利用動態粒子群演算法於三動力源複合動力系統之最佳能量管理
No Thumbnail Available
Date
2016
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
本研究採用動態粒子群演算法(Dynamic Particle Swarm Optimization, DPSO),用於三動力源複合動力車的能量管理系統。該車輛模型建製七個主要部分:內燃引擎、驅動馬達、起動發電機、鋰電池、能量管理系統、駕駛模式和變速箱。在能量管理系統,DPSO有五個演算步驟:(1)初始化、(2)確定適應度函數、(3)更新學習因子與慣性權重、(4)位置和速度更新、(5)輸出雙動力分配比。適應度函數粒子考慮運算效率,本研究設定四個。適應度函數於本研究為引擎、馬達、發電機之等效油耗,及懲罰值限制動力源之操作點於物理限制範圍內。可建製出三輸入輪胎轉速、電池電量、需求功率與雙輸出引擎與馬達動力分配比之DPSO能量管理系統並與整車模型作聯結。為了解本即時模擬之效益,本文與三種控制管理策略進行比較:(1)規則庫管理(Rule base):有五種控制模式(系統就緒、充電模式、電動模式、複合動力模式及煞車回充模式),根據內燃引擎的效率和馬達定轉速下找出最大扭矩;(2)最小等效油耗策略(ECMS):搭配全域搜尋(GSA)將所有的可能解都尋找,找出最小油耗時之動力源扭矩;(3)粒子群演算法(PSO):與DPSO相似,但DPSO會依據不同的型車狀態去改變學習值及慣性權重。模擬結果發現,在NEDC行車形態下, DPSO的效率比傳統的PSO改善率高0.09%,DPSO與規則庫控制相比,在等效油耗部分改善35.36%,而能量消耗改善為44.08%次於ECMS的最佳解 (改善37.3%等效油耗與46.95%能耗)。未來將本研究實際應用於三動力車輛並於動力計上測試實際能耗效果。
This study used dynamic particle swarm optimization (DPSO) to manage the energy system of engine/motor/generator hybrid electric vehicles. The vehicle featured seven major segments, including an internal combustion engine, a motor, a starter generator, a battery, an energy management system, a driver model and a gearbox. To manage the power distribution of triple power sources, the DPSO was equipped with five steps: (1) initialization; (2) tenacity of the fitness functions; (3) update on the learning factor and inertia weight; (4) modification of position and velocity; and (5) output of dual power distribution ratio. Considering the efficiency of fitness functions particle calculation, four fitness functions were included. They were the gasoline equivalent of engine, motor, generator, and the operating point of penalty value for limiting power sources under reasonable physical limitation. With these functions, we can build triple input (rotation speed of tire, power of battery and requiring power) and dual output (distribution ratio of engine and motor) from the energy management system of DPSO which can also connect the vehicle model. In order to figure out the efficiency of real time simulation, this study was compared with three different controlling management strategies: (1) rule base: Including five controlling modes (system standby, charging mode, electrical mode, hybrid power mode and regenerative breaking mode). According to efficiency of engine and the rotation speed of motor to find out the highest torque; (2) Equivalent Consumption Minimization Strategy (ECMS): figure out the power sources torque under minimal consumption with global searching algorithms (GSA); (3) Particle Swarm Optimization (PSO): which is similar to DPSO, but DPSO will change learning value and inertia weight according to different driving cycle. The results show that the efficiency of DPSO was 0.09% higher than traditional PSO under NEDC driving cycle. Comparing DPSO and rule base, the equivalent consumption was improved by 35.36%, and energy consumption was also improved 44.08%, which was better than the best solution of ECMS (improved 37.3% of equivalent consumption and 46.95% of energy consumption). The energy consumption in this study would be applied and tested on the triple power vehicle and dynamometer in the future.
This study used dynamic particle swarm optimization (DPSO) to manage the energy system of engine/motor/generator hybrid electric vehicles. The vehicle featured seven major segments, including an internal combustion engine, a motor, a starter generator, a battery, an energy management system, a driver model and a gearbox. To manage the power distribution of triple power sources, the DPSO was equipped with five steps: (1) initialization; (2) tenacity of the fitness functions; (3) update on the learning factor and inertia weight; (4) modification of position and velocity; and (5) output of dual power distribution ratio. Considering the efficiency of fitness functions particle calculation, four fitness functions were included. They were the gasoline equivalent of engine, motor, generator, and the operating point of penalty value for limiting power sources under reasonable physical limitation. With these functions, we can build triple input (rotation speed of tire, power of battery and requiring power) and dual output (distribution ratio of engine and motor) from the energy management system of DPSO which can also connect the vehicle model. In order to figure out the efficiency of real time simulation, this study was compared with three different controlling management strategies: (1) rule base: Including five controlling modes (system standby, charging mode, electrical mode, hybrid power mode and regenerative breaking mode). According to efficiency of engine and the rotation speed of motor to find out the highest torque; (2) Equivalent Consumption Minimization Strategy (ECMS): figure out the power sources torque under minimal consumption with global searching algorithms (GSA); (3) Particle Swarm Optimization (PSO): which is similar to DPSO, but DPSO will change learning value and inertia weight according to different driving cycle. The results show that the efficiency of DPSO was 0.09% higher than traditional PSO under NEDC driving cycle. Comparing DPSO and rule base, the equivalent consumption was improved by 35.36%, and energy consumption was also improved 44.08%, which was better than the best solution of ECMS (improved 37.3% of equivalent consumption and 46.95% of energy consumption). The energy consumption in this study would be applied and tested on the triple power vehicle and dynamometer in the future.
Description
Keywords
動態粒子群算法, 規則庫管理, 最小等效油耗策略, 混合動力, Dynamic Particle Swarm Optimization, rule base, Equivalent Consumption Minimization Strategy, Hybrid power