基于改进GA-BP算法的RFID天线参数优化方法

为了提高算法对天线参数的预测精度,提出了一种基于 Adagrad 优化器的改进遗传算法-反向传播(GA-BP)算法。 通过在迭代过程中引入 Adagrad 优化器与阈值策略,对发生退化的种群最优个体的位置信息进行重新引导,解决了 GA-BP 算法局部寻优能力不足等问题,大幅度减小了误差损失并且加快了收敛速度。 利用该方法对射频识别(RFID)标签天线的印刷品质和电磁参数进行了建模与分析。 结果表明,改进 GA-BP 算法在稳步搜索极值的同时可以避免陷入局部极值陷阱,在误差和收敛效率方面均优于传统的反向传播(BP)算法与GA-BP 算法,能够得到较高的预测精度,实现了 RFID 标签天线印刷品质的优化控制以及 S11 特征曲线的预测。 相比于 BP 算法与 GA-BP 算法,改进 GA-BP 算法在用于优化RFID 标签天线的印刷品质时,平均绝对误差分别降低了91. 92%和85. 64%。 在电磁参数预测应用时,分别降低了 13. 77%和 13. 19%。

RFID Antenna Paramter Optimization Method Based on Improved GA-BP Algorithm

In order to improve the prediction accuracy of the algorithm for antenna parameters, an improved genetic algorithm-back propagation (GA-BP) algorithm based on Adagrad optimizer is proposed. By introducing the Adagrad optimizer and threshold strategy in the iterative process, the position information of the optimal individual in the degenerated population is redirected, solving the problem of insufficient local optimization ability of the GA-BP algorithm, significantly reducing the error loss and accelerating the convergence speed. Based on this method, the printing quality and electromagnetic parameters of the radio frequency identification (RFID) tag antenna are modeled and analyzed. The results show that the improved GA-BP algorithm can search for extremes steadily while avoiding the local extremum trap, and outperforms the traditional BP algorithm and GA-BP algorithm in terms of error and con-vergence efficiency, and can achieve higher prediction accuracy for the optimal control of RFID tag antenna print quality and S11 char-acteristic curve prediction. Compared with the BP and GA-BP algorithms, the improved GA-BP algorithm can reduce the average ab-solute error in the application of printing quality of RFID tag antennas by 91. 92% and 85. 64%, respectively. The reductions are 13. 77% and 13. 19%, respectively, when the improved GA-BP algorithm is applied to electromagnetic parameter prediction.