VEC increase with frequency for #B & #C (life decreases)

100

1000

Failure time [h]

Life test results and life lines for sinusoidal tests (50 Hz-10 kHz ) in air for the three tested materials.

• #C good VEC and life times both at 50 Hz and 10 kHz

Voltage (peak-to-peak value) [V]

Test Results Life Test in AIR summary plot

#C #A #B

10000

#A

#B

#C

• Peak-to-peak voltage is the main stressing factor, besides pulse repetition frequency

#A 50 Hz SIN #A 10 kHz SIN #A 10 kHz UNIP #A 10 kHz BIP #B 50 Hz SIN #B 10 kHz SIN #B 10 kHz UNIP #C 50 Hz SIN #C 10 kHz SIN #C 10 kHz UNIP #C 10 kHz BIP

1000 0.001

0.01

• Effect of voltage shape negligible (experimental points fit the same line) 0.1

1

10

100

1000

10000

Failure time [h]

Life test results and life lines for sinusoidal (50 Hz-10 kHz ) and squarewave (unipolar and bipolar) tests in air for the three tested materials as function of peak-to-peak voltage.

Space Charge Meas. On #A and #B Tipo A - B valori di QM 12

#A #B

10

Material s [s-1] #A 2.5 #B 1.8

8

QM [C/m3]

Speed of charge expulsion, s, for 50 Hz squarewave voltage (1000 V peak)

6

4

2

0 DC

0.1Hz

50Hz

10kHz

Voltage frequency

Total absolute stored charge density, QM, as a function of frequency. Bipolar squarewave. Poling voltage: 1000 V peak

Note that:  QM decreases as the frequency increases  QM (#B) > QM (#A)  s (#B) < s (#A)

The CR solutions show different performance and behavior CR micro-composites must be evaluated carefully before they use D.Fabiani et al.”The Effect of Fast repetitive Pulses on the Degradation of Turn Insulation of Induction Motors”, Proc. of SDEMPED 2001, pp.289-293, Grado (I), 2001

Discussion  Pulsating voltage accelerates degradation both in air and in oil (i.e. with and without PD), but different VEC  CR materials frequencies

show

a

longer

life

at

higher

 The standard insulation, #A, seems to suffer significantly from PD activity and frequency increase  CR material, #B, withstands PD better than #A  PD increase due to PWM voltage waveforms  #B tends to accumulate much more charge than #A (>QM and 2 kHz si hanno scariche in testata) • Il campo elettrico cala all’aumentare Campo elettrico della conducibilità della gradatura Campo elettrico di scarica di scarica

Media conducibilità

Alta conducibilità

Typically: •

Insulation with no stress grading system (normally for Vn ≤ 4kV)

•

Insulation systems with anticorona coating within the slot ( 4kV ≤Vn ≤ 6kV)

•

Insulation with stress grading system (Vn ≥6kV)

Due to material discontinuity, high values of electric gradient affect the surface of the coil at the edge of the slot grading tape thus generating tangential surface discharges

The stress grading is designed solving the field equation

       ccU  ( * U )  0 t   and its solution allows to draw the electric field outside the magnetic core

 cosh k ( x  L) U ( x)  Va 1   cosh( kL)  

sinh k ( x  L )  E ( x )  Va k sinh( kL )

Hot spots due to PD, can be discovered considering the air breakdown strength (e.g., 2.3 kV/mm) and the electric field gradient

Stress grading materials are characterized by their resistance that can be constant or electric field dependent:

s  0 exp(nE ) 2/ 3

High values of n and low values of 0 increase the grading effect The stress grading is designed considering the number of layers and their coating length outside a slot portion After the material selection (0, n), the space distribution of the electric field is determined using FEM software tools considering the machine geometry and the groundwall insulation characteristics The correct choice of the number of layers and their length is evaluated checking the electric gradient (below of PDIV) and hot spots

Stress grading is currently designed for 50/60Hz applications Assuming the 0 reference the edge of the slot grading, the potential and the electric field distribution can be derived and analyzed The different behavior of a single layer stress grading with constant and exponential resistivity, is show

But the potential distribution is strongly related to frequency of the applied stress. Stress grading designed for ac is not able to operate at higher frequencies V [V] 36000 34000

30 kV f variabile

32000 30000 28000 26000 24000 22000 20000

50 Hz

18000

250 kHz

16000

1 kHz

14000 12000

20 kHz

10000 8000 6000 4000 2000 0 -80

-70

-60

-50

-40

-30

-20

-10

0

10

20

30

40

50

60

70

80

90

100 110 120 130 140 150

Distanza x dal termine del ricoprimento conduttivo [mm]

GEN.3.1

GEN.3.2

GEN.3.3

GEN.3.4

A multi-layer and longer stress grading having lower resitance, is required for higher frequencies

Vmax [V] 18000

4

17000

Mat.D

16 kV # 1,25 MHz Mod.1

16000 15000

3

14000

s = 2,7 mm

13000

Mat.A

12000

2

11000 10000 9000 8000 7000 6000

1

5000 4000 3000 2000 1000 0 -30

-25

-20

-15

-10

-5

0

5

10

15

20

25

Distanza x dal termine del ricoprimento conduttivo [mm]

SIM.1.2.1.A

SIM.1.2.1.B

30

35

40

45

50

Experimental Validation Two different stress grading configurations designed for standard (50 Hz) and PWM supply voltages have been tested by means of PD measurements

Both frames were supplied by HV rectangular wave-shape

PD measurements were performed using an antenna probe able to record PD pulses and the fundamental wave-shape adopted as the phase reference of PD

Moving the antenna probe, PD were localized at the edge of the 50Hz stress grading while only signals due to commutations were recorded on the other frame Thus confirming the validity of the stress grading design

Conclusions  Besides the advantages in using ASD, new problems rose due to the significant harmonic content of the power supply and the over-voltages generated by mismatch impedances in inverter/cable/drive connection  The insulation is subjected to increased electric, thermal and mechanical stresses and its life is shortened  Additional examined

stresses,

typical

of

ASD,

have

been

 New composite materials (micro, nano fillers) have been proposed  Specific test methods are developed  Specific standards are under discussion

Thank you for your attention!