Multiscale Modelling of a Metallized Film Capacitor for HVDC Applications

Abstract

Metallized dielectric films used for high-voltage direct current (HVDC) Light capacitors are subjected to voltage ripples in operation. This leads to heat generation and a temperature rise in the capacitor. The operating temperature, which affects the function and the lifetime of the capacitor. Designs of capacitor units are evaluated by thermal stability tests to ensure thermal stability under overload conditions. It is therefore important for the designer to know locations of hot spots, the average temperature and the maximum temperature. This thesis presents a multiphysics model of a HVDC Light capacitor unit simulating electric currents and heat transfer under the conditions of a thermal stability test. A capacitor unit consists of capacitor elements made of metallized films with large difference in scales. To overcome this, a multiscale approach have been used to model the capacitor unit. Effective properties and losses are computed on a small scale (several layers of metallized film) and utilized on a large scale (capacitor unit). The performed calculations yield the impedance of the capacitor unit corresponding to the rated impedance with the effective properties implemented. The loss computed on the small scale is lower than average measured loss of a capacitor element. Additionally, the loss computed on the large scale matches the average measured loss of a capacitor unit. The hot spot of the capacitor unit is located at the front. The elements located there are hotter than the elements at the rear. A multiscale modelling approach is utilized to model metallized film capacitors and to simulate thermal stability tests to estimate hot spots, average and maximum temperatures. The simulation results are compared to the results from a thermal stability test in a laboratory environment. This comparison shows that the simulated results are close to the test results. However, there are uncertainties related to heat transfer coefficients at different exterior boundaries of the unit and comparison with additional tests may improve the model.