Adaptive Polynomial Approximation-Based Data-Driven Model of Parameter Design for Bidirectional L-LLC Converter

The L-LLC resonant converters have excellent performance to realize bi-directional power flow due to the symmetrical circuit structure of the resonant tank. The traditional frequency-domain based design approach of L-LLC converter suffers inaccuracies, and the traditional time-domain based design approach usually utilizes discrete process to find the optimal resonant parameters, which requires large computation and cannot ensure the actual optimization. Therefore, an optimal design method of the L-LLC converter based on data-driven model is proposed in this paper. Based on the numerical calculation method in bidirectional power flow, a closed resonant parameter space is established by satisfying the constraints including soft switching, operational mode, switching frequency and resonant capacitor voltage stress. The optimization objective is to minimize the overall losses of the converter. To reduce computational complexity, the adaptive polynomial approximation (APA) method is employed to fit the constraints and optimization objectives, thus the surrogate model is constructed. Then, the nondominated sorting genetic algorithm (NSGA-II) is employed to search the pareto-based optimal solution, which guarantees the synthetical minimal power losses in bi-direction power flow. Finally, a 400V/1000W experimental prototype was built to validate the effectiveness of the proposed design method.