Developments for on-line monitoring of rotors and stator windings

30 April 2008



Monitoring the operation of hydrogenerators to determine maintenance requirements is not easy, but three developments are showing particular promise for the industry


Compared with turbine generators in fossil and nuclear plants, there are relatively few tools available to assess the condition of the rotor and stator windings during normal hydrogenerator operation, and thus to determine maintenance requirements before an outage is taken. Typically for hydrogenerators, any on-line data that are present are usually not monitored continuously to maximise the warning time given of an incipient problem with the windings[1]. Over several years the electric-power-research-institute (epri), together with the New York Power Authority (NYPA) and Canadian firm Iris Power, have endeavoured to address this situation with three research and development projects, involving the development of:

  • A means of detecting shorted turns in the rotor winding during normal operation;
  • A cost-effective continuous partial discharge (PD) monitor to find developing stator winding insulation problems; and,
  • An expert system software tool to integrate on-line data with hydrogenerator design and operation information. This will give both plant operators and maintenance engineers real-time assessments of the overall condition of the hydrogenerator and needs for maintenance or corrective action.

In this paper, three developments and their deployment at various NYPA plants are reviewed. The article first identifies the monitors that are presently available and the need for better tools.

Present Winding Monitoring

The only two quantities that may indicate the health of the windings that are both widely measured and continuously monitored are bearing vibration and stator winding temperature. The bearing vibration can indicate if the rotor winding turn insulation has failed in one or more locations since, if the shorts are numerous, there will be an unbalanced magnetic pull that will causes the rotor to “wobble”.

The stator winding temperature indicates if the winding is operating at a temperature that may limit the life of the stator winding insulation. However, if the temperature is increasing over time under the same load and ambient temperature conditions, it is also an indication that the winding is becoming contaminated with dirt or debris and/or the heat exchangers are losing their effectiveness. For example, the ventilation ducts in the stator may have become blocked, which would raise the stator temperature.

Thus monitoring the trend in temperature can provide valuable evidence of when to take an outage for cleaning the winding or performing heat exchanger maintenance. Yet, both vibration and temperature are usually only connected to an alarm to operate when dangerous conditions are reached. The trend in vibration or temperature over time is rarely observed automatically.

In some machines, the air gap between the rotor and stator is measured[2]. Although air gap monitoring mainly detects mechanical problems such as alignment and relative movement between components, it may also give some indication of rotor winding insulation problems. However, as with temperature and vibration monitoring, air gap monitoring produces data, which are often not directly trended, and usually requires some expertise on the part of the plant operator to interpret.

In many thousands of hydrogenerators, on-line partial discharge (PD) measurements are periodically made[1]. Such PD measurements can indicate a wide variety of situations that affect the high voltage electrical insulation on the stator winding. These include loose windings in the slot, long-term thermal degradation and electrical tracking due to partly conducive contamination of the stator winding. Most PD data are collected intermittently, and some expertise is needed for its interpretation.

Rotor Flux Monitoring

The salient pole rotor windings in hydrogenerators are very reliable. However, over the years the insulation in the rotor windings may age, eventually leading first to shorted turns, and then ground faults. Insulation aging occurs as a result of overheating, load cycling and/or contamination of the winding by partly conductive materials such as insects or dust mixed with oil or water[1]. Fibreglass reinforced epoxy and/or the film insulation on winding wire is used to insulate each copper turn from adjacent turns, as well as provide ground insulation between the copper and the rotor pole. Most plants will trip the generator once a rotor ground fault occurs.

When a shorted turn occurs, an unbalanced magnetic pull may result, which in turn may cause an increase in bearing vibration. Unfortunately, there are many causes of high vibration, of which rotor shorted turns is just one. Thus, bearing vibration is not an infallible way to detect rotor winding aging. Thus it would be also helpful to know that the cause of high vibration is not shorted rotor turns.

The most reliable and common way to detect shorted turns (and incipient ground faults) is to perform a “pole drop” test[1]. In the test, an AC voltage, for example 120V, is applied between the positive and negative slip rings when the hydrogenerator is shut down and partly disassembled. The voltage across each pole is then measured. If shorted turns are present, there will be a smaller than average voltage drop across that pole. But the test has three significant disadvantages:

  • It can only be performed with the generator shut down, implying a loss of revenue and power system reliability;
  • It is time-consuming to perform, especially on a large rotor with many dozens of poles; and,
  • Since the rotor is not rotating, the centrifugal forces are not occurring, and thus some shorts may not be present in the pole drop test, which nevertheless will be present at normal rotating speeds. Even worse, some shorts may appear when the rotor is shutdown that do not occur when it is spinning.

As organisations try to minimise the work (such as pole drop tests) performed during unit shutdowns due to restricted resources, and as they move to predictive maintenance to plan any repair work based on on-line condition monitoring, there is a need for an on-line tool that can replace the pole-drop test.

Rotor flux monitoring involves measuring the magnetic flux in the generator air-gap to determine if field winding shorts have occurred in the rotor poles. The radial magnetic flux is detected by means of a flat coil (or probe), developed by the US Bureau of Reclamation (USBR) and consisting of several dozen turns, and that is glued to stator teeth[3]. As each rotor pole sweeps by the flux probe, a voltage is induced in the coil that is proportional to the flux from the pole that is passing the coil. The voltage is measured by an analog-digital (A/D) converter. In a salient pole machine, the radial magnetic flux profile from each rotor pole depends on the MW and MVAr loading of the machine.

As each pole in the rotor passes, there will be a peak in the induced voltage caused by the change in magnetic flux from the pole. The voltage can then be recorded and the “average” flux across one rotor pole can be calculated. Any turn short in a pole reduces the effective ampere-turns of that pole and thus the signal from the flux probe associated with that pole. The recorded waveform data can then be analysed to locate the poles containing the fault, as long as the pole location is calibrated from a “start” site marked on the rotor shaft. An algorithm was developed to maximise the sensitivity to a pole with shorted turns. The algorithm involves integrating the data from each pole, applying autocorrelation, and comparing the integral from each pole to another pole. The flux probe and monitor (called FluxTrac) have now been installed on dozens of hydrogenerators and pumped storage generators. An example of the output is shown in Figure 1 where a simulated short circuit was applied to two of the poles (8 and 48) to ensure that shorted turns could be detected.

Continuous Stator Winding PD Monitoring Systems

The stator winding insulation deterioration process is relatively slow. The time between when significant PD can be detected and when there is a significant risk of winding failure is usually two or more years. Thus, periodic measurement – once every six months or so, as is the most common data collection means – is often adequate for detecting stator winding problems with sufficient warning to conveniently implement corrective action. However, there are several situations where continuous PD monitoring may be advantageous:

1. Expert systems. There are expert systems being developed to continuously monitor all the sensors in a generator to determine if any problems are occurring. Since the PD activity is an important source of information with regard to the stator winding condition, it is desirable that generators equipped with such expert systems have ready access to PD data.

2. Operating conditions. The stator winding PD activity is often affected by hydrogenerator operating conditions such as winding temperature, load and voltage, as well as environmental humidity[1]. For example, if a winding is loose (i.e., the stator coils can vibrate in the slot), an increase in generator load will cause the PD activity to dramatically increase. Although these operating-condition dependent results help in interpreting the type of deterioration occurring, they often make it difficult to trend PD data over time. To obtain such data, which is critical for interpretation, the machine must be tested under the same load, temperature and voltage conditions For peaking hydrogenerators or pumped-storage generating units, it is often difficult for the plant operators to recreate exact operating conditions from test to test. Continuous PD monitoring solves this problem by continuously measuring the motor/generator operating conditions, and then storing the PD measurement at which the proper operating conditions occur naturally during normal operation of the machine.

3. Remote operation. Many hydro generating plants are located in remote areas, making it expensive to transport test personnel to the site for conventional PD testing. Furthermore, many hydro generating stations are remotely operated, that is, there is no permanent on-site staff. A continuous monitoring system can allow the PD measurements to be made remotely, which will reduce overall operation and maintenance costs.

4. Failure risk. Sometimes periodic on-line PD monitoring has identified a stator winding that has a high risk of failure, but operating considerations prevent an immediate repair or rewind. By installing continuous PD monitoring, plant maintenance and operating personnel can keep a closer watch on the stator winding, possibly extending the winding life with a lower risk of a catastrophic in-service failure.

These reasons resulted in EPRI, NYPA and Iris Power working together to develop a continuous PD monitoring system called HydroTrac. The new continuous monitoring system – installed in most hydrogenerators in the NYPA’s portfolio and hundreds of others – employs the same noise separation methods that had been used for many years for periodic monitoring, which greatly reduces the risk of false indications. Also, great care was taken to make sure the same PD quantities such as Qm (peak PD magnitude) were produced, to facilitate seamless trending with past data obtained from portable instruments. A block diagram of the hardware is shown in Figure 2.

A key feature is how the generator operating and environmental data are collected. Instead of measuring the generator operating and environmental quantities such as kV, MW, MVAr, hotspot temperature and humidity using a direct analog feed from the measuring detectors in the hydrogenerator, this information is obtained from the plant computer. This greatly reduces costs, not the least of which is associated with running analog signal wiring to the monitoring system hardware.

Since every utility seems to be using different proprietary communications systems and protocols, the system had to be designed to be very flexible in both physical connection options and communications protocols. For local configuration and control, the acquisition units have an RS232 port. For remote operation and integration with other systems, an RS485 port is utilised. An addressing scheme on the RS485 bus allows for connection of up to 127 instruments (Figure 2).

The RS485 network can physically be fibre optic or shielded twisted pair cable. Ethernet communication is possible through any number of third-party terminal servers to convert either one or more monitors to TCP/IP Ethernet. At that point, additional communications options such as WiFi are possible.

Utilising any of these physical communications schemes, the system can be controlled via a remote Windows application. This control software is used to configure the monitors at each hydrogenerator, trigger PD data collection on a defined schedule, and archive the data in a database for later review. Even if communications should fail to the control application, the instrument continually collects and archives data in its local memory, storing up to 24 months of summary PD data. The data can be accessed later, either locally or by remote communications for trending and review.

The control software also has the ability to trigger PD data collection based on the generator state. PD levels fluctuate with parameters such as MW, MVAr, stator winding temperature, and humidity. Knowing these variables can smooth PD trends and provide additional diagnostic information on the stator winding condition. These parameters can be fed to system via a “PI” interface from a plant PI server (OSI SoftTM). Many other plant SCADA and diagnostic software packages also contain bridges to PI.

Some plants prefer to use 4-20mA analog outputs to facilitate trending the PD activity over time on plant SCADA, DCS, or other monitoring systems. In addition to the physical interfaces described above, external systems can connect using an “OPC” software interface. In this case, the OPC Server software must be running on a Windows computer connected to each continuous PD monitor. This server exposes an OPC Data Interface to any number of OPC clients. The OPC Server controls the PD monitors and exposes PD summary data and pulse magnitude analysis data as a set of OPC tags. Finally, the system has an alarm output relay, which can be wired to a remote system and will be activated when predetermined PD levels are exceeded.

Current technology advances as described above make available cost effective on-line monitors for critical components of hydrogenerators. Proper interpretation of this often complex information can save operating and maintenance expenses, in addition to reducing unscheduled outages and catastrophic failures. With other advanced monitors such as vibration and air-gap, as well as conventional sensors installed by OEMs, the array of on-line machine condition data is significant. The volume of available data from these sensors, and the extensive interpretation often necessary to evaluate the complex waveforms and spectrums, can overwhelm plant personnel and resources. Sophisticated software and algorithms are often necessary to correlate and interpret this data to establish the overall hydrogenerator and drive train condition.

HydroXTM (for “hydro expert”) is a knowledge-based expert system program for on-line monitoring of hydrogenerators. With the NYPA and Iris Power, and this time GE, working together over five years, the system was developed. After a further two years of prototype evaluation at NYPA’s St Lawrence power project on two 55MVA generators, the system is ready for commercial deployment.

The new system is a set of diagnostic rules (a RulePac) running on a commercial PC-based asset management tool called System 1 from GE. System1 is a distributed software product based on a SQL Server database and contains components for data collection from remote system via OPC, a production rule engine for processing user defined rules, and a design tool for developing and testing rules and developing custom user interfaces. The rules for the system were developed over several years using experts on machine design, monitoring instrumentation, and utility engineering and operations staff.

Use Of Mathematical Modelling

One significant advance in HydroXTM is the ability to use mathematical modelling to predict the expected value for sensors during various machine operating conditions. The predicted values are compared to the actual measured values and deviations are analysed by the rules to compute a condition diagnosis. For example, the predictions of stator winding temperatures are made based on the cooling air temperature and the MW load on the machine. During installation of the RulePac, this basic equation is then customised to account for heating/cooling time constants of the machine with load, and to the actual readings obtained at full load for each sensor which vary due to sensor location and other physical properties (Figure 3).

For many sensors, the alarm thresholds may be significantly different depending on the mode of the machine. HydroXTM has rules to determine the machine mode and where necessary, different thresholds and even rules are executed dependent on this mode. The machine mode is also used in several instances to calculate and alarm on the trend of a sensor values. For example, the trend of nominal air gap during field flashing can indicate a specific type of problem that just trending air gap at nominal machine load would not detect.

Industry trends are to move to more automated plants, with less on-site expertise and operations staff. As described, HydroXTM can calculate and trend key features as well as synthesise summary indications from complex data sets from monitors such as vibration, air-gap, PD, flux. Using these intermediate indicators, along with diagnostic rules, an expert system like HydroXTM can filter and focus attention to abnormal values, and provide diagnosis of specific faults and their remedies.

Lloyd, B.A., & Stone,G.C., Iris Power; Stein, J., EPRI; and, Jourdain, E., NYPA


Figure 1 Figure 1
Figure 3 Figure 3
Figure 2 Figure 2


Privacy Policy
We have updated our privacy policy. In the latest update it explains what cookies are and how we use them on our site. To learn more about cookies and their benefits, please view our privacy policy. Please be aware that parts of this site will not function correctly if you disable cookies. By continuing to use this site, you consent to our use of cookies in accordance with our privacy policy unless you have disabled them.