基于PSO-BP神经网络的平流层风场短期快速预测

平流层风场环境对临近空间低速飞行器驻空飞行性能有重要影响。研究了基于PSO-BP神经网络的平流层区域风场建模与快速预测方法，根据历史风场数据，采用主成分分析法对数据进行降维处理，通过BP神经网络对风场进行预测建模，利用粒子群优化（PSO）算法对其进行优化，采用Biharmonic样条曲面插值方法构建区域预测风场。以南海地区5年历史风场为对象，对比分析了基于BP神经网络和基于PSO-BP神经网络的风场预测模型，结果表明:使用具有全局寻优特性的PSO算法改进BP神经网络，能够有效避免传统BP神经网络易陷入局部最优的缺点，提高预测精度；通过结合PSO-BP神经网络预测与Biharmonic样条曲面插值，可实现区域风场的预测。研究结果可为临近空间低速飞行器的轨迹规划与区域驻留等任务的高精度区域快速预报风场提供解决途径。 

Short-term rapid prediction of stratospheric wind field based on PSO-BP neural network

The stratospheric wind field environment has an important influence on the flight performance of near space low speed aircraft. In this paper, the modeling and prediction methods of stratospheric regional wind field are investigated based on PSO-BP neural network. Firstly, the principal component analysis method is implemented to reduce the dimensions of the historical wind field data. Then, the BP neural network, which is trained by the processed data to predict the wind field, is optimized with particle swarm optimization (PSO) algorithm. Finally, the Biharmonic spline surface interpolation method using multi-point prediction wind fields is studied to construct the regional prediction wind field. Taking the 5-year historical wind field data of a certain place, comparative study of the wind field prediction model based on BP neural network and PSO-BP neural network is conducted. The results show that PSO algorithm, which is characterized global optimization, can improve BP neural network by avoiding the disadvantage of easily falling into local optimization, and enhance the prediction accuracy. The integration method of PSO-BP neural network prediction and Biharmonic spline surface interpolation can provide the prediction of the regional wind field. The proposed method can provide high precision regional prediction wind field for trajectory planning and station keeping of near space low speed aircraft.