A stochastic optimization technique for discrete DC capacitor bank design

Abstract

The DC capacitor bank is an expensive and bulky component, particularly in automotive applications. Its selection is subject to a number of competing constraints and is often sub-optimal. This paper proposes a generalized design methodology for the capacitor bank and introduces several optimization approaches under the umbrellas of convex, heuristic and pseudo optimization. Simulated annealing, a stochastic heuristic optimization algorithm, features prominently and shows good promise for automated selection. The discussed methods are assessed relative to one another for the design of the inverters and it was found that simulated annealing was able to reliably find the global minimum of the dataset, as well as returning solutions that were marginally less optimal than the ones obtained through careful manual optimization. The technique is general enough such that it can be applied to other components in a converter.