鯨魚演算法應用於三動力複合動力系統之最佳化能量管理

Optimal Energy Management Strategy Using Whale Optimization Algorithm for a Three-Power-Source Hybrid Powertrain

江明謙(Jiang Ming-Qian)

指導教授 ： 洪翊軒

國立臺灣師範大學/科技與工程學院/工業教育學系/碩士(2021年)

https://doi.org/10.6345/NTNU202101481

摘要

本研究旨於開發鯨魚演算法 (Whale Optimization Algorithm, WOA) 應用於三動力複合動力車系統之最佳化能量管理，並透過硬體嵌入式系統 (Hardware-In-the-Loop, HIL) 進行即時 (Real-time) 運算，驗證開發之能量管理系統於真實環境應用可行性。三動力源複合動力車之系統搭載43 kW的內燃機引擎、30 kW的馬達與15 kW的一體式啟動發電機 (Integrated Starter Generator, ISG)，搭配1.872 kW-h儲能鋰電池，整車重量為1,368 kg。於能量管理系統中，WOA透過三種行為模式進行最佳化搜索，分別為：(1) 探勘 (Exploration)、(2) 收縮環繞 (Shrinking Encircling)、(3) 螺旋更新 (Spiral Updating)，最大迭代次數為300次，共有80隻鯨魚進行最佳化能量管理。
本研究將開發之WOA與另外三種控制策略進行能耗比較：(1) 基本規則庫 (Rule-based)：依工程經驗與元件性能所撰寫模式切換之策略，共設計五種模式 (煞車回充、純電動、複合動力、純引擎、引擎回充)；(2) 最小等效油耗策略 (Equivalent Consumption Minimization Strategy, ECMS)：透過全域格點搜尋 (Global Grid Search) 各種行車條件的所有可行解，進而倒推最小等效油耗時之動力分配方式；(3) 人工蜂群演算法 (Artificial Bee Colony, ABC)：主要由三種蜜蜂角色分工進行最佳化搜尋，分別為：(i) 工蜂 (Employed bee)、(ii) 觀察蜂 (Onlooker Bee)、(iii) 偵查蜂 (Scout Bee)，即時運算當下行車需求之最佳動力分配方式。
各控制策略運行一次NEDC行車形態下，Rule-based、ECMS、ABC、WOA的等效燃油消耗量分別為[330.7g, 289.5g, 270.2g, 267.5g]；運行一次FTP-72行車形態下的等效燃油消耗量分別為[342.9g, 291.4g, 278.9g, 275.9g]。在一次NEDC中，以Rule-based為基底相比的能耗改善百分比是[12.458%, 18.294%, 19.110%]；在一次FTP-72中，能耗改善百分比是[15.018%, 18.664%, 19.539%]。Rule-based與WOA於10L燃油與起始電量90%的鋰電池耗盡下，於重複NEDC循環下的總行駛里程分別為[259.80 km, 276.66 km]；於重複FTP-72循環下的總行駛里程分別為[258.93 km, 282.56 km]。在重複NEDC循環下，以Rule-based為基底相比的里程改善百分比為6.489%；在重複FTP-72循環下，里程改善百分比為9.126%。由此可知，導入最佳化方法於複合動力車輛進行動力分配，可有效降低整車能耗，進而提高行駛里程。本研究透過兩台快速雛型控制器，建立一即時模擬平台。驗證由WOA為核心開發之能量管理系統於真實環境應用可行性，在兩種行車型態中，於電腦模擬與HIL環境運算之等效油耗結果有高達98%的相似度，藉此，將可實現未來於實車應用之願景。

關鍵字

複合動力系統 ； 能量管理控制策略 ； 鯨魚演算法 ； 人工蜂群演算法 ； 最小等校油耗策略

並列摘要

The purpose of this study is to develop the Whale Optimization Algorithm (WOA) for optimal energy management in a three-power-source hybrid powertrain. The real-time simulation and Hardware-In-the-Loop systems (HIL) were conducted to verify the feasibility of the developed energy management system (EMS) operates in real time. The subsystems of the vehicle include a 43kW internal combustion engine, a 30kW motor, 15kW integrated starter generator (ISG), and a 1.872kW-h lithium battery. The vehicle weight is 1,368 kilograms. About the EMS, WOA uses three behavior patterns to achieve the optimal search: (1) Exploration, (2) Shrinking Encircling, (3) Spiral Updating. The overall number of iterations was 300, and 80 whales were used to carry out optimal energy management.

This study will compare the energy consumption of the developed WOA with three other control strategies: (1) Rule-based：The strategy of mode switch designed by the engineering experience and component performance. There are five control modes which are brake regenerative, electric vehicle, hybrid power, engine only, and engine regeneration. (2) Equivalent Consumption Minimization Strategy (ECMS)：Use the grid to global search for feasible solutions under all driving conditions, and find the power distribution that minimizes the equivalent fuel consumption. (3) Artificial Bee Colony Algorithm (ABC)：Bees are divided into three roles to achieve optimal search: (1) Employed bee, (2) Onlooker bee, (3) Scout bee. The best power distribution for current driving demand was calculated in real time.
The equivalent fuel consumptions of Rule-based, ECMS, ABC and WOA were 330.7g, 289.5g, 270.2g and 267.5g, respectively, under one NEDC driving pattern. The equivalent fuel consumptions of each control strategy were 342.9g, 291.4g, 278.9g and 275.9g, respectively, under one FTP-72 driving pattern. The improvement percentage of energy consumption were 12.458%, 18.294% and 19.110%, respective, for ECMS, ABC and WOA compared to Rule-based control strategy under a NEDC driving pattern. The improvement percentage of energy consumption were 15.018%, 18.664% and 19.539%, respectively under a FTP-72 driving pattern. Before the 10 liters of fuel and initial SOC 0.9 of lithium battery run out, driving distances for the Rule-based and the WOA were 259.80km and 279.66km, respectively, under NEDC driving cycles repeatedly. And 258.93km and 282.56km were under FTP-72 driving cycles repeatedly. Under the NEDC, the improvement percentage of traveling mileage was 6.489% compared with Rule-based control, and 9.126% for FTP-72. From above, to import optimization methods to distribute power in hybrid vehicles which can reduce the energy consumption effectively, as well as improve the driving range. This study establishes a real-time simulation platform through two rapid prototype controllers. Under two driving pattern types, the similarity of equivalent fuel consumption results of computer simulation and HIL are as high as 98%. According to the experimental results, it proves that the control can be applied to actual vehicles in the future.

並列關鍵字

Hybrid Power System ； Energy Management Control Strategy ； Whale Optimization Algorithm ； Artificial Bee Colony ； Equivalent Consumption Minimization Strategy

參考文獻

[1] Bp plc, BP Statistical Review of World Energy, UK: British Petroleum, 2017.

[2] S. G. Osborn, et al. “Methane contamination of drinking water accompanying gas-well drilling and hydraulic fracturing,” in Proceedings of the National Academy of Sciences USA, vol. 108, no. 20, pp. 8172–8176, May 17, 2010.

[3] C. Mooney, “The truth about fracking,” Scientific American, pp.80-85, Nov, 2011.

[4] R. K. Pachauri, M. R. Allen, V. R. Barros, J. Broome, et al, “Climate change 2014 synthesis report,” Intergovernmental Panel on Climate Change, vol. 5, 2015.

[5] 李永展，永續發展: 大地反撲的省思，2003。