Behavioral Switching Loss Modeling of Inverter Modules

Kateryna Stoyka,Ricieri Akihito Pessinatti Ohashi,Nicola Femia
DOI: https://doi.org/10.1109/SMACD.2018.8434850
2019-06-11
Abstract:This paper presents a new behavioral model for switching power loss evaluation in phase-shifted full-bridge inverter Power Modules (PoMs). The proposed model has been identified by means of a Genetic Programming (GP) algorithm combined with a Multi-Objective Optimization (MOO) technique. A large set of loss data, evaluated by means of analytical loss formulas, has been considered for the identification of a compact behavioral model. The GP-MOO approach considers the inverter switching frequency, input voltage, duty-cycle and load resistance as model input variables, and the MOSFET gate driver voltage and resistance as parameters influencing the coefficients values of the identified loss formula. The behavioral model loss predictions confirm their reliability for a wide range of operating conditions.
Numerical Analysis,Signal Processing
What problem does this paper attempt to address?
The problem that this paper attempts to solve is the problem of accurately evaluating the switching power loss in full - bridge inverter power modules (PoMs). Specifically, the paper proposes a new behavioral model for evaluating the switching power loss of silicon carbide (SiC) MOSFET power modules under different operating conditions. Traditional analytical models may be inaccurate when the operating conditions of the device are inconsistent with the characterization conditions of the model, while numerical models can provide more accurate predictions but require a long simulation time and may have numerical instability problems. Therefore, the paper uses the genetic programming (GP) algorithm combined with multi - objective optimization (MOO) techniques to identify a compact behavioral model. This model takes the inverter's switching frequency, input voltage, duty cycle, and load resistance as input variables, and at the same time, the MOSFET gate - drive voltage and resistance as parameters affecting the values of the model coefficients. Through this method, the paper aims to provide a reliable and efficient design tool for power system efficiency prediction under a wide range of application conditions.