A method for optimizing the comprehensive efficiency of hybrid electric vehicles in transverse and longitudinal coordination

Abstract

The present invention relates to a method for optimizing the comprehensive efficiency of lateral and longitudinal coordination of a hybrid vehicle, and belongs to the field of energy-saving technology for new energy vehicles. The method comprises: S1: constructing a multi-lane dynamic traffic model to obtain the state information of the vehicle and surrounding traffic participants; S2: dividing the historical vehicle state information into a discarded set and a predicted set, using a dynamic mode decomposition algorithm to predict multi-lane traffic flow based on the predicted set, and obtaining the weighted value of the speed of each lane; S3: using the vehicle state information and lane traffic flow prediction information as the input of the priority experience playback DDQN algorithm to make lane change decisions and speed planning for the vehicle; S4: constructing a power system model of the vehicle according to the vehicle configuration, and establishing the energy flow distribution relationship between different power sources of the vehicle; S5: using the planned speed trajectory as the input of the deep reinforcement learning algorithm, calculating the vehicle's required power and required torque, and optimizing the distribution of energy flow in the power system of the vehicle under the constraint of minimizing fuel consumption.

Description

Technical Field

[0001] The invention belongs to the technical field of energy-saving of new energy vehicles and relates to a method for optimizing the lateral and longitudinal coordinated comprehensive efficiency of a hybrid vehicle. Background Art

[0002] Hybrid vehicles avoid the problems of high energy consumption and serious pollution of traditional fuel vehicles, slow charging speed and short driving range of pure electric vehicles. Compared with traditional fuel vehicles, there is almost no difference in use. At the same time, it achieves the goal of energy saving and environmental protection. It is the most ideal transitional model before high power density battery materials and super fast charging technologies are discovered or completely conquered. With the maturity of technologies such as intelligent networking and intelligent driving, collecting multi-source information and making full use of environmental characteristics to achieve intelligent energy-saving driving of hybrid vehicles has become a research direction that the industry pays close attention to. At present, researchers focus on hybrid vehicles combined with global road conditions information, such as traffic light phases, road congestion conditions, etc., to develop energy management strategies for vehicles in the global scope. However, the actual traffic conditions are dynamically changing, and global planning cannot guarantee the best at every moment. At the same time, inaccurate decision-making planning will deteriorate the overall energy-saving performance of hybrid vehicles. For example, road traffic accidents will cause overall traffic flow shocks. Rule-based methods cannot adapt to complex and changing traffic environments. Vehicle speed trajectories planned according to signal light timing cannot be realized because they do not consider actual traffic conditions. How to make hybrid electric vehicles achieve energy conservation and emission reduction while adapting to dynamic and complex traffic scenarios and improve the overall efficiency of vehicles is of great value and challenge. Existing research rarely combines vehicle speed planning at the lane level with energy management strategies. Therefore, there is an urgent need for a comprehensive efficiency optimization method for lateral and longitudinal coordinated control of hybrid electric vehicles in dynamic traffic scenarios. Summary of the invention

[0003] In view of this, the purpose of the present invention is to provide a method for optimizing the comprehensive efficiency of lateral and longitudinal coordination of hybrid vehicles, taking into account the impact of vehicle lane changing behavior on vehicle speed trajectory in multi-lane dynamic traffic scenarios, integrating traffic flow prediction in the lane changing decision process, and using deep reinforcement learning algorithms to achieve comprehensive performance optimization and improvement of the comfort, economy and traffic efficiency of hybrid vehicles in dynamic traffic.

[0004] In order to achieve the above object, the present invention provides the following technical solutions:

[0005] A method for optimizing the lateral and longitudinal coordinated comprehensive efficiency of a hybrid vehicle comprises the following steps:

[0006] S1: Build a multi-lane dynamic traffic flow model to obtain the status information of the vehicle and surrounding traffic participants, including position, speed and acceleration;

[0007] S2: Divide the historical vehicle status information into a discarded set and a predicted set, use the dynamic pattern decomposition algorithm to perform multi-lane traffic flow prediction based on the predicted set, and obtain the speed weighted value of each lane;

[0008] S3: The vehicle status information and lane traffic flow prediction information are used as the input of the priority experience playback DDQN algorithm, and the lane change decision and speed planning of the vehicle are made considering safety, comfort, economy and traffic efficiency;

[0009] S4: According to the vehicle configuration, construct the vehicle power system model and establish the energy flow distribution relationship between different power sources of the vehicle;

[0010] S5: The vehicle speed trajectory planned in step S3 is used as the input of a deep reinforcement learning (such as DDPG, i.e., deep deterministic policy gradient algorithm) algorithm to calculate the vehicle's required power and required torque, and optimize the distribution of energy flow in the vehicle's power system under the constraint of minimizing fuel consumption.

[0011] Further, step S1 specifically includes the following steps:

[0012] S11: Constructing a multi-lane road model: Using SUMO to build lane length Road len 、Lane widthRoad wid 、Road speed limit v_lim and lane priorityRoad per The multi-lane road model includes SUMO (Simulation of Urban Mobility), which is an open source, microscopic, multi-modal dynamic simulation software.

[0013] S12: Build includes vehicle type Veh typ 、Veh v_lim , maximum acceleration Veh acc_lim 、Maximum deceleration Veh dec_lim 、Entry point Veh ent_d and time Veh ent_t The dynamic traffic flow model including the traffic information of the vehicle and surrounding traffic participants Γ l :

[0014]

[0015] Among them, l is the vehicle number, t corresponds to the dynamic traffic time step, x, y, v, acc are the longitudinal position, lateral position, speed and acceleration information of the vehicle on the road respectively.

[0016] Further, step S2 specifically includes the following steps:

[0017] S21: Collect historical information of all traffic participants on the road to form a vehicle traffic data set, divide the data set into a discarded set and a predicted set, and extract the historical vehicle speed data set X of traffic participants in front of the vehicle in different lanes in the predicted set:

[0018] X=[X1,X2…X n ]

[0019] X i =[V0,V1…V k ]

[0020] V j =[v0,v1…v t ]

[0021] Among them, X i is the traffic information of vehicles on the lane, i∈{1,2…,n}, n is the number of lanes on the road; Vj is the historical speed information corresponding to each vehicle, the data collection time is 0~t time period, j∈{1,2…,k}, k is the number of vehicles on the road, and also corresponds to the vehicle number;

[0022] S22: Select the lane where the vehicle is located and the left and right adjacent lanes as the lane change decision range, extract the traffic information of the vehicle in front of the vehicle, and take the current moment to deduce u steps forward. Set as discard set, Set as prediction set;

[0023] S23: In the prediction set, establish a time-ordered traffic data matrix H:

[0024] H=[h1,h2,h3,…,h N ]

[0025] Among them, H contains N data vectors I∈{1,2,3...N}, N corresponds to the number of vehicles in the lane ahead of the vehicle;

[0026] S24: The delay matrix obtained after embedding the traffic data matrix with delay d is decomposed and reconstructed by the Hankel dynamic mode decomposition algorithm to obtain the observation point prediction function;

[0027] S25: Predict the traffic flow status on the left, middle and right lanes in front of the vehicle respectively to obtain speed weighted values.

[0028] Further, step S24 specifically includes: embedding the traffic data matrix into the delay d to obtain the delay matrix H d :

[0029]

[0030] The delay matrix H dDecomposed into two time-shift matrices H d1 and H d2 :

[0031]

[0032] The Koopman operator approximated by dynamic mode decomposition satisfies the following equation:

[0033] H d2 =KH d1 +r

[0034] Where K is the matrix representation of the finite-dimensional approximation of the Koopman operator, and l is the residual term;

[0035] Rewrite the above formula using singular value decomposition and minimize the residual to obtain:

[0036]

[0037] Among them, UΣW * H d1 The singular value decomposition of , U is the left singular matrix, Σ is the singular value diagonal matrix, W is the right singular matrix, U * , W * The conjugate transposed matrices of the corresponding matrices are obtained through similarity transformation to obtain the matrix L related to the K matrix:

[0038] U * H d2 WΣ -1 =U * KU≡L

[0039] Among them, Σ- 1 is the inverse matrix of Σ, and the characteristic pair of L is (κ I ,w i ), the characteristic pair of K is obtained as (κ I ,Uw i ), κ I is the eigenvalue, w i , Uw i are the eigenvectors of matrix L and matrix K respectively;

[0040] The eigenvalue in continuous time is T is the sampling rate, and the observation points can be obtained through Koopman mode decomposition. The reconstruction representation of

[0041]

[0042] Where r is the order of the U matrix, and the columns of the V matrix are composed of eigenvectors v k constitute, is the coefficient vector, V φ is the pseudo-inverse of V, is the initial data vector, e wt The diagonal elements are The matrix formed.

[0043] Further, step S3 specifically includes the following steps:

[0044] S31: Use the Priority Experience Replay DDQN algorithm to make lane change decisions for the ego vehicle, select the state space acquisition range of traffic participants, and define the state space of lane change decision and speed planning;

[0045] The state space State1 is defined as:

[0046]

[0047] Among them, i represents the i-th area around the vehicle, v ego 、a ego 、x ego They represent the speed, acceleration and relative position of the vehicle respectively. They represent the longitudinal relative distance, lateral relative distance, relative speed and relative acceleration between the traffic participant and the vehicle, They are the weighted values ​​of the traffic flow states of the left and right adjacent lanes and the lane where the vehicle is located obtained from traffic flow prediction, a total of 38 state quantities;

[0048] S32: When the ego vehicle meets the lane change conditions, it executes the lane change operation. At the same time, in order to ensure the stability of traffic flow, considering the vehicle following scenario, based on the characteristics of the priority experience replay DDQN algorithm, a discrete action space is used as the decision output to define the action space of lane change decision and speed planning;

[0049] The action space Action1 is defined as:

[0050] Action1 = {CL left ,CL right ,KL,Acc,Dec}

[0051] Among them, CL left Indicates that the vehicle changes lane to the left, CL right represents that the vehicle changes lanes to the right, KL represents that the vehicle maintains the current speed and does not change lanes, Acc represents that the vehicle accelerates, and Dec represents that the vehicle decelerates.

[0052] Further, step S4 specifically includes the following steps:

[0053] S41: Establish the power balance equation between different power sources of power-split hybrid electric vehicle:

[0054]

[0055] Among them, P req is the vehicle power demand, P final , P MG1 , P MG2 , P ice , P battery are the power of the main reducer output, motor MG1, motor MG2, engine and battery pack respectively, P a is the additional power, η final , η gear are the transmission efficiencies of the final drive and gear respectively, η c1 , η c2 are the working efficiency of motor MG1 and motor MG2, respectively, MG1 、w MG2 、T MG1 、T MG2 are the speed and torque of the motor MG1 and the motor MG2 respectively;

[0056] S42: The power coupling mechanism of the power split hybrid vehicle is a planetary gear, which establishes the speed torque relationship between different power sources:

[0057]

[0058] Among them, w ice 、w r are the speed output values ​​of the engine and the ring gear respectively, T ice 、T r are the torque output values ​​of the engine and the ring gear respectively, N is the number of teeth on the ring gear r Number of sun gear teeth N s Ratio.

[0059] Further, step S5 specifically includes the following steps:

[0060] S51: Use the DDPG algorithm to distribute the energy flow of the power components of a power-split hybrid electric vehicle and define the state space of the energy management control strategy;

[0061] The state space State2 is defined as:

[0062] State2={soc,fuelflw,spd,spderr,l / 100km,acc}

[0063] Among them, soc represents the SOC of the power battery, fuelflw represents the fuel consumption rate of the engine, spd represents the target speed of the vehicle, spderr represents the deviation between the actual speed of the vehicle and the target speed, acc represents the acceleration of the vehicle at this moment, and l / 100km represents the instantaneous fuel consumption per 100 kilometers of the vehicle;

[0064] S52: Use the DDPG algorithm to distribute the energy flow of the power components of a power-split hybrid electric vehicle and define the action space of the energy management control strategy;

[0065] The action space Action2 is defined as:

[0066] Action2 = {throttle}

[0067] Among them, throttle represents the throttle opening of the engine, and the engine power is controlled by adjusting this value;

[0068] S53: The optimization goal is to minimize fuel consumption. The objective function can be expressed as:

[0069]

[0070] Among them, Fuel total is the total fuel consumption, mf(·) is the instantaneous fuel consumption of the engine, and w ice is the engine speed, T ice is the engine torque, T is the total engine running time;

[0071] S54: Calculate the required power P of the vehicle when driving according to the planned vehicle speed req , the power P of the engine, the motor MG1 and the motor MG2 are calculated by the power balance equation between different power sources in step S41. ice , P MG1 and P MG2 .

[0072] The beneficial effects of the present invention are as follows: the present invention can collect traffic environment information, use deep reinforcement learning combined with traffic flow prediction methods to make lane change decisions for the vehicle, plan the optimized speed trajectory at the lane level in combination with safety, comfort and traffic efficiency, and at the same time ensure the optimization of the energy flow distribution of the hybrid vehicle power system under the speed trajectory, control the engine output power, and further improve the energy saving level. It is specifically reflected in the following aspects:

[0073] 1) Aiming at dynamic and complex traffic scenes, the present invention uses SUMO to establish a multi-lane road model and a traffic flow model, and the vehicle operation conditions are close to the actual traffic scenes.

[0074] 2) The present invention is based on the data-driven and priority experience playback DDQN algorithm, and combined with the Hankel dynamic mode decomposition algorithm to develop a vehicle lane change control and speed planning strategy with traffic flow prediction, which ensures the safety and rationality of the vehicle lane change operation, and at the same time achieves the optimization of comfort and traffic efficiency in the speed planning process.

[0075] 3) The present invention builds a power split hybrid vehicle power system model, and combines the DDPG algorithm to realize the energy flow distribution among multiple power sources of the hybrid vehicle, realizes energy-saving control during vehicle operation, and improves the overall efficiency of the vehicle.

[0076] Other advantages, objectives and features of the present invention will be described in the following description to some extent, and to some extent, will be obvious to those skilled in the art based on the following examination and study, or can be taught from the practice of the present invention. The objectives and other advantages of the present invention can be realized and obtained through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0077] In order to make the purpose, technical solutions and advantages of the present invention more clear, the present invention will be described in detail below in conjunction with the accompanying drawings, wherein:

[0078] Figure 1 It is a logical framework diagram for realizing the overall optimization method of the lateral and longitudinal coordinated comprehensive efficiency of a hybrid vehicle of the present invention;

[0079] Figure 2 During the lane change process, the vehicle obtains the 8-area division of other traffic participants;

[0080] Figure 3 This is a power split hybrid vehicle power system of the present invention. DETAILED DESCRIPTION

[0081] The following describes the embodiments of the present invention by specific examples, and those skilled in the art can easily understand other advantages and effects of the present invention from the contents disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and the details in this specification can also be modified or changed in various ways based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that the illustrations provided in the following embodiments only illustrate the basic concept of the present invention in a schematic manner, and the following embodiments and features in the embodiments can be combined with each other without conflict.

[0082] Among them, the drawings are only used for illustrative explanations, and they only represent schematic diagrams rather than actual pictures, and should not be understood as limitations on the present invention. In order to better illustrate the embodiments of the present invention, some parts of the drawings may be omitted, enlarged or reduced, and do not represent the size of actual products. For those skilled in the art, it is understandable that some well-known structures and their descriptions in the drawings may be omitted.

[0083] The same or similar numbers in the drawings of the embodiments of the present invention correspond to the same or similar parts; in the description of the present invention, it should be understood that if the terms "upper", "lower", "left", "right", "front", "rear", etc. indicate the orientation or position relationship, they are based on the orientation or position relationship shown in the drawings, which is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operate in a specific orientation. Therefore, the terms describing the position relationship in the drawings are only used for illustrative purposes and cannot be understood as limiting the present invention. For ordinary technicians in this field, the specific meanings of the above terms can be understood according to specific circumstances.

[0084] See also Figure 1 to Figure 3 The present invention proposes a hybrid vehicle lateral and longitudinal coordinated comprehensive performance optimization method. For dynamic and complex traffic scenes, considering the changes in vehicle speed trajectory and traffic flow fluctuations caused by the lane change operation of the vehicle at the lane level, a hybrid vehicle lateral and longitudinal coordinated comprehensive performance optimization method based on deep reinforcement learning algorithm and dynamic mode decomposition is designed to improve vehicle performance. A vehicle lane change control and speed planning strategy with traffic flow prediction properties that integrates the Hankel dynamic mode decomposition algorithm and the priority experience replay DDQN algorithm is constructed to improve the safety and rationality of vehicle lane changes, and at the same time, improve the comfort and traffic efficiency of vehicle speed planning. A power system simulation model of a power-split hybrid vehicle is built. The vehicle demand power is calculated by planning the vehicle speed, and the power distribution among multiple power sources of the power system is realized based on the DDPG algorithm. The engine output power is directly controlled, which improves the energy saving level of hybrid vehicles. The method specifically includes the following steps:

[0085] S1: Build a multi-lane dynamic traffic model to obtain the status information of the vehicle and surrounding traffic participants, including position, speed, and acceleration. The specific steps include:

[0086] S11: Constructing a multi-lane road model: Using SUMO to build lane length Road len 、Lane widthRoad wid 、Road speed limit v_lim , Lane Priority Road per Multi-lane road models including;

[0087] S12: Build includes vehicle type Veh typ 、Veh v_lim , maximum acceleration Veh acc_lim 、Maximum deceleration Veh dec_lim 、Entry point Veh ent_d and time Veh ent_tThe traffic flow model including the vehicle and surrounding traffic participants is used to obtain the traffic information Γ l :

[0088]

[0089] Among them, l is the vehicle number, t corresponds to the dynamic traffic time step, x, y, v, acc are the longitudinal position, lateral position, speed and acceleration information of the vehicle on the road respectively;

[0090] S2: Divide the historical vehicle status information into a discard set and a prediction set, use the Hankel dynamic mode decomposition algorithm to perform multi-lane traffic flow prediction in the prediction set, and obtain the speed weighted value of each lane, which specifically includes the following steps:

[0091] S21: Collect historical information of all traffic participants on the road to form a vehicle traffic data set, divide the data set into a discarded set and a predicted set, and extract the historical vehicle speed data set X of traffic participants in front of the vehicle in different lanes in the predicted set:

[0092] X=[X1,X2…X n ]

[0093] X i =[V0,V1…V k ]

[0094] V j =[v0,v1…v t ]

[0095] Among them, X i is the traffic information of vehicles on the lane, i∈{1,2…,n}, n is the number of lanes on the road, Vj is the historical speed information corresponding to each vehicle, the data collection time is 0~t time period, j∈{1,2…,k}, k is the number of vehicles on the road, and corresponds to the vehicle number;

[0096] S22: Select the lane where the vehicle is located and the left and right adjacent lanes as the lane change decision range, extract the traffic information of the vehicle in front of the vehicle, and take the current moment to deduce u steps forward. Set as discard set, Set as prediction set;

[0097] S23: In the prediction set, establish a time-ordered traffic data matrix H:

[0098] H=[h1,h2,h3,…,h N ]

[0099] Among them, H contains N data vectors I∈{1,2,3...N}, N corresponds to the number of vehicles in the lane ahead of the vehicle;

[0100] S24: The delay matrix H obtained after the traffic data matrix is ​​embedded with delay d d :

[0101]

[0102] The delay matrix H d Decomposed into two time-shift matrices H d1 and H d2 , that is:

[0103]

[0104] The Koopman operator approximated by dynamic mode decomposition satisfies the following equation:

[0105] H d2 =KH d1 +r

[0106] Where K is the matrix representation of the finite-dimensional approximation of the Koopman operator, and l is the residual term;

[0107] Rewrite the above formula using singular value decomposition and minimize the residual to obtain:

[0108]

[0109] Among them, UΣW * H d1 The singular value decomposition of , U is the left singular matrix, Σ is the singular value diagonal matrix, W * is a right singular matrix, U * , W * The conjugate transposed matrices of the corresponding matrices are obtained through similarity transformation to obtain the matrix L related to the K matrix:

[0110] U * H d2 WΣ -1 =U * KU≡L

[0111] Among them, Σ- 1 is the inverse matrix of Σ, and the characteristic pair of L is (κ I ,w i ), the characteristic pair of K is obtained as (κ I ,Uw i ), κ I is the eigenvalue, w i , Uw i are the eigenvectors of matrix L and matrix K respectively;

[0112] The eigenvalue in continuous time is T is the sampling rate, and the observation points can be obtained through Koopman mode decomposition. The reconstruction representation of

[0113]

[0114] Where r is the order of the U matrix, and the columns of the V matrix are composed of eigenvectors v k constitute, is the coefficient vector, V φ is the pseudo-inverse of V, is the initial data vector, e wt The diagonal elements are The matrix formed.

[0115] S25: Predict the traffic flow prediction matrix ahead of the vehicle on each lane They are the weighted values ​​of the traffic flow speeds on the left, middle and right lanes respectively.

[0116] S3: The vehicle status information and lane traffic flow prediction information are used as the input of the priority experience playback DDQN algorithm. Considering safety, comfort, traffic efficiency and economy, the ego vehicle lane change decision and longitudinal speed planning are carried out. Specifically, the following steps are included:

[0117] S31: When using the Priority Experience Replay DDQN algorithm to make lane change decisions for the vehicle, the state space State1 is defined as:

[0118]

[0119] Among them, i∈{1, 2, ..., 8}, represents the 8 regions around the vehicle, v ego 、a ego 、x ego Represent the speed, acceleration and relative position of the vehicle respectively. Respectively represent the longitudinal relative distance, lateral relative distance, relative speed and relative acceleration between the traffic participant and the vehicle. They are the weighted values ​​of the traffic flow states of the left and right adjacent lanes and the lane where the vehicle is located obtained from traffic flow prediction, totaling 38 state quantities.

[0120] S32: When the vehicle meets the lane change conditions, it executes the lane change operation. At the same time, in order to ensure the stability of traffic flow, considering the vehicle following scenario, according to the characteristics of the priority experience playback Double DQN algorithm, a discrete action space is used as the decision output. The action space Action1 is defined as:

[0121] Action1 = {CL left ,CL right ,KL,Acc,Dec}

[0122] Among them, CL left Indicates that the vehicle changes lane to the left, CL rightrepresents that the vehicle changes lanes to the right, KL represents that the vehicle maintains the current speed and does not change lanes, Acc represents that the vehicle accelerates, and Dec represents that the vehicle decelerates.

[0123] S4: According to the vehicle configuration, construct the vehicle power system model and establish the energy flow distribution relationship between different power sources, which specifically includes the following steps:

[0124] S41: Establish the power balance equation between different power sources of power-split hybrid electric vehicle:

[0125]

[0126] Among them, P final , P MG1 , P MG2 , P ice , P battery are the power of the main reducer output, motor MG1, motor MG2, engine, and battery pack, respectively. req is the vehicle power requirement, P a is the additional power, η final , η gear is the transmission efficiency of the final drive and gear, η c1 , η c2 are the working efficiency of motor MG1 and motor MG2, respectively, MG1 、w MG2 , T MG1 , T MG2 are the rotation speed and torque of the motor MG1 and the motor MG2 respectively.

[0127] S42: The power coupling mechanism of the power split hybrid vehicle is a planetary gear, which establishes the speed torque relationship between different power sources:

[0128]

[0129] Among them, w ice 、w r are the speed output values ​​of the engine and the ring gear respectively, T ice , T r are the torque output values ​​of the engine and the ring gear respectively, N is the number of teeth on the ring gear r Number of sun gear teeth N s Ratio.

[0130] S5: The vehicle speed trajectory planned in step S3 is used as the input of the DDPG algorithm to calculate the vehicle's required power and required torque, and optimize the distribution of the energy flow of the vehicle's power system under the constraint of minimizing fuel consumption, which specifically includes the following steps:

[0131] S51: Use the DDPG algorithm to distribute the energy flow of the power components of a power-split hybrid electric vehicle. The state space State2 is defined as:

[0132] State2={soc,fuelflw,spd,spderr,l / 100km,acc}

[0133] Among them, soc represents the SOC of the power battery, fuelflw represents the fuel consumption rate of the engine, spd represents the target speed of the vehicle, spderr represents the deviation between the actual speed of the vehicle and the target speed, acc represents the acceleration of the vehicle at this moment, and l / 100km represents the instantaneous fuel consumption per 100 kilometers of the vehicle.

[0134] S52: Use the DDPG algorithm to distribute the energy flow of the power components of the power split hybrid vehicle. The action space Action2 is defined as:

[0135] Action2 = {throttle}

[0136] Among them, throttle represents the throttle opening of the engine, and the engine power is controlled by adjusting this value.

[0137] S53: The optimization goal is to minimize fuel consumption. The objective function can be expressed as:

[0138]

[0139] Among them, Fuel total is the total fuel consumption, mf(·) is the instantaneous fuel consumption of the engine, and w ice is the engine speed, T ice is the engine torque, and T is the total running time of the engine.

[0140] S54: Calculate the required power P of the vehicle when driving according to the planned vehicle speed req , calculate the power P of the engine, motor MG1 and motor MG2 through the power balance equation between the power sources in S41 ice , P MG1 , P MG2 .

[0141] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solution of the present invention can be modified or replaced by equivalents without departing from the purpose and scope of the technical solution, which should be included in the scope of the claims of the present invention.

Claims

1. A method for optimizing the lateral and longitudinal coordinated comprehensive efficiency of a hybrid vehicle, characterized in that: The method specifically comprises the following steps: S1: Build a multi-lane dynamic traffic flow model to obtain the status information of the vehicle and surrounding traffic participants, including position, speed and acceleration; S2: Divide the historical vehicle status information into a discard set and a prediction set, use the dynamic pattern decomposition algorithm to perform multi-lane traffic flow prediction based on the prediction set, and obtain the speed weighted value of each lane, which specifically includes the following steps: S21: Collect historical information of all traffic participants on the road to form a vehicle traffic data set, divide the data set into a discarded set and a predicted set, and extract the historical vehicle speed data set X of traffic participants in front of the vehicle in different lanes in the predicted set: X=[X1,X2…X n ] X i =[V0,V1…V k ] V j =[v0,v1…v t ] Among them, X i is the traffic information of the vehicle on the lane, i∈{1,2…,n}, n is the number of lanes on the road; V j The historical speed information corresponding to each vehicle, the data collection time is 0~t time period, j∈{1,2…,k}, k is the number of vehicles on the road, and also corresponds to the vehicle number; S22: Select the lane where the vehicle is located and the left and right adjacent lanes as the lane change decision range, extract the traffic information of the vehicle in front of the vehicle, and take the current moment to deduce u steps forward. Set as discard set, Set as prediction set; S23: In the prediction set, establish a time-ordered traffic data matrix H: H=[h1,h2,h3,…,h N ] Among them, H contains N data vectors N corresponds to the number of vehicles in the lane ahead of the ego vehicle; S24: The delay matrix obtained after embedding the traffic data matrix with delay d is decomposed and reconstructed by the Hankel dynamic mode decomposition algorithm to obtain the observation point prediction function; S25: predicting the traffic flow states on the left, middle and right lanes in front of the vehicle respectively, and obtaining speed weighted values; S3: The vehicle status information and lane traffic flow prediction information are used as the input of the priority experience playback DDQN algorithm, and the lane change decision and speed planning of the vehicle are made considering safety, comfort, economy and traffic efficiency; S4: According to the vehicle configuration, construct the vehicle power system model and establish the energy flow distribution relationship between different power sources of the vehicle; S5: The vehicle speed trajectory planned in step S3 is used as the input of the DDPG algorithm to calculate the vehicle's required power and required torque, and optimize the distribution of energy flow in the vehicle's power system under the constraint of minimizing fuel consumption.

2. The method for optimizing the lateral and longitudinal coordinated comprehensive efficiency of a hybrid vehicle according to claim 1, characterized in that: Step S1 specifically includes the following steps: S11: Constructing a multi-lane road model: Using SUMO to build lane length Road len 、Lane widthRoad wid 、Road speed limit v_lim and lane priorityRoad per A multi-lane road model including SUMO, which is an open source, microscopic, multi-modal dynamic simulation software. S12: Build includes vehicle type Veh typ 、Veh v_lim , maximum acceleration Veh acc_lim 、Maximum deceleration Veh dec_lim 、Entry point Veh ent_d and time Veh ent_t The dynamic traffic flow model including the traffic information of the vehicle and surrounding traffic participants Γ l : Among them, l is the vehicle number, t corresponds to the dynamic traffic time step, x, y, v, acc are the longitudinal position, lateral position, speed and acceleration information of the vehicle on the road respectively.

3. The method for optimizing the lateral and longitudinal coordinated comprehensive efficiency of a hybrid vehicle according to claim 1, characterized in that: Step S24 specifically includes: embedding the traffic data matrix into the delay d to obtain the delay matrix H d : The delay matrix H d Decomposed into two time-shift matrices H d1 and H d2 : The Koopman operator approximated by dynamic mode decomposition satisfies the following equation: H d2 =KH d1 +l Where K is the matrix representation of the finite-dimensional approximation of the Koopman operator, and l is the residual term; Rewrite the above formula using singular value decomposition and minimize the residual to obtain: Among them, UΣW * H d1 The singular value decomposition of , U is the left singular matrix, Σ is the singular value diagonal matrix, W is the right singular matrix, U * , W * The conjugate transposed matrices of the corresponding matrices are obtained through similarity transformation to obtain the matrix L related to the K matrix: U * H d2 WΣ -1 =U * KU≡L Among them, Σ -1 is the inverse matrix of Σ, and the characteristic pair of L is (κ I ,w i ), the characteristic pair of K is obtained as (κ I ,Uw i ), κ I is the eigenvalue, w i , Uw i are the eigenvectors of matrix L and matrix K respectively; The eigenvalue in continuous time is T is the sampling rate, and the observation points are obtained by Koopman mode decomposition. The reconstruction representation of Where r is the order of the U matrix, and the columns of the V matrix are composed of eigenvectors v k constitute, is the coefficient vector, V φ is the pseudo-inverse of V, is the initial data vector, e wt The diagonal elements are The matrix formed.

4. The method for optimizing the lateral and longitudinal coordinated comprehensive efficiency of a hybrid vehicle according to claim 1, characterized in that: Step S3 specifically includes the following steps: S31: Use the Priority Experience Replay DDQN algorithm to make lane change decisions for the ego vehicle, select the state space acquisition range of traffic participants, and define the state space of lane change decision and speed planning; The state space State1 is defined as: Among them, i represents the i-th area around the vehicle, v ego 、a ego 、x ego They represent the speed, acceleration and relative position of the vehicle respectively. They represent the longitudinal relative distance, lateral relative distance, relative speed and relative acceleration between the traffic participant and the vehicle, are the weighted values ​​of the traffic flow states of the left and right adjacent lanes and the lane where the vehicle is located obtained by traffic flow prediction; S32: When the ego vehicle meets the lane change conditions, it executes the lane change operation. At the same time, in order to ensure the stability of traffic flow, considering the vehicle following scenario, based on the characteristics of the priority experience replay DDQN algorithm, a discrete action space is used as the decision output to define the action space of lane change decision and speed planning; The action space Action1 is defined as: Action1={CL left ,CL right ,KL,Acc,Dec} Among them, CL left Indicates that the vehicle changes lane to the left, CL right represents that the vehicle changes lanes to the right, KL represents that the vehicle maintains the current speed and does not change lanes, Acc represents that the vehicle accelerates, and Dec represents that the vehicle decelerates.

5. The method for optimizing the lateral and longitudinal coordinated comprehensive efficiency of a hybrid vehicle according to claim 4, characterized in that: Step S4 specifically includes the following steps: S41: Establish the power balance equation between different power sources of power-split hybrid electric vehicle: Among them, P req is the vehicle power demand, P final , P MG1 , P MG2 , P ice , P battery are the power of the main reducer output, motor MG1, motor MG2, engine and battery pack respectively, P a is the additional power, η final , η gear are the transmission efficiencies of the final drive and gear respectively, η c1 , η c2 are the working efficiency of motor MG1 and motor MG2, respectively, MG1 、w MG2 、T MG1 、T MG2 are the speed and torque of the motor MG1 and the motor MG2 respectively; S42: The power coupling mechanism of the power split hybrid vehicle is a planetary gear, which establishes the speed torque relationship between different power sources: Among them, w ice 、w r are the speed output values ​​of the engine and the ring gear respectively, T ice 、T r are the torque output values ​​of the engine and the ring gear respectively, N is the number of teeth on the ring gear r Number of sun gear teeth N s Ratio.

6. The method for optimizing the lateral and longitudinal coordinated comprehensive efficiency of a hybrid vehicle according to claim 5, characterized in that: Step S5 specifically includes the following steps: S51: Use the DDPG algorithm to distribute the energy flow of the power components of a power-split hybrid electric vehicle and define the state space of the energy management control strategy; The state space State2 is defined as: State2={soc,fuelflw,spd,spderr,l / 100km,acc} Among them, soc represents the SOC of the power battery, fuelflw represents the fuel consumption rate of the engine, spd represents the target speed of the vehicle, spderr represents the deviation between the actual speed of the vehicle and the target speed, acc represents the acceleration of the vehicle at this moment, and l / 100km represents the instantaneous fuel consumption per 100 kilometers of the vehicle; S52: Use the DDPG algorithm to distribute the energy flow of the power components of a power-split hybrid electric vehicle and define the action space of the energy management control strategy; The action space Action2 is defined as: Action2 = {throttle} Among them, throttle represents the throttle opening of the engine, and the engine power is controlled by adjusting this value; S53: The optimization goal is to minimize fuel consumption. The objective function is expressed as: Among them, Fuel total is the total fuel consumption, m f (·) is the instantaneous fuel consumption of the engine, w ice is the engine speed, T ice is the engine torque, T is the total engine running time; S54: Calculate the required power P of the vehicle when driving according to the planned vehicle speed req , the power P of the engine, the motor MG1 and the motor MG2 are calculated by the power balance equation between different power sources in step S41. ice , P MG1 and P MG2 .

Images