Design of experiments for characterization of a high voltage circuit generating a transient recovery voltage

Abstract:
Test circuits and procedures are a common method to stress components of the electrical power grid and prove their reliability according to standards. The build-up and characterization of precise test circuits is mostly an iteration process that requires experience. This paper presents Design of Experiments (DoE) methods cannot only be used for tests itself but apply for the building and verification of test circuits while considering parasitic elements. These methods are exemplary performed for the transient recovery voltage (TRV) stress at circuit breakers up to 72.5 kV. In the event of a fault, the circuit breaker disconnects all components and power lines to protect reliably all other grid participants. If this occurs, the circuit breaker is stressed by high short circuit currents and subsequently by TRV, which can be generated synthetically. The aim of this investigations are reliable statements about the required output values in respect to the predictions made by a created model. This investigation uses the DoE in combination with a system theory approach to create this model of output signals of the high voltage circuit by using modular components of the circuit. Short circuit and open circuit examinations are being performed to analyse a “black box” where the relation of output and input signals are unknown. The D-optimality design for experiments is used, which guarantee a statistical stable calculation but also minimize the volume of the step interval of the input signals. This results in a minimum number of experiments for creating a prediction model of a test circuit. It shows a method to determine the output signals in dependence on the input signals and the creation of a regression by the application of DoE method. The design and the calculation of a regression model is generating with the software tool “Cornerstone”. The resulting regression model can be used parameter variations without redesigning the test circuit. A valid interpolation was performed for peak value of the TRV of 50 kV to 130 kV and varying frequencies between 4.6 kHz and 19.6 kHz.