中文摘要：

采用变增益策略，用输入与稳态输出的映射表示系统的静态非线性，用一个增益为1的ARX模型表示系统的动态模型，代替多教文献中常用的分段线性多模型方法进行非线性系统的预测控制。文中通过对连续搅拌釜反应器（CSTR）的仿真，验证了本算法的有效性。

英文摘要：

With variable gain strategy, Nonlinear Model Predictive Control （NMPC） can be implemented by one offline identified model, consisting of a nonlinear static mapping form control input to steady-state output and an Auto Regressive with eXternal input （ARX） model, whose steady-state gain is constrained to be 1, representing the dynamic characteristic of the system, rather than linear multi-model. The control of Continuously Stirred Tank Reactor （CSTR） is studied and the simulation result demonstrates the effectiveness of the proposed method. The developed approach possesses great potential for the control of a general class of nonlinear systems whose steady-state gain can be obtained.