Power dividing type new energy automobile control method
By constructing a SAC energy management model with a vehicle queue model and an expert knowledge base, the speed and power distribution of new energy vehicles are optimized, and the problem of neglecting the interaction between vehicles in the existing technology is solved, and the fuel economy of new energy vehicles is improved.
Patent Information
- Application Number
- CN202411359985.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-09-27
- Publication Date
- 2024-12-06
AI Technical Summary
The fuel economy control strategy of existing new energy vehicles ignores the interaction between vehicles in the road environment, resulting in insufficient analysis and control of the fuel economy of automobiles.
By obtaining traffic information of new energy vehicles, building a vehicle queue model for speed prediction, and determining the torque and speed of the engine and motor based on the SAC energy management model of the expert knowledge base, thereby optimizing the power distribution of the car.
It effectively improves the fuel economy of power shunt new energy vehicles, and achieves more accurate fuel management by considering external factors in the vehicle queue.
Smart Images

Figure CN119078791A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to a vehicle control method, and in particular to a power split type new energy vehicle control method. Background Art
[0002] With the rapid development of intelligent connected vehicle technology, unprecedented innovative solutions have been provided for the real-time energy management and optimization of new energy vehicle fleets. At the same time, among new energy vehicles, power-split new energy vehicles can perform infinitely variable speeds and have huge energy-saving potential. On the road, if new energy vehicles can interact with adjacent vehicles in real time through information sharing, participate in vehicle fleets and achieve longitudinal control, and coordinately adjust driving status, then compared with single-operating vehicles, they can not only achieve significant improvements in safety, but also show greater advantages in economy.
[0003] However, in current technical practice, the control strategy of new energy vehicle fuel economy mainly relies on the real-time status of the vehicle itself, such as battery status, motor efficiency, and driver's operating mode, for local optimization. Although this strategy can improve vehicle performance to a certain extent, it ignores the interaction between vehicles in the road environment. In fact, the driving state of the car is deeply affected by factors such as the speed, acceleration, and position of the vehicles in front and behind. Therefore, isolated control strategies often fail to fully consider these external factors, resulting in inaccurate analysis and control of vehicle fuel economy.
[0004] Therefore, in order to solve the above technical problems, it is urgent to propose a new technical means. Summary of the invention
[0005] In view of this, the purpose of the present invention is to provide a power-split new energy vehicle control method, which can plan the speed according to the vehicle status information and distribute the vehicle power, so as to effectively improve the fuel economy of the power-split new energy vehicle.
[0006] The present invention provides a power split new energy vehicle control method, comprising the following steps:
[0007] S1. Obtaining traffic information of new energy vehicles, the traffic information including the position and speed information of the vehicle in the vehicle queue;
[0008] S2. Construct a vehicle queue model, predict the speed at the next moment based on the vehicle queue model, and output the optimal speed prediction result;
[0009] S3. Establish a SAC energy management model based on the expert knowledge base, input the optimal speed prediction result into the SAC energy management model to obtain the engine demand power, and determine the engine torque and speed as well as the motor torque and speed based on the engine demand power, and control the new energy vehicle according to the engine torque and speed as well as the motor torque and speed.
[0010] Furthermore, a vehicle queue model is constructed, and the speed at the next moment is predicted based on the vehicle queue model, and the optimal speed prediction result is output, specifically including:
[0011] Construct a nonlinear vehicle platooning model:
[0012]
[0013] Where: s i (t), ν i (t) are the displacement and velocity of vehicle i at time t, Δt is the time interval, m i is the mass of vehicle i, η i is the mechanical efficiency of the transmission system of vehicle i, τ i is the time delay constant of vehicle i, C D is the air resistance coefficient, A i is the frontal area of vehicle i, g is the acceleration due to gravity, T i (t) is the actual torque of vehicle i at time t, u i (t) is the expected required torque of vehicle i at time t;
[0014] Simplify the nonlinear vehicle platoon model:
[0015] The state variable of vehicle i is defined as The output variable is defined as in:
[0016]
[0017] in:
[0018]
[0019] Define the state matrix X(t), control matrix U(t), and output matrix Y(t):
[0020]
[0021] in: Γ=diag{γ,γ,...,γ}∈R 2N×3N ;
[0022] In a vehicle queue, the expected distance between adjacent vehicles is d0; the displacement and speed of the leading vehicle are s0(t) and v0(t) respectively. The expected state and expected output of any vehicle in the queue are expressed as:
[0023]
[0024] in: p i,des (t) = p0(t)-i·d0,v i,des (t) = v0, T i,des (t) = h i (v0), h i (v0) is the torque required to maintain the vehicle at a constant speed, r i is the wheel radius;
[0025] The information topology structure in the vehicle queue is represented by a directed graph G. N = {v N ,ε N} and the augmented directed graph G N+1 = {V N+1 ,E N+1} to indicate,
[0026] The set of vehicles that any vehicle i∈N in the queue can obtain information is N i ={i1,i2,…,i m} and the reachable set of the pilot car is:
[0027] p i =1 means vehicle i can obtain the information of the pilot vehicle, otherwise, p i =0;
[0028] definition and are the predicted state and the assumed state respectively, and They are the predicted control input sequence, the optimal predicted control input sequence and the assumed control input sequence. and is the predicted output sequence and the hypothesized output sequence, where k∈[0,1,2,…,N p -1,N p ];
[0029] Construct NDMPC optimization problem model:
[0030]
[0031] st
[0032]
[0033] |I i | represents the set I i =N i ∪P i The base number, represents the distance between vehicle i and vehicle j, and the objective function l i for:
[0034]
[0035] Where: Q i , R i 、F i , G i are weight matrices, and they are all positive semi-definite matrices.
[0036] Further, step S3 specifically includes:
[0037] Building an expert knowledge base, the expert knowledge base including an optimal fuel economy curve of an engine;
[0038] Construct a SAC energy management network, and the reward function of the SAC energy management network is:
[0039]
[0040] Among them: Reward i is the reward value of vehicle i, β is the weight coefficient of fuel consumption, t0 is the starting time, ε is the weight coefficient of SOC balance, is the instantaneous fuel consumption rate of the engine;
[0041] The constraints of the reward function are:
[0042]
[0043] Where: T e (t) represents the engine torque, ω e,min is the minimum speed of the engine, T MG1,min Represents the minimum torque of the generator, T MG2,min Represents the minimum torque of the motor, ω MG1 (t) represents the speed of the generator, ω MG2 (t) represents the speed of the motor, P batt (t) represents the battery charge and discharge cycle output power;
[0044] Define the state variables and action space of the SAC energy management network:
[0045] The state variable is: State i = {ν i (t),a i (t),SOC i (t)};
[0046] The action space is: Action i = {P ei (t)}; Action i is the action space of vehicle i, P ei (t) is the engine power of vehicle i at time t;
[0047] Construct the objective function:
[0048]
[0049] Where: r is the action value, s t is the state at time t, a t is the action at time t, ω is the weight coefficient, H(π(·|s t )) is the policy entropy value, π is the policy function;
[0050] The action value function Q(s) of the policy π t ,a t ) and value function V(s t ):
[0051]
[0052] Where: r t is the reward function, γ represents the learning rate;
[0053] In the value network, the value network parameter ψ is updated by minimizing the following loss function through the policy gradient descent method:
[0054]
[0055] In the action-value network, the network parameters are updated as follows:
[0056]
[0057] Where: Q′(s t ,a t ) is the target Q value, ψ′ is the target value function network parameter, and is updated using a soft update method as follows:
[0058] ψ'=τψ+(1-τ)ψ'; τ is the soft update weight coefficient;
[0059] In the policy network, the objective function for parameter update is as follows:
[0060]
[0061] The optimal speed is input into the trained SAC energy management network based on the expert knowledge base to output the optimal required power of the engine.
[0062] Further, determining the engine torque and the motor torque based on the engine demand power specifically includes:
[0063] According to the optimal required power, the power-torque correspondence table and the torque-speed correspondence table are searched to obtain the optimal torque and speed of the engine;
[0064] Model a planetary gear set:
[0065]
[0066] Where: MG1 is the speed of the generator, ω MG2 is the motor speed, ω eng is the engine speed, T eng is the engine torque, T MG1 is the torque of the generator, T MG2 is the torque of the motor; α is the ratio of the number of teeth of the ring gear to the number of teeth of the sun gear in the planetary gear set; g f is the main reducer transmission ratio, v * is the optimal speed.
[0067] Substitute the optimal torque and speed of the engine into the planetary gear set model to obtain the torque and speed of the motor.
[0068] Further, the state of charge of the new energy vehicle i is determined as follows:
[0069]
[0070] Where: V oc,i (t) is the open circuit voltage at time t, R 0,i is the internal resistance, I i (t) is the battery current at time t, P batt,i (t) is the charge-discharge cycle output power at time t, SOC i (t) is the battery state of charge at time t, Q 0,i is the initial battery capacity, Q i is the nominal capacity of the battery.
[0071] Beneficial effects of the present invention: Through the present invention, speed planning can be carried out according to the status information of the vehicle, and the power of the vehicle can be distributed, thereby effectively improving the fuel economy of power-split new energy vehicles. BRIEF DESCRIPTION OF THE DRAWINGS
[0072] The present invention will be further described below in conjunction with the accompanying drawings and embodiments:
[0073] Figure 1 It is a flow chart of the present invention.
[0074] Figure 2 This is the Actor-Critic architecture diagram. DETAILED DESCRIPTION
[0075] The present invention is further described in detail below:
[0076] The present invention provides a power split new energy vehicle control method, comprising the following steps:
[0077] S1. Obtaining traffic information of new energy vehicles, the traffic information including the position and speed information of the vehicle in the vehicle queue;
[0078] S2. Construct a vehicle queue model, predict the speed at the next moment based on the vehicle queue model, and output the optimal speed prediction result;
[0079] S3. Establish a SAC energy management model based on the expert knowledge base, input the optimal speed prediction result into the SAC energy management model to obtain the engine demand power, and determine the engine torque and speed as well as the motor torque and speed based on the engine demand power, and control the new energy vehicle according to the engine torque and speed as well as the motor torque and speed. Through the above method, speed planning can be carried out according to the vehicle status information, and the vehicle power can be allocated, thereby effectively improving the fuel economy of power-dividing new energy vehicles.
[0080] In this embodiment, a vehicle queue model is constructed, and the speed at the next moment is predicted based on the vehicle queue model, and the optimal speed prediction result is output, specifically including:
[0081] Construct a nonlinear vehicle platooning model:
[0082]
[0083] Where: s i (t), ν i (t) are the displacement and velocity of vehicle i at time t, Δt is the time interval, m i is the mass of vehicle i, η i is the mechanical efficiency of the transmission system of vehicle i, τ i is the time delay constant of vehicle i, C D is the air resistance coefficient, A i is the frontal area of vehicle i, g is the acceleration due to gravity, T i(t) is the actual torque of vehicle i at time t, u i (t) is the expected required torque of vehicle i at time t;
[0084] Simplify the nonlinear vehicle platoon model:
[0085] The state variable of vehicle i is defined as The output variable is defined as in:
[0086]
[0087] in:
[0088]
[0089] Define the state matrix X(t), control matrix U(t), and output matrix Y(t):
[0090]
[0091] in: Γ=diag{γ,γ,...,γ}∈R 2N×3N ;
[0092] In a vehicle queue, the expected distance between adjacent vehicles is d0; the displacement and speed of the leading vehicle are s0(t) and v0(t) respectively. The expected state and expected output of any vehicle in the queue are expressed as:
[0093]
[0094] in: p i,des (t) = p0(t)-i·d0,v i,des (t) = v0, T i,des (t) = h i (v0), h i (v0) is the torque required to maintain the vehicle at a constant speed, r i is the wheel radius;
[0095] The information topology structure in the vehicle queue is represented by a directed graph G. N = {v N ,ε N} and the augmented directed graph G N+1 = {V N+1 ,E N+1} to indicate,
[0096] The set of vehicles that any vehicle i∈N in the queue can obtain information is N i ={i1,i2,…,i m} and the reachable set of the pilot car is:
[0097] p i =1 means vehicle i can obtain the information of the pilot vehicle, otherwise, p i =0;
[0098] definition and are the predicted state and the assumed state respectively, and They are the predicted control input sequence, the optimal predicted control input sequence and the assumed control input sequence. and is the predicted output sequence and the hypothesized output sequence, where k∈[0,1,2,…,N p -1,N p ];
[0099] Construct NDMPC optimization problem model:
[0100]
[0101] st
[0102]
[0103] Among them, N p represents the prediction step length;
[0104] |I i | represents the set I i =N i ∪P i The cardinality is the number of vehicles that can receive information. represents the distance between vehicle i and vehicle j, and the objective function l i for:
[0105]
[0106] Where: Q i , R i 、F i , G i is a weight matrix, and all are semi-positive matrices, that is, 0≤Q i 、F i , 0≤R i ∈R.Q iThe value of is determined by whether vehicle i obtains the information of the pilot vehicle. If it obtains the information of the pilot vehicle, then Q i >0, otherwise, Q i = 0. i It is used to limit excessive changes in the control sequence to ensure that the vehicle can run as smoothly as possible to meet comfort requirements. i It is used to limit the difference between the predicted output and the assumed output to ensure that the vehicle can drive according to the assumed output. i It is used to limit the gap between vehicle i and the assumed output of the vehicle based on the information obtained, ensuring that the vehicles can travel according to their relative positions in the queue without changing the vehicle formation to meet safety requirements.
[0107] In this embodiment, step S3 specifically includes:
[0108] Building an expert knowledge base, the expert knowledge base including an optimal fuel economy curve of an engine;
[0109] Construct a SAC energy management network, where the SAC network adopts the existing Actor-Critic architecture. The Actor network in the Actor-Critic architecture is similar to the Critic network structure, with three hidden layers, each containing 600, 300, and 150 neurons respectively. Figure 2 As shown;
[0110] And the reward function of the SAC energy management network is:
[0111]
[0112] Among them: Reward i is the reward value of vehicle i, β is the weight coefficient of fuel consumption, t0 is the starting time, ε is the weight coefficient of SOC balance, is the instantaneous fuel consumption rate of the engine;
[0113] The constraints of the reward function are:
[0114]
[0115] The constraints come from the engine, motor and battery in the car, namely the torque and speed of the engine and motor, the SOC and charge and discharge power of the battery;
[0116] Where: T e (t) represents the engine torque, ω e,min is the minimum speed of the engine, T MG1,min Represents the minimum torque of the generator, T MG2,min Represents the minimum torque of the motor, ω MG1(t) represents the speed of the generator, ω MG2 (t) represents the speed of the motor, P batt (t) represents the battery charge and discharge cycle output power;
[0117] Define the state variables and action space of the SAC energy management network:
[0118] The state variable is: State i = { i (t),a i (t),SOC i (t)};
[0119] The action space is: Action i = {P ei (t)}; Action i is the action space of vehicle i, P ei (t) is the engine power of vehicle i at time t;
[0120] Construct the objective function:
[0121]
[0122] Where: r is the action value, s t is the state at time t, a t is the action at time t, ω is the weight coefficient, H(π(·|s t )) is the policy entropy value, π is the policy function;
[0123] The action value function Q(s) of the policy π t ,a t ) and value function V(s t ):
[0124]
[0125] Where: r t is the reward function, γ represents the learning rate;
[0126] In the value network, the value network parameter ψ is updated by minimizing the following loss function through the policy gradient descent method:
[0127]
[0128] In the action-value network, the network parameters are updated as follows:
[0129]
[0130] Where: Q′(s t ,a t) is the target Q value, ψ′ is the target value function network parameter, and is updated using a soft update method as follows:
[0131] ψ'=τψ+(1-τ)ψ'; τ is the soft update weight coefficient;
[0132] In the policy network, the objective function for parameter update is as follows:
[0133]
[0134] The optimal speed is input into the trained SAC energy management network based on the expert knowledge base to output the optimal required power of the engine.
[0135] In this embodiment, determining the engine torque and the motor torque based on the engine required power specifically includes:
[0136] According to the optimal required power, the power-torque correspondence table and the torque-speed correspondence table are searched to obtain the optimal torque and speed of the engine;
[0137] Model a planetary gear set:
[0138]
[0139] Where: MG1 is the speed of the generator, ω MG2 is the motor speed, ω eng is the engine speed, T eng is the engine torque, T MG1 is the torque of the generator, T MG2 is the torque of the motor; α is the ratio of the number of teeth of the ring gear to the number of teeth of the sun gear in the planetary gear set; g f is the main reducer transmission ratio, v * is the optimal speed.
[0140] Substitute the optimal torque and speed of the engine into the planetary gear set model to obtain the torque and speed of the motor.
[0141] In this embodiment, the charge state of the new energy vehicle i is determined as follows:
[0142]
[0143] Where: V oc,i (t) is the open circuit voltage at time t, R 0,i is the internal resistance, I i (t) is the battery current at time t, P batt,i (t) is the charge-discharge cycle output power at time t, SOC i (t) is the battery state of charge at time t, Q 0,iis the initial battery capacity, Q i is the nominal capacity of the battery.
[0144] 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 of the present invention, which should be included in the scope of the claims of the present invention.
Claims
1. A power split new energy vehicle control method, characterized in that: The following steps are involved: S1. Obtaining traffic information of new energy vehicles, the traffic information including the position and speed information of the vehicle in the vehicle queue; S2. Construct a vehicle queue model, predict the speed at the next moment based on the vehicle queue model, and output the optimal speed prediction result; S3. Establish a SAC energy management model based on the expert knowledge base, input the optimal speed prediction result into the SAC energy management model to obtain the engine demand power, and determine the engine torque and speed as well as the motor torque and speed based on the engine demand power, and control the new energy vehicle according to the engine torque and speed as well as the motor torque and speed.
2. The power split new energy vehicle control method according to claim 1, characterized in that: Construct a vehicle queue model, and predict the speed at the next moment based on the vehicle queue model, and output the optimal speed prediction results, including: Construct a nonlinear vehicle platooning model: Where: s i (t), ν i (t) are the displacement and velocity of vehicle i at time t, Δt is the time interval, m i is the mass of vehicle i, η i is the mechanical efficiency of the transmission system of vehicle i, r i is the wheel radius, τ i is the time delay constant of vehicle i, C D is the air resistance coefficient, A i is the frontal area of vehicle i, g is the acceleration due to gravity, T i (t) is the actual torque of vehicle i at time t, u i (t) is the expected required torque of vehicle i at time t; Simplify the nonlinear vehicle platoon model: The state variable of vehicle i is defined as The output variable is defined as in: in: Define the state matrix X(t), control matrix U(t), and output matrix Y(t): in: Γ=diag{γ,γ,...,γ}∈R 2N×3N ; In a vehicle queue, the expected distance between adjacent vehicles is d0; the displacement and speed of the leading vehicle are s0(t) and v0(t) respectively. The expected state and expected output of any vehicle in the queue are expressed as: Where: p i,des (t) = p0(t)-i·d0,v i,des (t) = v0, T i,des (t) = h i (v0), h i (v0) is the torque required to maintain the vehicle at a constant speed, r i is the wheel radius; The information topology structure in the vehicle queue is represented by a directed graph G. N = {v N ,ε N } and the augmented directed graph G N+1 = {V N+1 ,E N+1 } to indicate, The set of vehicles that any vehicle i∈N in the queue can obtain information is N i ={i1,i2,…,i m } and the reachable set of the pilot car is: p i =1 means vehicle i can obtain the information of the pilot vehicle, otherwise, p i =0; definition and are the predicted state and the assumed state respectively, and They are the predicted control input sequence, the optimal predicted control input sequence and the assumed control input sequence. and is the predicted output sequence and the hypothesized output sequence, where k∈[0,1,2,…,N p -1,N p ]; Construct NDMPC optimization problem model: |I i | represents the set I i =N i ∪P i The base number, represents the distance between vehicle i and vehicle j, and the objective function l i for: Where: Q i , R i 、F i , G i are weight matrices, and they are all positive semi-definite matrices.
3. The power split new energy vehicle control method according to claim 1, characterized in that: Step S3 specifically includes: Building an expert knowledge base, the expert knowledge base including an optimal fuel economy curve of an engine; Construct a SAC energy management network, and the reward function of the SAC energy management network is: Among them: Reward i is the reward value of vehicle i, β is the weight coefficient of fuel consumption, t0 is the starting time, ε is the weight coefficient of SOC balance, is the instantaneous fuel consumption rate of the engine; The constraints of the reward function are: Where: T e (t) represents the engine torque, ω e,min is the minimum speed of the engine, T MG1,min Represents the minimum torque of the generator, T MG2,min Represents the minimum torque of the motor, ω MG1 (t) represents the speed of the generator, ω MG2 (t) represents the speed of the motor, P batt (t) represents the battery charge and discharge cycle output power; Define the state variables and action space of the SAC energy management network: The state variable is: State i = {ν i (t),a i (t),SOC i (t)}; The action space is: Action i = {P ei (t)}; Action i is the action space of vehicle i, P ei (t) is the engine power of vehicle i at time t; Construct the objective function: Where: r is the action value, s t is the state at time t, a t is the action at time t, ω is the weight coefficient, H(π(·|s t )) is the policy entropy value, π is the policy function; The action value function Q(s) of the policy π t ,a t ) and value function V(s t ): Where: r t is the reward function, γ represents the learning rate; In the value network, the value network parameter ψ is updated by minimizing the following loss function through the policy gradient descent method: In the action-value network, the network parameters are updated as follows: Where: Q′(s t ,a t ) is the target Q value, ψ′ is the target value function network parameter, and is updated using a soft update method as follows: ψ'=τψ+(1-τ)ψ'; τ is the soft update weight coefficient; In the policy network, the objective function for parameter update is as follows: The optimal speed is input into the trained SAC energy management network based on the expert knowledge base to output the optimal required power of the engine.
4. The power split new energy vehicle control method according to claim 1, characterized in that: Determining the engine torque and motor torque based on the engine demand power specifically includes: According to the optimal required power, the power-torque correspondence table and the torque-speed correspondence table are searched to obtain the optimal torque and speed of the engine; Model a planetary gear set: Where: MG1 is the speed of the generator, ω MG2 is the motor speed, ω eng is the engine speed, T eng is the engine torque, T MG1 is the torque of the generator, T MG2 is the torque of the motor; α is the ratio of the number of teeth of the ring gear to the number of teeth of the sun gear in the planetary gear set; g f is the main reducer transmission ratio, v * is the optimal speed. Substitute the optimal torque and speed of the engine into the planetary gear set model to obtain the torque and speed of the motor.
5. The power split new energy vehicle control method according to claim 3 is characterized in that: The state of charge of the new energy vehicle i is determined as follows: Where: V oc,i (t) is the open circuit voltage at time t, R 0,i is the internal resistance, I i (t) is the battery current at time t, P batt,i (t) is the charge-discharge cycle output power at time t, SOC i (t) is the battery state of charge at time t, Q 0,i is the initial battery capacity, Q i is the nominal capacity of the battery.
