Usage of frequency response analysis for non-invasive diagnosis of the high-voltage transformers insulation

High voltage diagnostics methods short background:

Increasing the reliability and keeping the economical aspect at low levels are one of the upcoming trends in the field of a high voltage power machinery maintenance. Investments into the diagnostic processes could be returned rapidly, especially when these steps avoid cost-ineffective repairs. Periodical inspection together with the correct data representation are the key approaches to increase lifetime cycle of the high voltage machinery. Electrical testing relies on several methods which has been classified by tests and long-term data analysis as proper for the maintenance. Widely used methods are:

  • partial discharge measurement, localization - PD
  • capacitance and dissipation factor measurement - tgD
  • frequency response analysis - FRA
  • dielectric spectroscopy - DSP (FDS)

Mentioned methods are performed by specialized instruments which could be done fully analog or, in more modern devices, fully digital. By their complexity, they can be designed as on-line measuring system or as a standalone diagnostic system for periodical maintenance visits. It is common to build a poly-phase diagnostic instrument to produce complete data set of measured values at one time. This is very common for PD measurement for instance. Specialized algorithms for processing data underlying all these methods. Except real-time data visualization and representation more challenging is the long-term data comparison and estimation.

In other words some momentary critical value could not cause any failure but trends in whole datasets could predicate defects more accurately. To illustrate, some of the devices under test have their lifetime more the tens of years. For support this estimation, all measurements have to be done strictly in the same way in regular intervals with systematic data management for long time usage. Complete non-invasive diagnostic relies on very precise measurement device followed by data representation and comparison. 
 

Frequency response analysis:

The main idea behind the FRA process is to measure the transfer function of the inspected subsystem in some frequency span. Inputs and outputs of the particular transfer function could be freely chosen to obtain various data. Diagnostic result for the high voltage machinery FRA is typically a relation between the impedance of the DUT as a function of the excitation frequency. The typical frequency is from 10 Hz up to 5 MHz. It has been verified that even minor displacements in the geometric structure of the large power transformer windings or changes in the dielectric parameters of the insulation system will lead to the relevant changes in the FRA fingerprint.

Typical test connection for a low voltage SFRA
Fig. 1. Typical test connection for a low voltage SFRA

FPGA control board:

One of my main goals is the design of the digital control FPGA board which will suit FRA measurement together with data transfer and other tasks. Block diagram of this board is in Fig.2 and 3D model of the designed and developed HW unit is in Fig.3.

Block diagram of the FPGA board
Fig. 2. Block diagram of the designed FPGA board
3D modle of the FPGA board
Fig. 3. 3D model of the designed FPGA board

Awards:

Best Poster Award Competition at Poster 2016 conference for papers: FPGA-based System for Non-invasive Frequency Response Analysis Testing; An analog to ditial front-end for frequency domain spectroscopy analyzer

Related publications:

Tereň, O.; Tomlain, J.; Sedláček, R.; Vedral, J. "The design of an analog to digital front-end for frequency domain spectroscopy analyzer" In: 2016 15th Biennial Baltic Electronics Conference (BEC), Tallinn, 2016, pp. 155-158.

Tomlain, J.; Tereň, O. "FPGA-based System for Non-invasive Frequency Response Analysis Testing"
In: Proceedings of the 20th International Scientific Student Conferenece POSTER 2016. Praha: Czech Technical University in Prague, 2016. ISBN 978-80-01-05950-0.

Tereň, O.; Tomlain, J. "An analog to digital front-end for frequency domain spectroscopy analyze" In: Proceedings of the 20th International Scientific Student Conferenece POSTER 2016. Praha: Czech Technical University in Prague, 2016. ISBN 978-80-01-05950-0.