Design optimization of switching-cell-array-based power converters

Abstract

With the aim to increase standardization, reduce the cost, and obtain advanced performance features, the design of voltage-source power converter legs can be undertaken by combining several instances of a standard switching cell, properly connected in active neutral-point-clamped structures to reach the desired voltage and current ratings. These switching cells can be organized into switching-cell arrays. This design approach introduces several degrees of freedom into the design. Namely, the different options to interconnect the cells and the distribution of switching losses among these cells. This article aims to define an optimization problem to explore this design space. The design problem is formulated in different scenarios, involving different conversion configurations (dc-dc and dc-ac), different leg number of levels (two and three), and different types of available cells (standard and conduction-optimized in combination with switching-optimized). A weighted objective function is then defined in terms of leg simplicity, efficiency, and reliability. The value of the design variables that minimize the objective function with different sets of weighting factors are obtained under selected scenarios and operating conditions, to illustrate the flexibility of the converter design approach under study. The solution of the optimization problem is obtained using a surrogate optimization algorithm in MATLAB, well suited to quickly solve optimization problems involving a combination of integer design variables (the number of parallel switching cells in each converter leg position) and continuous design variables (the proportion of switching losses taken by each cell), together with linear and nonlinear constraints.