SF与RLS算法结合的时变系统参数辨识

Parameter Identification of Time-Varying System Based on RLS Algorithm Combined with Selective Forgetting

遗忘因子最小二乘算法(RLS)具有对时变系统参数在线估计的能力,而传统的遗忘策略对解决参数辨识矩阵过饱和问题具有一定局限性。为了拓展现有RLS算法在时变系统的适用范围,提出一种将选择遗忘机制(SF)与RLS算法结合的时变系统辨识算法。从而构造出一种基于参数矩阵特征值映射的有界函数,特征值映射函数能够根据系统数据传递过程中信息量的大小动态调整遗忘因子,解决了参数辨识过程中数据分布不均匀问题。仿真结果表明,相比传统的RLS算法,带有选择遗忘机制的RLS算法能够更加准确的跟踪系统参数的变化,同时保证系统不是2N阶持续激励信号的情况下,也能够对时变系统参数进行跟踪。

The recursive least squares algorithm(RLS)with traditional forgetting factor can estimate the parameter of time-varying system on-line,while the traditional forgetting strategy has some linitations in solving the super-saturation problem of parameter estimation matrix.In order to expand the usage of existing RLS algorithm,a modified RLS algorithm with selective forgetting was proposed.A bounded function based on the parameter matrix eigenvalue mapping was constructed.The eigenvalue mapping function can dynamically adjust the forgetting factor according to the amount of information in the system data transmission process,which can solve the problem of uneven distribution of data.The simulation results show that compared with the original algorithm,the RLS algorithm with selective forgetting can track the changes of system parameters more accurately,and is able to ensure the tracking ability for time-varying system with a non 2 n-order continuous excitation input signal.