Dynamic equivalent modeling for power converter based on LSTM neural network in wide operating range

Abstract

The internal parameters and topology of the power converter are unknown in some practical cases. Existing modeling methods based on impedance frequency scanning method can only guarantee that the dynamic modeling is effective at a single working point. To make the established dynamic model effective in a wide range, an equivalent modeling method for power converter based LSTM Neural Network is presented. At first, the equivalence of black-box modeling problem and deep loop neural network is studied. Then, dynamic modeling method black-box power converter on wide operating range by using LSTM neural network is proposed. Finally, the simulation results under large disturbance and multi-operating points show that the proposed method is effective under wide operation range.

Keywords

• Power converter
• Dynamic modeling
• Deep learning
• Neural network
• Equivalent modeling