基于GA-WPA优化的BP神经网络目标威胁估计  

Targets Threat Assessment Using BP Neural Network Optimized by GA—WPA Algorithm

作　　者：李槟槟 何广军[1] 尤晓亮 田德伟[1] 王俊[1] 

机构地区：[1]空军工程大学防空反导学院,西安710051

出　　处：《计算机测量与控制》2015年第12期4187-4190,共4页Computer Measurement &Control

摘　　要：在防空作战中,目标威胁估计是指挥控制过程的重要一环,是决策和指挥的重要依据;BP神经网络能够解决目标威胁估计问题,但存在收敛速度慢、易陷入局部最优等缺点;提出将遗传算法(genetic algorithm,GA)的选择、交叉和变异操作融入到狼群算法(wolf pack algorithm,WPA)中,提出了GA-WPA算法,以提高狼群算法的收敛速度;在此基础上,利用所提出的GA-WPA算法对BP神经网络进行优化,确定最优初始权值和阈值;最后,将优化后的BP神经网络解决地面防空系统目标威胁估计问题;仿真实验表明,所提算法能够有效克服BP神经网络收敛速度慢、易陷入局部最优等缺点,能够提高目标威胁估计的准确性和适应性。In air defense operation, Targets threat assessment is one of the most important links in the command and control process. Meanwhile it is an important evidence of decision--making and command. The back propagation （BP） neural network can solve the problem of targets threat assessment. It has the defects of slow convergence rate and local optimum. Based on the basic operations of selection, cross and aberrance in the genetic algorithm （GA）, the wolf pack algorithm （WPA） is modified to improve its convergence speed and calculation accuracy. Then, the proposed GA--WPA algorithm is utilized to train the weight and threshold value of the back propagation （BP） neural network. At last, the optimized BP neural network is applied to assess the threat of targets. Simulation results demonstrate that our method could overcome the drawbacks of BP neural network effectively, and the accuracy and steadiness of targets threat prediction are obviously im- proved.

关 键 词：目标威胁估计 遗传算法 狼群算法 BP神经网络 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]