Research On Multi-model Predictive Control For Bidirectional DC/DC Converter Of New Energy Vehicle

Hybrid electric vehicle of new energy has broad application prospects,as a key component of new energy vehicles,bi-directional DC/DC converters are currently one of the research hotspot in this field.This thesis aims to improve the dynamic performance and efficiency of the converter,the research about multi-model predictive control for bi-directional DC/DC converter of new energy Vehicle was conducted,and the main contents are as follows:The topology of the bi-directional DC/DC converter was designed.According to the automotive system power requirements,the parameters of the bi-directional DC/DC converter were determined,based on these,the topology of the converter was designed,then the main parameters of the power circuit were calculated,and key devices such as IGBT were selected,parameter and specification of the radiator was designed.Multi-model of bi-directional DC/DC converter was established.According to the circuit theory,the average state space equation was established,the prediction model under the discrete time can be derived form the predictive control theory.For the nonlinear characteristics of the bi-directional DC/DC converter,a scheme of approaching the real model by establishing multiple linearized sub-models was proposed,several sub-models of bi-directional DC/DC converter were established,this model lays the foundation for the design of multi-model predictive controllers.A multi-model predictive controller was designed.Two-layer control structure was adopted,the upper layer solves the nonlinear problem of the bidirectional DC/DC converter,the principle of multiple models was analyzed,parameter adaptive feedback correction was added in the controller,by calculating the model matching degree and control weight online to adapt the converter's model,a multi-model adaptive controller is designed.The lower layer solves the optimal control problem,according to the system control requirements,the objective function was designed,according to the actual situation,the optimization control problem of constrained predictive control algorithm was reduced to a quadratic programming problem,a recursive neural network optimization algorithm was used to solve the quadratic programming problem,a model predictive controller was successfully completed.Delay compensation plays an important role in predictive control,so a delay compensation scheme was proposed.A simulation platform was set up to verify the effectiveness of the multi-model predictive controller.Bi-directional DC/DC converter experiment platform based on DSP plus FPGA was designed,and it was verified by experiments.The controller adopts DSP plus FPGA hybrid architecture,FPGA is mainly responsible for the implementation of the predictive control algorithm,DSP implements other functions.Based on the hardware design,software for the main modules of DSP and FPGA was designed.A bidirectional DC/DC converter multi-model predictive control experimental platform was set up,an experimental study was conducted on the main research objectives of this thesis,base on the simulation platform verification theory,the experimental results proved the feasibility of multi-model predictive control algorithms and schemes.