Graphical Microwave Matching Networks Modeling and Design Optimization for Power Amplifiers

This paper introduces a graph neural network (GNN) approach to modeling microwave circuits for power amplifier (PA) design. By mapping matching circuit features to graphs, GNNs capture complex component relationships and predict circuit behavior. The proposed method consists of data labeling through simulations and GNN model training. Experiments demonstrate high accuracy in predicting circuit behavior compared to classical multilayer perception (MLP) models. The practical applicability of the proposed GNN model is showcased through integration into an optimization process for PA matching network design. This work promotes the adoption of machine learning (ML) techniques in microwave circuit modeling and design optimization.