Thermal-electric integrated optimization control method for intelligent connected fuel cell vehicles

Abstract

The present invention relates to a thermal-electric integrated optimization control method for an intelligent networked fuel cell vehicle, and the method comprises the following steps: establishing a speed prediction model for a rolling time domain; establishing a control-oriented battery thermal management system model; establishing an energy and heat optimization management problem; solving it using reinforcement learning; and a parallel computing control strategy based on an on-board domain controller. The thermal-electric coupling integrated optimization control method for the whole vehicle-system integrated thermal management system of an intelligent networked fuel cell vehicle proposed in the present invention has important innovative significance and application value in the field of thermal-electric coordinated optimization control of fuel cell vehicles. By introducing reinforcement learning control strategies and parallel computing control allocation strategies, accurate control and efficient utilization of the thermal-electric coordinated needs of fuel cell vehicles can be achieved, providing effective technical support for improving vehicle performance, reducing energy consumption and reducing emissions.

Description

Technical Field

[0001] The present invention relates to the technical field of thermal management control of fuel cell vehicles, and in particular to a thermal-electric integrated optimization control method for intelligent networked fuel cell vehicles. Background Art

[0002] As the global energy crisis and environmental problems become increasingly serious, the new energy vehicle industry has become the focus of the global automotive industry. Among them, hydrogen fuel cell vehicles are seen as an important development direction for future new energy clean power vehicles due to their high efficiency and environmental protection characteristics. Its core technology lies in the fuel cell system, which is the key to achieving its independent research and development and industrialization.

[0003] The automotive fuel cell system must be able to operate stably and efficiently in a sub-zero low-temperature environment, and has high requirements for thermal stability and energy consumption. Adaptability to low-temperature environments is a problem that must be solved for the commercialization of fuel cell vehicles, and the fuel cell thermal management system is one of the key technologies that need to be tackled. Controlling the reasonable intake temperature of the fuel cell is crucial to improving the efficiency and stability of the fuel cell stack. In addition, motors, batteries, charge air and other electronic components also require reasonable cooling and heating circuits. However, too many independent circuits will increase the weight and volume of the thermal management system, reducing the cruising range and energy economy. Therefore, a vehicle integrated thermal management system with a reasonable control strategy is of great significance to the stable operation and energy economy of fuel cell vehicles.

[0004] At present, the inventions disclosed about the thermal management system of fuel cell vehicles mainly focus on the integrated design of waste heat recovery and vehicle heating system. These designs have achieved certain results in ensuring that the battery temperature is within a safe range and improving energy utilization, but they have significant deficiencies in handling the dynamic operating condition control of complex thermal-electric multi-domain systems. Specifically, these solutions usually fail to fully consider the energy-saving needs of the battery system, resulting in low energy utilization efficiency in actual applications, causing unnecessary energy waste, and thus affecting the economy of the battery system.

[0005] Most existing studies on vehicle thermal-electric collaborative optimization control use a hierarchical control architecture. Although this architecture simplifies the complexity of the problem, it is difficult to ensure the optimal performance of the system as a whole due to the suboptimal solutions that may be generated between the layers. In addition, for complex thermal-electric multi-domain systems, general optimization algorithms often face the challenge of slow convergence, which not only affects the real-time response capability of the system, but also limits its performance in practical applications. Summary of the invention

[0006] The present invention proposes a thermal-electric integrated optimization control method for intelligent networked fuel cell vehicles, which not only takes into account the complex thermal coupling relationship between the fuel cell, the passenger compartment and the power battery, but also solves the electrical coupling problem of the heating of each subsystem. By adopting the reinforcement learning actor-critic algorithm, the present invention can effectively deal with the energy consumption minimization temperature control problem of the thermal management system.

[0007] Terminology explanation: Thermal management: Thermal management is the process of adjusting and controlling the temperature or temperature difference of a specific object by heating or cooling according to its requirements.

[0008] Reinforcement Learning: Reinforcement Learning (RL) is a field of machine learning that studies how an agent learns behavior strategies in an environment through trial-and-error to maximize the cumulative reward obtained from the environment.

[0009] Offline learning: Offline learning is a method of machine learning, as opposed to online learning. In offline learning, the model is trained using historical data, and after training, the model is used for tasks such as prediction or classification. In short, offline learning is performed without real-time data streams and relies on pre-collected data sets.

[0010] To achieve the above object, the present invention provides the following technical solutions:

[0011] A thermal-electric integrated optimization control method for an intelligent networked fuel cell vehicle, the method comprising the following steps:

[0012] 1) Establish a speed prediction model for the rolling time domain;

[0013] 2) Establish a control-oriented battery thermal management system model;

[0014] 3) Establish energy and heat optimization management issues;

[0015] 4) Use reinforcement learning to solve;

[0016] 5) Parallel computing control strategy based on vehicle-mounted domain controller.

[0017] As a further technical solution of the present invention, the steps of establishing a speed prediction model for the rolling time domain include:

[0018] First, the future speed of fuel cell vehicles is obtained through intelligent network technology. , through the automobile power balance formula, ignoring the slope resistance, the vehicle's driving demand power can be obtained ;

[0019] ;

[0020] Where: - vehicle quality, - rolling resistance coefficient, - air resistance coefficient, - windward area, - Rotational mass conversion factor, - vehicle acceleration;

[0021] The driving power required by the fuel cell and power battery is obtained through the power distribution relationship ;

[0022] Since the vehicle drive requires power Only and speed related, therefore Can be seen as The function of Similarly, the driving power required by fuel cells and power batteries is Can be seen as about speed function.

[0023] As a further technical solution of the present invention, the steps of establishing a control-oriented battery thermal management system model include:

[0024] Get the traction power required by the fuel cell and power battery Finally, a thermal model of the power battery is established for the thermal management system , fuel cell thermal model , passenger compartment thermal model and battery ;

[0025] 2.1) Thermal model of power battery :

[0026] Gas mass flow rate from the fan cooling duct to the power battery It is expressed as follows:

[0027] ;

[0028] In the formula, is the power of the power battery fan cooling system, It is the maximum power of the power battery fan cooling system, which is usually considered to be a fixed value. is the maximum gas mass flow rate flowing from the fan cooling duct into the power battery;

[0029] Through fan cooling, the heat transfer rate of the power battery to the outside air is for:

[0030] ;

[0031] In the formula, is the heat capacity of air, is the power battery temperature, is the ambient temperature. In this model, it is assumed that the ambient temperature is constant. is the heat transfer coefficient of air, is the contact area between air and battery;

[0032] Mass flow rate of water pump cooling pipe into power battery It is expressed as follows:

[0033] ;

[0034] In the formula, is the power of the power battery water pump cooling system, It is the maximum power of the power battery water pump cooling system, which is usually considered to be a fixed value. is the maximum mass flow rate from the water pump cooling pipe into the power battery;

[0035] Through the water pump cooling method, the heat transfer rate of the power battery to the coolant for:

[0036] ;

[0037] In the formula, is the heat capacity of the coolant, is the coolant temperature. In this model, the surrounding coolant temperature is assumed to be constant. is the heat transfer coefficient of the coolant, is the contact area between the coolant and the battery;

[0038] Total power of the power battery ,in is the power of the air conditioner, and the charge and discharge SOC of the power battery is expressed as:

[0039] ;

[0040] In the formula, Represents the internal resistance of the power battery, Output current for the power battery, is the open circuit voltage, is the total charge of the power battery, and Battery state of charge and temperature, which are usually modeled by experimentally generated MAP tables (i.e., lookup tables), so It is expressed as:

[0041] ;

[0042] The dynamic equation of power battery temperature is:

[0043] ;

[0044] ;

[0045] In the formula, The quality of the power battery; is the thermal capacity of the power battery;

[0046] Heat transfer rate of power battery for:

[0047] ;

[0048] 2.2) Thermal model of fuel cell :

[0049] Gas mass flow rate into the fuel cell from the fan cooling duct It is expressed as follows:

[0050] ;

[0051] In the formula, is the power of the fuel cell fan cooling system, is the maximum power of the fuel cell fan cooling system, which is usually considered to be a fixed value. is the maximum gas mass flow rate flowing into the fuel cell from the fan cooling duct;

[0052] Through fan cooling, the heat transfer rate of the fuel cell to the outside air is for:

[0053] ;

[0054] In the formula, is the fuel cell temperature, is the contact area between air and fuel cell;

[0055] Mass flow rate of water pump cooling pipe into fuel cell It is expressed as follows:

[0056] ;

[0057] In the formula, is the power of the fuel cell water pump cooling system, It is the maximum power of the fuel cell water pump cooling system, which is usually considered to be a fixed value. is the maximum mass flow rate from the water pump cooling pipe into the fuel cell;

[0058] Through water pump cooling, the heat transfer rate of the fuel cell to the coolant for:

[0059] ;

[0060] In the formula, is the contact area between the coolant and the fuel cell;

[0061] Fuel cell charge and discharge current is a function of the fuel cell power and is usually modeled by an experimentally generated MAP table (i.e., a lookup table);

[0062] The dynamic equation for the fuel cell temperature is:

[0063] ;

[0064] ;

[0065] In the formula, is the internal resistance of the fuel cell, is the mass of the fuel cell; is the heat capacity of the fuel cell;

[0066] Heat transfer rate of fuel cells for:

[0067] ;

[0068] 2.3) Passenger compartment heating model ;

[0069] The heat transfer rate of waste heat from the fuel cell system to the heat pump air conditioning system through the heat exchanger 1 for:

[0070] ;

[0071] In the formula, is the efficiency of heat exchanger 1;

[0072] The heat transfer rate of waste heat from the power battery system to the heat pump air conditioning system through heat exchanger 2 for:

[0073] ;

[0074] In the formula, is the efficiency of heat exchanger 2;

[0075] Heat transfer rate of heat pump air conditioner itself for:

[0076] ;

[0077] In the formula, is the power of the air conditioner, is the maximum power of the heat pump air conditioner. is the maximum cooling capacity of the heat pump air conditioner;

[0078] In a low temperature environment, the factors related to the temperature change of the passenger compartment include the heat pump air conditioner, fuel cell, residual heat of the passenger compartment and vehicle speed. Therefore, the dynamic equation of the heat in the passenger compartment is:

[0079] ;

[0080] In the formula, is the coefficient of influence of speed on the temperature of the passenger compartment;

[0081] The dynamic equation for the temperature of the passenger compartment is:

[0082] ;

[0083] ;

[0084] In the formula, is the density of air, is the volume of the crew compartment, is the heat capacity of the passenger compartment.

[0085] As a further technical solution of the present invention, the steps of establishing the energy and heat optimization management problem include:

[0086] Select control input variables, establish energy and heat optimization management problem description, and determine the constraints of the optimization problem;

[0087] 3.1) Establish energy and heat optimization management problem description:

[0088] This optimization process aims to achieve two goals: one is to keep the battery running within the optimal temperature range to improve its performance and safety; the other is to optimize the power consumption of the thermal management system to improve the energy efficiency of the entire vehicle; by comprehensively considering these two aspects, the battery thermal management system can be optimized to provide a strong guarantee for the reliability and economy of electric vehicles. Therefore, the present invention selects the temperature of the power battery, fuel cell and passenger compartment of the hydrogen fuel cell vehicle and the power battery SOC as state variables, selects the control quantity as the fan control power, water pump control power and heat pump air conditioning control power, and minimizes the optimization target within the global time range. Tracking error between temperature and optimal temperature and control energy consumption: ;

[0089] ;

[0090] in, is the starting time of the prediction time domain, is the end time of the prediction time domain, and the state variable , ,in They are the optimal operating temperatures of the power battery, fuel cell and passenger compartment, and the control input , Q and M are positive definite weight matrices, where Q is , M is , It is the thermal model of the power battery , fuel cell thermal model , passenger compartment thermal model and battery A collection of The prediction horizon refers to the length of future time considered in the control algorithm. Describes the virtual prediction time step From the current time step Starting state and control;

[0091] 3.2) Determine the constraints of the optimization problem:

[0092] The dynamic equations and safety constraints of the power battery temperature need to be met:

[0093] ;

[0094] in, It is the lowest temperature to ensure the safety of the power battery. It is the highest temperature to ensure the safety of the power battery. is the temperature of the power battery at time k, is the initial time, is the initial temperature of the power battery;

[0095] The dynamic equations and safety constraints of the hydrogen fuel cell temperature need to be met:

[0096] ;

[0097] in, It is the minimum temperature to ensure the safety of fuel cells. It is the highest temperature to ensure the safety of fuel cells. is the temperature of the fuel cell at time k, is the initial temperature of the fuel cell, is the initial time;

[0098] The dynamic equations and safety constraints of the cabin temperature need to be met:

[0099] ;

[0100] in, It is the minimum temperature to ensure the comfort of the passenger compartment. It is the highest temperature to ensure the comfort of the passenger compartment. is the temperature of the cabin at time k, is the initial temperature of the passenger compartment, is the initial time;

[0101] The fan cooling control power and water pump cooling control power constraints need to be met:

[0102] ;

[0103] in, is the power of the fuel cell water pump cooling system at time k, is the power of the fuel cell fan cooling system at time k, is the power of the power battery water pump cooling system at time k, is the power of the power battery fan cooling system at time k, is the maximum power of the fuel cell water pump cooling system, is the maximum power of the fuel cell fan cooling system, is the maximum power of the power battery water pump cooling system, is the maximum power of the power battery fan cooling system;

[0104] Need to meet the power battery constraint:

[0105] ;

[0106] in, It is a power battery Minimum value, It is a power battery The highest value, is the power battery at time k , Initial power battery , is the initial time.

[0107] As a further technical solution of the present invention, the steps of solving the problem using reinforcement learning include:

[0108] 4.1) Establish reinforcement learning problem description:

[0109] Combining the optimization problem and its constraints determined above, this complex system control problem is transformed into a problem to be solved under the reinforcement learning framework. Since this problem has multiple constraints and a high-dimensional state space, the actor-critic method is a suitable choice. The transformed reinforcement learning problem is in the form of:

[0110] ;

[0111] ;

[0112] ;

[0113] ;

[0114] yes A collection of It represents all the constraints.

[0115] ;

[0116] They correspond to the actor and critic, respectively, parameterized by θ and w, and usually appear in the form of a neural network;

[0117] ;

[0118] ;

[0119] In the formula, is the control quantity corresponding to time t, is the cost corresponding to time t.

[0120] 4.2) Solving reinforcement learning problems:

[0121] In order to solve the converted RL problem, a policy iteration framework is adopted, in which the two processes of policy evaluation and policy improvement are used to alternately update the actor and the critic. Since the critic update is an unconstrained problem that can be optimized by the ordinary gradient descent method, the main focus is on the actor update, which is challenging due to its large-scale parameter space, nonlinear properties and a large number of state constraints. In order to solve this problem, the present invention handles the constraints by the generalized external point method. The basic idea of ​​the generalized external point method is to transform the inequality constraint into an equality constraint by introducing a "penalty term" or "penalty function", thereby transforming the original constrained optimization problem into a new unconstrained optimization problem. This unconstrained problem can approximate the optimal solution by iteration, in which each step of iteration adds an "external point" to the local approximation of the original optimization problem to ensure that the iteration can gradually approach the optimal solution. The actor with a constrained problem is converted into an unconstrained problem by an external penalty function as follows:

[0122] ;

[0123] ;

[0124] ;

[0125] ;

[0126] in, is the penalty function, is the penalty factor, and then the policy parameters are optimized by performing gradient descent iterations;

[0127] 4.3) Offline training:

[0128] For each iteration number i, there is a critic network and actor networks With random parameters , learning rate , for each environment step execution, ,implement To obtain ,calculate , and finally update the parameters, , ;

[0129] 4.4) Online Application:

[0130] Get the optimal control strategy and optimal evaluation function trained offline , current status ,application Get the best fan power, water pump power and heat pump air conditioning power and the required price.

[0131] As a further technical solution of the present invention, the steps of the parallel computing control strategy based on the vehicle-mounted domain controller include:

[0132] In order to improve the real-time performance and computing efficiency of the algorithm, the present invention introduces a parallel computing control strategy based on the vehicle-mounted domain controller. Based on the vehicle-mounted domain controller of a multi-core processor, the thermoelectric coordinated optimization task of the fuel cell thermal management system is distributed to the three cores of the SoC and the two cores of the MCU in parallel, where SoC refers to System on a Chip, and MCU refers to Microcontroller Unit, also known as a microcontroller. Due to the complexity of the actor network and critic network structure and the limitations of the SoC computing power, directly using SoC for training may face challenges. Therefore, the present invention firstly uses the offline training method to pre-train the actor network and the critic network in a high-performance computing environment to obtain a network model with excellent performance. Next, the SoC is used to train new scenarios, and the speed prediction task and the power prediction task based on the intelligent network connection technology are assigned to the first core CPU0 of the SoC in parallel; the fuel cell integrated thermal management system thermoelectric coupling model is assigned to the second core CPU1 of the SoC in parallel; the actor-critic network training task is assigned to the third and fourth cores CPU2 and CPU3 of the SoC in parallel; at the same time, considering that the real-time requirements of autonomous driving are extremely high, the present invention deploys the trained actor network and critic network models to the MCU, and applies the actor network and critic network models in the MCU. The MCU, with its excellent real-time and low power consumption characteristics, can ensure that these network models respond quickly and accurately in practical applications, thereby meeting the real-time requirements. The actor network application task is assigned to the first core CPU0 of the MCU in parallel; the critic network application task is assigned to the second core CPU1 of the MCU in parallel.

[0133] Compared with the prior art, the present invention has the following beneficial effects:

[0134] Thermal-electrical coupling integrated optimization:

[0135] In low temperature environments, a new integrated thermal management system that considers waste heat recovery can achieve efficient use of the heat and energy generated by fuel cells, significantly reduce energy waste, and extend the service life of fuel cells. Comprehensively considering the coupled thermal and electrical requirements of fuel cell vehicles, the system performance is optimized to improve the vehicle's range, comfort, reliability, and safety.

[0136] Reinforcement Learning Control Strategy:

[0137] The introduction of reinforcement learning technology enables the system to have adaptive learning capabilities and can flexibly adapt to the thermal power coordination requirements under different working conditions and environments. It does not need to rely on accurate vehicle models, which reduces the dependence on model accuracy and reduces the complexity and workload of modeling. It improves the real-time responsiveness and stability of the system, ensuring efficient operation under various conditions.

[0138] Parallel computing control strategy:

[0139] Through efficient training and real-time application of actor networks and critic networks, the real-time performance and computational efficiency of the algorithm are significantly improved. Taking advantage of the vehicle domain controller's ability to simultaneously process and manage multiple different domains in the car, different tasks or task subsets are assigned to each processing core to achieve parallel processing and further improve system performance.

[0140] In summary, the thermal-electric coupling integrated optimization control method of the intelligent networked fuel cell vehicle system integrated thermal management system proposed in the present invention has important innovative significance and application value in the field of thermal-electric coordinated optimization control of fuel cell vehicles. By introducing reinforcement learning control strategy and parallel computing control allocation strategy, it is possible to achieve precise control and efficient utilization of the thermal-electric coordinated needs of fuel cell vehicles, providing effective technical support for improving vehicle performance, reducing energy consumption and reducing emissions. BRIEF DESCRIPTION OF THE DRAWINGS

[0141] Figure 1 A diagram of an integrated thermal management model for a fuel cell vehicle provided in an embodiment of the present invention.

[0142] Figure 2 This is a working mode diagram of the fuel cell integrated thermal management system provided in an embodiment of the present invention.

[0143] Figure 3 A flow chart of a thermal management control strategy for a fuel cell thermoelectric coupling system provided by an embodiment of the present invention.

[0144] Figure 4 This is a thermal management control block diagram of a fuel cell thermoelectric coupling system provided by an embodiment of the present invention.

[0145] Figure 5 A power allocation diagram provided by an embodiment of the present invention.

[0146] Figure 6 A parallel computing control strategy diagram provided for an embodiment of the present invention. DETAILED DESCRIPTION

[0147] The technical solution of the present invention is further described in detail below in conjunction with specific implementation methods.

[0148] The purpose of this invention is to develop a new integrated thermal management model for fuel cells based on a research group's published Figure 1 As shown in the figure, in order to reduce the energy consumed by controlling the battery temperature and improve the solution speed, a control strategy using reinforcement learning to solve the thermal management problem considering thermoelectric coordinated control is proposed. By controlling the power of different cooling methods, the battery temperature can be maintained within a reasonable range, and the battery cooling energy consumption is reduced. The main idea is to first establish a speed prediction model in the rolling time domain based on intelligent network technology, and then construct the optimal control problem of temperature control with the minimum energy consumption of the thermal management system. Due to the large number of system states and constraints, reinforcement learning is considered for solution; then the optimal control problem is converted into a reinforcement learning problem, and the actor-critic method of reinforcement learning is used for solution, and the constraints are handled by the generalized external point method. Among them, the actor is responsible for action output and is used for tracking, while the critic has the function of judging the state goodness and can output the cost of the action. The variables that need to be optimized in the reinforcement learning problem are the parameters of the actor and critic in the form of NN; the parameters of the actor and critic are trained by offline training; through online application, the temperature control problem with the minimum energy consumption can be solved quickly and accurately. At the same time, the present invention can realize efficient training and real-time application of actor network and critic network by introducing parallel computing control strategy, which not only significantly improves the real-time performance and computing efficiency of the algorithm, but also provides strong technical support for various application scenarios with high performance and strict real-time requirements.

[0149] The fuel cell integrated thermal management model is coupled with four subsystems and three loops, namely the fuel cell thermal management subsystem, the power battery thermal management subsystem, the electric drive thermal management subsystem, the passenger compartment thermal management subsystem, the heat pump air conditioning loop, the fuel cell thermal management loop, and the power battery and electric drive system thermal management loop. The working mode of the fuel cell integrated thermal management system is adjusted using logical threshold control. Based on the current ambient temperature, current operating conditions and driver intentions, the working mode is switched according to the temperature of each component, such as Figure 2 In order to improve the low temperature adaptability of the fuel cell, the present invention selects a mode in which the radiator dissipates heat for the fuel cell and the power battery and the passenger compartment is heated. , , , ,in is the temperature of the power battery, is the temperature of the fuel cell, is the temperature of the passenger compartment, The waste heat from the fuel cell system and the power battery system is transferred to the passenger compartment system through a heat exchanger. This design reduces the power consumption of the heat pump air conditioner at low temperatures.

[0150] See also Figure 3 and Figure 4 The invention proposes a thermal-electric integrated optimization control method for an intelligent networked fuel cell vehicle, the method comprising the following steps:

[0151] 1) Establish a speed prediction model for the rolling time domain;

[0152] First, the future speed of fuel cell vehicles is obtained through intelligent network technology. , through the automobile power balance formula, ignoring the slope resistance, the vehicle's driving demand power can be obtained ;

[0153] ;

[0154] Where: - vehicle quality, - rolling resistance coefficient, - air resistance coefficient, - windward area, - Rotational mass conversion factor, - vehicle acceleration;

[0155] The driving power required by the fuel cell and power battery is obtained through the power distribution relationship ,like Figure 5 As shown;

[0156] Since the vehicle drive requires power Only and speed related, therefore See The function of Similarly, the driving power required by fuel cells and power batteries is It's all about speed function.

[0157] 2) The steps to establish a control-oriented battery thermal management system model include:

[0158] Get the traction power required by the fuel cell and power battery Finally, a thermal model of the power battery is established for the thermal management system , fuel cell thermal model , passenger compartment thermal model and battery ;

[0159] 2.1) Thermal model of power battery :

[0160] Gas mass flow rate from the fan cooling duct to the power battery It is expressed as follows:

[0161] ;

[0162] In the formula, is the power of the power battery fan cooling system, It is the maximum power of the power battery fan cooling system, which is usually considered to be a fixed value. is the maximum gas mass flow rate flowing from the fan cooling duct into the power battery;

[0163] Through fan cooling, the heat transfer rate of the power battery to the outside air is for:

[0164] ;

[0165] In the formula, is the heat capacity of air, is the power battery temperature, is the ambient temperature. In this model, it is assumed that the ambient temperature is constant. is the heat transfer coefficient of air, is the contact area between air and battery;

[0166] Mass flow rate of water pump cooling pipe into power battery It is expressed as follows:

[0167] ;

[0168] In the formula, is the power of the power battery water pump cooling system, It is the maximum power of the power battery water pump cooling system, which is usually considered to be a fixed value. is the maximum mass flow rate from the water pump cooling pipe into the power battery;

[0169] Through the water pump cooling method, the heat transfer rate of the power battery to the coolant for:

[0170] ;

[0171] In the formula, is the heat capacity of the coolant, is the coolant temperature. In this model, the surrounding coolant temperature is assumed to be constant. is the heat transfer coefficient of the coolant, is the contact area between the coolant and the battery;

[0172] Total power of the power battery ,in is the power of the air conditioner, and the charge and discharge SOC of the power battery is expressed as:

[0173] ;

[0174] In the formula, Represents the internal resistance of the power battery, Output current for the power battery, is the open circuit voltage, is the total charge of the power battery, and Battery state of charge and temperature, which are usually modeled by experimentally generated MAP tables (i.e., lookup tables), so It is expressed as:

[0175] ;

[0176] The dynamic equation of power battery temperature is:

[0177] ;

[0178] ;

[0179] In the formula, The quality of the power battery; is the thermal capacity of the power battery;

[0180] Heat transfer rate of power battery for:

[0181] ;

[0182] 2.2) Thermal model of fuel cell :

[0183] Gas mass flow rate into the fuel cell from the fan cooling duct It is expressed as follows:

[0184] ;

[0185] In the formula, is the power of the fuel cell fan cooling system, is the maximum power of the fuel cell fan cooling system, which is usually considered to be a fixed value. is the maximum gas mass flow rate flowing into the fuel cell from the fan cooling duct;

[0186] Through fan cooling, the heat transfer rate of the fuel cell to the outside air is for:

[0187] ;

[0188] In the formula, is the fuel cell temperature, is the contact area between air and fuel cell;

[0189] Mass flow rate of water pump cooling pipe into fuel cell It is expressed as follows:

[0190] ;

[0191] In the formula, is the power of the fuel cell water pump cooling system, It is the maximum power of the fuel cell water pump cooling system, which is usually considered to be a fixed value. is the maximum mass flow rate from the water pump cooling pipe into the fuel cell;

[0192] Through water pump cooling, the heat transfer rate of the fuel cell to the coolant for:

[0193] ;

[0194] In the formula, is the contact area between the coolant and the fuel cell;

[0195] Fuel cell charge and discharge current is a function of the fuel cell power and is usually modeled by an experimentally generated MAP table (i.e., a lookup table);

[0196] The dynamic equation for the fuel cell temperature is:

[0197] ;

[0198] ;

[0199] In the formula, is the internal resistance of the fuel cell, is the mass of the fuel cell; is the heat capacity of the fuel cell;

[0200] Heat transfer rate of fuel cells for:

[0201] ;

[0202] 2.3) Passenger compartment heating model ;

[0203] The heat transfer rate of waste heat from the fuel cell system to the heat pump air conditioning system through the heat exchanger 1 for:

[0204] ;

[0205] In the formula, is the efficiency of heat exchanger 1;

[0206] The heat transfer rate of waste heat from the power battery system to the heat pump air conditioning system through heat exchanger 2 for:

[0207] ;

[0208] In the formula, is the efficiency of heat exchanger 2;

[0209] Heat transfer rate of heat pump air conditioner itself for:

[0210] ;

[0211] In the formula, is the maximum power of the heat pump air conditioner. is the maximum cooling capacity of the heat pump air conditioner;

[0212] In a low temperature environment, the factors related to the temperature change of the passenger compartment include the heat pump air conditioner, fuel cell, residual heat of the passenger compartment and vehicle speed. Therefore, the dynamic equation of the heat in the passenger compartment is:

[0213] ;

[0214] In the formula, is the coefficient of influence of speed on the temperature of the passenger compartment;

[0215] The dynamic equation for the temperature of the passenger compartment is:

[0216] ;

[0217] ;

[0218] In the formula, is the density of air, is the volume of the crew compartment, is the heat capacity of the passenger compartment.

[0219] 3) Establish energy and heat optimization management issues;

[0220] Select control input variables, establish energy and heat optimization management problem description, and determine the constraints of the optimization problem;

[0221] 3.1) Establish energy and heat optimization management problem description:

[0222] In the process of optimizing the battery thermal management system, a core concern is the temperature of the battery, because the heat generated by the battery will directly affect its temperature. The role of the integrated thermal management system is to regulate the temperature of the entire vehicle by consuming a certain amount of electrical energy to ensure that it operates within a safe and efficient range. Therefore, when optimizing this system, the battery temperature range and battery power must be strictly limited.

[0223] Specifically, the optimization problem should ensure that the battery temperature is maintained within a preset safety range to avoid overheating or overcooling that has a negative impact on battery performance, life, and safety. At the same time, the battery power limit should also be considered to ensure that the battery power is not over-consumed during the temperature control process, thereby affecting the vehicle's range.

[0224] This optimization process aims to achieve two goals: one is to keep the battery running within the optimal temperature range to improve its performance and safety; the other is to optimize the power consumption of the thermal management system to improve the energy efficiency of the entire vehicle; by comprehensively considering these two aspects, the battery thermal management system can be optimized to provide a strong guarantee for the reliability and economy of electric vehicles. Therefore, the present invention selects the temperature of the power battery, fuel cell and passenger compartment of the hydrogen fuel cell vehicle and the power battery SOC as state variables, selects the control quantity as the fan control power, water pump control power and heat pump air conditioning control power, and minimizes the optimization target within the global time range. Tracking error between temperature and optimal temperature and control energy consumption:

[0225] ;

[0226] ;

[0227] in, is the starting time of the prediction time domain, is the end time of the prediction time domain, and the state variable , ,in They are the optimal operating temperatures of the power battery, fuel cell and passenger compartment, and the control input , Q and M are positive definite weight matrices, where Q is , M is , It is the thermal model of the power battery , fuel cell thermal model , passenger compartment thermal model and battery A collection of The prediction horizon refers to the length of future time considered in the control algorithm. Describes the virtual prediction time step From the current time step Starting state and control;

[0228] 3.2) Determine the constraints of the optimization problem:

[0229] The dynamic equations and safety constraints of the power battery temperature need to be met:

[0230] ;

[0231] in, It is the lowest temperature to ensure the safety of the power battery. It is the highest temperature to ensure the safety of the power battery. is the temperature of the power battery at time k, is the initial temperature of the power battery, is the initial time;

[0232] The dynamic equations and safety constraints of the hydrogen fuel cell temperature need to be met:

[0233] ;

[0234] in, It is the minimum temperature to ensure the safety of fuel cells. It is the highest temperature to ensure the safety of fuel cells. is the temperature of the fuel cell at time k, is the initial temperature of the fuel cell, is the initial time;

[0235] The dynamic equations and safety constraints of the cabin temperature need to be met:

[0236] ;

[0237] in, It is the minimum temperature to ensure the comfort of the passenger compartment. It is the highest temperature to ensure the comfort of the passenger compartment. is the temperature of the cabin at time k, is the initial temperature of the passenger compartment, is the initial time;

[0238] The fan cooling control power and water pump cooling control power constraints need to be met:

[0239] ;

[0240] in, is the power of the fuel cell water pump cooling system at time k, is the power of the fuel cell fan cooling system at time k, is the power of the power battery water pump cooling system at time k, is the power of the power battery fan cooling system at time k, is the maximum power of the fuel cell water pump cooling system, is the maximum power of the fuel cell fan cooling system, is the maximum power of the power battery water pump cooling system, is the maximum power of the power battery fan cooling system;

[0241] Need to meet the power battery constraint:

[0242] ;

[0243] in, It is a power battery Minimum value, It is a power battery The highest value, is the power battery at time k , Initial power battery , is the initial time.

[0244] 4) Solve using reinforcement learning:

[0245] 4.1) Establish reinforcement learning problem description:

[0246] Combining the optimization problem and its constraints determined above, this complex system control problem is transformed into a problem to be solved under the reinforcement learning framework. Since this problem has multiple constraints and a high-dimensional state space, the actor-critic method is a suitable choice. The transformed reinforcement learning problem is in the form of:

[0247] ;

[0248] ;

[0249] ;

[0250] ;

[0251] yes A collection of It represents all the constraints.

[0252] ;

[0253] They correspond to the actor and critic, respectively, parameterized by θ and w, and usually appear in the form of a neural network;

[0254] ;

[0255] ;

[0256] In the formula, is the control quantity corresponding to time t, is the cost corresponding to time t.

[0257] 4.2) Solving reinforcement learning problems:

[0258] In order to solve the converted RL problem, a policy iteration framework is adopted, in which the two processes of policy evaluation and policy improvement are used to alternately update the actor and the critic. Since the critic update is an unconstrained problem that can be optimized by the ordinary gradient descent method, the main focus is on the actor update, which is challenging due to its large-scale parameter space, nonlinear properties and a large number of state constraints. In order to solve this problem, the present invention handles the constraints by the generalized external point method. The basic idea of ​​the generalized external point method is to transform the inequality constraint into an equality constraint by introducing a "penalty term" or "penalty function", thereby transforming the original constrained optimization problem into a new unconstrained optimization problem. This unconstrained problem can approximate the optimal solution by iteration, in which each step of iteration adds an "external point" to the local approximation of the original optimization problem to ensure that the iteration can gradually approach the optimal solution. The actor with a constrained problem is converted into an unconstrained problem by an external penalty function as follows:

[0259] ;

[0260] ;

[0261] ;

[0262] ;

[0263] in, is the penalty function, is the penalty factor, and then the policy parameters are optimized by performing gradient descent iterations;

[0264] 4.3) Offline training:

[0265] For each iteration number i, there is a critic network and actor networks With random parameters , learning rate , for each environment step execution, ,implement To obtain ,calculate , and finally update the parameters, , ;

[0266] 4.4) Online Application:

[0267] Get the optimal control strategy and optimal evaluation function trained offline , current status ,application Get the best fan power, water pump power and heat pump air conditioning power and the required price.

[0268] 5) Parallel computing control strategy based on vehicle-mounted domain controller.

[0269] In order to improve the real-time performance and computational efficiency of the algorithm, the present invention introduces a parallel computing control strategy based on the vehicle-mounted domain controller, such as Figure 6 As shown in the figure. The on-board domain controller based on a multi-core processor distributes the thermoelectric co-optimization task of the fuel cell thermal management system to the three cores of the SoC and the two cores of the MCU in parallel. SoC refers to System on a Chip, and MCU refers to Microcontroller Unit, also known as microcontroller. Due to the complexity of the actor network and critic network structure and the limitations of SoC computing power, direct training using SoC may face challenges. Therefore, the present invention firstly uses the offline training method to pre-train the actor network and the critic network in a high-performance computing environment to obtain a network model with excellent performance. Next, the SoC is used to train new scenarios, and the speed prediction task and the power prediction task based on the intelligent network connection technology are assigned to the first core CPU0 of the SoC in parallel; the fuel cell integrated thermal management system thermoelectric coupling model is assigned to the second core CPU1 of the SoC in parallel; the actor-critic network training task is assigned to the third and fourth cores CPU2 and CPU3 of the SoC in parallel; at the same time, considering that the real-time requirements of autonomous driving are extremely high, the present invention deploys the trained actor network and critic network models to the MCU, and applies the actor network and critic network models in the MCU. The MCU, with its excellent real-time and low power consumption characteristics, can ensure that these network models respond quickly and accurately in practical applications, thereby meeting the real-time requirements. The actor network application task is assigned to the first core CPU0 of the MCU in parallel; the critic network application task is assigned to the second core CPU1 of the MCU in parallel.

[0270] The present invention aims to make full use of the respective advantages of SoC and MCU through parallel computing control strategies, and realize parallel processing by allocating different tasks or task subsets to each processing core, thereby significantly improving the overall performance and the real-time performance of control. The key to parallel computing task allocation is how to efficiently schedule and manage the resources of multiple processing cores to ensure that the system can handle various tasks in the most optimized way. The present invention realizes efficient training and real-time application of actor networks and critic networks. This not only improves the real-time performance and computing efficiency of the algorithm, but also provides strong technical support for various high-performance and strict real-time application scenarios.

[0271] It should be noted that, in this article, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, an element defined by the sentence "comprises a ..." does not exclude the existence of other identical elements in the process, method, article or device including the element.

[0272] The above are only preferred embodiments of the present invention, and are not intended to limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A thermal-electric integrated optimization control method for an intelligent networked fuel cell vehicle, characterized in that: The method comprises the following steps: 1) Establish a speed prediction model for the rolling time domain; 2) Establish a control-oriented battery thermal management system model; 3) Establish energy and heat optimization management issues; 4) Use reinforcement learning to solve; 5) Parallel computing control strategy based on vehicle-mounted domain controller; The steps to establish a speed prediction model for the rolling time domain include: First, the future speed v of the fuel cell vehicle is obtained through intelligent network connection technology, and the driving demand power p of the vehicle is obtained through the vehicle power balance formula. req ; Where: m-vehicle mass, f-rolling resistance coefficient, C D - Air resistance coefficient, A-frontal area, δ-rotational mass conversion factor, - vehicle acceleration; The driving power P required by the fuel cell and power battery is obtained through the power distribution relationship fc-req , P bat-req ; Since the vehicle driving power required P req It is only related to the speed v, so P req As a function of v, P req =f(v), similarly, the driving power P required by the fuel cell and power battery fc-req , P bat-req are all considered as functions of velocity v; The steps to establish a control-oriented battery thermal management system model include: Get the traction power P that the fuel cell and power battery need to provide fc-req , P bat-req Finally, a thermal model of the power battery is established for the thermal management system. bat , fuel cell thermal model fc 、Civil compartment thermal model cab and battery soc ; 2.1) Thermal model of power battery bat : Gas mass flow rate from the fan cooling duct to the power battery It is expressed as follows: Where P fan-bat is the power of the power battery fan cooling system, P fan-bat-max is the maximum power of the power battery fan cooling system, is the maximum gas mass flow rate flowing from the fan cooling duct into the power battery; Through fan cooling, the heat transfer rate of the power battery to the outside air is for: In the formula, C air is the heat capacity of air, T bat is the power battery temperature, T air is the ambient temperature, h air is the heat transfer coefficient of air, A air is the contact area between air and battery; Mass flow rate of water pump cooling pipe into power battery It is expressed as follows: Where P pump-bat is the power of the power battery water pump cooling system, P pump-bat-max is the maximum power of the power battery water pump cooling system, is the maximum mass flow rate from the water pump cooling pipe into the power battery; Through the water pump cooling method, the heat transfer rate of the power battery to the coolant for: In the formula, C water is the heat capacity of the coolant, T water is the coolant temperature, h water is the heat transfer coefficient of the coolant, A water is the contact area between the coolant and the battery; The total power of the power battery P bat =P fan-bat +P pump-bat +P air +P bat-req , where P air is the power of the air conditioner, and the charge and discharge SOC of the power battery is expressed as: In the formula, R bat Represents the internal resistance of the power battery, I bat is the power battery output current, U oc is the open circuit voltage, N bat is the total charge of the power battery, R bat and U oc They are all functions of the battery state of charge SOC and temperature. The functions are modeled by the MAP table generated by the experiment, so the SOC is expressed as; The dynamic equation of power battery temperature is: In the formula, m bat is the mass of the power battery; C bat is the thermal capacity of the power battery; Heat transfer rate of power battery for: 2.2) Thermal model of fuel cell fc : Gas mass flow rate into the fuel cell from the fan cooling duct It is expressed as follows: Where P fan-fc is the power of the fuel cell fan cooling system, P fan-fc-max is the maximum power of the fuel cell fan cooling system, is the maximum gas mass flow rate flowing into the fuel cell from the fan cooling duct; Through fan cooling, the heat transfer rate of the fuel cell to the outside air is for: Where, T fc is the fuel cell temperature, A air-fc is the contact area between air and fuel cell; Mass flow rate of water pump cooling pipe into fuel cell It is expressed as follows: Where P pump-fc is the power of the fuel cell water pump cooling system, P pump-fc-max is the maximum power of the fuel cell water pump cooling system, is the maximum mass flow rate from the water pump cooling pipe into the fuel cell; Through water pump cooling, the heat transfer rate of the fuel cell to the coolant for: In the formula, A water-fc is the contact area between the coolant and the fuel cell; Fuel cell charge and discharge current I fc is a function of the fuel cell power, modeled by an experimentally generated MAP table; The dynamic equation for the fuel cell temperature is: In the formula, R fc is the internal resistance of the fuel cell, m fc is the mass of the fuel cell; C fc is the heat capacity of the fuel cell; Heat transfer rate of fuel cells for: 2.3) Passenger compartment heating model f cab ; The heat transfer rate of waste heat from the fuel cell system to the heat pump air conditioning system through the heat exchanger 1 for: In the formula, ε fc is the efficiency of heat exchanger 1; The heat transfer rate of waste heat from the power battery system to the heat pump air conditioning system through heat exchanger 2 for: In the formula, ε bat is the efficiency of heat exchanger 2; Heat transfer rate of heat pump air conditioner itself for: Where P air-max is the maximum power of the heat pump air conditioner, w air is the maximum cooling capacity of the heat pump air conditioner; In a low temperature environment, the factors related to the temperature change of the passenger compartment include the heat pump air conditioner, fuel cell, residual heat of the passenger compartment and vehicle speed. Therefore, the dynamic equation of the heat in the passenger compartment is: In the formula, ε v is the coefficient of influence of speed on the temperature of the passenger compartment; The dynamic equation for the temperature of the passenger compartment is: Where ρ is the density of air, V is the volume of the passenger compartment, and C p is the heat capacity of the crew compartment; The steps to formulate the energy and heat optimization management problem include: Select control input variables, establish energy and heat optimization management problem description, and determine the constraints of the optimization problem; 3.1) Establish energy and heat optimization management problem description: The optimization process aims to achieve two goals: one is to keep the battery running within the optimal temperature range to improve its performance and safety; the other is to optimize the power consumption of the thermal management system to improve the energy efficiency of the whole vehicle. The temperature of the power battery, fuel cell and passenger compartment of the hydrogen fuel cell vehicle and the power battery SOC are selected as state variables, and the control variables are selected as fan control power, water pump control power and heat pump air conditioning control power. The optimization goal is to minimize the global time range [t0, t f Tracking error between temperature and optimal temperature and control energy consumption: x i+1|t =F(x i|t ,u i|t ,v); Among them, t0 is the starting time of the prediction time domain, t f is the end time of the prediction time domain, the state variable x i|t =[T bat , T fc , T cab , SOC] T , in are the optimal operating temperatures of the power battery, fuel cell and passenger compartment, respectively, and the control input u i|t =[P pump-fc , P fan-fc , P pump-bat , P fan-bat , P air ] T , Q and M are positive definite weight matrices, where Q is M is F is the thermal model of the power battery bat , fuel cell thermal model fc 、Civil compartment thermal model cab and battery soc The set of T is the prediction time domain, which refers to the future time length considered in the control algorithm, and x i|t ,u i|t Describes the state and control starting from the current time step t in the virtual prediction time step i; 3.2) Determine the constraints of the optimization problem: The dynamic equations and safety constraints of the power battery temperature need to be met: Among them, T bat-min It is the lowest temperature to ensure the safety of the power battery. bat-max It is the highest temperature to ensure the safety of the power battery. bat (k) is the temperature of the power battery at time k, T bat-init is the initial temperature of the power battery, t0 is the initial time; The dynamic equations and safety constraints of the hydrogen fuel cell temperature need to be met: Among them, T fc-min is the minimum temperature to ensure the safety of the fuel cell, T fc-max is the maximum temperature to ensure the safety of the fuel cell, T fc (k) is the temperature of the fuel cell at time k, T fc-init is the initial temperature of the fuel cell, t0 is the initial time; The dynamic equations and safety constraints of the cabin temperature need to be met: Among them, T cab-min is the minimum temperature to ensure the comfort of the passenger compartment, T cab-max is the maximum temperature to ensure the comfort of the passenger compartment, T cab (k) is the temperature of the cabin at time k, T cab-init is the initial temperature of the passenger compartment, t0 is the initial time; The fan cooling control power and water pump cooling control power constraints need to be met: Among them, P pump-fc (k) is the power of the fuel cell water pump cooling system at time k, P fan-fc (k) is the power of the fuel cell fan cooling system at time k, P pump-bat (k) is the power of the power battery water pump cooling system at time k, P fan-bat (k) is the power of the power battery fan cooling system at time k, P pump-fc-max is the maximum power of the fuel cell water pump cooling system, P fan-fc-max is the maximum power of the fuel cell fan cooling system, P pump-bat-max is the maximum power of the power battery water pump cooling system, P fan-bat-max is the maximum power of the power battery fan cooling system; The power battery SOC constraints need to be met: Among them, SOC min It is the minimum SOC value of the power battery. max is the maximum SOC value of the power battery, SOC(k) is the SOC of the power battery at time k, SOC init is the initial SOC of the power battery, t0 is the initial time; The steps to solve the problem using reinforcement learning include: 4.1) Establish reinforcement learning problem description: The actor-critic method is a suitable choice. The transformed reinforcement learning problem is: s.t.x i+1|t =f(x i|t ,π θ (x i|t ),v); g e (x i|t )≥0; f is f bat 、f fc 、f cab and f soc The collection of g e (x i|t ) represents all the constraints, Where Q and M are positive definite weight matrices, where Q is M is π θ and v w They correspond to the actor and critic, respectively, parameterized by θ and w; the t =π θ (x) i|t ,v); J=v w (x i|t ,u t ,v); In the formula, u t is the control amount corresponding to time t, J is the cost corresponding to time t; 4.2) Solve the reinforcement learning problem: A policy iteration framework is adopted, in which the two processes of policy evaluation and policy improvement are used to update the actor and critic alternately. Since the critic update is an unconstrained problem optimized by the ordinary gradient descent method, the main focus is on the actor update, which is challenging due to its large-scale parameter space, nonlinear properties and a large number of state constraints; the constraints are handled by the generalized external point method. The basic idea of ​​the generalized external point method is to transform the inequality constraints into equality constraints by introducing a "penalty term" or "penalty function", thereby transforming the original constrained optimization problem into a new unconstrained optimization problem. This unconstrained problem approximates the optimal solution through iteration, in which each iteration adds an "external point" to the local approximation of the original optimization problem to ensure that the iteration gradually approaches the optimal solution. The external penalty function is used to transform the actor with a constrained problem into an unconstrained problem as follows: s.t.x i+1|t =f(x i|t ,π θ (x i|t ),v); in, is the penalty function, ρ is the penalty factor, and then the policy parameters are optimized by performing gradient descent iterations; 4.3) Offline training: For each iteration number i, there is a critic network v w and actor network π θ with random parameters w, θ, and learning rate β w and β θ , for each environment step execution, u t =π θ (x i|t ,v),J t =v w (x i|t ,v), execute u t To get x t+1 , calculate J ρ and J critic , and finally update the parameters, 4.4) Online Application: Get the optimal control strategy and optimal evaluation function trained offline and Current state x i|t ,application and J * =v w (x i|t ,v) Obtain the optimal fan power, water pump power and heat pump air conditioning power and the required cost.

2. The thermal-electric integrated optimization control method for an intelligent networked fuel cell vehicle according to claim 1, characterized in that: The steps of the parallel computing control strategy based on the vehicle domain controller include: A parallel computing control strategy is introduced based on the vehicle domain controller. Based on the vehicle domain controller of a multi-core processor, the thermoelectric coordinated optimization task of the fuel cell thermal management system is assigned to the four cores of the SoC and the two cores of the MCU in parallel, where SoC refers to System on a Chip and MCU refers to Microcontroller Unit. The actor network and the critic network are pre-trained in a high-performance computing environment by offline training. The SoC is used to train new scenarios, and the speed prediction task and the power prediction task based on intelligent network connection technology are assigned to the first core CPU0 of the SoC in parallel; the thermoelectric coupling model of the fuel cell integrated thermal management system is assigned to the second core CPU1 of the SoC in parallel; the actor-critic network training task is assigned to the third and fourth cores CPU2 and CPU3 of the SoC in parallel; at the same time, the trained actor network and critic network models are deployed on the MCU, and the actor network and critic network models are applied in the MCU, and the actor network application task is assigned to the first core CPU0 of the MCU in parallel; and the critic network application task is assigned to the second core CPU1 of the MCU in parallel.

Images