对于在线辨识非线性系统，传统的小波网络缺乏有效的算法。本文在正交小波网络的基础上，提出一种利于实时应用的可变正交小波网络，给出了其结构变化和参数调整的策略，并对建模误差的变化进行了分析。最后用于辨识非线性动态系统，仿真结果验证了该方法的可行性和有效性。 For the online identification of nonlinear systems, the traditional wavelet network lacks an effective algorithm. Based on Orthogonal Wavelet Network, this paper proposes a variable orthogonal wavelet network that is beneficial to real-time applications. The strategy of structural change and parameter adjustment is given, and the changes of modeling errors are analyzed. Finally, it is used to identify nonlinear dynamic system. The simulation results verify the feasibility and effectiveness of the proposed method.