Please use this identifier to cite or link to this item:
Adaptive Neural Predictive Control with Applications toa Class of Industrial Processes
|關鍵字:||predictive control;預估控制;recurrent neural network;Neural-based Linearized Identifier;迴授式類神經網路;類神經辨識器||出版社:||電機工程學系||摘要:||
本論文旨的在發展迴授類神經網路(Recurrent Neural networks)為基礎的適應預測控制方法及其工業程序之應用技術。為了準確的估測工業程序系統的系統參數與輸出預測值，本文提出類神經線性化驗證器(Neural-based Linearized Identifier)及長時間類神經預測器(Long-Range Predictor)。為達到良好的控制成效，本文提出二種適應預測控制策略：類神經線性化模型參考預估控制 (Neural-based Linearized Model Predictive Control)及適應多步類神經預估控制 (Adaptive Multi-step Neural Predictive Control)。經由電腦模擬和變頻油溫冷卻實驗結果證明本論文所提之的控制策略，皆能有效地在工業程序上得到優異之控制性能。
This thesis develops methodologies and techniques for predictive control of industrial processes using recurrent neural networks. The neural-based linearized identifier and long-range predictor are presented in order to estimate the parameters of industrial processes and to obtain output predictions. To achieve satisfactory control performance, two adaptive predictive control strategies, neural-based linearized predictive control and adaptive multi-step neural predictive control are proposed. Numerous simulation results are provided to show the effectiveness and feasibility of proposed control methods. Experimental results for a variable-frequency oil-cooling process have been conducted to show that the presented control approaches would perform well for the industrial processes with nonlinear, time-delay and nonminimum-phase properties.
|Appears in Collections:||電機工程學系所|
Show full item record
TAIR Related Article
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.