改良式人工蜂群演算法應用於雙動力混合動力系統之最佳化能量管理

Abstract

本研究為加強人工蜂群演算法之運算效率，對其進行改良並開發改良式人工蜂群演算法 (Improved Artificial Bee Colony，IABC) 應用於雙動力混合動力系統的能量管理策略。除在電腦環境進行模擬測試外，更應用硬體嵌入式系統 (Hardware-In-The-Loop, HIL) 進行即時 (Real-Time) 驗證演算法的可行性。於本研究之系統架構中，車輛之重量為1368 kg，其子系統包含雙動力元件43 kW內燃引擎及15 kW驅動馬達及1.872 kWh儲能鋰電池。於能量管理系統中，本研究首先針對 ABC 之初始化參數進行訓練，結果顯示蜜蜂數10隻、疊代次數60次及最大開採次數20次為本研究最有效率的初始化參數。經參數訓練後，本研究更結合細菌覓食演算法 (Bacterial Foraging Algorithm, BFA) 之搜尋邏輯於ABC中，延伸設計出IABC，並將ABC與IABC兩者之搜尋結果與計算時間進行比較，驗證IABC有更好的運算效率。最終經由IABC之 (1) 工蜂、(2) 觀察蜂及 (3) 偵查蜂三階段的搜尋任務尋找最佳動力輸出扭力進行控制。本研究將開發之IABC與另外兩種控制策略進行能耗比較：(1) 基本規則庫控制策略 (Rule-base Control), 依照元件效率之工程經驗撰寫模式切換之規則，可分為四種模式 (充電模式、純電動模式、純引擎模式、混合動力模式) ；(2) 最小等效油耗策略 (Equivalent Consumption Minimization Strategy, ECMS) 係利用窮舉法搜尋邊界條件內之所有可能解，找出最小油耗時之動力輸出結果。為驗證ECMS之變數步階幅度(Step size)與計算結果之關係，本研究亦規劃相關實驗證實。最後，本研究透過 HIL 模擬 IABC 於車輛控制單元 (Vehicle Control Unit, VCU) Real-time 之可行性與油耗效益驗證。在執行1次NEDC行車形態下，基本規則庫、ECMS 、 IABC 的等效燃油消耗量為 [352.8 g , 315.7 g , 313 g]，ECMS 與 IABC 比較基本規則庫控制法之能耗改善率為 [10.5 % , 11.2 %]。相同以IABC控制策略下，電腦模擬與即時模擬(HIL)之相似度為99.6 %；再者是執行5次NEDC行車型態，三者等效燃油消耗量為 [1887 g , 1736 g , 1752 g]，改善率為 [8% , 7.1%]。相同以IABC控制策略下，電腦模擬與即時模擬 (HIL) 之相似度為96 %。綜合上述可知，IABC之能耗改善率與ECMS極度相似，且皆有顯著改善。於HIL測試結果與電腦模擬結果相似，未來將會實施於真實之雙動力源混合動力車輛。In order to enhance ABC (Artificial Bee Conly) computational efficiency, this study was applied to dual-power hybrid vehicle energy management strategy with improving ABC. In addition to simulate in computer environment, it was applied to the HIL (Hardware-In-The-Loop) system to verify the feasibility of algorithms in real-time condition. The 1368 kg dual-power hybrid vehicle configuration includes a 43kW combustion engine, a 15 kW traction motor and a 1.872 kW lithium battery was employed. In energy management strategy, IABC initial parameters was set and defined as 10 bees (SN), 60 maximum cycle number (MCN) and 20 maximum number of repeated collecting (Limit). After the tuning process, BFA (Bacterial Foraging Algorithm) searching method was combined with the ABC in this study and designed an improved method which is called IABC. By comparing the searching result and convergence time of ABC and IABC, IABC has the better calculation efficiency with similar fuel consumption. After completing three steps including 1) employed bee phase, 2) onlooker bee phase and 3) scout bee phase, IABC was designed in energy management strategy. It recorded the best solution and output the distributed torque to control elements.IABC and two control strategy were used to carry out a comparison of equivalent fuel consumption: 1) the rule-base control, according to the elements effectiveness and the engineering experience, the fore modes were set as charging mode, EV mode, engine mode and hybrid mode. 2)the ECMS (Equivalent Consumption Minimization Strategy) was used for the exhausted search method to search all the possible solutions within the boundary conditions and find the output torque result with the minimum fuel consumption. Finally, the HIL system was used to verify the feasibility of IABC in the vehicle-control-unit (VCU).For 1 time NEDC, the equivalent fuel consumption of rule-base, ECMS and IABC are [352.8 g, 315.7 g, 313 g], and the improvement rate of ECMS and IABC compared with rule-base are [10.5%, 11.2%]. For 5-times NEDC, their three equivalent fuel consumption are [1887 g, 1736 g, 1752 g], the improvement rates are [8.0%, 7.1%]. Therefore, the improvement rate of IABC is extremely similar to that of ECMS and it achieves significant improvement. It will be applied to actual dual-power hybrid vehicle in the future.