Probabilistic Approach to InsulationCoordination

Abstract: The present work was performed at HVDC ABB as an initial study on how to adopt probabilistic concepts into the VCSHVDC insulation coordination. Due to large voltage levels in HVDC applications the corresponding insulation need to be properly addressed to ensure a safe, economical and reliable operation. Traditionally, only the maximum overvoltage is considered, where no adoption to the shape of the overvoltage distribution is regarded. Use of probabilistic concepts in the insulation coordination procedure can ideally reduce insulation margins with a maintained low risk of flashover. Analysis and understanding of probabilistic concepts of AC systems is needed in order to implement the concepts into VSC-HVDC. With use of advanced VSC-HVDC models, faults are simulated with varied fault insertion time in PSCAD. The resulting overvoltages from the simulation is gathered using different statistical methods in order to obtain the approximated overvoltage distribution. It was found from the simulation results that use of a Gaussian distribution is inappropriate due to shape variety in the overvoltage distributions. Instead, Kernel Density Estimate can serve as a flexible tool to approximate overvoltage distributions with a variety in number of modes and shape. The retrieved approximated overvoltage distributions are compared with the insulation strength in order to calculate the risk of flashover. The comparison shows that the insulation can be tuned in order to match set requirements. The thesis work should be seen as pilot study, where key problems have been pointed out and recommended further studies are proposed.