Multi-objective GPR-based design optimization of high frequency transformers

FEA-based design optimization of high frequency transformers (HTSs) are by far more accurate but they come at huge computational cost. This paper presents a Gaussian Process Regression-based multiobjective design optimization technique for HFTs. The GPR model is formulated mathematically and trained using extensive FEA-based multiphysics simulations to predict three optimization objectives which include the HFT’s power loss, cost and power density. The GPR model is coupled with a Non-dominated Sorting Genetic Algorithm multiobjective optimization algorithm to determine the Pareto Optimal Solutions (POS). The optimization variables are the geometrical parameters of the HFT core and are constrained based on practical amorphous cores designs. The HFT is integrated into dual active bridge DC–DC resonant converter. The GPR-based multi-objective optimization results are compared with those obtained more accurate FEA-based solutions. The results obtained are validated experimentally. The results show that the proposed GPR-based HFT design optimization is a very promising framework in determining the optimal design parameters of the HFTs with acceptable level of accuracy.