Power Matching Strategy Modeling and Simulation of PHEV Based on Multi‐agent Limin Niu*1, Lijun Ye2 School of Mechanical Engineering, Anhui University of Technology, Ma’anshan 243032, China *1 niulmdd@163.com; 2ylj860119@163.com Abstract Multi-agent and Its Application The power matching strategy of Parallel Hybrid Electric Vehicle based on Multi‐agent (MA) has been proposed aimed at the inadaptability of current strategy faces on complicated driving conditions. The models of engine agent, motor agent, generator agent can be modeled and so on. For resolved the problem of power matching in complicated driving conditions, the cooperation solving mechanism based on multi‐agent were applied. MA control strategy has been proposed by using software MATLAB/Simulink and Advisor. Simulation research was carried out in order to verify the validity of control strategy. Results show that MA system coordinate power matching demand of power train components effectively. It has a better adaptation on the requirement of complicated driving conditions. Vehicle fuel economy and power performance can be further improved. Keywords Hybrid Electric Vehicle (HEV); Multi‐agent; Power Matching Strategy; Modeling and Simulation Introduction Parallel Hybrid Electric Vehicle (PHEV) is the most widely used Hybrid Electric Vehicle. The fuel consumption and exhaust emissions of HEV is depends on the composition, the ratio of the power from fuel converter and motor and the control strategy (Titina Banjac et al, 2009). Automotive engineer have been trying to develop a power flow control strategy to reduce the fuel consumption and exhaust emissions according to the composition of HEV. Various power flow control strategies have been used in HEV (Qin Datong et al, 2014). Although the existing strategies can reduce fuel consumption and exhaust emissions, but they were developed based on specific driving cycle. So these strategies cannot fit the complex driving cycles for HEV (Montazeri‐Gh M et al, 2008 and Tian Yi et al, 2011). In view of the above question, this paper proposes a cooperative control framework and researches the power control strategy based on multi‐agent in order to improve the intelligence control of PHEV. Multi‐agent Agent was proposed by professor Minsky earliestly, which is used to description a hardware, a software or other entity with self‐adaption and autonomy in order to recognize and simulate human intelligent behavior (Minsky M, 1991). Multi‐agent System (MAS) is composed by muti‐agent, which can cooperate the behavior of a group of agents in order to get the solution cooperatively. MAS have characteristics of autonomy, distribution and coordination. Component agents model of HEV power strain can be structure by MATLAB/Simulink. Intelligence and coordination attributes of agent and MAS will be applied by simulation modeling. MAS Frame of Power Assembly Multi‐agent cooperative control frame is established as shown in Figure 1 based on layering principle. The frame is composed of three layer: system layer, cooperation layer and execution layer. FIGURE 1. MAS FRAME OF POWER ASSEMLY. System layer acts as a system agent role in the centralized architecture. It is responsible for the overall planning, scheduling and understanding the intention of driver. The cooperation layer is composed of power International Journal of Engineering Practical Research, Vol. 4 No. 1‐April 2015 5 2326‐5914/15/01 005‐07 © 2015 DEStech Publications, Inc. doi: 10.12783/ijepr.2015.0401.02 6 Limin Niu, Lijun Ye strain agents, such as fuel converter agent, motor agent, and so on. The local architecture is a distributed structure and all of the power component agents is at the same level. Every component agent can communicate with each other. The execution layer is responsible for the implementation of the superior instructions to complete the final action. PHEV Power Matching Strategy Based on MAS cooperation control frame, power matching strategy model of PHEV is built up in MATLAB/Simulink, as shown in Figure 2. This model consists of system agent and power component agents such as fuel converter agent(FC agent), motor agent, etc. Engine speed and torque, motor speed and torque, vehicle speed are detected in real‐time. System agent processes sensor signal and assigns power matching tasks. Component agents control respective actuator to complete power matching and driving task. System Agent System agent is used to identify the driving cycle and determine the matching scheme. Its model is shown as Figure 3. System agent judges current driving cycle and determines power matching scheme according to vehicle speed, acceleration and lookup table, then transmits signal to torque coupler. Driving cycle is divided into rural driving cycle, city driving cycle and high speed driving cycle. System agent identify the driving cycle based on average speed, average acceleration and standard deviation of acceleration. Big average acceleration means that driver focuses on dynamic performance and small average acceleration means that driver focuses on fuel economy. It may make mistakes only focusing on the FIGURE 2. PHEV POWER MATCHING STRATEGY MODEL. FIGURE 3. THE SYSTEM AGENT MODEL. Power Matching Strategy Modeling and Simulation of PHEV Based on Multi‐agent 7 FIGURE 4. DRIVING CYCLE MODEL OF SYSTEM AGENT. FIGURE 5. POWER MATCHING MODEL OF SYSTEM AGENT. average acceleration because the average acceleration cannot represent the degree of dispersion of acceleration (Wang Qingnian et al, 2012). Average and standard deviation of acceleration are entered into a fuzzy logic controller which as two input signals. Controller output signal is the numerical driving style which is dynamic performance or fuel economy. Driving style and average speed are entered into another fuzzy logic controller to determine the driving cycle after determining the driving style. Figure 4 shows the driving cycle model. from the system agent and control torque coupler changing power matching type in order to distribute power for engine and motor reasonably. Torque coupler agent model is shown as Figure 6. Power matching scheme is formulated when the driving cycle has been identified, according to the current vehicle speed and acceleration. Power matching scheme includes start model, idle model, driving model (excepting pure electric driving model), braking model and pure electric driving model. Power matching model is shown as Figure 5. Five power matching scheme decide how the power components work in each driving cycle. Torque Coupler Agent Torque coupler agent receives power matching scheme FIGURE 6. TORQUE COUPLER AGENT MODEL. According to engine, motor and gearbox output shaft working state, the vehicle working mode is divided into single motor working mode, single engine driving, 8 Limin Niu, Lijun Ye engine idling power mode and combined driving mode. Single motor working mode includes pure electric driving and braking power mode. Torque coupler agent decides the vehicle working mode according to the power matching scheme from the system agent. Then it according to the required rotation rate and gearbox torque, distributes rotation rate and torque for engine and motor/generator to complete the driving task. Fuel Convertor Agent Fuel convertor agent model is shown as Figure 7. It makes engine operating in high efficient region aimed at high fuel economy. Motor Agent process of motor agent can be described as follows. According to the rotation rate and torque from the torque coupler agent, motor agent compares the rotation rate received with its maximum rotation rate and takes the smaller value as the output rotation rate. Then it looks up the speed‐torque table to find the torque at the output rotation rate. Add the torque of coupler and the required torque of engine. Then Compares with the torque at output rotation rate, takes the smaller value as the output torque. Battery Agent Figure 9 shows the battery agent model. The main function of the battery agent is to keep the SOC in a certain range, that is SOClow≤ SOC≤ SOChigh, Thus, it is mainly composed of comparing element. Figure 8 shows the motor agent model. The working FIGURE 7. FUEL CONVERTOR AGENT MODEL. FIGURE 8. THE MOTOR AGENT MODEL. Power Matching Strategy Modeling and Simulation of PHEV Based on Multi‐agent 9 strategies. TABLE 2 THE PERFORMANCE COMPARISON TABLE OF TWO CONTROL STRATEGY Performance Dynamic performance Fuel economy Exhaust emissions Parameter MACS EACS 0‐96.6km/h acceleration time t(s) 12.4 9 Fuel consumption Q(L/100km) HC(g/km) CO(g/km) NOx(g/km) PM(g/km) 4.9 0.302 1.703 0.133 0 5.9 0.344 1.795 0.195 0 Figure 11 and Figure 12 show the speed following curves, SOC track, and exhaust emissions curves. FIGURE 9. THE BATTERY AGENT MODEL. Simulation and Results Analysis Power matching strategy was combined with software Advisor and the vehicle simulation model is built up, which is shown as Figure 10. TABLE 1 THE SIMULATION VEHICLE PARAMETERS Parameter Value Parameter Value Vehicle mass 1350kg Wheel radius 0.282m Front area 2.0m2 Rolling resistance coefficient 0.009 Air resistance coefficient 0.335 Maximum engine power 41KW Energy capacity 25Ah Maximum motor power 41KW The NEDC driving cycle is similitude with city driving cycle and the high speed driving cycle, so it was adopted as the driving cycle in the simulation usually. The simulation vehicle parameters refer to some vehicle’s parameters, which is shown as Table 1. The simulation results of the power matching strategy based on MA (MACS) is compared with that of the electric assist control strategy (EACS). Table 2 shows the dynamic performance, the fuel economy and the exhaust emissions performance of the two control Table 2 reflects that the fuel economy and the exhaust emissions of the MACS is better than EACS. The fuel consumption reduces 16.9%, HC reduces 12.2%, CO reduces 5.1%, and NOx reduces 31.8%. The dynamic performance of MACS is worse than EACS. The acceleration time of MACS in 0‐96.6 km/h is 12.4s but is less than 20s. This is a reasonable value according to the opinion of Timothy C. Moore (Timothy C. Moore et al, 1995). Figure 11 and Figure 12 indicate that SOC fluctuation of MACS is bigger than SOC fluctuation of EACS, however the SOC is still in the reasonable region. We also found that actual speed almost coincide with the required speed. This result shows that MACS can satisfy the required speed of vehicle. Figure 13 and Figure 14 show the efficiency of engine and motor of the two control strategy. Figure 13(a) and (b) show that engine operating points of MACS are less but more concentrated in efficient region than those of EACS in the first 800s. 800th second later, engine operating points of the two control strategy both concentrated in efficient region. Figure 14(a) and (b) show that the motor operating points of MACS are more and more concentrated in efficient region than those of EACS in the first 800s. After 800th second, there are little operating points. It means that motor was almost participate in operation. FIGURE 10. THE VEHICLE SIMULATION MODEL. 10 Limin Niu, Lijun Ye FIGURE 11. THE CURVES OF SPEED FOLLOWING, SOC AND EXHAUST EMISSIONS OF MACS. FIGURE 12. THE CURVES OF SPEED FOLLOWING, SOC AND EXHAUST EMISSIONS OF EACS. (b) ENGINE EFFICIENCY OF EACS FIGURE 13. ENGINE EFFICIENCY COMPARE. (a) MOTOR EFFICIENCY OF MACS (b) MOTOR EFFICIENCY OF EACS (a) ENGINE EFFICIENCY OF MACS FIGURE 14. MOTOR EFFICIENCY COMPARE. Power Matching Strategy Modeling and Simulation of PHEV Based on Multi‐agent 11 Conclusions (2014): 1550‐1555. 1) Fuel economy and exhaust emissions of MACS are better than those of EACS in NEDC driving cycle. Dynamic performance and SOC stability are worse than those of EACS. However, these values are still in reasonable region. The reasons is that MACS have the function of recognizing driving cycle. 2) Dynamic performance reduction but fuel economy and exhaust emissions increase of MACS are due to focusing on fuel economy and exhaust emissions in city driving cycle. It shows the intelligence of control strategy based on multi‐agent. Dynamic performance reduction indicates that current MACS should be further improved. Tian Yi, Zhang Xin, Zhang Liang, Zhang Xin. “Fuzzy control strategy for hybrid electric vehicle based on neural network identification of driving conditions.” Control Theory & Applications 28(3) (2011): 363‐369. Titina Banjac, Ferdinand Trenc, Tomaz Katrašnik . “Energy Conversion Efficiency of Hybrid Electric Heavy‐duty Vehicles Operating According to Diverse Drive Cycles.” Energy Conversion and Management 50 (2009) : 2865‐2878. Wang Qingnian, Tang Xianzhi, Wang Pengyu, Sun Lei. “Control Strategy of Hybrid Electric Vehicle Based on Driving Intention Identification.” Journal of Jilin University(Engineering and Technology Edition) 42(4) ACKNOWLEDGMENT The research was supported by the National Science Foundation of China under grant 51275002. (2012): 789‐795. Yu Zhisheng. Automobile Theory. Beijing: China Machine Press, 2009: 1‐2. REFERENCES Minsky M. “The Society of Mind.” Artificial Intelligence 48(3) (1991):335‐339. Momoh, Omonowo David, Omoigui, Michael Osaretin. “An Overview of Hybrid Electric Vehicle Technology.” Vehicle Power and Propulsion Conference, New York, USA: IEEE, 2009: 1286‐1292. Montazeri‐Gh M, Ahmadi A, Asadi M. “Driving Condition Recognition for Genetic‐fuzzy HEV Control.” Paper presented at the meeting for 3rd International Workshop on Genetic and Evolving Fuzzy Systems, Witten‐Bommerholz, 2008: 65‐70. Qin Datong, Peng Zhiyuan, Liu Yonggang. “Dynamic Energy Management Strategy of HEV Based on Driving Pattern Recognition.” China Mechanical Engineering 25(11) Limin Niu received the Ph.D. degree from the Jiangsu University, Zhenjiang, China, in 2008. And he worked as a postdoctoral in Nanjing University of Aeronautics and Astronautics, Nanjing, China, in 2010. His research interests are in the areas of optimal estimation and control of vehicle chassis, new energy vehicles, and intelligent control. Lijun Ye received the B.S. degree in metallurgical engineering from Anhui University of Technology, Maanshan, China, in 2009. He is currently working toward the M.S. degree in mechanical engineering as University of Technology. His current research interests include new energy vehicles, and intelligent control.