Design of Fuzzy Decision Adaptive Optimal Controller for Multi-objective Optimization Problems

多目標優化考慮系統有兩個或兩個以上優化目標的問題，傳統求解多目標優化問題大都採用權重法或基因演算法，但是權重法無法處理不同偏好的情況，基因演算法很難應用於即時優化控制，本論文採用模糊決策適應型最佳控制法求解多目標優化問題，適應型最佳控制法用成本函數表示優化的目標，再根據極小值原理和強化學習架構順向求解最佳控制訊號。適應型最佳控制法採用成本函數引導控制器的優化方向，多目標優化問題可以用許多子目標組成成本函數，子目標的相對權重表示其重要程度，相對權重會影響優化的結果，但是固定的相對權重不能適應環境改變或決策因子改變的情況。本論文採用模糊邏輯調整子目標的相對權重，可以針對環境改變或決策因子改變自動調整相對權重，然後針對多目標最佳化問題建立模糊決策適應型最佳控制法的求解步驟，並以混合能源燃料電池電動車的燃料利用最佳化和儲電系統的蓄電狀態監控做驗證，子目標的相對權重隨著車速和儲電系統的蓄電狀態自動調整，使儲電系統在不同車速的情況下都能夠提供加速所需的電功率或減速時提供吸收再生電功率所需的儲能空間，使混合能源電動車達成高效率的操作。Multi-objective optimization concerns about optimization problems involving more than one objective function to be optimized simultaneously. Conventionally, multi-objective optimization problems are solved by the weighted-sum approach or the genetic algorithm. However, the weighted-sum approach is not able to deal with situations concerning different preference, and the genetic algorithm is hard to use in real-time optimization and control. This thesis proposes the fuzzy decision adaptive optimal control algorithm to solve the multi-objective optimization problem. The objective of optimization is defined to minimize a cost function which is the fuzzy association of many individual cost functions. The fuzzy association can adapt the weight of each individual cost function to changes in the environmental condition or decision making factor. Then, optimization is achieved by implementing the adaptive optimal control algorithm to minimize the cost-to-go until getting a convergent result. The proposed fuzzy decision adaptive optimal control algorithm is verified in the energy management of a fuel-cell hybrid vehicle. It is shown that, by introducing the fuzzy association, the cost function for energy management can adapt to the vehicle speed and the state of charge (SoC) of the energy storage system (ESS). Thus, the optimal energy management strategy will maintain a higher SoC at a lower speed that prepares the ESS to supply power for acceleration. Conversely, the optimal energy management strategy will maintain a lower SoC at a higher speed that prepares the ESS to retrieve regenerated power