Parallel tolerance analysis of a power electronic converter by the genetic algorithm with the island model

It is necessary to consider variations of circuit element values for a power electronic converter design to satisfy a given specification. This analysis is so-called tolerance or worst-case analysis. The analysis is a sort of nonlinear minimum and maximum (min/max) problem with respect to circuit parameters. However its computation cost is large and practically the problem is converted to a simpler and easier one with various techniques. This paper proposes an approximate method which finds a quasimin/max solution by applying the genetic algorithm (GA). The GA models an evolution process of life forms for an engineering use and it is mainly applied to an optimization problem to utilize its selection process. This paper proposes the GA method which is based on the island model and suitable for the parallel tolerance analysis to find quasioptimal solutions. Output voltage variations, for example, of a Cuk converter for element parameter variations are analyzed.