Modular Instrumentation for Online Partial Discharge Monitoring

André Tomaz de Carvalho, Eletrobras Cepel

"Because IMA-DP is based on the PXI architecture, the system is scalable, flexible, and fully integrable with other monitoring systems. With PD measurement, the monitoring of large hydrogenerators can now be carried out exclusively using NI hardware, as is the case of the two largest Brazilian hydroelectric plants: Tucuruí and Itaipu. Using NI hardware has aided the development of an effective tool with competitive cost and total control of the technology, a product of great success."

- André Tomaz de Carvalho, Eletrobras Cepel

The Challenge:

Cepel needed to implement an online partial discharge (PD) monitoring and analysis system for hydroelectric power plant generators to aid the predictive diagnosis of stator electric insulation.

The Solution:

Cepel developed a modular instrumentation system (IMA-DP) for online monitoring of PD in hydrogenerators using PXI modular instrumentation from NI and signal processing algorithms implemented with LabVIEW. The company has adopted the system as an effective predictive maintenance tool.

André Tomaz de Carvalho - Eletrobras Cepel
Hélio de Paiva Amorim Júnior - Eletrobras Cepel
Caio Fleming Ferreira de Carvalho Cunha - Eletrobras Cepel



The Brazilian Electric Energy Research Center (Cepel), considered the largest electric energy research center in the Southern Hemisphere, has engaged in research and development for more than 40 years. Our efforts are focused on generation, transmission, and distribution of electric power. The Department of Transmission Lines and Electrical Equipment supports the maintenance engineering teams of various companies, trains professionals, and develops new technologies for predictive diagnosis and equipment prognosis.


Generators are prominent among other electrical system equipment and continuous monitoring is recommended. In the 1990s, we began developing the Asset Oriented Monitoring System (SOMA) for online monitoring of mechanical, thermal, and operational parameters in rotary machines. We adopted the PXI platform due to the ruggedness, flexibility, and modularity requirements of the DAQ hardware. However, to keep up with other state-of-the-art equipment, there was still a demand for monitoring the stator insulation of generators.


Electrical insulation has always been the weak spot of any high-voltage equipment. In addition to failures in maintenance and possible contingent events, the equipment’s own normal duty cycle causes insulation materials subject to vibration and thermal cycling to age and lose their dielectric properties. It is therefore necessary to assess the health of the insulation in such equipment with a view to the continuity of power supply and to the reduction of failures.


Partial discharge (PD) monitoring is an effective way to assess insulation integrity in high-voltage electrical equipment. In relation to other techniques traditionally recommended for monitoring insulation in rotary machines, PD measurement presents the highest sensitivity, permits the localization of defects, and is the only technique that can monitor generators online. Theoretically, PD measurement could have detected the estimated 89 percent of failures that occurred in insulation.


To meet the demand for online monitoring of stator insulation in rotary machines, we needed to implement an effective online PD monitoring system that was also compatible with the previously adopted generator monitoring hardware platform using the PXI instrumentation standard and taking advantage of its flexibility and modularity.



Measurement Features

PDs are localized dielectric breakdowns of a small portion of a solid or fluid electrical insulation system under high voltage stress, which does not bridge the space between two conductors. They occur in regions of gaseous insertions that represent imperfections in the dielectric (Figure 1). A PD can indicate defects that may evolve into insulation failures with serious consequences.





Online PD measurement in generators by the electrical method requires the prior installation of coupling capacitors and appropriate measurement impedances, as suggested in Figure 2. The number of coupling capacitors per phase can vary depending on the dimensions of the monitored unit.




PD signals are pulses with frequency components that can reach hundreds of MHz, stochastic by nature, with variable amplitude, and heavily immersed in noise (Figure 2). The acquired pulses must have amplitudes recorded and be correlated with the phase of the voltage cycle. The products of the measurement are two-dimensional histograms, in which the repetition rate of the pulses is grouped as a function of their amplitudes and the phase angle (Figure 3).





Digital PD measurement systems comprise digital signal processing (DSP) units implemented in an FPGA and in conventional processors, with some additional signal conditioning circuits and A/D converters (Figure 4)



We developed the PD Analysis and Instrumentation System (IMA-DP) to measure the PD in the HF band (<30 MHz) according to the technical standards IEC 60034-27 and IEEE Standard 1434-2014. We could develop the system proposed here due to the availability of suitable modular hardware components in NI’s PXI platform.


The key to development was selecting the NI PXIe-5122 digitizer that features a 20 V dynamic range, 14-bit resolution, 100 MHz sampling rate, and 40 MHz of bandwidth. The module also includes circuits for impedance matching, selectable anti-aliasing analog filters, amplifiers, and attenuators.




We added a PXIe-2593 NI Switch module (Figure 4a) to the input of the PXIe-5122 (Figure 4b-e), which can expand the system for sequential measurement of up to 16 channels, significantly reducing monitoring costs per channel. The PXIe-5122 digitizer sends the acquired raw data to the FPGA of the PXIe-7965R module (Figure 4f) for heavy processing in real time. Sending data uses the PXI Express bus over a peer-to-peer connection. Final processing and consolidation of results happens in the PXIe-8135 embedded controller (Figure 4g). Figure 5 shows the complete measurement path. Figure 6 shows the built hardware and process flow diagram.




We used LabVIEW as the only programming language for the data acquisition, FPGA programming, signal processing and analysis, database, and reporting interfaces (Figure 7).













After the instrumentation we developed had proved its effectiveness in the laboratory and in the field, Eletronorte adopted IMA-DP as a predictive diagnostic tool in the evaluation of PD in hydrogenerators. The company has conducted continuous online monitoring since 2004 in two 350 MW generators in Tucuruí, and since 2013 throughout the Coaracy Nunes power plant (78 MW). Gradually, all companies of the Eletrobras group have adopted the system. And Itaipu Binacional recently selected IMA-DP to monitor its generators. Eletronorte has published several success stories in which the IMA-DP detected insulation failures and contributed effectively to generator safety.




We achieved an important result in 2009 when the IMA-DP successfully detected an early defect during the reconditioning process of one generator of the Coaracy Nunes hydroelectric plant. IMA-DP indicated abnormal PD patterns distinct from the other phases on the V phase of Generating Unit 3 (see Figure 9). An inspection on the stator coils revealed deterioration of surface points of bars between the front bar and the bottom bar on the bottom of the magnetic core (see Figure 10 and Figure 11). After the manufacturer made the repair, a subsequent measurement revealed that the defect had been corrected because PD levels were uniform and within acceptable values in the three phases. IMA-DP had then proved to be an effective tool for PD monitoring and diagnosis of insulation of stator windings of hydrogenerators.







Another interesting result is that continuous PD monitoring has helped us study the correlation between PD levels and other monitored quantities. Figure 12a shows the correlation between the instantaneous value of generated active power and the monitored PD levels, and Figure 12b shows the correlation between PD levels and relative vibration on a generator at Coaracy Nunes power plant.






By integrating modular PXI instruments, we could design a PD monitoring system based on signal processing algorithms developed by Cepel. Because it is based on the PXI architecture, the system is scalable, flexible, and fully integrable with other monitoring systems. With PD measurement, the monitoring of large hydrogenerators can now be carried out exclusively using NI hardware, as is the case of the two largest Brazilian hydroelectric plants: Tucuruí and Itaipu. Using NI hardware has aided the development of an effective tool with competitive cost and total control of the technology, a product of great success.



Author Information:
André Tomaz de Carvalho
Eletrobras Cepel
Av. Horácio Macedo, 354 - Cidade Universitária Rio de Janeiro
Rio de Janeiro
Tel: +55 21-2666-6446

Figure 1. Cross Section of a Generator Bar With Cavities in Which PD May Occur
Coupling Capacitors and PD Measurement Impedances Installed in the Three Phases of a Stator and Raw Signals Acquired
Figure 2. PD Raw Signals Immersed in Nois
Figure 3. PD Pattern or Phase-Resolved Partial Discharge (PRPD) Graph
Figure 4. Block Diagram of a Digital PD Measurement System
Figure 5. IMA-DP Measurement Path Through Selected PXI Express Modules Used in Implementation
Figure 6. (a) Measurement Hardware for Online Monitoring of 16 PD Channels; (b) General Flow Chart of the Measurement System
Figure 7. Analysis and Report Generation Interface Screens of IMA-DP
Figure 8. Partial Discharges Monitoring at Eletronorte Tucuruí Hydro Power Plant
Figure 9. PD Pattern Acquired From Phase V of the UG 03 of Coaracy Nunes Hydro Power Plan
Figure 10. Bar on the Bottom of Slot 134 and Signals of Heating on the Painting Surface, Possibly Due to Surface PD
Figure 11. Stator Bars in Slot 214 and Leakage of Impregnation Resin, Possibly Due to Deficient Curing Process; and Bottom Bar With Signals of Surface Erosion, Possibly Due to Surface PD or Corona
Figure 12. Correlation Between PD Levels and Generated Power on a Hydrogenerator of Coaracy Nunes Plant
Correlation Between PD Levels and Relative Vibration in the Guide Bearing of the Turbine on the Same Generato