Evaluation of Diagnostic Measurements Used on Sta-tor Insulation of Hydrogenerators

Thomas Brügger

High Voltage Laboratory, ETH Zürich, Switzerland thomabru@eeh.ee.ethz.ch

Abstract- Measurements of Dissipation Factor (tanδ/Δtanδ), Po-larization Index (PI) and DC High Potential Insulation Currents are widely used diagnostic tools to assess the condition of stator insulation in hydro generators. This paper presents long-term diagnostic data gathered from three 60 MVA generators in a Swiss hydropower station. The measurements were performed in 5-year intervals as part of regular maintenance since the rewind of the stators with a mica/epoxy insulation system in the early seventies. Measurement results of Dissipation Factor, Polariza-tion and DC High Potential tests are discussed with regard to their effectiveness in assessing the ageing condition of the stator insulation. After a brief review of the theoretical background, the strengths of each diagnostic method in detecting symptoms of insulation deterioration are identified. It is concluded that Pola-rization Index and DC High Potential Insulation measurements are best suited for the detection of humidity and contamination problems. Dissipation Factor measurements below partial dis-charge inception voltage are a robust indicator of long-term age-ing caused by oxidative degradation.

I. INTRODUCTION

While there exists a plethora of literature on the theoretical background of diagnostic measurements used to assess the ageing condition of stator windings in large rotating electrical machines, actual data that show the evolution of specific diag-nostic values during long-term operation are somewhat scarce.

This paper presents diagnostic data gathered from three 60 MVA generators with a rated voltage Un = 15.5 kV in a Swiss hydropower station over a range of 25 years. As part of regu-lar maintenance with an interval of 5 years (starting in 1976), electrical measurements including Dissipation Factor (tanδ/Δtanδ), Polarization Index and DC High Potential Insu-lation Current tests were performed on the stator windings to assess their ageing condition. Fig. 1 shows a quite constant operating time of about 4,000 hours per year until 1996 and a sharp increase to an average of about 8,000 hours per year thereafter. The number of annual starts averaged to 250 for all three machines and 90% of the production occurred during winter season.

II. DC HIGH POTENTIAL CURRENTS

DC high potential (hipot) current measurements are a com-

mon tool to detect problems like cracks and fissures, moisture absorption, tape separations, conductive surface contamina-tion, lack of epoxy cure, and internal delamination in mi-ca/epoxy insulation systems [1, 2].

Fig. 1. Cumulative hours of operation since the stator rewind in 1973

When a DC voltage is applied to the groundwall insulation of

a stator winding the measured current consists of three com-ponents. First there is a capacitive charging current which is determined by the winding capacitance and the rate of change of the applied voltage. After an applied voltage step, the ca-pacitive charging current decays exponentially to zero in seconds. The second component is an absorption current which is associated with the polarization of the dielectric ma-terial and can be expressed as ia = KCVt-n, where V is the ap-plied voltage step, t is time, and K and n are material-dependent parameters. The third component is a continuous leakage current which passes through the groundwall insula-tion and the stress-grading system of the endwindings and is an indicator for defects [3, 4].

The three generators discussed in this paper were subjected

to a graded step-stress test during which the hipot DC voltage is applied to one phase at a time with the other two phases grounded. The voltage is increased stepwise after precalcu-lated intervals such that the absorption current at the end of each interval is a linear function of the test voltage [5, 6]. This method allows for a sensitive detection of significant leakage current as the leakage current clearly manifests in a strong non-linear growth of the total current above a threshold test voltage UT.

Fig. 2 shows the test results of generator 1. In the curve for

1981 (8 years after commissioning) the measured current grows linearly due to the increasing absorption current until a voltage of approximately 1.3 Un where the non-linear leakage

978-1-4244-6301-5/10/$26.00 @2010 IEEE

current starts dominating the subsequent increase. In the 1991 curve the absorption current grows faster and the threshold voltage decreases to 1.2 Un. However, this change cannot be attributed to irreversible insulation ageing since the current measured in 2001 virtually matches the 1981 characteristics again. The reason for this temporary increase of conductivity is most probably a higher moisture content of the insulation in 1991. The test results for generator 2 (see Fig. 3) and genera-tor 3 (see Fig. 4) provide further evidence that currents are a poor indicator for irreversible ageing as it is caused, for exam-ple, by thermal degradation which would express in a conti-nuous increase of the measurement values.

Remarkable for an epoxy/mica insulation system are the ra-

ther high leakage currents as well as a low threshold voltage, especially in the 2001 values of generator 2 (see Fig. 3). A probable explanation for this phenomena is conductive surface contamination of the endwindings combined with specific ageing properties of the endwinding stress-grading system [7].

A further parameter used in the analysis of hipot DC currents

is the ratio of the current growth rate in the non-linear (lea-kage) respectively the linear (absorption current) region, which can be quantified with a specific current growth factor as defined in (1).

Fig. 2. Insulation current of generator 1 during DC hipot test (Un = 15.5 kV)

Fig. 3. Insulation current of generator 2 during DC hipot test (Un = 15.5 kV)

Fig. 4. Insulation current of generator 3 during DC hipot test (Un = 15.5 kV)

Current 2Un 2Un 5kV

10kV 5kV

I IGrowth Factor

I I−−

=−

(1)

Fig. 5 shows the values for the three generators. The in-

crease after 1991 as consequence of the higher leakage current growth rates at voltages above Un is another indicator for an increased surface conductivity of the endwindings due to the reasons mentioned earlier.

III. POLARIZATION INDEX

The measurement of polarization currents is another common diagnostic test for stator winding insulation and is based on time domain analysis of insulation currents after the applica-tion of a DC voltage step. After the fast decay of the capaci-tive charging current (as already explained in the previous paragraph) the remaining measured current consists of polari-zation and leakage portions [8, 9]. The polarization current decreases with the increasing degree of polarization of the dielectric material whereas the leakage current is time inde-pendent. In the actual test the current is measured just after the capacitive charging current has decayed to a negligible value but the polarization current is still high and again when the polarization current has decayed to near-zero (common practice are 1 and 10 minutes).

Fig. 5. Growth Factor history

Absorption currents

Leakage currents

UT

The Polarization Index (PI) can then be calculated as the ra-tio of the two measured currents:

1min

10 min

IPI

I= (2)

Obviously, if the leakage current is small compared to the

polarization current the PI will be greater than 2. The PI val-ues (see Fig. 6) are all in a range that indicates a healthy insu-lation with a low moisture content.

Except from generator 1 the values remain almost constant during nearly 30 years of operation which leads to the conclu-sion that PI measurements are rather insensitive to thermo-mechanical ageing. Furthermore, the high PI values imply that, in contrast to DC hipot tests, there are no significant lea-kage currents across the endwinding surface at the applied test voltage of 500 VDC. In practice it is thus advisable to use PI measurements as a preliminary test in order to detect any se-rious insulation problems before the application of test voltag-es above the generator’s rated voltage.

IV. DISSIPATION FACTOR

The Dissipation Factor (tanδ) quantifies the dielectric losses in the stator winding insulation caused by conduction losses, polarization losses and, during the application of a voltage above partial discharge (PD) inception voltage, ionization losses. As seen in chapter 2, the conduction (leakage) losses of a healthy insulation are negligible up to the rated voltage. With a partial discharge inception voltage of an estimated 0.4 Un the dielectric dissipation at 0.2 Un is dominated by polari-zation losses. Polarization losses are caused by reorientation movements of polar molecules under the influence of the al-ternating electric field [10]. With an increasing age of the insulation, the number of polar molecules augments due to the oxidative break-up of the original epoxy chains into shorter fragments which are polar themselves. The higher number of polar molecules causes higher dielectric losses in the insula-tion and leads to an increase of dissipation factor values at 0.2 Un. However, the dissipation factor is also influenced by tem-perature and humidity as both higher insulation tempera-

Fig. 6. Polarization Index history

tures and a higher content of moisture increase the dielectric losses [11]. Fig. 7 shows the development of tanδ at 0.2 Un and the winding temperatures during the measurements. Al-though the temperatures during the measurements in 1996 and 2001 were significantly lower than in 1986 and 1991 the re-spective values of tanδ are increasing. This can only be ex-plained by thermal ageing as the nearly constant Polarization Index values exclude a significantly higher moisture content of the insulation in 1996 and 2001. In summary the measure-ment of tanδ below partial discharge inception voltage proves to be a robust indicator of ageing caused by oxidative degrada-tion in a modern epoxy-mica insulation system. Moreover, variation in tanδ data from the three different generators is rather small compared to DC hipot and PI measurements.

Whereas tanδ below PD inception voltage is dominated by

polarization losses, the value at Un also contains a contribution from PD. PD is a symptom of the presence of gaseous voids in the groundwall insulation which can develop due to delami-nation at interfaces (conductor-epoxy/mica-epoxy) as a conse-quence of thermal-mechanical stresses. The increase (tip-up, Δtanδ) of tanδ between a low voltage (usually 0.1-0.2 Un) and the rated voltage (3) is considered to be a significant parame-ter related to the total volume of voids within the insulation [12].

Un 0.2Un

0.2Un

tan tantan

tanδ − δ

Δ δ =δ

(3)

Fig. 8 shows the history of the dissipation factor tip-up.

From 1976 (after the first 4 years of operation) until 1986 the tip-up values don’t change significantly because the void vo-lume inside the insulation remains quite stable after the initial delaminations which occur in an early stage of service life as a result of thermal-mechanical stress cycling. After the mid-eighties there is a slightly decreasing tendency which can

Fig. 7. History of Dissipation Factor (tanδ) at 0.2 Un with winding tempera-tures during the measurement

Fig. 8. History of Dissipation Factor Tip-Up

be explained with a reduction of PD because some of the voids are shorted-out by conductive epoxy degradation products due to PD corrosion inside the voids.

V. CONCLUSION

The analysis of the long-term diagnostic measurement data

in this paper shows a good correlation with the underlying theoretical background. Measurements of insulation currents at high DC voltages up to 2 Un are particularly suitable for the detection of endwinding surface currents due to contamination with conductive particles and/or ageing of the endwinding stress grading system. The insulation current caused by the relatively low DC voltage during polarization current mea-surements is dominated by polarization and leakage currents in the groundwall insulation and is a sensitive indicator for an elevated moisture content. Both DC hipot and PI measure-ments show rather high variation between different machines and are insensitive to insulation degradation due to thermo-mechanical ageing. Dissipation Factor (tanδ) measurements below partial discharge inception voltage are a robust indica-tor of long-term ageing caused by oxidative degradation. Var-iation in tanδ measurement values from different generators is rather small compared to DC hipot and PI measurements and therefore well suited as an integral insulation condition indica-tor.

ACKNOWLEDGMENT

The author appreciates the continued and constructive inter-

est of Prof. Dr. K. Fröhlich, Head of the High Voltage Labora-tory of ETH Zürich, in this work. A special thanks goes to Dr. Yvan Michellod of Hydro Exploitation SA, Sion, Switzerland, for valuable contributions. This work was supported by swisselectric research.

REFERENCES [1] B.K. Gupta, G.C. Stone, J. Stein, “Stator Winding Hipot (High Poten-

tial) Testing,” IEEE Electrical Insulation Conference, Montreal, 31 May-3 June 2009.

[2] G.C. Stone, E.A. Boulter, I. Culbert, H. Dhirani, “Electrical Insulation

for Rotating Machines,” IEEE Press, Piscataway, John Wiley & Sons, INC., ISBN 0-471-44506-1, 2004.

[3] L. Rux, W. McDermid, "Assessing the Condition of Hydrogenerator

Stator Winding Insulation Using the Ramped High Direct-Voltage Test Method," Electrical Insulation Magazine, IEEE , vol.17, no.6, pp.27-33, Nov.-Dec. 2001.

[4] W. McDermid, B.G. Solomon, "Significance of defects found during

high direct-voltage ramp tests," Electrical Insulation Conference and Electrical Manufacturing & Coil Winding Conference, 1999. Proceed-ings, pp.631-636, 1999.

[5] F.R. Schleif, “Corrections for Dielectric Absorption in High Voltage D-

C Insulation Tests,” AIEE Transactions (Power Apparatus and Systems), vol. 75, pt. III, pp. 513-517, 1956.

[6] L.M. Rux, “High-Voltage DC Tests for Evaluating Stator Winding

Insulation: Uniform Step, Graded Step, and Ramped Test Methods,” IEEE Conference on Electrical Insulation and Dielectric Phenomena, Minneapolis, October 1997.

[7] A.M. Gully, J.C.G. Wheeler, “The Performance of Aged Stress Grading

Materials for Use in Electrical Machines,” IEE Dielectric Materials, Measurements and Applications Conference, Pub. no. 473, 2000.

[8] G.R. Soltani, E. David, "Condition Assessment of Rotating Machine

Winding Insulation by Analysis of Charging and Discharging Currents," Electrical Insulation, 2006. Conference Record of the 2006 IEEE Inter-national Symposium on Electrical Insulation, pp.336-339, 11-14 June 2006.

[9] E. David, L. Lamarre, D.N. Nguyen, "Measurements of Polariza-

tion/Depolarization Currents for Modern Epoxy-Mica Bars in Different Conditions," Electrical Insulation Conference and Electrical Manufac-turing Expo, 2007, pp.189-193, 22-24 Oct. 2007.

[10] W.S. Zaengl, "Dielectric Spectroscopy in Time and Frequency Domain

for HV Power Equipment. I. Theoretical Considerations," Electrical In-sulation Magazine, IEEE , vol.19, no.5, pp.5-19, Sept.-Oct. 2003.

[11] C. Sumereder, H.M. Muhr et al., “Observation of Dielectric Parameters at Generator Stator Windings under Changing Environmental Condi-tions,” Conference Record of the International Symposium on High Vol-tage Engineering ISH, pp. 190-190, Capetown, 2009.

[12] M. Farahani, H. Borsi, E. Gockenbach, “Study of Capacitance and Dis-

sipation Factor Tip-Up to Evaluate the Condition of Insulating Systems for High Voltage Rotating Machines,” Electrical Engineering, Springer, vol. 89, no. 4, pp. 263-270, March, 2007.