An Artificial Neural Network-Based System-Level Modeling of Power Converters for Real-Time Simulation

FPGA-based real-time simulation is a powerful tool to test and validate power converter designs and their associated controls before field application. Half-bridge modules are the basic building blocks of many power converters. However, they are usually simply modeled by a piecewise linearized model, which neglects the nonlinear characteristics of power switches. Therefore, to improve the half-bridge model's accuracy while maintaining a high computational efficiency, this paper proposes an artificial neural network (ANN)-based half-bridge real-time system-level modeling approach. In this model, the nonlinear static characteristics in relation to the current, voltage and junction temperature of power switches are considered. Furthermore, a floating interleaved boost converter (FIBC) is modeled under the proposed modeling approach and simulated with a 200 nanoseconds (ns) time-step on a National Instrument (NI) PXIe-7975R FlexRIO real-time platform. The accuracy of the proposed model is validated against the results from the Simulink Simscape model and compared with the Simulink SimPowerSystem model as well. The effectiveness of the proposed model is also verified by hardware-in-the-loop (HIL) experimental tests.