Parameter Optimization of Balise Circuit Based on Fusion of BNN and Genetic Algorithm

Abstract

The optimization of the parameters of the components related to the radio frequency (RF) transmission circuit of the balise can keep the balise working normally under low power consumption and increase the reliability and stability of the high-speed railway vehicle-ground communication. However, the circuit has high complexity, many parameters need to be considered in optimization, and the constraint relationship is complex. Optimizing a single objective is very difficult and time-consuming. Therefore, this paper proposes a ground transponder design and optimization method based on deep learning. Firstly, the functional modules of the balise RF circuit are decomposed, and the influencing factors of circuit start-up conditions and load quality factors are analysed, and the component parameters that need to be optimized are extracted as decision variables. The objective function of the model is established from the perspective of circuit cost and static power consumption, and a multi-objective optimization model is established through its overall circuit scheme. Finally, in order to reduce the time cost, the multi-objective optimization model is processed by the fusion of neural network and genetic algorithm. Among them, the experimental results show that the optimization effect of Bayesian neural network (BNN) is the most significant, and the static power consumption and cost of the circuit can be reduced by 55% and 42%, respectively, with less time overhead.