Multi-objective Optimization of High-frequency Transformer for Dual-Active-Bridge Using Multilayer Perceptron-Genetic Algorithm

This paper presents an optimal transformer design methodology tailored for Dual-Active-Bridge converters, lever-aging the capabilities of the Multilayer Perceptron and the Non-Dominated Sorting Genetic Algorithm II for multi-objective optimizations. The primary focus is on minimizing both the physical weight and the energy losses of the transformer. We develop a detailed and accurate loss model that forms the basis of our optimization process. The methodology involves intricate algorithmic development, ensuring precise and reliable performance in transformer design. To validate our approach, we empirically characterize a prototype transformer, demonstrating improvements in efficiency and compactness. The results underscore the effectiveness of combining Multilayer Perceptron and Non-Dominated Sorting Genetic Algorithm II in achieving optimal balances between size and loss, offering valuable insights for designing high-performance DAB converters.