The dual active bridge converter design with artificial intelligence

• Main Author: Lin, Fanfan
• Other Authors: Cai Wenjian
• Format: Thesis-Doctor of Philosophy
• Language: English
• Published: Nanyang Technological University 2022
• Subjects: Engineering::Electrical and electronic engineering::Power electronics; Engineering::Computer science and engineering::Computing methodologies::Artificial intelligence
• Online Access: https://hdl.handle.net/10356/163850
• Institution: Nanyang Technological University

The dual active bridge (DAB) topology has been recognized as the key circuit for the next generation of high-frequency-link power conversion systems. It plays an important role in the hybrid AC-DC microgrid to accommodate renewable energy resources. To realize the optimal operating performance of the DAB converter in a designated application scenario, the converter requires careful design, which includes circuit parameter design and modulation strategy design. Various performance metrics will be usually considered in the design process. However, there are some challenges in the DAB converter design that slow down the penetration of this useful converter. The challenges can be summarized as the heavy computational burden and the low design accuracy, attributable to the number of designable circuit parameters, the degrees of freedom in modulation strategies as well as the performance metrics considered. Thanks to the development of artificial intelligence (AI), this emerging technology provides potential solutions for those challenges. Therefore, this thesis mainly focuses on the design of the DAB converter with the application of artificial intelligence to automate and speed up the design process. To deal with this research scope, the second chapter of this thesis provides a comprehensive literature review. The review firstly covers the working principles, the resonant tanks, and the modulation strategies of the DAB converter. After these, the state-of-the-art artificial intelligence techniques and their applications in the design of power converters are also reviewed. The review of relevant research works indicates the great potential and significance of the application of artificial intelligence in the design of the DAB converter. How to design the parameters in the resonant tank for the DAB converter is one of the concerns. For example, the CLLC resonant tank has multiple designable parameters, and the choices will affect the operating performance. The unpredictive fluctuations of parameter values due to temperature changes make the design harder. To decide the CLLC parameters for the sake of comprehensively optimal performance, a CLLC resonant tank parameter design approach with the particle swarm optimization algorithm is proposed. This design approach takes three objectives into consideration: robust voltage conversion ratio, high efficiency, and stability in a cascaded system. The designed DAB converter is able to meet requirements in all of the three objectives even if the parameters fluctuate in practical operation. The experimental results have proven the effectiveness of this proposed CLLC parameter design approach. The issue pertaining to the modulation strategy design is also discussed in this thesis. The triple phase shift (TPS) modulation strategy has three degrees of freedom. The modeling for the TPS modulation strategy faces an unbalance between computational simplicity and modeling accuracy. To deal with this challenge, an AI-based TPS modulation strategy is proposed for the sake of optimal efficiency. Instead of the conventional manual modeling process, the neural network is adopted to build a surrogate model automatically. And then the particle swarm optimization algorithm finds optimal modulation parameters for the best efficiency performance. To perform TPS modulation, the traditional look-up table is replaced by a fuzzy inference system to avoid discrete values in the look-up table. Experiments have been conducted to verify the effectiveness of this proposed modulation design strategy. Besides single modulation strategy design, this thesis also covers hybrid modulation strategy design. Extended phase shift and double phase shift are both considered in this hybrid modulation strategy to combine their advantages. However, this hybrid modulation increases difficulties in modeling compared to the single modulation strategy. Therefore, an advanced machine learning technique is leveraged to design this hybrid modulation strategy. A surrogate model is built by the machine learning technique. After this, the differential evolution algorithm is used to optimize modulation parameters with the objective of smaller current stress. Hardware experiments have validated this proposed hybrid modulation strategy. In addition to the aforementioned research works, this thesis proposes several potential research directions in this area. More design aspects in the DAB converter can be investigated like the transformer design and the controller design. And the modeling process proposed in this thesis can be further improved with real-time modeling. Moreover, a scenario-based design system is also worth researching to lower the requirements on the designers’ expertise in the objective-based design system.