Hydro project owners can use 80-pF line couplers, installed in many rotating machines, to measure partial discharge and detect signals from deeper inside the winding. This method can provide a more reliable assessment of insulation condition than traditional monitoring systems.
By Stefan Kornhuber, Matthias Boltze and Greg Topjian
Within the stator windings of rotating machines, partial discharge (PD) signals that come from deep inside the machine, away from installed measuring points (or couplers), are attenuated in amplitude because of the low pass behavior of the signal path along the winding.1 Thus, the couplers only pick up the lower-frequency contents of the signal.
Many rotating machines come equipped with 80-pF line couplers that can be used to perform on-line PD measurements.2 These PD detectors have a 50-Ω load, forming a high pass filter starting at about 40 MHz. PD is measured in the frequency range of several tens of MHz. Because of this, the internal discharges are measured with poor sensitivity by the 80-pF line couplers, and the resulting PD pattern is dominated by the end winding discharges.
A more promising methodology is measuring PD in picocoulomb (pC) according to International Electrotechnical Commission standard 60270.3 With this methodology, reading of PD activity depends less on the pulse shape. That provides more sensitivity to internal discharges, leading to a better and more reliable assessment of the entire insulation state.4
By signal conditioning, it is possible to maintain the signal low pass processing necessary for pC measurement. Additionally, the test voltage signal can be derived for synchronization, enabling phase-resolved PD pattern analysis. The benefit of this is to also acquire inner PD and measure this PD against a charge that can be calibrated according to IEC 60270, allowing the plant owner to compare measurements with quality assurance checks.
Comparative measurements at a generator stator with larger coupling capacitors illustrate the feasibility of charge measurements via 80-pF line couplers.5 This article presents further ideas for characterizing the integration behavior and discusses calibration questions.
Understanding the situation
Measuring PD activity of the stator winding in terms of pC provides the advantage of being sensitive to PD inside the winding, away from the decoupling point. On the other hand, the PD detector as well as the whole signal chain including the PD coupler must provide an appropriate integration behavior.1,4
Figure 1 shows a calibration signal and a true PD signal, both gathered at the phase terminal of a generator stator winding (11 kV, 41 MVA) in the manufacturer test lab, single phase voltage applied. In comparison to the calibration impulse, the true PD signal has a much longer pulse duration and is an indication of the internal attenuation behavior in the generator winding itself, depending on the fault position.
With respect to the permanently installed 80-pF line couplers, the high pass filter behavior would lead to a remarkable loss of signal area (charge) when the PD signal passes that coupler. This can be further illustrated if the high pass time constant of 80 pF x 50 Ω = 4 ns is considered in relation to the pulse PD duration of about 1 μs. The mentioned line coupler shows a strong differentiating behavior.2 That hampers the measurement of the apparent charge and reduces the sensitivity to PD signals from deep inside the stator winding.
The basic idea to overcome the loss of signal area is to compensate for the high pass filter behavior by an appropriate low pass filtering of higher order in the signal path. This signal processing allows the charge to be recovered and the integration error minimized.
For that purpose, Doble developed a line coupler matching unit in 2010 that includes a low pass filter with a frequency of 500 kHz. That module provides an electronic op-amp circuit with low impedance input that integrates the capacitive coupler current and gives out the corresponding low-voltage signal of the high-voltage signal for synchronizing the phase-resolved PD pattern. This matching unit can easily be connected to the low-voltage part of the 80-pF couplers simply using screws and cables.
Generator stator measurements
Deeper comparative measurements of the generator stator mentioned previously were conducted to validate the feasibility of apparent charge measurements even with 80-pF line couplers and subsequent signal conditioning (i.e. proper low pass filtering). The first main target was to compare the measuring results in the frequency range up to 500 kHz. The calibration procedure was performed according to standard by means of a 10 nanocoulomb calibrator in parallel to the phase terminal and ground.
Coupler arrangements that were investigated are: 100-nF, 2-nF and 80-pF with line coupler matching unit (low pass active).
Figure 2 shows the resulting phase resolved PD patterns and the associated reading of the apparent charge. The good correlation of the obtained patterns and the measured apparent charge values is the first indicator of the feasibility of charge measurements using the 80-pF line couplers.
Investigation with larger bandwidths
Under real measuring conditions in a powerhouse, it is sometimes impossible to apply the measuring bandwidth restrictions of IEC 60270 because of noise issues. To improve the signal-to-noise ratio, the upper frequency limit has to be raised up to the MHz region. In that case, it is possible to deactivate the internal low pass filter of the line coupler matching unit.
To evaluate the consequences for the charge measurements, the investigations at the generator stator winding were continued. Table 1 gives an overview of the results. The first column shows the coupler arrangement, and the upper line contains the upper frequency limit adjusted at the PD detector.
A comparison of the indicated charge values shows a general trend of measured charge decreasing with a higher upper frequency limit. The exception is the last row of the table. Independent of the settings of the PD detector, the upper frequency limit of the total signal chain is maintained at 500 kHz provided by the low pass.5
At the 80-pF coupler without the low pass filter, the charge reading is lower even in the IEC frequency range of the detector as a result of the high pass filter behavior and the loss of signal-time area. These results give a first impression, of which integration errors can be expected under real conditions where the measuring frequencies have to be matched according to the actual circumstances of the machine. Note that these results are only an example because they are dependent on the individual PD signal shape (see Figure 1 on page 39) and must not be generalized.
Determining the integration behavior and errors
The next target was to apply a simple solution to characterize the integration behavior of the entire measuring arrangement coupler/PD detector. The low pass filter components of R1/C1 shape a low pass, via the time constant T1. Hence, the rise time of the step response on the incoming step signal can be adjusted. By means of the connected oscilloscope, the rise time can be determined and read.
The elements C0/RL form a high pass. The small time constant, provided by T2, in this part of the circuit can be considered as a pure differentiation stage. The resistance value of R1 is small enough to avoid interactions from the connected PD coupler. Thus, the resulting voltage drop across R1 (Usig) represents the desired needle pulse with constant area (charge) and variable duration.
The pulses from the above circuit were injected into the corresponding couplers, and the pulse duration was increased until the charge reading at the detector reached the arbitrarily selected value of 63%. In that way, the integration time (TI) was defined and determined for each individual measuring configuration (see Table 2).
The results of this analysis show that TI is very short at the 80-pF coupler, finally resulting in an integration error that is not negligible. The charge reading at the PD detector would deviate, especially in cases where the actual PD impulse shapes are much longer than the calibration pulses.
The 80-pF coupler in combination with low pass filtering shows, again, much better results. By this means, it can be demonstrated that it is possible to perform proper charge measurements even with small coupling capacitance. The integration error of these measurements would stay within reasonable limits. Integration times (TI) of about 1 μs correspond to the apparent charge measurement according to IEC standard 60270.
Determining apparent lumped capacitance of the stator winding
In general, the application of small coupling capacitances leads to a decreased measuring sensitivity.2 According to the classical capacitive equivalent circuit of IEC standard 60270, the charge amount transferred to the measuring circuit increases as the coupling capacitance increases in relation to the capacitance of the test object.
In the case of a generator winding, it makes no sense to refer to the total winding capacitance, because it does not act as lumped capacitance in the relevant frequency range of PD measurements.4 Furthermore, the question arises about the size of the equivalent lumped capacitance of the generator winding that appears in the total measuring arrangement — including coupler, detector and calibrator — depending on the measuring configuration, bandwidth, and other factors. To determine the mentioned capacitance, a simple substitution method was applied.
After each calibration of the measuring configuration, the stator winding was disconnected and substituted by a variable low voltage capacitance. Its capacitance value was tuned to get the same reading as before, with the generator winding connected. Table 3 shows the resulting capacitance values.
Taking into account the dimension and price of PD coupling capacitors, it can be stated that the benefit is limited if the coupling capacitance exceeds that of the test object, in the actual case of the lumped capacitance of the generator stator winding. The measuring sensitivity would not increase.
The determined capacitance values are specific for the stator winding under test and the measuring configuration. Nevertheless, we can generalize as follows: 100 nF as coupling capacitance is not necessary, and 10 nF seems to be adequate but is in reality not the optimum compromise considering the dimensions and price.
The authors find that PD couplers around 1 to 2 nF represent the most efficient solution; 80-pF-couplers are not recommended for new installation but are often preinstalled on machines. To cope with the 80-pF couplers, the matching unit solution was created.
For most on-line PD measurements, there is the principle challenge that the test object terminals are not accessible for calibration purposes. The installation of on-line calibrators is not common.
But for trending purposes, there is the need to have the measuring results comparable in terms of pC. Finally, it does not matter for trending if the results are the same as calibrated according to IEC standard 60270, but the results of repetitive measurements must be based on the same scale factor of the measuring PD detector. So it is recommended to perform at a minimum an indirect calibration in parallel with the input of the PD detector connection to the coupler terminal box.
Taking the previous results into account, it is further recommended not to inject charge pulses directly from the calibrator. To get a scale factor in the same range roughly of the IEC calibration, the pulse should be injected via a capacitive charge divider. This should consist of a parallel capacitor of 10 nF to simulate the lumped capacitance of the stator winding and a 80-pF-capacitor to simulate the coupler capacitance.
The proposed calibration procedure does not guarantee the measurements are in accordance to apparent charge calibration of IEC standard 60270, but it fulfills the main task of PD calibration in general. Using this recommended calibration procedure, the measuring results will be reproducible and comparable.
The detection of PD deeper inside generator stator insulation requires the transition to charge measurements with respect to IEC 60270. This means a sufficient integration behavior of the entire measuring arrangement must be provided. On many generators, 80 pF on-line couplers are already installed with a strong high pass behavior that hampers the charge measurement. An additional module with switchable low pass of higher order can be applied to overcome this obstacle.5
Comparative PD measurements at a real test object verified the feasibility of comparable charge measurements even with a coupler of small capacitance. TI was introduced as measure to evaluate the integration ability of the measuring setup. That value was determined for different measuring configurations with different couplers and processing bandwidths. The results revealed that the integration time decreases and the integration error increases if the total upper frequency limit increases and the coupling capacitance decreases.
In principle, it is possible to measure the PD activity in terms of pC even in the MHz region. With some concessions to the integration error, the measured value might be below 50% of the actual one.4 This error is less critical if the calibration is performed thoroughly before each measurement. A calibration adaptor is proposed for an indirect calibration for getting scale factors similar to that of the IEC calibration.
A method was applied to determine the equivalent lumped capacitance of a generator winding. Basing on the generalized results, recommendations were given for the selection of capacitance values of generator on-line PD couplers.
1. Train, D., "The Measurement of Partial-Discharge Phenomena in Stator Windings," Doble Client Committee Fall Meetings, Doble Engineering, Watertown, Mass., USA, 1994.
2. Zhu, H., V. Green, M. Sasic, and S. Halliburton, "Increased Sensitivity of Capacitive Couplers for In-Service PD Measurement in Rotating Machines," IEEE Transactions on Energy Conversion, Vol. 14, No. 4, 1999, pages 1184-1192.
3. "High-Voltage Test Techniques – Partial Discharge Measurements," International Electrotechnical Commission International Standard 60270, Third Edition 2000.
4. Wilson, A., "Partial Discharge Measurements on Rotating Machinery: Apparent Charge Techniques," Doble Client Committee Fall Meetings, Doble Engineering, Watertown, Mass., USA, 1994.
5. "Line Coupler Matching Unit," User Manual Version 00, Doble Lemke GmbH Germany, 2011.
Stefan Kornhuber is engineering manager and Matthias Boltze is service manager with Doble Lemke GmbH in Germany. Greg Topjian is solutions manager partial discharge with Doble US.