Real-world Research at Wanapum Dam

By Yannick Baril, Maxim Bergeron, Jose Figueroa and Jeff Niehenke

Washington’s Grant County Public Utility District, Quebec’s Laval University and Alstom Renewable Power Canada collaborated on a full-scale investigation of Wanapum’s 95-MVA generator in Unit 09 before its scheduled replacement.

The opportunity opened a window for rotor and stator measurements in operation that could provide valuable data for ongoing research projects, including detailed modeling of pole face and damper winding behaviors, generator characteristic identification by stand-still frequency response tests, machine vibration prediction, improvement of ventilation models and improvements of stator winding copper temperature estimation.

A unique opportunity

To innovate and improve knowledge on hydro generators requires extensive modelling using the most advanced tools available.

However, even with state-of-the-art tools, measurements are essential to validate the models, measure the influence of formerly unconsidered parameters and develop an intimate knowledge of the machines.

Most of the time, validations and measurements are performed on laboratory setups because they allow for controlled studies on particular aspects that cannot be done in a power station.

However, a measurement in a full-scale generator in a power station setting adds the indispensable validation and insight required to complete the laboratory measurements and fully understand the problems arising in operation. In this context, measurements performed on real hydro generators are invaluable.

They allow researchers to validate their models, manufacturers to improve their designs and utilities to better identify the operational limits of existing machines.

For these reasons, Grant County PUD, Laval University and Alstom – now GE Renewable Energy – decided to collaborate in performing research during ongoing rehabilitation of the 1,038-MW Wanapum plant, with successful cooperation allowing many challenges in instrumentation implementation, logistics and scheduling to be overcome.

Conducting the research

One main goal of the study was to measure the damper bar currents in a rotor pole. Even simple temperature measurements performed on the rotor can be challenging to implement. Measuring the damper bar currents’ waveforms is even more complex since poles are not designed to allow current sensor installation. To reach this goal, a spare pole was modified and custom current sensors were installed.

Because the unit was to be dismantled afterward, and to make the investment even more worthwhile, the testing team performed additional measurements during the same testing outage: rotor vibrations, pole face magnetic flux, stator copper temperatures, stator vibration, and end winding and cooler air flow.

Research at Wanapum was conducted when Unit 09 was stopped for two weeks at the beginning of August 2014. During the first week, all the instrumentation was installed by all parties on the unit. During the nights of that first week, the standard stand still frequency response (SSFR) test was carried out automatically by an automated equipment. This test is an alternate method to sudden short circuit tests to evaluate the generator impedances and takes roughly 30 hours to complete. The second week was reserved for various rotating tests with measurements in the stator and rotor.

The most audacious goal of the campaign was to measure the damper bar current waveforms in each of the five damper bars of one pole. The utility provided a spare pole that could be modified and instrumented by Laval University. The modification allowed the installation of a custom current sensor on each damper bar on one end of the pole.

The damper bar current sensors were designed by Laval University. In the design process, the most promising current sensing technologies were investigated. The sensors had to:

  1. Have a good frequency response (DC to 10 times the slotting frequency);
  2. Be able to measure small and large currents (10 A to 30 kA);
  3. Be immune to air gap end region magnetic field;
  4. Reject noise from the sensor to the acquisition box;
  5. Resist the mechanical stresses;
  6. Resist the rapid temperature variations and be temperature compensated;
  7. Limit the impedances modification of the damper winding; and
  8. Limit the modification to the pole (it must be acceptable modifications).

The final solution is a non-inductive coaxial shunt with a very small insertion resistance of 4% (5.22μOhm compared to 142μOhm for the bar resistance). The small resistance imposes the use of amplifying electronics that are contained in the sensor enclosure to maximize noise rejection.

We were required to perform pole modifications in order to install the sensors. Installation of the sensors required grooving the pole end plate and tapping the bars. Effects of the modifications on the end plate stresses were analyzed by Alstom. Because the current sensors were of a complex assembly, they were tested on a rotating table.

For pole face flux measurements, a flexible printed circuit board (PCB) was designed with many search coils printed in it. The flexible PCB allowed local flux variation measurements. This provided additional information to validate the pole face steel losses.

On the rotor, we also installed triaxial accelerometers to measure vibrations and oscillations within the rotor for validation of forces calculated by simulation. One was installed on the pole end plate, one on the rotor rim and one close to the shaft.

Results obtained

Rotor measurements

The Wanapum measurement campaign provided the waveforms of the current in all five damper bars on the research pole. The current sensors were equipped with integrated temperature sensors, allowing approximation of the damper bars’ edge temperature during the transients. A search coil glued on the pole surface provided the detailed pole face flux variations during the same events. We also measured accelerations on the research pole, on the rim and close to the shaft. The measurements were performed during the following steady state and transient operations:

  • Open circuit and short-circuit saturation curve;
  • Load tests (active and reactive);
  • Three phases sudden short-circuit (max 70%);
  • Two phases sudden short-circuit (max 50%);
  • Voltage recovery;
  • Synchronization to the grid with slight voltage of frequency deviation; and
  • Load rejection tests under partial reactive and active loading.

Figure 1 presents the reaction of the damper bars during a “not so smooth” synchronization to the grid. During usual synchronization events, if the synchronization is well done, the induced currents are limited. During the Wanapum research campaign, a few manual synchronizations were performed: smooth synchronization and a few less smooth ones.

On the left of Figure 1, the low-frequency oscillating response of the damper winding can be seen over many cycles. At the right, one can see that in addition to the damping response, there is a superimposed high-frequency content linked to time and space harmonics of the generator (for example the slot harmonics).

Stand still frequency response (SSFR) testing

The rotor has to be positioned twice for that test (in the direct and quadrature magnetic axis). GCPUD used a slow rotation device (turning gear) operated very softly. A 42 pole-pairs generator with a 10.36m diameter rotor has the equivalent of one electrical degree on 2.15 mm at the end of the rotor radius.

The positioning is done by the operator, under the rotor, looking at the relative position of the rotor to the stator and looking at the field voltage as explained in IEEE 115. We easily achieved a good precision with the turning gear. Alternatively, we could have pushed the rotor with a jack screw nut for a better precision.

Generator SSFR measurements were performed in several steps. The raw measurements consisted of applying the required sinusoidal excitation to the generator terminals and measuring the response. As the frequency increased and the voltage increased accordingly, we were required to change the acquisition system (see Figure 2 on page 34) gains to improve the signal-to-noise ratio.

The processing of the signals’ phase and amplitude showed there is a discontinuity in the operational impedance measured at 1 Hz. Also, at frequencies lower than 0.01 Hz, the measured Ld and Lq inductances are increasing. Instead, it is expected that the inductances should remain constant in this frequency range.

Measurement of the inductance at low frequency is a challenge. One factor affecting measurement results is that the Hall Effect sensors used in this measurement campaign present an offset that drifts with time. At low frequency, the variation of the offset in time may coincide with the very low frequency signal being measured, making it difficult to differentiate them.

A possible solution for the future is to use two types of sensors at the same time, like shunts and a Hall Effect sensor. This combination should help to get an acceptable response at low frequency and high frequency. This measurement shows that even if the SSFR technique is known, applying it to industrial equipment requires further work.

Because traditional short-circuit tests were performed during the study, we can compare the parameters identified with the SSFR to classical tests and to finite element electromagnetic simulations.

This will provide a starting point for comparison between both methods. Additional measurements will be required to properly compare the SSFR with the sudden short-circuit tests to confirm it as a viable alternative.

The Wanapum experience gave valuable clues and shows the need to accelerate the SSFR test and correctly measure the low frequency signals. Alternatives to the low frequency measurement will also be explored. This is important because even with all the preliminary planning and preparation, in practice, only a limited number of low-frequency tests were realized so researchers could continue to prepare for the remaining tests.

Copper temperature

Direct copper temperature measurement was done on two points in a coil of the stator. The water head available at the time of the measurement campaign was lower than the nominal one and it was not possible to attain the full nameplate rating loading conditions. The results show that in the coil measured, the copper temperature reading is very similar to the temperature reading of the generator resistance temperature detector (RTD). We would intuitively expect the copper temperature to be higher than the RTD reading. But this data makes us consider the following factors:

  • RTDs are installed in the axial middle of the stator core.
  • Instead, the fiber optic is installed near the upper part of the stator coil.
  • The thermal camera reading on the winding overhang showed them colder than RTD reading.
  • In addition, fiber optic readings on the winding overhang facing the airgap were colder by 5 K than the thermal camera reading.
  • The overhang copper temperature readings are colder than the RTD.
  • The machine has axial-radial ventilators.

The above factors, together with the data obtained, made us question if placement of the fiber optic measurement on the copper has not been the optimal one to measure a copper temperature approaching the hot spot in this particular case. This was due mainly to the limited access we had to the slot for the installation. Better access would have required removal of the rotor, which was not justified in this case. The obtained data was used to improve our measurement methods and analysis tools.

Ventilation results

Before the measurement campaign, Grant PUD and Alstom measured the airflow using a vane anemometer and wood frame. Total air flow measured was 45.1 m3/s. The values obtained with the hot-wire setup are near this reference value proving the soundness of the technique. The measurements also showed variability depending on the exact generator operating point that will be investigated further. A second version of hot wire flow meter is currently in development.

The variation of the air flow against the speed was also measured. It showed a linear relation, serving as a second validation that the hot-wire setup was effectively measuring the air flow, as expected. The advantage of being able to measure the air flow continuously while avoiding having personnel near the working generator encourages the further development of this measurement method for generators.

The air speed measured on the end winding section showed there is a difference between the radial and tangential speeds. The difference became attenuated as we moved away from the top of the core. The air speed variation observed was due to the speed sweep being done at the time of the recordings.


Preparation of this measurement campaign took about two years. The data obtained will need several man-years to be fully analyzed, though results will be published in an Institute of Electrical and Electronics Engineers (IEEE) paper in the future. The results and experience gained will be shared between the participating parties.

The measurements have shown we have several things still to learn from hydro generators and that there are several open questions that will need further research. The cooperation between Grant PUD, Laval University and Alstom allowed a unique opportunity to deepen our understanding of phenomena occurring in generators, and its results will be used to improve Wanapum’s operations, Alstom designs, and general knowledge on synchronous machines.

In the short term, measurements and analysis reports will allow for hands-on experience of new measurement techniques and analysis available. The SSFR technique, when fully validated, will provide a low-stress alternative to obtain machine parameters for old or in-service hydro generators.

The rotor data generated in the week of testing provided essential data that will be used by Laval University’s electrical and industrial laboratory, LEEPCI.

In addition to specific studies and technical challenges, the laboratory has the mandate of training the next generation of electrical engineers. The research collaboration with Grant PUD and Alstom provides insight on the industry’s needs.

Measurements on existing machines are necessary to further improve our knowledge and validate calculation models.

This provides a much richer insight on the subtleties of research and development in the large hydro generator domain than laboratory prototyping.


The authors would like to acknowledge Jerome Cros, Simon Frutiger, Mourad Heniche and Carl Messier for contributing to this article.

Editor’s Note: This article was taken from “Wanapum Dam Research Campaign: A Success Story Between Manufacturing, Research, and Plant Operations.” The original paper, which includes further details about the study and the process used to gather data, is available as part of the HydroVision International 2015 online access to conference papers and proceedings at

Yannick Baril and Jose Figueroa are research and development engineers at Alstom’s Global Technology Center in Sorel-Tracy, Canada, which is now GE Renewable Power. Maxim Bergeron is a Ph.D. candidate at Laval University. Jeff Niehenke is a project manager for Grant County Public Utility District.

This article has been evaluated and edited in accordance with reviews conducted by two or more professionals who have relevant expertise. These peer reviewers judge manuscripts for technical accuracy, usefulness, and overall importance within the hydroelectric industry.

More HR Current Issue Articles
More HR Archives Issue Articles
Previous articlePTFE Thrust Bearings Improve Operations at Sir Adam Beck
Next articleTech Briefs

No posts to display