Research On Optimized Modulation Strategy Of Dual Active Bridge DC-DC Converter

In recent years,with the development of the power electronics dominated power system,distributed renewable energy generation has become a research hotspot.In order to realize the collection,transformation and transmission of the renewable energy resources,the Dual-Active-Bridge(DAB)DC-DC converter has become a key energy conversion device.Due to many nonlinear parameters are contained in the DAB DC-DC converter,it is difficult to be modeled accurately by using the traditional mathematical modeling methods.Moreover,this converter contains many switching devices and its operating conditions are very complex in practical application,which leads to the defects of heavy calculation burden,high complexity,and difficulty in optimization when solving the optimization modulation strategy.Faced with such a high-dimensional optimization solution system with complex and changing modeling,artificial intelligence(AI)technology has shown its advantages in many system optimization decisions with large data volume,complex modeling and uncertainty,and is suitable for fast optimization solving problems of high-dimensional complex systems.Based on this,dissertation applies AI technology to the comprehensive optimization of the modulation strategy of the DAB DC-DC converter,focusing on the following three aspects of research work.Part Ⅰ: Optimization modulation strategy based on the linear piecewise timedomain modelAiming to improve the power efficiency of the DAB DC-DC converter,dissertation proposes a linear piecewise time-domain model based minimum-current-stress scheme.Firstly,the Q-learning as a typical algorithm of the reinforcement learning(RL)method,is adopted for offline training.The aim of the Q-learning algorithm is to solve the optimized modulation strategy based on the triple-phase-shift(TPS).More specifically,the ZVS constraints and each effective operation modes are taken into consider during the training process of the Q-learning algorithm.Thus,the cumbersome process for selecting the optimal operation mode in the conventional schemes can be circumvented successfully.After that,the training results of the Q-learning algorithm is used to train an artificial neural network(ANN),in order to reduce the computational time and memory allocation.The proposed TPS modulation strategy based on RL algorithm and ANN can effectively reduce peak current and rms current,and achieve the soft-switching performances over the entire operating range,thereby effectively improving the conversion efficiency and steady state performances of the DAB DC-DC converter.Part Ⅱ: Optimization modulation strategy based on the unified harmonic analysis modelAiming to overcome the disadvantages of the linear piecewise time-domain model,such as complex and not universal,dissertation proposes a unified harmonic analysis modelb based efficiency optimization modulation strategy for the DAB DC-DC converter.With the aim of reducing the reactive power of the DAB DC-DC converter,dissertation proposes a deep reinforcement learning(DRL)aided minimum reactive power modulation scheme.Specifically,as an advanced algorithm of the DRL method,the Deep-Deterministic-Policy-Gradient(DDPG)is used to train an agent off-line to solve the optimized modulation strategy.During the training of DDPG algorithm,the TPS modulation is adopted and the ZVS constraints are considered.The proposed DDPG algorithm aided TPS modulation strategy can effectively reduce the reactive power and achieve the soft-switching performance over the entire operating range of the DAB DCDC converter.Aiming to further promote the conversion efficiency of the DAB DC-DC converter,dissertation demonstrates a variable-frequency phase shift based minimum power losses modulation scheme.As an improved algorithm of the DDPG,the Twin Delayed Deep Deterministic policy gradient(TD3)algorithm is adopted to solve the optimized modulation strategy under the variable-frequency TPS modulation with the aim of minimum power losses and ZVS constrains.The proposed variable-frequency TPS modulation strategy based on the TD3 algorithm possesses four optimization degrees of freedom,which can further improve the conversion efficiency and steady-state performances of the DAB DC-DC converter with the soft switching performance.Part Ⅲ: Online efficiency optimization self-learning modulation strategy without prior knowledge about the circuit modelAs yet,the efficiency optimization of the power electronics converters needs to rely on their circuit models,and the accuracy of the power electronics converter model is of great importance for the efficiency optimization.However,the existing modeling methods cannot provide accurate power electronics converter models for the purpose of efficiency optimization,since the parasitic parameters of a converter are closely related to the components,their layout,and the device structure size.In order to solve above problem,dissertation presents an online efficiency selfoptimization method for the DAB DC-DC converter without prior knowledge about its circuit model.An online efficiency optimization self-learning system which can adapt to the environmental parameters are obtained,by integrating practical circuit equipment with DDPG algorithm for automatic exploration experimentation.Thus,the DAB DCDC converter can perform automatic exploration experiments in the constructed sixdimensional space.Based on this,an optimized TPS modulation strategy can be obtained,which can greatly improve the conversion efficiency compared with the latest existing optimal phase shift modulation approach.Besides,this online efficiency self-optimization approach can be deployed in the conventional power electronics converters for a range of operation performance optimization problems beyond the DAB DC-DC converter.In dissertation,a DAB DC-DC converter prototype platform is built and the corresponding experimental verification is undergone.The experimental results have well verified the feasibility and validity of the theoretical analysis.