本篇論文提出兩種具有適應性的外力干擾降低架構，而此兩架構可降低未知的低頻外力干擾對線性微小延遲系統的影響。第一種架構是利用類神經演算法的概念設計而成。第二種架構中包含輸入干擾降低控制器與殘餘干擾降低控制器。輸入干擾降低控制器可以降低未知低頻外力干擾與系統不確定項的影響。因為輸入干擾無法完全經由輸入干擾降低控制器消除，所以殘餘的外力干擾則經由殘餘干擾降低控制器處理。不同於常見的干擾消除方法，提出的兩個架構在處理外力干擾時，不需要估測得到任何外力干擾的資訊，例如，外力干擾的頻率。此外，本篇論文提出的兩種外力干擾降低架構可與改良型Smith估測器分別作結合，以得到更佳的性能於微小延遲系統的控制上。而且，提出的控制架構可在微小延遲系統中降低週期或非週期的低頻外力干擾。 This dissertation proposes two adaptive disturbance reduction schemes for linear small delay systems with unknown low-frequency load disturbances. One of the proposed disturbance reduction schemes is based on an artificial neural network (ANN). The artificial-neural-network disturbance reduction controller (ANNDRC) is proposed for small delay systems with unknown low-frequency load disturbances. Another proposed scheme contains an input disturbance reduction controller (IDRC) and a residual disturbance reduction controller (RDRC). The IDRC using the ANN is used to reduce the unknown low-frequency load disturbances and modeling uncertainties. Residual disturbances and residual uncertainties are reduced by the RDRC based on a disturbance observer. Unlike other methods, both of the proposed schemes do not require disturbance frequencies to be known. The proposed schemes are respectively applied to a modified Smith predictor for the control of small delay systems. Simulation examples are illustrated to show the effectiveness of the proposed disturbance reduction schemes for linear small delay uncertain systems with periodic or non-periodic unknown low-frequency load disturbances.
