麻雀算法改进BP神经网络的4D航迹预测

4D Trajectory Prediction Based on SSA Optimizing BP Neural Network

提出基于麻雀算法(sparrow search algorithm,SSA)优化BP(back propagation)神经网络的改进算法(SSA-BP)进行4D航迹预测,在一定程度上改进了单靠神经网络预测航迹的精度。首先利用麻雀算法计算寻找全局最优解,提高初始解质量,增强算法的全局搜索能力。其次,将最优解作为BP神经网络输入层,进行权值和阈值更新,优化了BP网络神经的自学习和自适应能力。通过ADS-B(automatic dependent surveillance-broadcast)数据插值后形成航空器4D航迹并作为麻雀算法输入变量进行SSA-BP模型的4D航迹预测。最后将SSA-BP算法模型、单纯BP网络模型的预测结果与航空器真实航迹进行比较。实验得出,在规定时间序列内,SSA-BP算法模型的经度、纬度和高度的三种误差比单纯BP神经网络模型更小,能够实现高精度的4D航迹预测。

An improved algorithm(SSA-BP)based on sparrow search algorithm(SSA)to optimize BP(back propagation)neural network was proposed for 4D track prediction,which improved the accuracy of track prediction by neural network alone to a certain extent.Firstly,the sparrow algorithm was used to find the global optimal solution,improve the quality of the initial solution,and enhance the global search ability of the algorithm.Secondly,the optimal solution was used as the input layer of BP neural network,and the weights and thresholds were updated to optimize the self-learning and adaptive ability of BP neural network.Then,the 4D trajectory of the aircraft was formed by interpolating ADS-B(automatic dependent surveillance-broadcast,)data and used as input variables for the SSA-BP model's 4D trajectory prediction using the sparrow algorithm.Finally,the prediction results of the SSA-BP algorithm model and the simple BP network model were compared with the real trajectory of the aircraft.Experiments show that in the specified time series,the three errors of longitude,latitude and height of SSA-BP algorithm model are smaller than those of simple BP neural network model,and can achieve high-precision 4D track prediction.