Abstract:
Abstract: The control systems of hybrid electric vehicles (HEVs), which greatly affect their power performances, fuel economy and emissions are nonlinear, multi-variable, time-varying, and not differentiable, and they are the key objectives and the most difficult parts for HEVs. One of the most popular concerns about HEVs is the realization of real-time control. Researchers have come up with multiple architectures about the optimization of the energy management control of HEVs, and the most commonly used algorithms are dynamic programming (DP), equivalent consumption minimization strategy (ECMS), Pontryagin's minimum strategy (PMP), as well as some algorithms developed from them. In order to achieve the optimal solutions with DP, a given route or velocity profile is required, which is yet normally impossible. Therefore, some researchers come up with the method of model predictive control (MPC) to obtain the predicted driving cycle, so that the need of prior-knowledge can be met. As the combination of DP and ECMS is time-consuming, some researchers finally find ECMS and PMP can be alternatives to achieve global optimal solutions or approximate to them. However, these methods mentioned above cannot realize real-time control, even though many efforts have been done to develop more time-efficient algorithms, and the controlled objects were simplified in different degrees and some details were ignored, such as driver's intention, road conditions, signal lights and mutual effects among vehicles under real driving environment. With the development of intelligent traffic system (ITS), the real-time control based on vehicle connection becomes possible. In the connected vehicle environment, vehicles can communicate with each other and with traffic infrastructure, and they can also share their position and velocity information within the vehicle group with the help of the technologies of dedicated short range communication (DSRC), radio frequency identification devices (RFID), Bluetooth, Wi-Fi and cellular network. This paper focused on the hierarchical control of a group of 4 connected vehicles. The upper level controller integrated signal phase and timing (SPAT), multi-island genetic algorithm and nonlinear model predictive control (NMPC), which dealt with the generation of the primary velocity profiles, the best locations and target velocities, and the optimal target velocity sequences respectively. The lower level controller dealt with the energy management of HEV with adaptive equivalent consumption minimization strategy (A-ECMS), and obtained the optimal power splits of the engine and motor at a given time. One thing that should be addressed was that the higher level controller and lower level controller were both virtual controllers embedded in the distant server, which communicated with the driver and vehicle parts by wireless communications. Because of the super calculation ability of the server or cloud computing, the real-time control of HEVs could be achieved. In the upper level controller, the mathematic description of the problem was presented, which consisted of the vehicle longitude dynamics equation, the power request equation and the cost minimization function, and was a weighted sum of the fuel consumption, the velocity deviations between the vehicle and the one at its immediate back, the control variables of traction or braking force per unit mass, and the relative distance between the 2 vehicles. The aim of the higher level controller was to get the optimal velocity profile as well as the avoidance of red light stopping. Another method based on Gipp's car following model was also presented as a baseline method. With the best positions and target velocity sequences obtained from the optimal control problem, the optimal target velocity was calculated using NMPC over a given time horizon. The hierarchical control strategy was validated using hardware-in-the-loop and Clemson Palmetto server. The simulation results showed that the control method proposed in this paper could realize the energy management of the HEVs, avoid red light stopping, and achieve good velocity tracking as well as reduce fuel consumptions. Compared with the baseline method, the average fuel economy improvements for the A-ECMS, ECMS and rule-based controllers were 27.2%, 28.2%, and 29.5%, respectively. When the higher level controller was based on V2X (vehicle to vehicle, and vehicle to infrastructure), compared with ECMS and rule-based controllers, the average fuel economy improvements for A-ECMS-based controller were 10.3% and 14.8%, respectively. Above all, the research method could be a new architecture to the energy management of HEVs. With the help of the architecture and some developed algorithms, the realization of the real-time control for HEVs will be likely possible.