研究了一类基于动态神经网络的未知非线性多变量系统的鲁棒辨识问题．用Ｌｙａ－ｐｕｎｏｖ稳定性理论获得了具有保护策略的鲁棒调权律．从理论上证明了被辨识的系统是鲁棒稳定的，辨识误差按建模误差和未建模动态收敛到一个稳定区域．该策略的特点是不需要离线学习又不需要对象的状态完全可测．仿真结果验证了提出的动态网鲁棒辨识策略的有效性 A class of robust identification problem based on dynamic neural network for unknown nonlinear multivariable systems is studied. Robust transfer law with protection strategy is obtained by Lya-Punov stability theory. It is proved theoretically that the identified system is robust and stable, and the identification error converges to a stable region dynamically according to modeling errors and unmodeled dynamics. The strategy is characterized by the fact that it does not require off-line learning and does not require the state of the object to be completely measurable. The simulation results verify the validity of the proposed dynamic network robust identification strategy