本研究採用智慧型穩健參數設計方法進行質子交換膜燃料電池堆多品質性能分析，其內容包括：實驗計劃法進行篩選實驗、田口法(Taguchi Method)多品質分析與離散模型預測、主成份分析法(Principal components analysis，PCA)進行統計上的多目標決策，以及倒傳遞神經網路結合蒙地卡羅法之智慧型參數設計；然後，利用品質損失減少百分率(PRQL)判定最佳參數水準組合，其中探討的操作參數計有電池溫度、陰極與陽極加濕溫度及反應氣體計量比，以達供氣最少，同時輸出電功率最大之多目標性能，最後，由智慧型參數設計、主成份分析法與田口法分別比較電功率最大參數水準組合提升平均品質損失減少百分率(APRQL)為32.35%、31.61%、32.03%。

This paper presents an approach of the intelligent-robust parameter design to improve the performance of a PEM fuel cell stack with multi-objective cases. Firstly, a screen experiment has to be carried out by using a fractional factorial design; then the Taguchi multi-quality method can be conducted to predict the discrete model; the principal component analysis (PCA) can then be performed on the multi-objectives. The intelligent parameter design is developed via the definition of the percentage reduction of quality loss (PRQL) combined with the S/N ratio models that can be performed by a Backpropagation Neural Network (BPNN), in order to supply a fitness function to the Monte Carlo method. Finally, the prediction model created by this approach can be verified through a confirmation experiment. In this work, a combined approach is employed to determine the optimal combination of five operating parameters that include the temperature of a fuel cell, the anode and cathode humidification temperatures, the stoichiometric flow ratios of the reaction gas etc. for a PEM fuel cell stack. The results indicate that the intelligent parameter design via the average PRQL was improved by 32.35%. However, the Taguchi method and the PCA were improved by 32.03% and 31.61%, respectively.