cGANET-Enhanced Voltage Gain Modeling: Elevating CLLC Converter Accuracy

The voltage gain model of resonant converters is crucial for parameter design and optimization. Common methods, such as Fundamental Harmonic Approximation (FHA), lack sufficient accuracy, and time-domain analysis (TDA) is computationally intensive and varies across different operating modes. Therefore, this paper proposes a voltage gain prediction model based on conditional Generative Adversarial Networks (cGANET). The new voltage gain model is developed by training the cGANET with a dataset of voltage gain data from simulations under various conditions, enabling accurate predictions of the complex, non-linear interactions within the resonant converter without relying on simplifying assumptions or mode-specific formulas like FHA and TDA. Leveraging its powerful extrapolation and interpolation abilities, the cGANET model maintains high accuracy for voltage gain within and beyond the training range, even with limited training data, and adapts to new conditions without extensive reconfiguration. To validate the proposed model, a prototype CLLC converter was constructed and tested across various switching frequencies. The results demonstrates that the proposed model improves accuracy by 60% under the worst conditions compared to FHA, beneficial for enhancing design and optimization of resonant converters.