質子交換膜燃料電池之自動學習溫度控制的方法

Abstract

在環保與石油即將枯竭的趨勢下，發展綠色能源已成為世界各國勢在必行的政策。綠色能源中的燃料電池是發電效率最好的，但燃料電池在運作時，需要注意燃料電池本身的狀況，否則運作的效能就會降低，而燃料電池的運作溫度是影響到燃料電池發電效率的指標之一。 然而，通常在實現燃料電池溫度控制前，必須花費許多人力與時間進行資料的收集與分析，接著再開發相關的控制方法進行實驗，之後一直不斷的實驗與調整參數，直到能將燃料電池的溫度控制在目標範圍內。由於上述的程序非常繁瑣，人力與時間的花費是非常高的，所以必須找到一個自動學習的方法，減少人力與時間的花費，也能夠達到燃料電池穩定控制溫度的功能。 本研究提出以智慧代理人結合自適應網路模糊推論系統來處理自動學習溫度控制的問題，來減少人力在實驗與分析數據時所產生的錯誤和花費的時間。而實驗結果顯示，系統透過自動學習後，燃料電池的溫度可以控制在目標範圍內。除此固定輸出的燃料電池溫度控制研究外，本研究的方法也可以適用在變動輸出的燃料電池溫度控制研究上。

Under the trend of environmental protection and the exhausted of the petroleum, developing green energy has become one policy that countries in the world must carry out. Among green energies, fuel cell has the best generating efficiency; however, the cell situation needs to be paid attention when operating it or the efficiency will be decreased. The operating temperature is one of the indications that influence the generating efficiency. However, normally before controlling the temperature of the fuel cell, the data collecting and analyzing is labor and time consuming. Later, it needs to develop experiments on related control method and keep doing experiments and adjusting parameters until the temperature of the fuel cell is within the target. Due to that the above mentioned procedure is extremely complicated, the cost of labor and time is high; therefore, we need to find an automatic learning method to reduce the cost of labor and time to reach the function of stabilizing the temperature of fuel cell. This research proposes to combine intelligent agent and ANFIS to process the automatic learning temperature control to decrease the error and time in experiments and data analyze. The results show that the temperature of fuel cell can be controlled within targets through automatic learning. Besides exporting the research of the temperature control of fuel cell, this research can be used on the different loadings of fuel cell.