| dc.description.abstract | Estimation of impulse waveform parameters (viz. test voltage, front time, tail time) from digital data acquired during high voltage impulse tests-particularly on transformers, cables, and reactors-poses difficulties due to the presence of superimposed oscillations and/or overshoot. Solutions proposed so far show large scatter in the estimated parameters. This is because procedures in IEC 60060 1 and IEEE Std 4 (originally formulated to suit manual evaluation) are somewhat ambiguous, inconsistent, and therefore difficult to convert into an algorithm.
Motivated by this, an EU funded project (involving five HV laboratories) was undertaken, whose objectives were:
Experimentally establish (for air, oil, XLPE, and SF media) a correlation between test voltage and the parameters of oscillation/overshoot under different electric stress and polarity conditions.
Consequently, propose a method for estimating test voltages, valid for all types of lightning impulses.
Eventually, revise IEC 60060 1 and 61083 2 to eliminate inconsistency, empiricism, and ambiguity.
The goals set by the project group were quite remarkable. The project finally resulted in a method (called the k factor) for estimation of impulse test voltage, involving a filtering based implementation. Some details of the k factor method were published on behalf of CIGRE WG 33 03 in Electra in October 2002, along with a recommendation for revision of IEC 60060 1 and 61083 2.
Subsequently, drawbacks in the k factor method were found and reported in a Letter to the Editor in Electra in April 2003, along with a response from the project group. From these two publications (and other internal working group documents), it emerges that the k factor proposal contains infirmities and therefore is unacceptable. However, these infirmities are somewhat masked in the publications, either due to brevity imposed or false logic being employed, thereby giving the method an artificial sense of acceptability.
To date, it is not certain what the final version of the k factor method would consist of, nor the steps involved in its implementation. While work still continues to resolve core issues related to it, efforts are also underway to include it in international standards-perhaps oblivious of its shortcomings. In anticipation of such a revision, two reputed companies selling HV impulse analysis systems have already included this method in their existing software and are ready to market it. Such a contrasting scenario necessitates a detailed evaluation of the k factor method to identify its lacunae. This forms the motivation and objective of this thesis.
This thesis presents a detailed analysis based on an assessment of various aspects that have influenced its formulation, namely:
Experimental results
Treatment of breakdown data
A single curve representing behaviour of all dielectrics
Myths and realities about filtering
Issues in filter realisation
Residual and global filtering, etc.
Specific attention is focused on matters related to the implementation of the k factor method. In fact, the very basis of implementing it as a filtering operation is strongly criticised, and it is proven (by simulation) that it can lead to inconsistencies in the definition of the k factor. This is one of its major drawbacks. Problems arising due to filtering a transient waveform and those associated with obtaining a single mean curve (using the Levenberg–Marquardt algorithm) in the “residual filtering” scheme are examined. Inconsistency in the definition of overshoot arises while processing practical data, thereby questioning the viability of the residual filtering scheme.
Supported by simulations and studies on practical data, answers to some important questions are attempted, namely:
Does the k factor method, per se, satisfy the requirements expected of a standard method
Were the initial goals-removal of empiricism, ambiguity, etc.-actually achieved
How can the experimental k factor data be used more meaningfully
In summary, it emerges that unless all identified drawbacks in the k factor method are adequately resolved, it would be premature to recommend its adoption into IEC standards. Details are presented in the thesis. | |
