A Physics-informed Neural Network Method for LC Parameter Estimation in Three-Phase Inverter

The DC-link capacitance and the AC-side inductance parameters can be used as feedback for control optimization and component degradation monitoring. This paper proposes a parameter estimation method based on the combination use of artificial neural network and circuit analytical models, e.g., physics-informed neural network (PINN), for a three-phase inverter application. It does not require any additional hardware circuitry and can be well-trained based on a small training dataset. A three-phase inverter case study is presented with theoretical analyses, simulations, and experimental verifications. The results show that satisfactory accuracy can be achieved for the estimation of DC-link capacitance and AC-side inductance parameters.