A Simultaneous Prediction Method for Multiple Characteristics of a Power Converter using Multiple-Output Hybrid Model

This paper describes a modeling method for rapidly computing the multiple characteristics of a DCDC converter. The proposed method rapidly predicts the gain and phase characteristics of a feedback-controlled plant that closely match those of actual circuits. Although a conventional method such as circuit simulation can achieve high accuracy, it requires significant computation time. The proposed method referred to as the Multiple-Output Hybrid Model (MOHM) combines physics-based model (physics model) and surrogate model. The surrogate model is based on an Artificial Neural Network (ANN) with multiple nodes in the output layer. The test results demonstrate that the computation time of the proposed method decreases to less than 4% of that of the circuit simulation. Compared to using only an ANN, the proposed method improves prediction performance in more than 68% of the evaluated test scenarios.