Parameter Estimation of DAB Converter Using Intelligent Algorithms and Steady-State Modeling Considering Nonidealities
Abstract
Parameter estimation of dual active bridge (DAB) converters is a powerful tool in the applications of fault location, modulation strategy optimization, and preventive maintenance. However, most of the existing parameter estimation techniques often rely on costly experimental measurements and only estimate some converter parameters. To address these problems, we apply intelligence techniques to propose a flexible, generalized, and reliable multiparametric estimation methodology. Specifically, the circuit nonidealities, including the dead-time effect and the power loss, are considered to derive accurate open-loop and closed-loop steady-state models of the DAB converter. Then, the dynamic behavior and stability of the DAB converter are investigated based on those theoretical models. Next, a back-propagation (BP) neural network optimized by the genetic algorithm (GA) is employed to establish training models. Notably, the datasets used in training are generated from the derived DAB models, eliminating the need for parameter data collection from different converters. Finally, the accuracy, robustness, and generalization ability of the GA-BP are compared with traditional estimation methodologies and the pretrained BP model. Illustrative results demonstrate that the proposed intelligent approach is an effective approach for parameter estimation.
