IET Activities Diagnosis Of The Condition Of Electric Components Using Transfer Funktions (transformer analyze)

by Søren Damsgaard Mikkelsen
RÉSUMÉ
In this report the performance of transfer functions in diagnosis of faults and ageing phenomena in components in electrical power systems has been tested. The basic principles of transfer function based diagnosis are explained in the following:

A low voltage test signal is applied to the device to be analyzed through a coupling circuit. This input signal is measured together with an output signal from the component and the obtained digitized data areused in a computer to calculate a dynamic input/output model of the device - the transfer function.

The calculated transfer function is subsequently compared with a reference model obtained when the device was new to determine whether differences has emerged. The differences, if they appear, are the key to evaluate the condition of the component and predict the time to breakdown.
Two transformer types are used as test basis in the research work....
The first step in the investigation is to survey the incipient failures and ageing phenomena to be detected by transfer functions.

Based on a systematic description of the most commonly appearing failures intransformers and ZnO surge arresters in electrical power systems a common pattern of symptoms for allthese irregularities is given.For transformers in electrical power systems four characteristic symptoms of incipient failures and ageing phenomena are set up:
a. Raised eddy current/magnetic losses.
b. Displaced turns/windings.
c. Decomposed or stressed winding insulation.
d. Acid, moisture and sludge in the insulating oil.

For ZnO surge arresters a single symptom is set up:a. Increased leakage current. These symptoms are the basic symptoms to be detected by transfer functions, if the method is going to be an effective tool in diagnosis of transformers and ZnO surge arresters in electrical power systems.

The second step in the research work is to set up the basic theoretical models describing the transformer and the ZnO surge arrester dynamic. An important demand for a descriptive model to be used in diagnosis is that - the model is closely connected to the physical operation of the component.

A mathematical description like a s-domain transfer function is thus less suited for diagnosis purposes than an equivalent circuit with RLC parameters related directly to the physical operation of specific parts within the component. Two basic equivalent circuit configurations describing the dynamic of single phase transformers and ZnO surge arresters, respectively, are set up for the project.The third step in the research work is to find the identification methods to identify the dynamic behaviourof transformers and ZnO surge arrester.

The most important demand for an identification method to be used in diagnosis is that- the identification result is stable and reproducible under unchanged conditions, i.e. the stochastic deviation of the identified model is small.

Basically, the available identification methods can be divided in two groups: non-parametric methods and parametric methods.The non-parametric identification methods determine a non-parametric model of the device under test.

The non-parametric model is usually a frequency domain model showing the amplitude gain and phasedisplacement of the component versus frequency. Such a frequency domain model forms together with an equivalent circuit a nearly complete basis to evaluate the dynamic (like the existence of resonant frequencies, impedance level, phase angle etc.) of a component.

The two most commonly used methods to determine the frequency response of a device is frequency sweep analysis and FFT analysis.

Both methods are thoroughly tested in this project by impedance identifications performed on the two singlephase transformer types and in the leakage region of the two ZnO surge arrester types. The result of these fundamental modelling tests are:- Frequency sweep analysis gives highly stable and reproducible models of transformers as well as of ZnO surge arresters.- FFT analysis gives stable and reproducible models of transformers, while the dynamic range of the available test equipment is too small to obtain reliable results on ZnO devices in the leakage region.

The parametric identification methods determines the parameter values of a parametric component model. The parametric model can be an equivalent circuit with RLC parameters or a general s-domain transfer function with a suitable number of coefficients. Several methods to solve the problem ofparameter identification in a given model from recorded test data are available.

In this project LeastSquares method is tested by impedance identifications performed on transformers and in the leakage andnormal operation region of ZnO surge arresters. The results of the these preliminary identification testsare:- Parametric identification gives stable and reproducible results on transformers and on ZnO surgearresters in the normal operation region. Contrary to this are the results of parametric identification in leakage region of ZnO arresters useless due to the same reason as stated for FFT analysis, i.e.limitations in the test equipment.

Three different identification methods are thus concluded to produce stable and reproducible results on transformers: frequency sweep analysis, FFT analysis and parametric identification.Two identification methods are similarly determined to give satisfactory results on ZnO surge arresters:frequency sweep analysis in the leakage region and parametric identification in the normal operation region.

The final but very important step in the research work is to determine the relation between the identified models and the condition of the devices concerned.Two test series are performed on the available transformers: sensitivity tests and ageing tests. The former test series is used to determine the sensitivity of the identified models to the condition of the insulating oil/grease, the magnetic circuit and the winding insulation. The latter test series is performed inorder to see whether or not the dynamic model of a transformer changes due to ageing phenomena andto determine whether or not it is possible to detect a development towards breakdown.

A single test series is performed on the available ZnO surge arresters: ageing tests. As the performed ageing tests on transformers, the ageing test series on ZnO surge arresters serves the purpose of investigating the development in the identified models towards breakdown.

The following results are obtained via the performed tests:- Detection of faults and ageing phenomena in a transformer is possible from the transfer function implied that changes in the lumped winding capacitances, the total losses, the reluctance of the core orthe geometry of the windings occurs.

However, a complete and precise condition evaluation of a transformer from the transfer function is impossible to obtain. The number of parameters included in the dynamic transfer function is thus insufficient.- Detection of faults and ageing phenomena in a ZnO surge arrester is also possible. However, in this situation it is the question, whether or not it is reasonable to use the described relatively complicated method.

A much simpler alternative is thus proposed in the project based on the experiences gained.

To view the full abstract.
To order the paper:
http://www.iet.aau.dk/