基于系统状态空间模型的系统辨识方法的一个困难在于算法具有较大的运算量和存储量.该文将基于APEX算法的子空间跟踪方法引入辨识算法.APEX算法的神经网络实现可以有效地减少辨识算法的运算量和存储量.该文将得到的算法应用于时变系统的辨识,仿真结果和应用于实际数据的结果验证了方法的有效性.“,”A dificulty in state spase model system identification approaches is the high computation and storage burden. In this paper, a subspace tracking approach based on APEX algorithm is applied in state space model identification. The Neural network model of APEX algorithm can effectively reduce the computation cost andmake the state spce model identification algorithm available for real time application. We apply the state spacemodel identification algorithm obtained in time varying system identification.This approach is demonstrated in thesimulation and application in real data.