Application of Artificial Neural Networks Coupled with Sequential Pseudo-Uniform Design to Optimization of Membrane Reactors for Hydrogen Production

應用類神經網路與接續式擬均勻設計進行氫氣薄膜反應器最適化生產

張志雄(Jyh-Shyong Chang) ； 楊紹民(Shao-Min Yang)

《Journal of the Chinese Institute of Chemical Engineers》 37卷4期 (2006/07) Pp. 395-400

https://doi.org/10.6967/JCICE.200607.0395

摘要

搭配燃料電池組的重組反應器需要質量輕體積小的元件。薄膜反應器因提供可同時進行甲醇重組反應的管狀反應器與供氫氣滲透分離的薄膜並具備質量輕體積小的優點而常被採用。以緻密堆積金屬製成的薄膜可在大範圍的壓力與進料負荷下均可分離出燃料電池所需純度的氫氣。在薄膜反應器實際應用上，若能有效地決定其最適操作條件而不仰賴耗時才能建構的物理模式將是非常值得開發的操作策略。在本研究中我們首先建立供甲醇水蒸氣重組之鈀膜反應器之模擬器以供測試驗證我們所提的最適化方法的可用性。我們採用接續式擬均勻設計來安排實驗取得少量但充足的數據並以類神經網路來建構鈀膜反應器之代表性模式。以此鑑別之代表性模式可以正確地提供系統之最適操作條件。

並列摘要

Fuel cells with on board reforming require compact and lightweight components. A membrane reactor (MR) that combines hydrogen permeable membranes with a methanol steam reformer promises considerable weight and space savings. Its dense metal membranes produce high purity hydrogen over a wide range of pressure and load. For a real application of MR, there is much incentive to determine optimal operating conditions of a membrane reactor without resorting to the time consuming knowledge-based modeling work. In this work, a Pd membrane reactor (PMR) for carrying out the methanol steam reforming was simulated and adopted as the test process for verification of the applicability of the proposed optimization method. The artificial neural networks (ANN) with back propagation algorithms coupled with the sequential pseudo-uniform design (SPUD) was applied and demonstrated successfully to the modeling of the PMR system using limited but adequate experimental data. The optimum operating conditions determined from the identified ANN model were applied precisely.

並列關鍵字

Steam reforming ； Membrane reactor ； Neural networks ； Sequential pseudo-uniform design