A Method for optimizing Impedance Bandwidth of Vivaldi Antenna Based on Random Forest-Differential Evolution Algorithm

In this paper, an intelligent optimization method is presented to optimize impedance bandwidth of Vivaldi antenna based on random forest-differential evolution (RF-DE) algorithm. The input to the random forest regressor algorithm is the Vivaldi antenna geometrical parameters and the output is the S-parameter, from which an accurate and efficient surrogate model is constructed. Then differential evolution algorithm is utilized to search for the antenna geometry information with maximum impedance bandwidth of a reflection coefficient less than −15dB based on the surrogate model. The surrogate model shows that mean absolute percentage error (MAPE) of parameter prediction is 3.96% for the test dataset. After the optimization using the differential evolution algorithm, the reflection coefficient is less than −15dB from 3.38 to 9.82 GHz and less than -12dB in the range of3-3.38 GHz and 9.82-10 GHz. The presented optimization method using the RF-DE algorithm is rapid and trustworthy, and thus could be applied to optimize antennas and arrays.