Research On The Intelligent Optimization Of Electric Vehicle Driving Behavior

Driving behavior is one of the main factors that affect the driving energy consumption of electric vehicles.Different driving behaviors on the same driving condition will lead to different driving energy consumptions.The optimization of driving behavior can effectively reduce the driving energy consumption of electric vehicles.However,the driving behavior of vehicles is affected by many factors,such as the states of vehicles themselves,traffic environment,weather and geography information,etc.The current driving behavior analysis method is difficult to analyze and optimize the driving behavior according to various complicated influencing factors in a unified framework.In this paper,we study on the energy-saving optimization of electric vehicles,consider the various complicated factors that affect the driving behavior,optimize the driving behavior of electric vehicles,reduce the vehicle energy consumption and improve the driving range.This paper proposes an energy-saving optimization method for driving behavior of electric vehicles based on deep reinforcement learning.Considering various factors that affect driving behavior under the framework of deep reinforcement learning,a vehicle intelligent control method based on multi-source high-dimensional heterogeneous information fusion is constructed.The experimental results show that the proposed method can greatly reduce the driving energy consumption of electric vehicles.The main work and research results are as follows:（1）The optimization method of driving behavior of electric vehicles based on deep reinforcement learning is proposed,which realizes the autonomous energy saving optimization of electric vehicle driving behavior.The deep reinforcement learning algorithm is used to fully exploit the potential correlation between the optimal energy-saving driving behavior of electric vehicles and various complex influencing factors.Under the simulation experiment platform,the superiority of the driving behavior optimization method is verified.The test of sport-type driving behavior shows that the electric vehicle agent can learn the acceleration,lane changing and other behaviors autonomously and drive as fast as possible.In the energy-saving driving behavior optimization test,the electric vehicle intelligent body can by itself optimize its own control strategy,reduce driving energy consumption by 40%,about 7 kwh power saving per 100 kilometers.The above experiments fully demonstrates that the algorithm agent can by itself optimize the driving behavior and achieve more energy-efficient driving.（2）A multi-source high-dimensional heterogeneous information representation extraction method is proposed to solve the multi-source high-dimensional heterogeneous information fusion problem in vehicle driving environment.According to different types of multi-source high-dimensional heterogeneous information,different deep learning networks are respectively used to extract information representation vectors,and L2 norm normalized concatenation is used to realize multi-source high-dimensional heterogeneous data fusion,which makes the optimization model of electric vehicle driving behavior fully exploit the characteristics of the multi-source high-dimensional heterogeneous information in an uniform framework.For traffic information,the long-term memory neural network is used to extract time-series traffic information representation;the denoise autoencoder is used to extract other multi-source high-dimensional heterogeneous information features such as vehicle states and geographic information.（3）A driving style modeling method based on generative adversarial network is proposed,which solves the problem of abstract driving style modeling and description.The driving energy saving characteristic of driving style can be modeled by generative adversarial network.An objective function of driving behavior optimization for electric vehicles considering driving style energy saving characteristic was established.Energy saving optimization of driving behavior and driving style transfer based on optimal energy consumption were realized.