Electric vehicle air conditioner and passenger cabin thermal management control method based on td3 algorithm

Abstract

The present application relates to a method for controlling the air conditioner and passenger cabin thermal management of an electric vehicle based on the TD3 algorithm, belonging to the field of vehicle thermal management technology, comprising the following steps: S1: establishing a thermal coupling model of the air conditioner and passenger cabin; S2: establishing a passenger thermal comfort evaluation calculation model; S3: designing the TD3 algorithm and training and verifying the agent; S4: combining the passenger thermal comfort evaluation model to establish a control strategy based on the TD3 algorithm that matches the thermal coupling model of the air conditioner and passenger cabin. The present application adopts a control strategy based on the TD3 algorithm of reinforcement learning, which is more suitable for complex and variable environmental control problems, making the control more energy-efficient, and ensuring the comfort and safety of passengers throughout the vehicle driving process.

Description

Technical Field

[0001] This invention belongs to the field of vehicle thermal management technology, and relates to a thermal management control method for electric vehicle air conditioning and passenger compartment based on the TD3 algorithm. Background Technology

[0002] The comfort and health of the passenger cabin environment in electric vehicles directly impacts the driving experience, which in turn reduces the risk of traffic accidents and improves driving safety. Air conditioning is essential for maintaining a comfortable temperature in the passenger cabin, and its energy consumption is a significant portion of an electric vehicle's energy consumption, greatly affecting its driving range: under the same operating conditions, turning on the air conditioning will drastically reduce the driving range. Therefore, a more precise and intelligent controller is needed to regulate the thermal balance of the passenger cabin.

[0003] Precise control of compressor speed is a prerequisite for ensuring accurate temperature control in the passenger compartment. Electric vehicle compressors primarily operate driven by electric motors, unaffected by vehicle speed or engine speed, resulting in flexible and precise control. In the field of control, various control methods are currently available. The mainstream control method for automotive air conditioning systems is rule-based switching controllers that use lookup tables, including PID control, fuzzy control, or combinations of PID and fuzzy control—some of the more traditional methods. These traditional control methods primarily rely on the feedback signal from the system, using the difference between the target value and the feedback value for control and adjustment.

[0004] While the aforementioned control methods can address the basic requirements for comfort control, they are not ideal for automotive systems where speed changes drastically and operating conditions are highly variable. Relying solely on traditional feedback signals often has limitations. For example, there is a time lag in signal feedback; the feedback signal is actually based on the operating conditions of the previous moment, while the operating conditions of the next moment may differ significantly from the previous moment. This leads to control actions based on feedback signals that do not match the current actual operating conditions. Such control is called ineffective control. Due to the complexity of automotive operating conditions, this ineffective control will gradually accumulate and cause significant energy consumption. Summary of the Invention

[0005] In view of this, the purpose of this invention is to provide an electric vehicle air conditioning and passenger cabin thermal management control method based on the TD3 algorithm, with the goal of comfortable temperature control and energy saving. On the basis of more precise balance between electric vehicle air conditioning and energy consumption control, it can better adapt to the complex operating conditions in actual electric vehicle driving and reduce the electrical energy consumed by control.

[0006] To achieve the above objectives, the present invention provides the following technical solution:

[0007] Includes the following steps:

[0008] S1. Establish a one-dimensional simulation model of the automotive air conditioning system based on the moving boundary method and the lumped parameter method. The automotive air conditioning system includes a compressor, condenser, evaporator and expansion valve. Establish a dynamic thermal model of the automotive passenger compartment system and couple it with the air conditioning system model to establish a thermal coupling model between the electric vehicle air conditioning and the passenger compartment.

[0009] S2. Establish a calculation model for passenger thermal comfort assessment;

[0010] S3. Design the TD3 algorithm;

[0011] S31. Introduce a target policy smoothing mechanism to the TD3 algorithm;

[0012] The TD3 algorithm is as follows:

[0013]

[0014] Where y1 represents right Optimization is performed, y2 represents right Optimize; θ i For Critic network parameters, These are the parameters of the Actor network;

[0015] The updated formula for the target policy is:

[0016]

[0017] Wherein, ∈~clip(N(0,σ) represents that the added noise follows a Gaussian distribution and the absolute value of the noise is less than or equal to the hyperparameter c;

[0018] S32. Design the agent for the TD3 algorithm; select state space elements, including the passenger cabin temperature T. cab The absolute value of the temperature difference from the target temperature, the PMV value, and the air conditioning energy consumption; air conditioning energy consumption includes compressor energy consumption E. AC and heat exchanger fan energy consumption T evap,wall ;

[0019] State={|T cab -25℃ | PMV, E AC ,T evap,wall}

[0020] Compressor speed and heat exchanger fan speed are selected as the control actions of the intelligent agent:

[0021] action = {N comp N fan}

[0022] The reward function is designed by combining main reward and auxiliary reward, with PMV, an important indicator of human thermal comfort, and air conditioning energy consumption designed as the main reward.

[0023] S33. Set the agent parameters of the TD3 algorithm and perform reinforcement learning training. When the passenger cabin temperature exceeds the preset high temperature threshold or falls below the preset low temperature threshold, the reinforcement learning training stops.

[0024] S4. Conduct agent training and verification to establish a control strategy that matches the thermal coupling model of the car air conditioning and passenger compartment.

[0025] Furthermore, in S1, the one-dimensional simulation model of the automotive air conditioning system includes:

[0026] S11. Establish a one-dimensional dynamic model of the refrigerant inside the compressor, as follows:

[0027]

[0028]

[0029] η v =f1(N comp ,P c / P e )

[0030] η is =f2(N comp ,P c / P e )

[0031] Where, η v ρ represents the volumetric efficiency of the compressor. r N represents the density of the refrigerant at the compressor inlet. comp V represents the compressor's rotational speed; d This represents the compressor's displacement. (h) c,o h represents the specific enthalpy at the compressor outlet. c,i h represents the specific enthalpy at the compressor inlet. is,o η represents the outlet specific enthalpy of the refrigerant under isentropic compression. is f1 represents the isentropic efficiency of the compressor under isentropic compression conditions; f2 represents the fitting function of the compressor's volumetric efficiency and isentropic efficiency to the compressor's inlet pressure difference; f3 represents the fitting function of the compressor's volumetric efficiency and isentropic efficiency to the compressor's outlet pressure difference; P represents... c / P e This represents the pressure ratio at the compressor inlet and outlet, where P c P represents the pressure of the condenser. e This represents the pressure in the evaporator;

[0032] S12. Evaporator modeling: The change in the two-phase region length le of the evaporator is obtained by the following formula:

[0033]

[0034] Where, ρ le h represents the density of the refrigerant. lge A represents the latent heat absorbed by the refrigerant during a phase change. e This represents the cross-sectional area of ​​the evaporator flat tube structure. This represents the average vapor proportion of the refrigerant during the two-phase stage of the evaporator. The refrigerant mass flow rate at the evaporator inlet, h ge h represents the enthalpy of the refrigerant at the evaporator outlet. ie This represents the enthalpy of the refrigerant at the evaporator inlet, a. ie D represents the heat transfer coefficient between the evaporator wall and the refrigerant. ie T represents the inner diameter of the evaporator tube. we T represents the temperature of the evaporator wall. re This represents the refrigerant temperature under the current pressure condition;

[0035] The density of refrigerant vapor in the evaporator is expressed by the following formula:

[0036]

[0037] Among them, L e ρ represents the total length of all the flat tubes in the evaporator. ge P represents the density of a saturated refrigerant in its gaseous state. e This represents the current pressure of the refrigerant. The refrigerant mass flow rate at the evaporator outlet and the temperature change at the evaporator wall are:

[0038]

[0039] a oe =f p2 (N fan )

[0040]

[0041] Among them, C p The specific heat capacity of the evaporator is represented by m, and the mass of the evaporator is represented by a. ish T represents the heat transfer coefficient between the refrigerant and the wall during the evaporator superheating stage. ie Represents the inlet refrigerant temperature, a oe f represents the heat transfer coefficient of the evaporator on the air side, where f p2 N fanLet A represent the polynomial fitting formula and the evaporator fan speed, respectively. oe T represents the air-facing area of ​​the evaporator. ae This represents the ambient air temperature around the evaporator inlet. C represents the air mass flow rate on the air side. p,air,mix T represents the specific heat capacity of air. air,mix This represents the temperature of the air entering the evaporator after the mixing damper has been activated. This represents the previous air quality flow rate in the passenger cabin. C represents the new air mass flow rate in the external environment. p,air,cab and C p,air,amb These are the specific heats of air corresponding to the two; and The sum of these is the total air mass flow rate after mixing.

[0042] The calculation method is as follows:

[0043]

[0044] Where, γ cycle ρ represents the proportion of old air in the mixed airflow. air,cab ρ represents the density of the old wind. air,amb V represents the density of fresh air. air This represents the total air intake of the mixing damper, and its size is affected by the fan speed.

[0045] S13. Condenser Modeling:

[0046]

[0047] Assuming no refrigerant leakage in the entire refrigeration cycle piping of the air conditioning system, the total mass of refrigerant in the system remains constant. Therefore, the total mass of refrigerant in the evaporator and condenser can be considered constant, hence:

[0048]

[0049] Where, ρ lc h is the density of the liquid refrigerant in the condenser. lgc A is the latent heat of vaporization of the refrigerant in the condenser. c This represents the total cross-sectional area of ​​the flat tube microchannels in the condenser. h is the average porosity of the two-phase region of the condenser. gc h lc and h ic These represent the specific enthalpy values ​​of the gas, liquid, and inlet refrigerant in the condenser at the current pressure, respectively. icD is the heat transfer coefficient between the condenser inner wall and the refrigerant in the two-phase region. ic The diameter of the inside of the condenser flat tube, T wc T is the condenser wall temperature. rc It is the saturation temperature of the refrigerant at the current pressure of the condenser, L. c It is the total length of the condenser flat tube, (C p m) wc a represents the specific heat and mass of the condenser material. oc It is the heat transfer coefficient between air and the condenser wall, A oc T represents the frontal area of ​​the condenser. ac It is the current ambient temperature around the condenser, and ∑ represents a constant.

[0050] While the car is moving, the external wind speed of the condenser is affected by the vehicle speed. oc The relationship with vehicle speed is as follows:

[0051] a oc =f p2 (V car )

[0052] Among them, V car Vehicle speed is the disturbance input controlled solely by the driver.

[0053] S14. Expansion valve modeling: During the dynamic process, the refrigerant mass flow rate through the expansion valve... The relationship between it and the pressure drop ΔP of its expansion valve is:

[0054]

[0055] Among them, C q ρ is the flow coefficient of the expansion valve. v Let A be the refrigerant density at the inlet of the expansion valve. v ΔP is the flow area of ​​the expansion valve, and ΔP is the pressure drop, which is the pressure difference between the inlet and outlet of the expansion valve.

[0056] Furthermore, in S1, establishing a dynamic thermal model of the vehicle passenger compartment system and coupling it with the air conditioning system model specifically involves:

[0057] Total heat load of the car passenger compartment Represented as:

[0058]

[0059] During vehicle operation, heat exchange occurs between the vehicle cabin and the outside environment via convection. Subject to vehicle speed V car and ambient temperature T ac The input disturbances, which are unaffected by the controller and are not controlled by the controller, are present in the heat transfer model. Calculated by the following formula:

[0060]

[0061] Where T s Given the temperature of the structure surrounding the crew compartment, based on the law of conservation of energy, the dynamic temperature change of the surrounding structure is as follows:

[0062]

[0063] The dynamic changes in air temperature in the crew cabin are represented as follows:

[0064]

[0065] in, For heat exchange of the vehicle body surface structure, For solar radiation heat load, The heat load caused by ventilation For the human body's heat load, For mechanical and instrumentation heat load, T cab For the temperature of the passenger cabin, M represents the amount of cooling capacity delivered to the cabin by the air conditioning system per unit time. a For the air quality within the passenger cabin volume range, cp a For the specific heat of air, h o It is the heat transfer coefficient between the outer side of the external structure of the passenger compartment and the air side, which is determined by the vehicle speed. S is the total surface area of ​​the passenger compartment's outer surface structure, h i M is the heat transfer coefficient between the passenger cabin interior surfaces and the air. s and C ps These are the mass and specific heat of the enclosed structure surrounding the vehicle cabin.

[0066] Furthermore, the passenger thermal comfort assessment calculation model includes a PMV calculator, a learning regulator, and a T... comfort calculator;

[0067] The PMV calculator module is used to calculate the average rating of people's comfort temperature prediction. The formula for calculating PMV is as follows:

[0068] PMV = (0.303e -0.036M +0.028)×(M-φ1-φ2-φ3-φ4-φ5-φ6)

[0069] Where M represents human metabolism, the metabolic rate of passengers is set to 1 and the metabolic rate of drivers is set to 1.5.

[0070] The specific calculation methods for Φ1 to Φ6 are as follows:

[0071] Φ1=3.05e -3 ·(5733-6.99MPw )

[0072]

[0073] Φ3=1.7e -5 M(5867-P w )

[0074] Φ4=1.4e -3 M(34-t cab )

[0075] Φ5=3.96e -8 f cl [(T cl +273) 4 -(T r +273) 4 ]

[0076] Φ6=f cl h c (T cl -T cab )

[0077] Among them, h c The calculation method is as follows:

[0078]

[0079] T cl =35.7-0.028MI cl (φ5+φ6)

[0080] Among them, P w T represents the partial pressure of water vapor. cab and T r h represents the average radiant temperature inside the passenger cabin and the average radiant temperature inside the cockpit, respectively. c T represents the heat transfer coefficient of the human body surface. cl I represents the surface temperature of a person's clothing. cl V represents the thermal resistance of clothing. a Represents the airflow velocity inside the vehicle;

[0081] The learning regulator analyzes past data on temperature regulation and calculates the average PMV after n manual temperature adjustments:

[0082]

[0083] Where i represents the adjustment sequence number of the manual adjustment, and N represents the PMV adaptive sampling period. When n exceeds N, the PMV... a Calculated based on the average of N PMV values;

[0084] The Tcomfort The calculator is used to calculate PMV a The value is used to calculate the target comfortable temperature T at this time. comfort .

[0085] I cl In summer, when temperatures are high, this value is generally 0.7; since humidity has a relatively small impact on thermal comfort calculations, the humidity is set to 50%.

[0086] Furthermore, in S32, the main storyline reward is:

[0087]

[0088] Where α and β are target weight coefficients, α being greater than β, reflecting the importance the reward function places on thermal comfort and energy efficiency, and m and n are proportionality coefficients;

[0089] The auxiliary line reward is:

[0090]

[0091] Where i, j, p, q are proportionality coefficients, and T is the proportionality coefficient. evap,wall For the energy consumption of the heat exchange fan, T cab For passenger cabin temperature, T target The target temperature.

[0092] Furthermore, the agent parameters include Gaussian action exploration noise model parameters, target network parameters, and training parameters; the Gaussian action exploration noise model parameters include standard deviation, decay rate, and minimum decay value; the target network parameters include update frequency and smoothing factor; and the training parameters include the number of experience pool samples, soft update parameters, discount factor, Actor network learning rate, Critic network learning rate, and minimum sample set sample size.

[0093] Furthermore, in S4, the agent training and verification specifically includes:

[0094] S41. Observe the changes in actual return, average return and predicted return at each step through the dynamic return curve;

[0095] S42. Select the agent with the highest actual return for training, and analyze it after simulation testing;

[0096] S43. Determine whether the agent training meets all the qualification criteria at the same time. If yes, the training is qualified; otherwise, the training is unqualified.

[0097] S44. Analyze the unqualified agent and related training results, and readjust the control action, observation state and reward function;

[0098] S45. Perform hyperparameter tuning and then retrain.

[0099] Furthermore, in S43, the qualification criteria include: the cumulative discount reward during training and whether there is a convergence trend; whether the control action changes; whether the PMV value can always remain within the human comfort range; and whether the passenger cabin temperature gradually decreases to the target temperature within a preset time.

[0100] The beneficial effects of this invention are as follows:

[0101] In this control scenario, a thermally coupled model of the automotive air conditioning system and passenger cabin was constructed as the training environment for a reinforcement learning algorithm. An intelligent control strategy for air conditioning and passenger cabin thermal management, based on the TD3 algorithm and combined with passenger thermal comfort calculation methods, was designed. In the design of the reinforcement learning agent, a reward function combining primary and secondary rewards was adopted. Through training, learning, and verification analysis, the TD3 control strategy can automatically adjust the passenger cabin temperature according to changes in the external environment and the PMV value calculated based on passenger thermal comfort to ensure human thermal comfort. This effectively protects the thermal comfort of both passengers and the driver, and achieves safe, energy-efficient, and healthy intelligent automatic adjustment of the air conditioning system, making the entire air conditioning control more human-centered and intelligent.

[0102] Specific advantages include:

[0103] (1) The TD3 algorithm was selected to effectively suppress the overestimation problem by using the Critic network to estimate the minimum value of the action value function; the network delay update was used to ensure the stability of the Actor network training; the added exploration noise effectively improved the convergence of the algorithm.

[0104] (2) A passenger thermal comfort assessment calculation model was established. Using this model, when the vehicle is in motion, the passenger's thermal preference temperature can be estimated and the passenger's thermal comfort level can be calculated in real time with relatively accurate accuracy, so that the intelligent agent can learn better control strategies more efficiently.

[0105] (3) The agent is trained by using a reward design that combines main reward and auxiliary reward. This effectively realizes the communication between the target and the algorithm, which is conducive to the learning of the agent's strategy. The agent's reward can quickly enter the convergence state after about 1000 training sessions.

[0106] Compared to existing feedback control algorithms, this invention combines the improved TD3 algorithm with a passenger thermal comfort assessment calculation model to more accurately balance the air conditioning and energy consumption control of electric vehicles, resulting in more intelligent and efficient control. This better meets the complex operating conditions in actual driving of electric vehicles, enhances the energy-saving control of the intelligent connected vehicle thermal management system, and reduces the number of times ineffective control occurs.

[0107] Other advantages, objectives, and features of the invention will be set forth in part in the description which follows, and in part will be apparent to those skilled in the art from the following examination, or may be learned from practice of the invention. The objectives and other advantages of the invention can be realized and obtained through the following description. Attached Figure Description

[0108] To make the objectives, technical solutions, and advantages of the present invention clearer, the preferred embodiments of the present invention will be described in detail below with reference to the accompanying drawings, wherein:

[0109] Figure 1 This is a simplified diagram of the overall control logic of the intelligent control strategy of the present invention;

[0110] Figure 2 A simplified structural diagram of the dynamic coupling thermal model of the automotive air conditioning and passenger cabin systems;

[0111] Figure 3 A calculation model for assessing passenger thermal comfort;

[0112] Figure 4 This is a schematic diagram of the TD3 algorithm.

[0113] Figure 5 A schematic diagram illustrating the training and validation process of the TD3 algorithm agent;

[0114] Figure 6 A simplified diagram of the intelligent control logic for air conditioning and passenger cabin thermal management of electric vehicles based on the TD3 algorithm;

[0115] Figure 7 This is a comparison chart of the energy consumption of the TD3 algorithm control strategy with traditional PID controllers and ON-OFF controllers. Detailed Implementation

[0116] The following specific examples illustrate the implementation of the present invention. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and various details in this specification can be modified or changed 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 are only schematic representations of the basic concept of the present invention. Unless otherwise specified, the following embodiments and features can be combined with each other.

[0117] The accompanying drawings are for illustrative purposes only and are schematic diagrams, not actual pictures. They should not be construed as limiting the invention. To better illustrate the embodiments of the invention, some parts in the drawings may be omitted, enlarged, or reduced, and do not represent the actual product dimensions. It is understandable to those skilled in the art that some well-known structures and their descriptions may be omitted in the drawings.

[0118] In the accompanying drawings of the embodiments of the present invention, the same or similar reference numerals correspond to the same or similar components. In the description of the present invention, it should be understood that if terms such as "upper," "lower," "left," "right," "front," and "rear" indicate the orientation or positional relationship based on the orientation or positional relationship shown in the drawings, they are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, the terms used to describe positional relationships in the drawings are only for illustrative purposes and should not be construed as limiting the present invention. For those skilled in the art, the specific meaning of the above terms can be understood according to the specific circumstances.

[0119] Please see Figures 1-6 This describes a control method for air conditioning and passenger cabin thermal management in electric vehicles based on the TD3 algorithm. The method specifically includes the following steps:

[0120] S1: Establish as follows Figure 2 The dynamic thermal coupling model of the automotive air conditioning and passenger compartment shown includes:

[0121] A one-dimensional simulation model of an automotive air conditioning system is established based on the moving boundary method and the lumped parameter method. The main components of the automotive air conditioning system include the compressor, condenser, evaporator and expansion valve.

[0122] The one-dimensional simulation thermal model of an electric vehicle air conditioning system includes:

[0123] S11. Establish a one-dimensional dynamic model of the refrigerant inside the compressor, as follows:

[0124]

[0125] ηv ρ represents the volumetric efficiency of the compressor. r N represents the density of the refrigerant at the compressor inlet. comp V represents the compressor's rotational speed, which is primarily controlled by the drive motor; d This represents the compressor's displacement. (h) c,o h represents the specific enthalpy at the compressor outlet. c,i h represents the specific enthalpy at the compressor inlet. is,o η represents the outlet specific enthalpy of the refrigerant under isentropic compression. is This represents the isentropic efficiency of the compressor under isentropic compression conditions. The enthalpy values ​​at the compressor's outlet and inlet are related to the state of the R134a refrigerant in this paper. The compressor's volumetric efficiency and isentropic efficiency are related to the compressor's outlet pressure ratio, inlet pressure ratio, and compressor speed. The specific calculation formulas are shown below:

[0126] η v =f1(N comp ,P c / P e )

[0127] η is =f2(N comp ,P c / P e )

[0128] In the above calculation formula, f1 represents the fitting function of compressor volumetric efficiency and isentropic efficiency on compressor inlet pressure difference, f2 represents the fitting function of compressor volumetric efficiency and isentropic efficiency on compressor outlet pressure difference, and P c / P e This represents the pressure ratio at the compressor inlet and outlet, where P c P represents the pressure of the condenser. e This represents the pressure in the evaporator.

[0129] S12. Evaporator modeling: The gas and liquid refrigerant inside the evaporator satisfy the law of conservation of mass. Therefore, the change in the two-phase region length le of the evaporator is obtained by the following formula:

[0130]

[0131] Where, ρ le h represents the density of the refrigerant. lge A represents the latent heat absorbed by the refrigerant during a phase change. e This represents the cross-sectional area of ​​the evaporator flat tube structure. This represents the average vapor proportion of the refrigerant during the two-phase stage of the evaporator. The refrigerant mass flow rate at the evaporator inlet, h ge h represents the enthalpy of the refrigerant at the evaporator outlet.ie This represents the enthalpy of the refrigerant at the evaporator inlet, a. ie D represents the heat transfer coefficient between the evaporator wall and the refrigerant. ie T represents the inner diameter of the evaporator tube. we T represents the temperature of the evaporator wall. re This represents the temperature of R134a under the current pressure condition.

[0132] The density of refrigerant vapor in the evaporator can be described by the following formula:

[0133]

[0134] In this formula, L e ρ represents the total length of all the flat tubes in the evaporator. ge P represents the density of a saturated refrigerant in its gaseous state. e This represents the current pressure of the refrigerant. This represents the refrigerant mass flow rate at the evaporator outlet. From the perspective of mass conservation, this formula illustrates that the change in the refrigerant vapor mass within the evaporator can be considered as the sum of the refrigerant mass flow rate at the evaporator inlet, the mass flow rate at the outlet, and the amount of vapor refrigerant generated.

[0135] Since the superheating stage accounts for a small proportion of the entire evaporation process, after neglecting the heat generated during this stage, the heat exchange or temperature change of the evaporator wall can be described by the following formula:

[0136]

[0137] a oe =f p2 (N fan )

[0138] In the above formula, C p The specific heat capacity of the evaporator is represented by m, and the mass of the evaporator is represented by a. ish T represents the heat transfer coefficient between the refrigerant and the wall during the evaporator superheating stage. ie Represents the inlet refrigerant temperature, a oe f represents the heat transfer coefficient of the evaporator on the air side, where f p2 N fan Let A represent the polynomial fitting formula and the evaporator fan speed, respectively. oe T represents the air-facing area of ​​the evaporator. ae This represents the ambient air temperature around the evaporator inlet, and it is calculated as follows:

[0139]

[0140] In the formula C represents the air mass flow rate on the air side. p,air,mix T represents the specific heat capacity of air. air,mix This represents the temperature of the air entering the evaporator after the mixing damper has been activated, and it has a functional relationship with the specific heat capacity of the air side of the evaporator. This represents the previous air quality flow rate in the passenger cabin. C represents the new air mass flow rate in the external environment. p,air,cab and C p,air,amb These are the specific heats of air corresponding to the two. and The sum is the total air mass flow rate after mixing. The calculation is as follows:

[0141]

[0142] Where, γ cycle ρ represents the proportion of old air in the mixed airflow. air,cab ρ represents the density of the old wind. air,amb V represents the density of fresh air; both are affected by air temperature. air This represents the total air intake of the mixing damper, and its size is affected by the fan speed. S13. For the condenser, its heat exchange principle is similar to that of the evaporator, utilizing the laws of conservation of mass and energy; therefore:

[0143]

[0144] To simplify the model, we assume that there is no refrigerant leakage in the entire refrigeration cycle piping of the air conditioning system. Therefore, the total mass of refrigerant in the system remains constant, and the total mass of refrigerant in the evaporator and condenser is considered constant. Hence:

[0145]

[0146] Where, ρ lc It is the density of the liquid refrigerant in the condenser, h lgc A is the latent heat of vaporization of the refrigerant in the condenser. c This represents the total cross-sectional area of ​​the flat tube microchannels in the condenser. h is the average porosity of the two-phase region of the condenser. gc h lc and h ic These represent the specific enthalpy values ​​of the gas, liquid, and inlet refrigerant in the condenser at the current pressure, respectively. ic D is the heat transfer coefficient between the condenser inner wall and the refrigerant in the two-phase region. ic The diameter of the inside of the condenser flat tube, T wc T is the condenser wall temperature. rc It is the saturation temperature of the refrigerant at the current pressure of the condenser, L.c It is the total length of the condenser flat tube, (C p m) wc a represents the specific heat and mass of the condenser material. oc It is the heat transfer coefficient between air and the condenser wall, A oc T represents the frontal area of ​​the condenser. ac It is the temperature of the air around the condenser, i.e., the ambient temperature, and ∑ represents a constant.

[0147] condenser air-side heat transfer coefficient a oc The condenser's external wind speed is mainly affected by the vehicle's speed while the car is in motion, therefore a oc The relationship between the speed and the vehicle speed was obtained through experimental fitting and is expressed as:

[0148] a oc =f p2 (V car )

[0149] Among them, vehicle speed V car The driver decides the speed, rather than the air conditioning system controller. In the air conditioning system control, vehicle speed can be regarded as a disturbance input.

[0150] S14. For the expansion valve modeling, during its dynamic process, the refrigerant mass flow rate through the expansion valve... Its relationship with the pressure drop ΔP of the expansion valve is expressed by the following formula:

[0151]

[0152] Among them, C q ρ is the flow coefficient of the expansion valve. v Let A be the refrigerant density at the inlet of the expansion valve. v ΔP is the flow area of ​​the expansion valve, and ΔP is the pressure drop, which is the pressure difference between the inlet and outlet of the expansion valve.

[0153] Based on the relevant knowledge of heat transfer, a dynamic thermal model of the automobile passenger compartment system is established and coupled with the air conditioning system model to establish a thermal coupling model of electric vehicle air conditioning and passenger compartment.

[0154] Total heat load of the car passenger compartment Represented as:

[0155]

[0156] During vehicle operation, heat exchange occurs between the vehicle cabin and the outside environment via convection. Mainly affected by vehicle speed V car and ambient temperature T ac The impact, and these two variables are input disturbances that are not controlled by the controller, in the heat transfer model, Calculated by the following formula:

[0157]

[0158] Where T s Let be the temperature of the structure surrounding the crew compartment. Based on the law of conservation of energy, the dynamic change of the temperature of the surrounding structure is expressed by the following formula:

[0159]

[0160] According to the law of conservation of energy, the dynamic change of air temperature in the crew cabin can be expressed as:

[0161]

[0162] in, For heat exchange of the vehicle body surface structure, For solar radiation heat load, The heat load caused by ventilation For the human body's heat load, For mechanical and instrumentation heat load, T cab For the temperature of the passenger cabin, M represents the amount of cooling capacity delivered to the cabin by the air conditioning system per unit time. a For the air quality within the passenger cabin volume range, cp a For the specific heat of air, h o It is the heat transfer coefficient between the outer side of the external structure of the passenger compartment and the air side, which is mainly determined by the vehicle speed. S is the total surface area of ​​the passenger compartment's outer surface structure, h i M is the heat transfer coefficient between the passenger cabin interior surfaces and the air. s and C ps These are the mass and specific heat of the enclosed structure surrounding the vehicle cabin.

[0163] S2: Establish a calculation model for passenger thermal comfort assessment, specifically including:

[0164] S21: Establish as follows Figure 3 The passenger thermal comfort assessment calculation model shown;

[0165] Factors influencing human thermal comfort include air humidity, thermal radiation, and airflow velocity. To maintain a comfortable temperature in the passenger compartment while the vehicle is in motion, a passenger thermal comfort assessment calculation model was developed. This model consists of three parts: a "PMV calculator," a "learning regulator," and a "T..." comfort The "Calculator" section.

[0166] The PMV calculator is primarily used to calculate the average comfort temperature prediction rating for people. The PMV value indicates how well a person currently feels about heat. The calculation formula is shown below:

[0167] PMV = (0.303e -0.036M +0.028)×(M-φ1-φ2-φ3-φ4-φ5-φ6)

[0168] In the formula, M represents human metabolism. Here, the metabolic rate of passengers is set to 1, and the metabolic rate of drivers is set to 1.5. The specific calculations for Φ1 to Φ6 are shown in the formulas below:

[0169] Φ1=3.05e -3 ·(5733-6.99MP w )

[0170]

[0171] Φ3=1.7e -5 M(5867-P w )

[0172] Φ4=1.4e -3 M(34-T cab )

[0173] Φ5=3.96e -8 f cl [(T cl +273) 4 -(T r +273) 4 ]

[0174] Φ6=f cl h c (T cl -T cab )

[0175] in,

[0176]

[0177] T cl =35.7-0.028MI cl (φ5+φ6)

[0178] In the above formula, P w T represents the partial pressure of water vapor. cab and T r These represent the passenger cabin temperature and the average radiant temperature inside the cockpit, respectively. To ensure consistency in the subsequent use of the same values ​​for both, h... c T represents the heat transfer coefficient of the human body surface. cl I represents the surface temperature of a person's clothing. cl This represents the thermal resistance of clothing; in the high temperatures of summer, this value is typically 0.7. aThis represents the airflow velocity inside the vehicle. Since humidity has a relatively small impact on thermal comfort calculations, it is set to 50%.

[0179] The learning regulator aims to learn different people's thermal preferences through continuous calculation and analysis. By analyzing past data on temperature regulation, it calculates a moving average PMV value to make the passenger cabin temperature more closely match people's preferences.

[0180]

[0181] When the calculated PMV value is not satisfactory to passengers, they will manually adjust to their own comfortable temperature and record the temperature value T at that moment. i And with the adjustment sequence number i, the learning regulator will first calculate PMV. i Then, the average PMV after these n artificial temperature adjustments is calculated, where N represents the PMV adaptive sampling period. When n exceeds N, the PMV... a The calculation is based on the average of the last N PMV values, which ensures that people can adapt to changes in air conditioning temperature and makes the results more accurate.

[0182] S3: Design the TD3 algorithm:

[0183] The TD3 algorithm, short for Dual-Delay Deep Deterministic Strategy, is a deep reinforcement learning algorithm developed by Scott et al., based on the DDPG algorithm, for solving continuous control problems. The TD3 algorithm introduces a dual Critic network to estimate the action value function on top of the DQN algorithm. It then compares the target values ​​calculated by the two networks and takes the minimum to estimate the action value of the next state action, as shown in the following formula:

[0184]

[0185] Where y1 and y2 represent respectively right Optimize, right Optimization is performed because of certain states in reinforcement learning. It will be greater than the already estimated. To avoid this situation, it will be done through Come to By imposing restrictions, the overestimation problem caused by high variance is effectively solved.

[0186] For deterministic policies, the agent's exploration of the policy space is limited, and it is prone to getting trapped in local optima when updating the Critic network and calculating the action value function. Therefore, to avoid these problems, the TD3 algorithm introduces a target policy smoothing mechanism during training, adding random noise to make the Q-value function smoother and improve training stability. The target policy update formula is shown below:

[0187]

[0188] Here, ∈~clip(N(0,σ) represents that the added noise follows a Gaussian distribution and is restricted; the absolute value of the noise cannot exceed the hyperparameter c (usually taken as 1). This ensures that the actions do not exceed the allowable range, effectively improving the stability of the network training process. The TD3 algorithm uses a gradient update method similar to DDPG to update the Critic network parameters θ. i And delay updating the Actor network parameters after step d. The specific process of the TD3 algorithm is as follows: Figure 4 As shown.

[0189] S32: Design an agent for the TD3 algorithm;

[0190] Information related to passenger cabin temperature and air conditioning energy consumption is selected as elements of the state space, namely the absolute value of the difference between the passenger cabin temperature and the target temperature, the PMV value, and the air conditioning energy consumption. Among them, the passenger cabin temperature can provide real-time information about the temperature environment in which the driver is located; the PMV value can directly reflect people's thermal perception of the temperature environment; and the air conditioning energy consumption mainly includes compressor energy consumption and heat exchanger fan energy consumption.

[0191] State={|T cab -25℃ | PMV, E AC ,T evap,wall}

[0192] For the electric vehicle air conditioning system studied, the compressor plays a primary role in cooling, while the heat exchanger fan improves the radiator's heat dissipation efficiency, thus contributing to air conditioning cooling and improving the overall cooling effect. Based on this, compressor speed and heat exchanger fan speed will be selected as the control actions of the intelligent agent:

[0193] action = {N comp N fan}

[0194] To reduce the learning difficulty of the algorithm, this paper adopts a combination of primary and secondary rewards in the design of the reward function. The primary rewards are designed based on PMV (partial thermal comfort) and air conditioning energy consumption, key indicators of human thermal comfort, as shown below:

[0195]

[0196] Where α and β are target weight coefficients, different α and β reflect the different degrees of importance the reward function attaches to thermal comfort and energy saving. Since ensuring human comfort is the main control objective, α should be greater than β. m and n are proportional coefficients, and here α, m, n and β are set to 4000, 2000, 2000 and 1000 respectively.

[0197] The auxiliary reward can effectively reduce the learning difficulty of the agent and improve the algorithm performance. Its specific design is as follows:

[0198]

[0199]

[0200] Where i, j, p, q are proportionality coefficients, which are -2000, 2000, 2000, and -2000, respectively.

[0201] S33: Set the parameters of the TD3 algorithm agent;

[0202] To ensure the agent has a large action exploration space, a Gaussian action exploration noise model will be added during training. The standard deviation of the noise will be set to 100, and it will decay at a rate of 1e-5 until it reaches a minimum of 0.005. During training, the target network will be updated once every 10 agent steps, with a smoothing factor of 0.005. Furthermore, when the passenger cabin temperature exceeds 55°C or falls below 15°C, it indicates that the agent has learned an effective control strategy, and reinforcement learning training will automatically stop at this point. Other key parameter settings for the agent are shown in Table 1.

[0203] Table 1 Agent Parameters

[0204]

[0205] S4: Conduct agent training and validation, and establish a control strategy based on the TD3 algorithm that matches the thermal coupling model of the automotive air conditioning and passenger compartment. Specifically, this includes:

[0206] S41. Observe the changes in actual return, average return and predicted return at each step through the dynamic return curve;

[0207] S42. Select the agent with the highest actual return for training, and analyze it after simulation testing;

[0208] S43. Determine whether the agent training meets all the qualification criteria at the same time. If yes, the training is qualified; otherwise, the training is unqualified.

[0209] S44. Analyze the unqualified agent and related training results, and readjust the control action, observation state and reward function;

[0210] S45. Perform hyperparameter tuning and then retrain.

[0211] In this embodiment, agent training and verification are performed. The training and verification process for the TD3 algorithm agent is as follows: Figure 5 As shown. First, the training code for the agent is set up, executing the policy every 10 seconds. The total simulation time for one training session is 6000 seconds, where the sampling time and simulation time can be adjusted according to the actual training process. The number of training iterations is set to 3000. For the trained agent, the changes in its actual reward, average reward, and predicted reward at each step are observed through a reward dynamic curve. Then, the agent with the highest actual reward is selected as the training example, the training mode is turned off, and the selected example is loaded into the environment. After simulation testing, the results are analyzed to evaluate whether they meet the expected content, i.e., whether the policy control objective has been achieved. If they meet the expectations, the agent's training is deemed successful; otherwise, it is deemed unsuccessful. Subsequently, the unsuccessful agents and related training results are analyzed, and the control actions, observation states, reward functions, etc., are readjusted. Based on this, hyperparameter tuning is performed, and then retraining is conducted. The evaluation criteria for whether the agent's training is successful are as follows: whether the cumulative discounted reward during the training process has a convergence trend; whether the control actions have a change process; and whether the PMV value can always remain within the human comfort range, i.e., between 1 and -1. Whether the passenger cabin temperature can drop significantly in a short period of time and gradually decrease to the target temperature (around 25°C).

[0212] Establish as Figure 6 The method shown is an intelligent control method for electric vehicle air conditioning and passenger compartment thermal management based on reinforcement learning algorithm (TD3). According to steps S3 and S41, a trained and qualified intelligent agent is obtained. Combining the thermal coupling model of the vehicle air conditioning and passenger compartment established in steps S1 and S2, and the passenger thermal comfort evaluation calculation model, a system can finally be established as follows: Figure 5 The method for intelligent control of electric vehicle air conditioning and passenger cabin thermal management based on reinforcement learning algorithm (TD3) is shown.

[0213] In summary, this solution, based on the TD3 algorithm, is a thermal management control method for the air conditioning and passenger compartment of electric vehicles. By applying the TD3 algorithm to the thermal management of the air conditioning and passenger compartment, the intelligent agent only needs to continuously interact with the environment. Its internal policy network can then make corresponding actions based on current environmental information and rewards to achieve the control objective indefinitely. This strategy's control decision-making method is closer to that of a real-world living organism. Compared to traditional PID controllers and ON-OFF controllers, its control is more intelligent in complex environments such as the coupled air conditioning and passenger compartment model. It can interact with the environment in real time, reduce the number of ineffective control actions, and effectively reduce passenger compartment temperature and air conditioning energy consumption. Figure 7 As shown, the energy consumption of this scheme in a 5000s WLTC cycle (Worldwide Harmonized Light Vehicles Test Cycle) is 2.268 kWh, while the energy consumption of the traditional feedback control algorithm ON-FF control is much lower.

[0214] The PID control results are 2.884 kWh and 2.693 kWh, respectively, saving approximately 22% and 16% of energy.

[0215] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for controlling the air conditioning and passenger compartment thermal management of electric vehicles based on the TD3 algorithm, characterized in that, Includes the following steps: S1. Establish a one-dimensional simulation model of the automotive air conditioning system based on the moving boundary method and the lumped parameter method. The automotive air conditioning system includes a compressor, condenser, evaporator and expansion valve. Establish a dynamic thermal model of the automotive passenger compartment system and couple it with the air conditioning system model to establish a thermal coupling model between the electric vehicle air conditioning and the passenger compartment. S2. Establish a calculation model for passenger thermal comfort assessment; S3. Design the TD3 algorithm; S31. Introduce a target policy smoothing mechanism to the TD3 algorithm; The TD3 algorithm is as follows: Where y1 represents right Optimization is performed, y2 represents right Optimize; θ i For Critic network parameters, These are the parameters of the Actor network; The updated formula for the target policy is: Wherein, ∈~clip(N(0,σ) represents that the added noise follows a Gaussian distribution and the absolute value of the noise is less than or equal to the hyperparameter c; S32. Design the agent for the TD3 algorithm; select state space elements, including the passenger cabin temperature T. cab The absolute value of the temperature difference from the target temperature, the PMV value, and the air conditioning energy consumption; air conditioning energy consumption includes compressor energy consumption E. AC and heat exchanger fan energy consumption T evap,wall ; State={|T cab -25℃|,PMV,E AC ,T evap,wall } Compressor speed and heat exchanger fan speed are selected as the control actions of the intelligent agent: action) {N comp ,N fan } The reward function is designed by combining main reward and auxiliary reward, with PMV, an important indicator of human thermal comfort, and air conditioning energy consumption designed as the main reward. S33. Set the agent parameters of the TD3 algorithm and perform reinforcement learning training. When the passenger cabin temperature exceeds the preset high temperature threshold or falls below the preset low temperature threshold, the reinforcement learning training stops. S4. Conduct agent training and verification to establish a control strategy that matches the thermal coupling model of the car air conditioning and passenger compartment.

2. The electric vehicle air conditioning and passenger cabin thermal management control method based on the TD3 algorithm according to claim 1, characterized in that: In S1, the one-dimensional simulation model of the automotive air conditioning system includes: S11. Establish a one-dimensional dynamic model of the refrigerant inside the compressor, as follows: η v =f1(N comp ,P c / P e ) η is =f2(N comp ,P c / P e ) Where, η v ρ represents the volumetric efficiency of the compressor. r N represents the density of the refrigerant at the compressor inlet. comp V represents the compressor's rotational speed; d h represents the compressor's displacement. c,o h represents the specific enthalpy at the compressor outlet. c,i h represents the specific enthalpy at the compressor inlet. is,o η represents the outlet specific enthalpy of the refrigerant under isentropic compression. is f1 represents the isentropic efficiency of the compressor under isentropic compression conditions; f2 represents the fitting function of the compressor's volumetric efficiency and isentropic efficiency to the compressor's inlet pressure difference; f3 represents the fitting function of the compressor's volumetric efficiency and isentropic efficiency to the compressor's outlet pressure difference; P represents... c / P e This represents the pressure ratio at the compressor inlet and outlet, where P c P represents the pressure of the condenser. e This represents the pressure in the evaporator; S12. Evaporator modeling: The change in the two-phase region length le of the evaporator is obtained by the following formula: Where, ρ le h represents the density of the refrigerant. lge A represents the latent heat absorbed by the refrigerant during a phase change. e This represents the cross-sectional area of ​​the evaporator flat tube structure. This represents the average vapor proportion of the refrigerant during the two-phase stage of the evaporator. The refrigerant mass flow rate at the evaporator inlet, h ge h represents the enthalpy of the refrigerant at the evaporator outlet. ie This represents the enthalpy of the refrigerant at the evaporator inlet, a ie D represents the heat transfer coefficient between the evaporator wall and the refrigerant. ie T represents the inner diameter of the evaporator tube. we T represents the temperature of the evaporator wall. re This represents the refrigerant temperature under the current pressure condition; The density of refrigerant vapor in the evaporator is expressed by the following formula: Among them, L e ρ represents the total length of all the flat tubes in the evaporator. ge P represents the density of a saturated refrigerant in its gaseous state. e This represents the current pressure of the refrigerant. The refrigerant mass flow rate at the evaporator outlet and the temperature change at the evaporator wall are: Among them, C p The specific heat capacity of the evaporator is represented by m, and the mass of the evaporator is represented by a. ish T represents the heat transfer coefficient between the refrigerant and the wall during the evaporator superheating stage. ie Represents the inlet refrigerant temperature, a oe f represents the heat transfer coefficient of the evaporator on the air side, where f p2 N fan Let A represent the polynomial fitting formula and the evaporator fan speed, respectively. oe T represents the air-facing area of ​​the evaporator. ae This represents the ambient air temperature around the evaporator inlet. C represents the air mass flow rate on the air side. p,air,mix T represents the specific heat capacity of air. air,mix This represents the temperature of the air entering the evaporator after the mixing damper has been activated. This represents the previous air quality flow rate in the passenger cabin. C represents the new air mass flow rate in the external environment. p,air,cab and C p,air,amb These are the specific heats of air corresponding to the two; and The sum of these is the total air mass flow rate after mixing. The calculation method is as follows: Where, γ cycle ρ represents the proportion of old air in the mixed airflow. air,cab ρ represents the density of the old wind. air,amb V represents the density of fresh air. air This represents the total air intake of the mixing damper, and its size is affected by the fan speed. S13. Condenser Modeling: Assuming no refrigerant leakage in the entire refrigeration cycle piping of the air conditioning system, the total mass of refrigerant in the system remains constant. Therefore, the total mass of refrigerant in the evaporator and condenser can be considered constant, hence: Where, ρ lc h is the density of the liquid refrigerant in the condenser. lgc A is the latent heat of vaporization of the refrigerant in the condenser. c This represents the total cross-sectional area of ​​the flat tube microchannels in the condenser. h is the average porosity of the two-phase region of the condenser. gc h lc and h ic These represent the specific enthalpy values ​​of the gas, liquid, and inlet refrigerant in the condenser at the current pressure, respectively. ic D is the heat transfer coefficient between the condenser inner wall and the refrigerant in the two-phase region. ic The diameter of the inside of the condenser flat tube, T wc T is the condenser wall temperature. rc It is the saturation temperature of the refrigerant at the current pressure of the condenser, L. c It is the total length of the condenser flat tube, (C p m) wc a represents the specific heat and mass of the condenser material. oc It is the heat transfer coefficient between air and the condenser wall, A oc T is the frontal area of ​​the condenser. ac It is the current ambient temperature around the condenser, and ∑ represents a constant. While the car is moving, the external wind speed of the condenser is affected by the vehicle speed. oc The relationship with vehicle speed is as follows: a oc =f p2 (V car ) Among them, V car Vehicle speed is the disturbance input controlled solely by the driver. S14. Expansion valve modeling: During the dynamic process, the refrigerant mass flow rate through the expansion valve... The relationship between it and the pressure drop ΔP of its expansion valve is: Among them, C q ρ is the flow coefficient of the expansion valve. v Let A be the refrigerant density at the inlet of the expansion valve. v ΔP is the flow area of ​​the expansion valve, and ΔP is the pressure difference between the inlet and outlet of the expansion valve.

3. The electric vehicle air conditioning and passenger cabin thermal management control method based on the TD3 algorithm according to claim 1, characterized in that: In S1, establishing a dynamic thermal model of the vehicle passenger compartment system and coupling it with the air conditioning system model specifically involves: Total heat load of the car passenger compartment Represented as: During vehicle operation, heat exchange occurs between the vehicle cabin and the outside environment via convection. Subject to vehicle speed V car and ambient temperature T ac The input disturbances, which are unaffected by the controller and are not controlled by the controller, are present in the heat transfer model. Calculated by the following formula: Where T s Given the temperature of the structure surrounding the crew compartment, based on the law of conservation of energy, the dynamic temperature change of the surrounding structure is as follows: The dynamic changes in air temperature in the crew cabin are represented as follows: in, For heat exchange of the vehicle body surface structure, For solar radiation heat load, The heat load caused by ventilation For the human body's heat load, For mechanical and instrumentation heat load, T cab For the temperature of the passenger cabin, M represents the amount of cooling capacity delivered to the cabin by the air conditioning system per unit time. a For the air quality within the passenger cabin volume range, cp a For the specific heat of air, h o It is the heat transfer coefficient between the outer side of the external structure of the passenger compartment and the air side, which is determined by the vehicle speed. S is the total surface area of ​​the passenger compartment's outer surface structure, h i M is the heat transfer coefficient between the passenger cabin interior surfaces and the air. s and C ps These are the mass and specific heat of the enclosed structure surrounding the vehicle cabin.

4. The electric vehicle air conditioning and passenger cabin thermal management control method based on the TD3 algorithm according to claim 1, characterized in that: The passenger thermal comfort assessment calculation model includes a PMV calculator, a learning regulator, and a T... comfort calculator; The PMV calculator is used to calculate the average rating of people's comfort temperature prediction. The formula for calculating PMV is as follows: PMV=(0.303e -0.036M +0.028)×(M-Φ1-Φ2-Φ3-Φ4-Φ5-Φ6) Where M represents human metabolism, the metabolic rate of passengers is set to 1, and the metabolic rate of drivers is set to 1.5; the specific calculation method for Φ1 to Φ6 is as follows: Φ1=3.05e -3 ·(5733-6.99M-P w ) Φ3=1.7e -5 M(5867-P w ) Φ4=1.4e -3 M(34-T cab ) Φ5=3.96e -8 f cl [(T cl +273) 4 -(T r +273) 4 ] Φ6=f cl h c (T cl -T cab ) Among them, h c The calculation method is as follows: T cl =35.7-0.028M-I cl (Φ5+Φ6) Among them, P w T represents the partial pressure of water vapor. cab and T r h represents the average radiant temperature inside the passenger cabin and the average radiant temperature inside the cockpit, respectively. c T represents the heat transfer coefficient of the human body surface. cl I represents the surface temperature of a person's clothing. cl V represents the thermal resistance of clothing. a Represents the airflow velocity inside the vehicle; The learning regulator analyzes past data on temperature regulation and calculates the average PMV after n manual temperature adjustments: Where i represents the adjustment sequence number of the manual adjustment, and N represents the PMV adaptive sampling period. When n exceeds N, the PMV... a Calculated based on the average of N PMV values; The T comfort The calculator is used to calculate PMV a The value is used to calculate the target comfortable temperature T at this time. comfort .

5. The electric vehicle air conditioning and passenger compartment thermal management control method based on the TD3 algorithm according to claim 1, characterized in that, In S32, the main storyline reward is: Where α and β are target weight coefficients, α being greater than β, reflecting the importance the reward function places on thermal comfort and energy efficiency, and m and n are proportionality coefficients; The auxiliary line reward is: Where i,j,p,q are proportionality coefficients, and T evap,wall For the energy consumption of the heat exchange fan, T cab For passenger cabin temperature, T target The target temperature.

6. The electric vehicle air conditioning and passenger compartment thermal management control method based on the TD3 algorithm according to claim 1, characterized in that, The agent parameters include Gaussian action exploration noise model parameters, target network parameters, and training parameters; the Gaussian action exploration noise model parameters include standard deviation, decay rate, and minimum decay value; the target network parameters include update frequency and smoothing factor; and the training parameters include the number of experience pool samples, soft update parameters, discount factor, Actor network learning rate, Critic network learning rate, and minimum sample set sample size.

7. The electric vehicle air conditioning and passenger compartment thermal management control method based on the TD3 algorithm according to claim 1, characterized in that, In step S4, the agent training and verification specifically includes: S41. Observe the changes in actual return, average return and predicted return at each step through the dynamic return curve; S42. Select the agent with the highest actual return for training, and analyze it after simulation testing; S43. Determine whether the agent training meets all the qualification criteria at the same time. If yes, the training is qualified; otherwise, the training is unqualified. S44. Analyze the unqualified agent and related training results, and readjust the control action, observation state and reward function; S45. Perform hyperparameter tuning and then retrain.

8. The electric vehicle air conditioning and passenger compartment thermal management control method based on the TD3 algorithm according to claim 7, characterized in that, In S43, the qualification criteria include: the cumulative discount reward during the training process and whether there is a convergence trend; whether the control action changes; whether the PMV value can always be within the human comfort range; and whether the passenger cabin temperature gradually drops to the target temperature within a preset time.

Images