針對一些復雜的非線性系統在神經網絡逆動態控制方法下控制效果不理想的問題，本文提出對被控對象進行直接多步預測，利用多步預測性能指標函數對系統實現基于神經網絡的逆控制。并將仿真結果與利用遞推多步預測方法的結果進行了比較，得到直接多步預測控制較遞推多步預測更準確的結論。
關鍵詞多步預測性能指標函數神經網絡逆動態控制非線性系統
Abstract To against the problem of unsatisfied control effects are obtained in neural network reverse dynamic control method in complicated non-linear
systems，the author states the measure that multiple-step prediction is taken directly for the controlled object，and reverse control based on neural network
is implemented by using multi-step performance function to the system. In addition，by comparing results of simulation and recurrence multi-step prediction，
the conclusion of direct multi-step predicted control is more accurate than recurrence multi-step prediction is obtained.
Keywords Multi-step prediction Performance function Neural network Reverse dynamic control Non-linear system