Research On Energy-saving Optimization Of Coordinated Control Strategy For The Platoon Of Connected Hybrid Electric Vehicles

With the rebound of energy demand after the pandemic of COVID-19,the global carbon dioxide emissions reached the largest single increase recorded in history.The supply-demand imbalance led to the spread of the energy crisis around the world,which alarms again for the schedule of ecologicalization.As the main field of low-carbon transformation,green transportation takes an important role of achieving carbon peaking and carbon neutrality.Among them,the new energy vehicles with clean energy sources become the main carrier to realize green transportation.Recently,the promotion of electronic control and internet of vehicles further digs out the energy-saving optimization potential of the new energy vehicles.Accordingly,this paper focuses on energy-saving optimization by combining artificial intelligence and communication technology for new energy vehicles.Besides,the platooning is an efficient multi-vehicle grouping mode,which effectively improves the overall traffic efficiency of road traffic and reduces energy consumption.Among them,the platoon composed of hybrid electric vehicles(HEVs)has the advantages of long driving range and low-carbon emissions,which has become the mainstream development trend.However,it is difficult to achieve multi-vehicle coordination due to the complex powertrain and various driving scenarios.The current state-of-art framework still failed to give full play to the information-interconnected and global-planning power of the internet of vehicles.Therefore,the systematic solution has not been appropriately formed for the ecological driving of a platoon of HEVs.Thus,the energy-saving and emission reduction optimization potential of the connected platoon of HEVs has not been fully released.This paper takes the connected hybrid electric platoon as the studied object,aiming to break the information barrier among the connected HEVs.With the integrated optimization of driving behaviors and powertrain,the proposed methods could fully release the energy consumption optimization potential from the three levels of powertrain,vehicle,and platoon.Thus,it is capable to realize multi-vehicle collaborative energy-saving optimization control of the HEV platoon.The main works are summarized as follows:The state transition of longitudinal motion and powertrain execution during the driving process is studied for the platoon of HEVs.The control response model of the platoon of HEVs is constructed considering the engine and motor as the basic control unit,and the energy consumption model of the powertrain is established to calculate the fuel and battery consumption.The objective function of energy-saving optimization of the platoon of HEVs is proposed and the coupling relationship of control variables in the objective function is further analyzed.Based on that,the overall architecture of energy-saving optimization is divided into energy management,eco-driving,and platoon control.The global state of charge(SOC)planning strategy is studied for the energy management of HEV.The frequency of forced battery charging by the engine in the future trip is estimated based on the Bayesian neural network,and the advantage function involved in the conditional information entropy is constructed by introducing the observation of conditional probability.The action selection is evaluated according to the advantage function,further guiding parameters update of the deep neural network-based energy management strategy.The proposed method,adaptive policy optimization(APO),adaptively adjusts the engine power according to the remaining driving distance and SOC,so as to reduce the fuel efficiency reduction while avoiding battery shortage.Experiments show that,compared with rule-based and standard reinforcement learning strategies,the proposed energy management strategy achieves stable convergence to the approximate global optimal value regardless of the initial state,and the power-feeding frequency under battery shortage is reduced by90% or more.Moreover,the adaptability and robust performance get validated over the Dublin bus trajectories data.With the comprehensive experiments in this paper,the proposed model exhibits enhanced fuel economy and more suitable SOC planning in comparison with the existing energy management strategies.The results indicate that APO respectively outperforms the compared online strategies by 9.8% and 2.6% and reaches 98% energy-saving rate of the global optimum.The safe reinforcement learning based training framework for energy management strategy is studied,and a Watchdog mechanism is introduced into the controller to avoid systemic risks caused by free exploration and insufficient optimization in the early stage of training.With the help of the internet of vehicles and cloud computing to optimize parameters and real-time control decoupling,a reinforcement learning training method based on Lagrangian relaxation is proposed.By using Lagrangian relaxation,the optimization for the constrained Markov decision process transforms into an unconstrained dual problem to minimize the energy consumption while minimizing the Watchdog participation.The experimental results under the driving conditions of the Dublin bus show that the strategy training effectively reduces the trigger frequency of auxiliary rules under the safety constraints from 11% to below 0.5%,which ensures the safety of policy training and avoids relying on auxiliary rules,the energy management strategy after training can adjust the working time of the engine within the global operating conditions according to the power system and operating conditions while ensuring the high fuel efficiency of the engine.The economy-oriented automated HEV is studied in this chapter to save energy by optimizing both driving behaviors and power distribution.A hierarchical reinforcement learning based ACC-EMS strategy is proposed with a hierarchical policy and non-hierarchical execution.The upper layer learns to plan stateof-charge and time-headway trajectories,while the low layer policy learns to achieve the expected goals by outputting control variables executed by the host vehicle.The proposed ACC-EMS strategy was self-learning by interaction in a car-following scenario constructed with GPS data on the I-880 highway.Comprehensive experiments show the proposed strategy has significantly improved the training speed and stability,achieving a 98% energy-saving rate with less than 600 times of computational load compared to the global optimum.The multi-vehicle collaborative energy-saving optimization architecture for the platoon control of connected HEVs is studied in this chapter,and a multi-agent reinforcement learning based centralized training distributed controlling framework is proposed.Whereas the value function of the joint action is estimated with the LSTM network leading the HEVs of platoon to cooperate with each other.Asynchronous strategy optimization is presented for accelerating the training process of lower-layer policies and improving robustness and adaptability.The results under the HWFET driving cycle show that the proposed method effectively reduces the transmission of shock waves through multi-vehicle coordination.Compared with the online baseline,the fuel consumption of the platoon is reduced by 19.2% through the coordination of connected HEVs.