Hybrid electric vehicle energy and motor heat management strategy optimization method

Abstract

The invention discloses a hybrid electric vehicle energy and motor heat management strategy optimization method, and relates to a hybrid electric vehicle technology. The method comprises the following steps that S11, state information of the hybrid power vehicle is obtained, and a model of the vehicle is built; s12, extracting a state space and an action space according to the state information of the hybrid vehicle in the driving process; s13, according to the vehicle model built in the S11, the fuel economy, the motor temperature and the battery SOC of the hybrid electric vehicle are used as optimization item construction, an integrated energy and motor heat management reward function based on a PPO-ER algorithm interacts with the environment in the state space and the action space extracted in the S12, and off-line training is conducted through driving data; and S14, the inheritable network parameters obtained in the offline mode are updated to a vehicle control unit of the hybrid electric vehicle. According to the invention, collaborative optimization of energy management and motor thermal management control under one algorithm architecture is realized, and compared with a traditional decoupling independent architecture, the development difficulty of each system is reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of hybrid vehicles, and in particular to a method for optimizing energy and motor thermal management strategies of hybrid vehicles. Background Art

[0002] Global warming, air pollution, and energy crises have brought great uncertainty to the earth's environment, and the use and emissions of fossil fuels in traditional internal combustion engine vehicles have attracted widespread social concern. The shift to electric vehicles has become a major and necessary shift towards increasingly stringent emission regulations and consumers' growing demand for environmentally friendly transportation options. Among the various types of electric vehicles, hybrid electric vehicles are considered a promising and efficient solution that can extend driving range, reduce tailpipe emissions, and improve fuel economy.

[0003] The characteristic of hybrid vehicles is that the hybrid system includes at least two energy sources, usually fuel-based primary energy and electric auxiliary energy. The purpose of the energy management strategy is that hybrid vehicles can select different operating modes according to different power requirements to achieve efficient cooperation of multiple energy sources. The motor thermal management strategy is to achieve efficient operation of the motor by dynamically controlling the motor temperature within the optimal operating range while preventing motor aging.

[0004] Currently, real-time control strategies can only optimize motor temperature by controlling motor power. By imposing rule-based physical constraints on the motor, the potential of hybrid vehicles in fuel economy and vehicle dynamics is greatly limited, making it difficult to achieve global optimality.

[0005] In order to achieve the global optimal performance of energy management strategy and motor thermal management strategy, integrating energy management and motor thermal management into one model and developing an integrated energy and motor thermal management strategy optimization method based on deep reinforcement learning has become a very promising solution. At the same time, it is of great significance for new energy vehicles to achieve further energy conservation and emission reduction and prevent motor degradation. Summary of the invention

[0006] In view of this, the present invention aims to propose a hybrid vehicle energy and motor thermal management strategy optimization method to solve the problems in the prior art.

[0007] To achieve the above object, the technical solution of the present invention is achieved as follows:

[0008] The present invention discloses a hybrid vehicle energy and motor thermal management strategy optimization method, the method comprising the following steps:

[0009] S11, obtaining status information of the hybrid vehicle and building a vehicle model;

[0010] S12, extracting a state space and an action space according to state information of the hybrid vehicle during driving;

[0011] S13, based on the vehicle model built in S11, takes the fuel economy, motor temperature and battery SOC of the hybrid vehicle as optimization items, and interacts with the environment in the state space and action space extracted in S12 based on the integrated energy and motor thermal management reward function of the PPO-ER algorithm, and uses driving data for offline training;

[0012] S14, updating the inheritable network parameters obtained offline to the vehicle controller of the hybrid electric vehicle, so as to realize the multi-objective optimization of the integrated energy and motor thermal management strategy of the plug-in hybrid electric vehicle.

[0013] Furthermore, in S11, the state information of the hybrid vehicle obtained includes vehicle mass m, gravity acceleration g, wheel rolling resistance coefficient f, road slope θ, air density ρ a , frontal area A, air resistance coefficient C d , vehicle speed v, effective wheel radius r, mass coefficient δ caused by the inertia moment of the effective wheel and powertrain components.

[0014] Furthermore, the vehicle model in S11 includes a longitudinal dynamics model, an engine model, a motor thermodynamics model, a motor thermal management model, and a power battery equivalent circuit model.

[0015] Furthermore, the longitudinal dynamic model is:

[0016]

[0017] Among them, P req is the driving power required by the hybrid vehicle, m is the vehicle mass, g is the acceleration of gravity, f is the wheel rolling resistance coefficient, θ is the road slope, ρ a is the air density, A is the windward area, C d is the air resistance coefficient C d , v is the vehicle speed, r is the effective wheel radius, δ is the mass coefficient caused by the inertia moment of the effective wheel and powertrain components, and a is the vehicle acceleration;

[0018] The engine model is:

[0019]

[0020] Among them, m fuel is the fuel consumption rate per unit time, T eng is the engine torque, W eng Engine speed, b eng It is the engine fuel rate status;

[0021] The thermodynamic model of the motor is:

[0022]

[0023] Among them, C th is the equivalent specific heat capacity of the motor stator winding, H th is the motor equivalent operation heat dissipation coefficient, T m is the temperature rise of the motor above the ambient temperature, The motor temperature rises above the ambient temperature T m Differentiate with respect to time t, I m is the motor equivalent current, R m is the motor equivalent resistance;

[0024] The motor thermal management model is:

[0025]

[0026] Among them, Q fan is the heat dissipated by the motor cooling system, h is the convection heat transfer coefficient, ρ is the fluid density, c p is the specific heat capacity of the fluid, ΔT is the temperature change of the motor, P fan is the power of the motor air cooling system;

[0027] The battery equivalent circuit model is:

[0028]

[0029] Among them, P batt is the output power in the charge and discharge cycle, V OC is the open circuit voltage, R0 is the internal resistance of the battery, I batt is the battery current, Q0 is the initial capacity of the battery, Q batt The nominal capacity of the battery, SOC is the battery state of charge, t is the working time, and dt is the differential of time t;

[0030] The constraints of the vehicle model are:

[0031]

[0032] in, are the upper and lower limits of engine speed, are the upper and lower limits of engine torque, are the upper and lower limits of the motor speed, is the upper and lower limits of the motor torque, SOC max , SOC min are the upper and lower limits of the power battery charge state value, are the upper and lower limits of the power battery current, The upper and lower limits of the power battery.

[0033] Further, in S12, the state space S is:

[0034] S={SOC,T,P fan ,v,T mot ,W mot ,T eng ,W eng}

[0035] Among them, S is the state space set, SOC is the state of charge of the power battery, T is the motor temperature, P is fan is the air cooling system power, v is the driving speed, T mot is the motor torque, W mot is the motor speed, T eng is the engine torque, W eng is the engine speed.

[0036] Further, in S12, the action space A is:

[0037] A={P fan ,T eng ,W eng}

[0038] Among them, A is the action space set, and P in the action space fan Indicates the working power of the air cooling system, T eng Indicates the engine working torque, W eng is the engine operating speed.

[0039] Further, in S13, the integrated energy and motor thermal management reward function is as follows:

[0040] r=―[β1(eff) 2 +β2|SOC―SOC tar |+ω(T―T0)]

[0041] Among them, r is the total reward value, eff is the fuel efficiency of the vehicle at the time of collection, SOC is the battery state of charge at the current corresponding time t, SOC tar is the target state of charge of the power battery, T is the motor temperature at the current corresponding time t, T0 is the initial temperature of the motor, β1, β2, ω are the corresponding reward coefficients respectively.

[0042] Furthermore, in S13, the offline training includes the following steps:

[0043] S131, initialize the Actor network, Critic network and target network;

[0044] S132, collect the trajectory D = {s0, a0, ..., a T―1 ,s T}, where D is the trajectory set, {s0,a0,…,a T―1 ,s T} respectively represent the state and action at the time step {0,…,T―1,T}, T is the total running time. Under the given trajectory collected, the Critic network outputs the advantage value, which is obtained by the following formula:

[0045]

[0046] in, is the advantage value, T is the total time of the environment interaction, t is the time index in [0,T], r t ,r t+1 ,…,r T―1 is the reward value at each time step t, t+1,…,T―1, γ is the discount factor, V is the current value function, V(s t ) means in state s t The expected total reward of continuing to perform actions according to the strategy is V(s T ) means in state s T The expected total reward of continuing to perform actions according to the strategy;

[0047] S133, according to the advantage value in S132, the objective function L(θ) of the strategy is obtained as:

[0048]

[0049] in, The empirical expectation at time step t, π θ and π θold They are the new strategy and old strategy of the PPO-ER algorithm, is the estimated value of the advantage function, clip is to maintain the probability ratio within the interval [1-ε, 1+ε] for truncation, min means to find the minimum value, ε is the truncation coefficient, which is used to control the degree to which the new strategy is allowed to deviate from the old strategy;

[0050] S134, using stochastic gradient ascent method to update network parameters, obtained by the following formula:

[0051] θ k+1 =argmaxE[L(θ)]

[0052]

[0053] Among them, θ k+1is the strategy parameter after the k+1th iteration of the Actor network, argmax means finding the network parameter with the maximum value, E[K(θ)] represents the expected value of the objective function, L(θ) is the objective function in S133, is the policy parameter of the Critic network after the k+1th iteration, argmin means finding the network parameter with the minimum value, E[V(s t )―R t ] represents the expected value of the error, V(s t ) means in state s t The cumulative reward for continuing to perform actions according to the strategy, R t Indicates the real cumulative reward;

[0054] S135, repeat S132-S134, when the total cumulative reward function is in a stable maximization convergence state, the training ends, and then save and download the neural network parameters.

[0055] Furthermore, in S134, entropy regularization is introduced, β is the entropy regularization coefficient, H(θ) is the entropy regularization term formula:

[0056]

[0057] Among them, s t is the state, a t To select an action, For all possible states s t and action a t The summation represents the expected value calculation, P(a t |s t ) means in state s t Next select action a t The probability of , log represents the logarithmic operation.

[0058] Compared with the prior art, the present invention has the following advantages:

[0059] In the present invention, energy management and motor thermal management control are implemented in a coordinated optimization under an algorithm framework, which reduces the development difficulty of each system compared with the traditional decoupled independent architecture. In addition, the present invention implements active thermal management of the motor through a cooling system, which improves the motor temperature control effect compared with the traditional physical rule constraint to limit the motor operating power, and effectively prevents motor degradation. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] The accompanying drawings constituting a part of the present invention are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the accompanying drawings:

[0061] Figure 1 It is the overall algorithm framework structure diagram of the present invention;

[0062] Figure 2 A structural diagram of a hybrid vehicle of the present invention;

[0063] Figure 3 The battery characteristic curve diagram of the vehicle power battery of the present invention;

[0064] Figure 4 The PPO-ER algorithm flow of the present invention; DETAILED DESCRIPTION

[0065] It should be noted that, in the absence of conflict, the embodiments of the present invention and the features in the embodiments may be combined with each other.

[0066] In the description of the present invention, it should be noted that the directions or positional relationships indicated by the terms "upper", "lower", "inner", "back", etc. are based on the directions or positional relationships shown in the drawings, and are 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 direction, be constructed and operated in a specific direction, and therefore cannot be understood as limiting the present invention. In addition, the terms "first" and "second" are only used for descriptive purposes and cannot be understood as indicating or implying relative importance.

[0067] In addition, in the description of the present invention, unless otherwise clearly defined, the terms "installed", "connected", "connected", and "connector" should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be an indirect connection through an intermediate medium, or it can be the internal communication of two components. For those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood in combination with specific circumstances.

[0068] The following will refer to the attached Figures 1 to 4 The present invention is described in detail with reference to embodiments.

[0069] In general, the present invention proposes a hybrid vehicle energy and motor thermal management strategy optimization method, such as Figure 1 and Figure 2 As shown, the method comprises the following steps:

[0070] S11, obtaining the status information of the hybrid vehicle and building a vehicle model, the vehicle model includes a longitudinal dynamics model, an engine model, a motor thermodynamics model, a motor thermal management model, and a power battery equivalent circuit model. ;

[0071] S12, extracting a state space and an action space according to state information of the hybrid vehicle during driving;

[0072] S13, based on the vehicle model built in S11, takes the fuel economy, motor temperature and battery SOC of the hybrid vehicle as optimization items, and interacts with the environment in the state space and action space extracted in S12 based on the integrated energy and motor thermal management reward function of the PPO-ER algorithm, and uses driving data for offline training;

[0073] S14, updating the inheritable network parameters obtained offline to the vehicle controller of the hybrid electric vehicle, so as to realize the multi-objective optimization of the integrated energy and motor thermal management strategy of the plug-in hybrid electric vehicle.

[0074] In S11, the state information of the hybrid vehicle is obtained, including the vehicle mass m, gravity acceleration g, wheel rolling resistance coefficient f, road slope θ, air density ρ a , frontal area A, air resistance coefficient C d , vehicle speed v, effective wheel radius r, mass coefficient δ caused by the inertia moment of the effective wheel and powertrain components.

[0075] Based on the above settings, the longitudinal dynamics model is:

[0076]

[0077] Among them, P reg is the driving power required by the hybrid vehicle, m is the vehicle mass, g is the acceleration of gravity, f is the wheel rolling resistance coefficient, θ is the road slope, ρ a is the air density, A is the windward area, C d is the air resistance coefficient C d , v is the vehicle speed, r is the effective wheel radius, δ is the mass coefficient caused by the inertia moment of the effective wheel and powertrain components, and a is the vehicle acceleration.

[0078] The engine model is:

[0079]

[0080] Among them, m fuel is the fuel consumption rate per unit time, T eng is the engine torque, W eng Engine speed, b eng It is the engine fuel rate status.

[0081] The thermodynamic model of the motor is:

[0082]

[0083] Among them, C th is the equivalent specific heat capacity of the motor stator winding, Hth is the motor equivalent operation heat dissipation coefficient, T m is the temperature rise of the motor above the ambient temperature, The motor temperature rises above the ambient temperature T m Differentiate with respect to time t, I m is the motor equivalent current, R m is the motor equivalent resistance.

[0084] The motor thermal management model is:

[0085]

[0086] Among them, Q fan is the heat dissipated by the motor cooling system, h is the convection heat transfer coefficient, ρ is the fluid density, c p is the specific heat capacity of the fluid, ΔT is the temperature change of the motor, P fan is the power of the motor air cooling system.

[0087] The battery equivalent circuit model is:

[0088]

[0089] Among them, P batt is the output power in the charge and discharge cycle, V OC is the open circuit voltage, R0 is the internal resistance of the battery, I batt is the battery current, Q0 is the initial capacity of the battery, Q batt The battery is rated capacity, SOC is the battery state of charge, t is the working time, and dt is the differential of time t. Figure 3 As shown, according to the functional relationship between the internal resistance and open circuit voltage of the power battery and the SOC value, the relationship between the internal resistance and the SOC value in the charging state, the relationship between the internal resistance and the SOC value in the discharging state, and the relationship between the open circuit voltage and the SOC value are input, and a univariate linear interpolation fitting is performed, so that the above functional relationship can be used to solve the battery SOC value at any time and state.

[0090] The constraints of the above vehicle model are:

[0091]

[0092] in, are the upper and lower limits of engine speed, are the upper and lower limits of engine torque, are the upper and lower limits of the motor speed, is the upper and lower limits of the motor torque, SOC max , SOC min are the upper and lower limits of the power battery charge state value, are the upper and lower limits of the power battery current, The upper and lower limits of the power battery.

[0093] It should be noted that the state space S in S12 is:

[0094] S={SOC,T,P fan ,v,T mot ,W mot ,T eng ,W eng}

[0095] Among them, S is the state space set, SOC is the state of charge of the power battery, T is the motor temperature, P is fan is the air cooling system power, v is the driving speed, T mot is the motor torque, W mot is the motor speed, T eng is the engine torque, W eng is the engine speed.

[0096] Further, in S12, the action space A is:

[0097] A={P fan ,T eng ,W eng}

[0098] Among them, A is the action space set, and P in the action space fan Indicates the working power of the air cooling system, T eng Indicates the engine working torque, W eng is the engine operating speed.

[0099] In S13, the integrated energy and motor thermal management reward function is as follows:

[0100] r=―[β1(eff) 2 +β2|SOC―SOC tar |+ω(T―T0)]

[0101] Among them, r is the total reward value, eff is the fuel efficiency of the vehicle at the time of collection, SOC is the battery state of charge at the current corresponding time t, SOC tar is the target state of charge of the power battery, T is the motor temperature at the current corresponding time t, T0 is the initial temperature of the motor, β1, β2, ω are the corresponding reward coefficients respectively.

[0102] It should be further explained that in S13, the vehicle data is driving data on vehicle speed v and time t collected under actual driving conditions.

[0103] It should be understood that the goal of the PPO-ER algorithm is to achieve stable policy updates by limiting the update amplitude of the policy, preventing the model from undergoing drastic policy changes during training. At the same time, through the exploration and learning of the entropy regularization balance strategy, the strategy π is continuously improved to maximize the reward, thereby obtaining the best strategy π. * , where the policy π represents a series of actions taken by the agent.

[0104] Specifically, assume that the finite state set of the agent is S = {s1, s2, ..., s n}, the finite action sequence is A = {a1, a2, ..., a n}, when the current state of the agent is s t ∈S, take action a t ∈A, the state of the agent is transferred to a new state s under the influence of the environment. t+1 ∈S, the immediate reward generated is r(s t ,a t ).

[0105] Based on the above settings, in this embodiment, in S13, if Figure 4 As shown, offline training includes the following steps:

[0106] S131, initialize the Actor network, Critic network and target network; the goal of the Actor network is to approximate the optimal strategy π θ (a|s). For a continuous action space, the output is the mean and standard deviation of the action, and for a discrete action space, the output is the probability distribution of the action. The goal of the Critic network is to approximate the state value function V φ (s), the output is a scalar value representing the estimated value of a given state.

[0107] Define a storage space R as the experience replay pool, set the capacity to N; set the maximum number of agent loops M.

[0108] S132, information T generated by the interaction between the agent and the environment t =(s t , a t , r t ,s t+1 ) in the form of experience replay pool, and store T t The sampling probability P(t) is used as the training data set for the Actor and Critic networks.

[0109] Advantage function It is used to measure the pros and cons of an action a in state s. The generalized advantage estimate is used to calculate the advantage function to reduce the variance. By using the policy network environment interaction to collect the trajectory D = {s0, a0, ..., a T―1 ,s T}. Where D is the trajectory set, {s0,a0,…,a T―1 ,s T} respectively represent the state and action at the time step {0,…,T―1,T}, T is the total running time. Under the given trajectory collected, the Critic network outputs the advantage value, which is obtained by the following formula:

[0110]

[0111] in, is the advantage value, T is the total time of the environment interaction, t is the time index in [0,T], r t ,r t+1 ,…,r T―1 is the reward value at each time step t, t+1,…,T―1, γ is the discount factor, V is the current value function, V(s t ) means in state s t The expected total reward of continuing to perform actions according to the strategy is V(s T ) means in state s T The expected total reward of continuing to perform actions according to the strategy;

[0112] S133, through the priority experience replay technology, sampling is performed from the experience replay pool R according to the sampling probability to obtain a small batch of samples S, and the Actor and Critic networks are trained.

[0113] According to the advantage value in S132, the objective function L(θ) of the strategy is obtained as:

[0114]

[0115] in, The empirical expectation at time step t, π θ and π θold They are the new strategy and old strategy of the PPO-ER algorithm, is the estimated value of the advantage function, clip is to maintain the probability ratio within the interval [1-ε, 1+ε] for truncation, min means to find the minimum value, ε is the truncation coefficient, which is used to control the degree to which the new strategy is allowed to deviate from the old strategy;

[0116] In order to improve the exploration ability of the strategy, entropy regularization is introduced, β is the entropy regularization coefficient, H(θ) is the entropy regularization term expressed as:

[0117]

[0118] Among them, s t is the state, a t To select an action, For all possible states s t and action a t The summation represents the expected value calculation, P(a t |s t ) means in state s t Next select action a t The probability of , log represents the logarithmic operation.

[0119] S134, use stochastic gradient ascent to update the Actor network parameters, the gradient of the Actor network parameters θ, and use the optimizer to perform a one-step gradient update to maximize the objective of the Actor network:

[0120] θ k+1 =argmaxE[L(θ)]

[0121] Among them, θ k+1 is the policy parameter after the k+1th iteration of the Actor network, argmax represents the search for the network parameter with the maximum value, E[L(θ)] represents the expected value of the objective function, and L(θ) is the objective function in S133. In other words, the parameter update method of the stochastic gradient ascent method is:

[0122]

[0123] Among them, θ k+1 is the strategy parameter after the k+1th iteration of the Actor network, θ k is the strategy parameter after the kth iteration of the Actor network, α is the learning rate, Represents the gradient of the objective function L(θ) with respect to the policy parameter θ. The goal of the update is to maximize the expected value of the loss function, so we choose the Actor network parameter θ that can maximize L(θ).

[0124] Update the Critic network parameters using the gradient descent method:

[0125]

[0126] is the policy parameter of the Critic network after the k+1th iteration, argmin means finding the network parameter with the minimum value, E[V(s t )―R t ] represents the expected value of the error, V(s t ) means in state s tThe cumulative reward for continuing to perform actions according to the strategy, R t Represents the actual cumulative reward. It should be further explained that the parameter update method of the stochastic gradient descent method is

[0127]

[0128] in, is the strategy parameter after the k+1th iteration of the Critic network, is the strategy parameter after the kth iteration of the Actor network, α is the learning rate, Represents the objective function For strategy parameters The goal of updating is to maximize the expected value of the loss function, so we choose Critic network parameters that reach the maximum value

[0129] S135, repeat S132-S134, when the total cumulative reward function is in a stable maximization convergence state, the training ends, and then save and download the neural network parameters.

[0130] In the present invention, the network parameters obtained by offline training are downloaded to the vehicle controller of the plug-in hybrid electric vehicle and applied online in real time, thereby effectively improving fuel economy and preventing motor degradation.

[0131] In summary, in the present invention, energy management and motor thermal management control are implemented in a coordinated optimization under an algorithm architecture, which reduces the development difficulty of each system compared to the traditional decoupled independent architecture. In addition, the present invention implements active thermal management of the motor through a cooling system, which improves the motor temperature control effect compared to the traditional physical rule constraint to limit the motor operating power, and effectively prevents motor degradation.

[0132] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A hybrid vehicle energy and motor thermal management strategy optimization method, characterized by: The method comprises the following steps: S11, obtaining status information of the hybrid vehicle and building a vehicle model; S12, extracting a state space and an action space according to state information of the hybrid vehicle during driving; S13, based on the vehicle model built in S11, takes the fuel economy, motor temperature and battery SOC of the hybrid vehicle as optimization items, and interacts with the environment in the state space and action space extracted in S12 based on the integrated energy and motor thermal management reward function of the PPO-ER algorithm, and uses driving data for offline training; S14, updating the inheritable network parameters obtained offline to the vehicle controller of the hybrid electric vehicle, so as to realize the multi-objective optimization of the integrated energy and motor thermal management strategy of the plug-in hybrid electric vehicle.

2. A hybrid vehicle energy and motor thermal management strategy optimization method according to claim 1, characterized in that: In S11, the state information of the hybrid vehicle obtained includes vehicle mass m, gravity acceleration g, wheel rolling resistance coefficient f, road slope θ, air density ρ a , frontal area A, air resistance coefficient C d , vehicle speed v, effective wheel radius r, mass coefficient δ caused by the inertia moment of the effective wheel and powertrain components.

3. A hybrid vehicle energy and motor thermal management strategy optimization method according to claim 2, characterized in that: The vehicle model in S11 includes a longitudinal dynamics model, an engine model, a motor thermodynamics model, a motor thermal management model, and a power battery equivalent circuit model.

4. A hybrid vehicle energy and motor thermal management strategy optimization method according to claim 3, characterized in that: The longitudinal dynamics model is: Among them, P req is the driving power required by the hybrid vehicle, m is the vehicle mass, g is the acceleration of gravity, f is the wheel rolling resistance coefficient, θ is the road slope, ρ a is the air density, A is the windward area, C d is the air resistance coefficient C d , v is the vehicle speed, r is the effective wheel radius, δ is the mass coefficient caused by the inertia moment of the effective wheel and powertrain components, and a is the vehicle acceleration; The engine model is: Among them, m fue l is the fuel consumption rate per unit time, T eng is the engine torque, W eng Engine speed, b eng It is the engine fuel rate status; The thermodynamic model of the motor is: Among them, C th is the equivalent specific heat capacity of the motor stator winding, H th is the motor equivalent operation heat dissipation coefficient, T m is the temperature rise of the motor above the ambient temperature, The motor temperature rises above the ambient temperature T m Differentiate with respect to time t, I m is the motor equivalent current, R m is the motor equivalent resistance; The motor thermal management model is: Among them, Q f an is the heat dissipated by the motor cooling system, h is the convection heat transfer coefficient, ρ is the fluid density, c p is the specific heat capacity of the fluid, ΔT is the temperature change of the motor, P f an is the power of the motor air cooling system; The battery equivalent circuit model is: Among them, P batt is the output power in the charge and discharge cycle, V OC is the open circuit voltage, R0 is the internal resistance of the battery, I batt is the battery current, Q0 is the initial capacity of the battery, Q batt The nominal capacity of the battery, SOC is the battery state of charge, t is the working time, and dt is the differential of time t; The constraints of the vehicle model are: in, are the upper and lower limits of engine speed, are the upper and lower limits of engine torque, are the upper and lower limits of the motor speed, is the upper and lower limits of the motor torque, SOC max , SOC min are the upper and lower limits of the power battery charge state value, are the upper and lower limits of the power battery current, The upper and lower limits of the power battery.

5. A hybrid vehicle energy and motor thermal management strategy optimization method according to claim 4, characterized in that: In S12, the state space S is: S={SOC,T,P fan ,v,T mot ,W mot ,T eng ,W eng } Among them, S is the state space set, SOC is the state of charge of the power battery, T is the motor temperature, P is fan is the air cooling system power, v is the driving speed, T mot is the motor torque, W mot is the motor speed, T eng is the engine torque, W eng is the engine speed.

6. A hybrid vehicle energy and motor thermal management strategy optimization method according to claim 5, characterized in that: In S12, the action space A is: A={P fan ,T eng ,W eng } Among them, A is the action space set, and P in the action space fan Indicates the working power of the air cooling system, T eng Indicates the engine working torque, W eng is the engine operating speed.

7. The hybrid vehicle energy and motor thermal management strategy optimization method according to claim 1, characterized in that: In S13, the integrated energy and motor thermal management reward function is as follows: r=―[β1(eff) 2 +β2|SOC―SOC tar |+ω(T―T0)] Among them, r is the total reward value, eff is the fuel efficiency of the vehicle at the time of collection, SOC is the battery state of charge at the current corresponding time t, SOC tar is the target state of charge of the power battery, T is the motor temperature at the current corresponding time t, T0 is the initial temperature of the motor, β1, β2, ω are the corresponding reward coefficients respectively.

8. The hybrid vehicle energy and motor thermal management strategy optimization method according to claim 7, characterized in that: In S13, the offline training includes the following steps: S131, initialize the Actor network, Critic network and target network; S132, collect the trajectory D = {s0, a0, ..., a T―1 ,s T }, where D is the trajectory set, {s0,a0,…,a T―1 ,s T } respectively represent the state and action at the time step {0,…,T―1,T}, T is the total running time. Under the given trajectory collected, the Critic network outputs the advantage value, which is obtained by the following formula: in, is the advantage value, T is the total time of the environment interaction, t is the time index in [0,T], r t ,r t+1 ,…,r T―1 is the reward value at each time step t, t+1,…,T―1, γ is the discount factor, V is the current value function, V(s t ) means in state s t The expected total reward of continuing to perform actions according to the strategy is V(s T ) means in state s T The expected total reward of continuing to perform actions according to the strategy; S133, according to the advantage value in S132, the objective function L(θ) of the strategy is obtained as: in, The empirical expectation at time step t, π θ and π θold They are the new strategy and old strategy of the PPO-ER algorithm, is the estimated value of the advantage function, clip is to maintain the probability ratio within the interval [1-ε, 1+ε] for truncation, min means to find the minimum value, ε is the truncation coefficient, which is used to control the degree to which the new strategy is allowed to deviate from the old strategy; S134, using stochastic gradient ascent method to update network parameters, obtained by the following formula: <h2 style=";text-align:left;direction:ltr">θ<h2 style=";text-align:left;direction:ltr"> k+1 <h2 style=";text-align:left;direction:ltr"> =argmaxE[L(θ)] Among them, θ k+1 is the policy parameter after the k+1th iteration of the Actor network, argmax means finding the network parameter with the maximum value, E[L(θ)] represents the expected value of the objective function, L(θ) is the objective function in S133, is the policy parameter of the Critic network after the k+1th iteration, argmin means finding the network parameter with the minimum value, E[V(s t )―R t ] represents the expected value of the error, V(s t ) means in state s t The cumulative reward for continuing to perform actions according to the strategy, R t Indicates the real cumulative reward; S135, repeat S132-S134, when the total cumulative reward function is in a stable maximization convergence state, the training ends, and then save and download the neural network parameters.

9. A hybrid vehicle energy and motor thermal management strategy optimization method according to claim 8, characterized in that: In S134, entropy regularization is introduced, β is the entropy regularization coefficient, H(θ) is the entropy regularization term formula: Among them, s t is the state, a t To select an action, For all possible states s t and action a t The summation represents the expected value calculation, P(a t |s t ) means in state s t Next select action a t The probability of , log represents the logarithmic operation.

Images