基于自适应Kalman滤波的改进PSO算法  

Improved PSO Algorithm Based on Adaptive Kalman Filter

作　　者：戴邵武[1] 王克红[2] 戴洪德[1] 

机构地区：[1]海军航空工程学院控制工程系,山东烟台264001 [2]海军航空工程学院研究生管理大队,山东烟台264001

出　　处：《海军航空工程学院学报》2015年第1期1-6,共6页Journal of Naval Aeronautical and Astronautical University

基　　金：航空科学基金资助项目(20135184007);中国博士后科学基金资助项目(2013m532173)

摘　　要：针对基于Kalman滤波的PSO算法在设计与应用过程中存在的不足,提出了基于自适应Kalman滤波的改进PSO算法。利用粒子群状态空间Markov链模型,建立粒子群系统状态方程;采用粒子的速度和位置作为观测量,构建观测方程;引入记忆衰减因子动态调整Kalman滤波模型参数及噪声方差阵,降低模型误差,提高粒子的位置估计精度。仿真实验表明:改进的PSO算法无论在优化精度、收敛速度,还是在稳定性方面都有很大的改进和提高,这就有效避免了粒子的"早熟"收敛问题;尤其在处理复杂多峰问题上,改进算法表现出很明显的优越性。For the shortcomings which existed in the designing and application process of Particle Swarm Optimization （PSO） algorithm based on Kalman filter, in this paper, an improved PSO algorithm based on adaptive Kalman filter was proposed. Using the particle swarm state space Markov chain model, a state equation of particle swarm system was estab- lished. Then taking the speed and position of particle swarm as observation, the observation equation was constructed. Memory fading factor was introduced into the Kalman filter to dynamically adjust model parameters and noise covariance matrix, reduce the model errors and improve the estimation accuracy of the position of the particle. Simulations showed that improved PSO algorithm had great improvement in both optimization accuracy, convergence speed and stability and it could effectively avoid the premature convergence problem of particle. Especially in dealing with the complex multi-peak problems, the improved algorithm showed prominent superiority.

关 键 词：粒子群优化 马尔科夫链 自适应卡尔曼滤波 基准函数 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]