Polynomial chaos expansion for sensitivity analysis of two types of transmission line models

Abstract

This paper presents two sensitivity analyses, one of which is for the time-domain transmission line (TL) model using the latest weight-based algorithm that is a combination of the coherence-optimal sampling (COS) and the l1-minimum problem (l1-M) and the other is for the high-frequency field line coupling model based on the algorithm that combines the Latin hypercube sampling (LHS) with the l1-M for the first time, where the COS and the LHS are employed to draw samples and the polynomial chaos expansion (PCE) coefficients are obtained by solving the l1-M. Both algorithms are within the PCE framework. The results show that for the time-domain TL model, the capacitance has the greatest impact on the voltage response, followed by the resistance. The height and the radius have the least impact, while there is little difference between them. For the high-frequency field line coupling model, the azimuth angle has the greatest impact on the current in the TL, followed by the elevation angle. The terminal load, the conductivity, the field amplitude, the relative permittivity, and the TL radius have the least effect, and there is little difference among them. Additionally, both algorithms only require no more than 0.60% of the total Monte Carlo (MC) computation time when yielding solutions with almost identical accuracy. Compared to performing a sensitivity analysis using the MC method, the use of two algorithms significantly improves efficiency.