Neural network based model-free predictive control method for dual active bridge converter
By constructing a hyperlocal equivalent model and a neural network observer, a model-free predictive controller was designed, which solved the problems of model accuracy dependence and disturbance observation in dual active bridge converters, and achieved rapid following and stable control of the output voltage.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- 西安航科创星电子科技有限公司
- Filing Date
- 2024-06-07
- Publication Date
- 2026-06-12
AI Technical Summary
The existing dual active bridge converter control suffers from the dependence on model accuracy and the difficulty of the extended state observer to handle complex disturbances, which affects control performance.
A circuit model of a dual active bridge converter is constructed. Based on the actual value of external disturbances in the system, a neural network observer is designed for online fitting. A model-free predictive controller is designed based on the hyperlocal equivalent model and the estimated value of external disturbances in the system. Control is achieved by driving an insulated gate bipolar transistor through a PWM generator.
It can achieve rapid following of the output voltage reference value of the dual active bridge converter without prior information, ensuring stable operation, simplifying the design and providing strong anti-disturbance capability.
Smart Images
Figure CN118763904B_ABST