Attenuation adaptive energy management method for intelligent networked fusion fuel cell vehicle

Abstract

This invention relates to the field of fuel cell vehicle energy management technology, providing a degradation-adaptive energy management method for fuel cell vehicles that integrates intelligent connectivity. It collects driving information and constructs a vehicle speed prediction model based on a long short-term memory neural network; establishes a health state estimation model for the fuel cell and lithium-ion battery; builds a deep reinforcement learning DDPG network and constructs a multi-objective reward function that integrates durability and economy; deploys the converged optimal strategy model to the vehicle controller to achieve real-time power allocation; and continuously collects real-vehicle operating data through a cloud platform, periodically iteratively optimizing the strategy model offline. This method achieves adaptive energy management strategy to dynamic operating conditions and self-adjustment to power system degradation, significantly improving the economy and durability of fuel cell vehicles, and has significant theoretical and engineering application value.

Description

Technical Field

[0001] This invention belongs to the field of fuel cell vehicle energy management technology, and particularly relates to a degradation adaptive fuel cell vehicle energy management method integrating intelligent connectivity. Background Technology

[0002] Fuel cell vehicles offer advantages such as being green and clean, having high energy conversion efficiency, and short refueling time, but they still suffer from poor durability and high cost. Because fuel cells themselves have a slow dynamic response speed, they struggle to quickly respond to sudden changes in operating conditions and high power output. Therefore, hydrogen fuel cell vehicles often incorporate power batteries (i.e., lithium-ion batteries) to create an electric-electric hybrid power system, combining the long range of fuel cells with the fast response of lithium-ion batteries.

[0003] The energy management strategy (EMS) of fuel cell vehicles, as the core control strategy of the powertrain, is responsible for the rational allocation of power among different power sources, directly determining the vehicle's energy economy and powertrain lifespan. Current energy management strategies are mainly divided into rule-based, optimization-based, and learning-based strategies. However, existing technologies still have the following shortcomings:

[0004] (1) Poor passivity and poor adaptability of strategies: Existing strategies are mostly based on the real-time operating parameters of the vehicle to make passive power distribution. When facing harsh conditions such as rapid acceleration and deceleration and long-distance slopes, the fuel cell is prone to frequent load changes and high power instantaneous output, and the lithium-ion battery is deeply charged and discharged. This not only increases hydrogen consumption and reduces efficiency, but also accelerates the aging of the fuel cell stack and lithium-ion battery, resulting in an imbalance between economy and durability.

[0005] (2) Lack of quantitative control over durability: Existing strategies for durability optimization mostly remain at the level of qualitative physical constraints, without introducing core quantitative evaluation indicators for durability, and lack accurate estimation of the state of health (SOH) of fuel cell stacks. This makes it impossible for strategies to actively control the stack degradation rate, and also impossible to quantitatively assess the impact of power distribution on lithium-ion battery aging, which easily leads to the problem of local optima in economics but significant degradation in durability.

[0006] (3) Insufficient multi-objective collaborative optimization capability: Energy management of fuel cell vehicles needs to meet multiple objective constraints such as minimizing hydrogen consumption, suppressing the degradation of fuel cells and lithium-ion batteries, and stabilizing the state of charge (SOC) of lithium-ion batteries. Existing strategies have shortcomings in multi-objective optimization, either focusing only on hydrogen consumption or not including SOH in the core optimization objective, and the weighting coefficient design fails to take into account the dynamic characteristics of intelligent connected vehicle operation, making it difficult to achieve multi-objective collaborative optimization.

[0007] Therefore, there is an urgent need for an intelligent energy management strategy that can adapt to complex operating conditions, take into account multi-objective optimization, and comply with the physical constraints of the power system, so as to improve the overall performance of fuel cell vehicles. Summary of the Invention

[0008] The purpose of this invention is to provide an energy management method for adaptive fuel cell vehicles with integrated intelligent connectivity, aiming to solve the problems mentioned in the background art.

[0009] The present invention is implemented as follows: A degradation-adaptive energy management method for fuel cell vehicles integrating intelligent connectivity is adopted, comprising the following steps:

[0010] Step 1: Vehicle speed prediction model construction;

[0011] The vehicle collects driving information through intelligent connected vehicle wireless communication technology, and builds a vehicle speed prediction model based on long short-term memory neural network. The vehicle speed prediction model is used to predict the future driving speed and then calculate the vehicle's power demand.

[0012] Step 2: Decaying adaptive health state estimation;

[0013] The health status of fuel cells and lithium-ion batteries is estimated. The cumulative degradation of fuel cells is estimated based on an empirical model of fuel cell efficiency degradation, and their health status is defined. At the same time, the total aging degradation rate of lithium-ion batteries is estimated based on calendar aging model and cycle aging model.

[0014] Step 3: Building the Deep Reinforcement Learning DDPG Network Architecture;

[0015] An energy management strategy model is constructed based on the Deep Deterministic Policy Gradient (DDPG) algorithm, and the health state of the fuel cell is introduced as a state variable into the state space of the model. The state space also includes the vehicle speed, the real-time state of charge of the lithium-ion battery, and the real-time power change rate of the fuel cell.

[0016] Step 4: Constructing the multi-objective reward function;

[0017] Construct a multi-objective weighted reward function that includes equivalent hydrogen consumption cost, fuel cell stack degradation cost, lithium-ion battery degradation cost, and lithium-ion battery state of charge deviation cost;

[0018] Step 5: Energy allocation and iterative optimization;

[0019] The optimal strategy model after training convergence is deployed to the vehicle controller, which collects vehicle information in real time and inputs the vehicle speed prediction model and the optimal strategy model to output the optimal output power command of the fuel cell to complete the power allocation; and uses the cloud to perform offline iterative training and updates of the optimal strategy model.

[0020] A further technical solution involves preprocessing and noise reduction of the collected driving information in step 1: using singular spectrum analysis to eliminate random noise from sensors and communication, and removing outliers; and determining the speed prediction time step and prediction steps based on vehicle speed and road conditions.

[0021] A further technical solution is to select the empirical model "four-factor model" as the fuel cell efficiency degradation model in step 2. The expression corresponding to this empirical model is shown in formula (1):

[0022] (1)

[0023] In the formula, The degradation rate of the fuel cell, The upper limit of degradation defined for fuel cells from initial use to the end of their lifespan. For acceleration coefficient, , , and The attenuation rate is characterized under typical operating conditions such as dynamic loading, start-stop process, idling operation, and high power output. , , and These are the number of times dynamic loading is performed, the number of start-stop cycles, the idling time, and the high-power output time, respectively.

[0024] In a further technical solution, in step 2, the calendar aging model and the cyclic aging model are as follows:

[0025] (2)

[0026] (3)

[0027] (4)

[0028] In the formula, Calendar aging decay rate; Thermodynamic temperature; This refers to the state of charge of a lithium-ion battery. The rate of degradation due to cycle aging of lithium-ion batteries; This refers to the current of a lithium-ion battery. This is the current factor for lithium-ion batteries, with a value of 1. This represents the total aging degradation rate.

[0029] A further technical solution involves, in step 2, determining the cumulative degradation of the fuel cell based on an empirical model of fuel cell efficiency degradation. The health state of a fuel cell is determined by its operating conditions, as shown in formula (5). As in formula (6):

[0030] (5)

[0031] (6)

[0032] The polarization curve of the fuel cell voltage is described by the following empirical formula equation, see formula (7):

[0033] (7)

[0034] In the formula, For Nernst voltage, Activation voltage, For ohmic voltage, Concentration voltage, This represents the total output voltage of the fuel cell stack. This refers to the number of individual cells in a fuel cell stack. The operating voltage of a single fuel cell unit;

[0035] As the health of the fuel cell declines, the internal resistance of the fuel cell increases. Will follow The changes are shown in formula (8):

[0036] (8)

[0037] In the formula, The initial internal resistance, This represents the maximum increase in internal resistance caused by fuel cell degradation; Substituting the polarization curve, we obtain The voltage expression under the influence is shown in formula (9):

[0038] (9)

[0039] As the health of the fuel cell declines, its maximum output power, efficiency and variable load capacity will decrease, and the hydrogen consumption rate will increase. The hydrogen consumption rate is shown in formula (10), the fuel cell output power is shown in formula (11), and the adaptive calculation of the fuel cell system efficiency after degradation is shown in formula (12).

[0040] (10)

[0041] (11)

[0042] (12)

[0043] In the formula, This represents the theoretical hydrogen consumption mass rate. Here is the molar mass of hydrogen. It is Faraday's constant. For real-time power output of fuel cells, This refers to the output voltage of a fuel cell under specific current and health conditions. Because of the low calorific value of hydrogen, This refers to the operating efficiency of the fuel cell system.

[0044] A further technical solution is that, in step 3, the state space Defined as:

[0045] (13)

[0046] in, For the total vehicle speed, For the health status of the power system of fuel cell vehicles, This refers to the real-time state of charge of a lithium-ion battery. This represents the real-time power change rate of the fuel cell.

[0047] Action space Select the real-time output power of the fuel cell As an action variable, it forms part of a continuous action space and is defined as follows:

[0048] (14)

[0049] in, This refers to the idle power of the fuel cell. The rated maximum power of the fuel cell, the output power of the lithium-ion battery Power demand of the whole vehicle in real time Real-time output power of fuel cells The difference is determined, that is ,in, For fuel cell DC / DC converter efficiency.

[0050] A further technical solution involves defining a reward function in step 4. for:

[0051] (15)

[0052] in, This is the equivalent hydrogen consumption cost; To account for the degradation cost of fuel cell stacks, the degradation characteristics under low power, high power, and power-variable loads are considered. , , and These are the degradation coefficients for low power, high power, and power-variable loads, respectively. , These refer to the operating times of the fuel cell at low and high power, respectively. The cost of lithium-ion battery degradation. , The charge / discharge rate of lithium-ion batteries, and These are the curve fitting coefficients; This is the reference SOC value for lithium-ion batteries. The deviation of SOC from the reference value is used to constrain SOC to remain stable within a reasonable range; As a weighting factor for hydrogen consumption, As a weighting factor for fuel cell degradation, As a weighting factor for lithium-ion battery degradation, This is the weighting factor for battery SOC fluctuations.

[0053] The energy management method for adaptive fuel cell vehicles with integrated intelligent connectivity provided in this invention has the following beneficial effects:

[0054] (1) Improve adaptability and economy of operating conditions: By integrating intelligent connected vehicle speed prediction, the energy management strategy is forward-looking and can optimize the power output of fuel cells in advance according to future operating conditions. This avoids the inefficient power distribution of traditional passive strategies under rapid acceleration and deceleration conditions, effectively reducing the hydrogen consumption of the whole vehicle and improving fuel economy.

[0055] (2) Extend the life cycle durability of the power system: For the first time, the health status of the fuel cell stack and lithium-ion battery is introduced and quantified within the DDPG framework, and is taken as the core optimization target. Through adaptive degradation modeling and targeted reward function design, the frequent load changes and high-power operation of the fuel cell are actively reduced, the overcharging and over-discharging of the lithium-ion battery are suppressed, the aging rate of the power system is significantly slowed down, and the maintenance and replacement costs throughout the life cycle are reduced.

[0056] (3) Achieving multi-objective synergistic optimization of economy and durability: A multi-objective reward function integrating hydrogen consumption, fuel cell degradation, lithium-ion battery degradation and SOC stability was constructed and harmonized by weighting factors. This solved the problem of the decline in the overall system performance caused by the single objective optimization in the traditional strategy and achieved the optimal overall performance of the vehicle.

[0057] (4) Possesses continuous learning and evolution capabilities: Through the iterative optimization mechanism of “vehicle-cloud” collaboration, the energy management strategy can continuously learn from the accumulated real vehicle operation data, adapt to the driving habits of different users and the driving conditions in different regions, and the generalization ability and practicality of the strategy are significantly enhanced. Attached Figure Description

[0058] Figure 1 A flowchart illustrating the energy management method for attenuation adaptive fuel cell vehicles integrating intelligent connectivity, as provided in an embodiment of the present invention.

[0059] Figure 2 A diagram showing the vehicle speed prediction process;

[0060] Figure 3 The output characteristic curves of the fuel cell are shown below (where a is the polarization curve of the fuel cell, b is the operating efficiency curve of the fuel cell, and c is the hydrogen consumption rate curve). Detailed Implementation

[0061] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0062] The specific implementation of the present invention will be described in detail below with reference to specific embodiments.

[0063] like Figure 1 As shown, an embodiment of the present invention provides a degradation adaptive fuel cell vehicle energy management method integrating intelligent connectivity, which includes the following steps:

[0064] Step 1: Information collection and speed prediction model construction for intelligent connected vehicles using wireless communication technology;

[0065] Fuel cell vehicles interact with roadside equipment via onboard wireless communication modules and sensors such as millimeter-wave radar and cameras to collect driving information. The collected information undergoes preprocessing and noise reduction: singular spectrum analysis is used to eliminate random noise from sensors and communications, removing outliers to ensure the accuracy of forward-looking information; based on vehicle speed and road conditions, the speed prediction time step and number of prediction steps are determined, providing standardized input for subsequent multi-step speed predictions.

[0066] Vehicle speed information reflects the driving status of fuel cell vehicles under different operating conditions. Knowing the vehicle speed in advance allows for adjustments to the energy management strategy when allocating energy, thus improving fuel cell efficiency and reducing hydrogen consumption. Therefore, vehicle speed information plays a crucial role in improving the fuel economy of fuel cell vehicles during the formulation of energy management strategies. Furthermore, Long Short-Term Memory (LSTM) neural networks, as a novel type of recurrent neural network structure, compare and analyze current inputs with past inputs, retaining or updating network parameters, making them more suitable for processing and predicting time series problems.

[0067] At the vehicle speed prediction level (see details for process) Figure 2First, vehicle speed information from the intelligent connected vehicle information system is collected to establish a database. Then, a vehicle speed prediction model based on a long short-term memory neural network is constructed based on the database. Road driving data collected in the connected vehicle information system is used as the training and testing set for the prediction algorithm. The historical vehicle speed and acceleration sequences are used as network inputs, and the output is the predicted future vehicle speed. Through multiple training optimizations, the prediction accuracy of the algorithm is improved. After obtaining the future speed sequence, the vehicle's power demand is calculated by combining the predicted speed information.

[0068] Step 2: Decaying adaptive health state estimation;

[0069] In fuel cell vehicle energy management research, improper energy allocation not only accelerates the aging and degradation of fuel cells and lithium-ion batteries but also significantly impacts the health status of fuel cells. If power allocation is performed based on an energy management strategy assuming perfect health, the fuel cell's economy and durability will inevitably decline, with output voltage and operating efficiency decreasing to varying degrees, leading to increased hydrogen consumption. Therefore, to ensure that the power system of fuel cell vehicles always operates at its optimal state, it is necessary to estimate the health status of the power system. This is crucial for improving the durability and economy of fuel cell vehicles throughout their entire lifecycle.

[0070] The empirical model "four-factor model" is selected as the fuel cell efficiency degradation model. It is based on the degradation rate of the fuel cell under four operating conditions: dynamic loading, start-stop cycle, idling, and high power output. The expression corresponding to this empirical model is shown in formula (1):

[0071] (1)

[0072] In the formula, The degradation rate of the fuel cell, The upper limit of degradation defined for fuel cells from initial use to the end of their lifespan. For acceleration coefficient, , , and The attenuation rate is characterized under typical operating conditions such as dynamic loading, start-stop process, idling operation, and high power output. , , and These are the number of dynamic loading cycles, the number of start-stop cycles, the idling time, and the high-power output time, which are determined by specific operating conditions.

[0073] The calendar aging model related to the storage temperature and SOC of the lithium-ion battery and the cycle aging model related to the charge and discharge current of the lithium-ion battery are selected as the lithium-ion battery degradation models, as shown in formulas (2), (3) and (4):

[0074] (2)

[0075] (3)

[0076] (4)

[0077] In the formula, Calendar aging decay rate; Thermodynamic temperature; This refers to the state of charge of a lithium-ion battery. The rate of degradation due to cycle aging of lithium-ion batteries; This refers to the current of a lithium-ion battery. This is the current factor for lithium-ion batteries, with a value of 1. This represents the total aging degradation rate of lithium-ion batteries.

[0078] Based on an empirical model of fuel cell efficiency degradation, the cumulative degradation of fuel cells... The health state of a fuel cell is determined by its operating conditions, as shown in formula (5). As in formula (6):

[0079] (5)

[0080] (6)

[0081] The polarization curve of the fuel cell voltage can be described by the following empirical formula equation, see formula (7):

[0082] (7)

[0083] In the formula, For Nernst voltage, Activation voltage, For ohmic voltage, Concentration voltage, This represents the total output voltage of the fuel cell stack. This refers to the number of individual cells in a fuel cell stack. The operating voltage of a single fuel cell unit;

[0084] As the health of the fuel cell declines, the internal resistance of the fuel cell increases. Will follow The changes are shown in formula (8):

[0085] (8)

[0086] In the formula, The initial internal resistance, This represents the maximum increase in internal resistance caused by fuel cell degradation; Substituting the polarization curve, we can obtain... The voltage expression under the influence is shown in Equation (9), which is used to describe the adaptive model of voltage decay.

[0087] (9)

[0088] As the health of the fuel cell declines, its maximum output power, efficiency and variable load capacity will decrease, and the hydrogen consumption rate will increase. The hydrogen consumption rate is shown in formula (10), the fuel cell output power is shown in formula (11), and the adaptive calculation of the fuel cell system efficiency after degradation is shown in formula (12).

[0089] (10)

[0090] (11)

[0091] (12)

[0092] In the formula, This represents the theoretical hydrogen consumption mass rate. Here is the molar mass of hydrogen. It is Faraday's constant. For real-time power output of fuel cells, This refers to the output voltage of a fuel cell under specific current and health conditions. Because of the low calorific value of hydrogen, This refers to the operating efficiency of the fuel cell system.

[0093] Therefore, once the current health state of the fuel cell is determined, the output characteristics of the stack at the current health level can be obtained. Based on the current adaptive decay calculation formula, the fuel cell can be calculated under different conditions. The output characteristic curve below, such as Figure 3 This is crucial for the proposed adaptive health state energy management strategy. By periodically estimating the health state of the fuel cell and lithium-ion battery, a strategy that aligns with current energy status can be achieved. The external output characteristics, such as Figure 3 The polarization curve of fuel cell a Figure 3 The operating efficiency curve of fuel cell b Figure 3 The hydrogen consumption rate curve, etc., are shown in Figure c. When the on-board controller detects the fuel cell vehicle powertrain... After a certain percentage decrease, the energy management strategy is automatically adjusted to match the current health status of the powertrain based on the reduced output characteristics. Simultaneously, the powertrain system under the new strategy is continuously monitored. The process of leveling up and continuously cycling through this process enables an adaptive energy management strategy that gradually decreases.

[0094] Step 3: Building the Deep Reinforcement Learning DDPG Network Architecture;

[0095] A long short-term memory neural network is used to predict vehicle speed and obtain future driving conditions, which serve as the input to the state space of the energy management strategy. Then, based on the DDPG energy management strategy, the optimal power allocation for the fuel cell vehicle is obtained. (Fuel cell stack...) The level is a key factor affecting power energy distribution; the operating efficiency and output power of a fuel cell stack are directly affected by its current level. The impact of horizontal movement. After a period of operation, the fuel cell stack... The energy level will decline, and if the energy management strategy is not tailored to the current health status of the power system, it may lead to the fuel cell entering a vicious cycle of overload or low efficiency, thereby accelerating fuel cell aging. Therefore, the DDPG model needs to account for the fuel cell's energy efficiency. As one of the state variables, the training characteristics of the Actor-Critic dual network are optimized to improve convergence speed and generalization ability. State space Four-dimensional feature parameters capable of comprehensively reflecting the vehicle's operating conditions and powertrain system status are selected to provide a complete decision-making basis for the DDPG algorithm. The state space is defined as follows:

[0096] (13)

[0097] in, For the total vehicle speed, For the health status of the power system of fuel cell vehicles, This refers to the real-time state of charge of a lithium-ion battery. This is the real-time power change rate of the fuel cell, reflecting the dynamic power change characteristics of the fuel cell and avoiding the aggravation of degradation due to frequent load changes.

[0098] Action space Select the real-time output power of the fuel cell As an action variable, it forms part of a continuous action space and is defined as follows:

[0099] (14)

[0100] in, This refers to the idle power of the fuel cell. The core advantage of lithium-ion batteries lies in their rated maximum power, continuous operating space, and adaptation to the DDPG algorithm for handling continuous operations. Power demand of the whole vehicle in real time Real-time output power of fuel cells The difference is determined, that is ,in, For fuel cell DC / DC converter efficiency.

[0101] Step 4: Construct a multi-objective reward function that integrates durability and economy;

[0102] As feedback from the vehicle to the intelligent agent, the reward function significantly impacts the algorithm's results. Energy consumption and the durability of fuel cells and lithium-ion batteries are crucial considerations for fuel cell vehicles. This paper incorporates fuel cell power fluctuations and lithium-ion battery overcharging and over-discharging factors into the reward function, employing multi-objective optimization to mitigate fuel cell power variations, suppress excessively high and low battery SOC states, and extend the lifespan of both the fuel cell and lithium-ion batteries. To address the multi-objective optimization requirements of energy management, a reward function weighted by multiple objectives—hydrogen consumption, battery degradation, and SOC stability—is constructed. To enable the DDPG algorithm to consider multi-objective collaborative optimization during training and avoid the shortcomings of single-objective optimization, a reward function is defined. for:

[0103] (15)

[0104] in, This is the equivalent hydrogen consumption cost; To account for the degradation cost of fuel cell stacks, the degradation characteristics under low power, high power, and power-variable loads are considered. , , and These are the degradation coefficients for low power, high power, and power-variable loads, respectively. , These refer to the operating times of the fuel cell at low and high power, respectively. The degradation cost of lithium-ion batteries is determined by the charge / discharge rate. , The charge / discharge rate of lithium-ion batteries, and The curve fitting coefficients are used; SOC represents the state of charge of the lithium-ion battery. This is the reference SOC value for lithium-ion batteries. The deviation of SOC from the reference value is used to constrain SOC to remain stable within a reasonable range; As a weighting factor for hydrogen consumption, As a weighting factor for fuel cell degradation, As a weighting factor for lithium-ion battery degradation, This is the battery SOC fluctuation weighting factor. The reward function is negative to transform the optimization objective of "cost minimization" into the reward maximization objective of the DDPG algorithm, thereby gradually reducing hydrogen consumption and system degradation during algorithm training, while simultaneously bringing the SOC closer to the reference value.

[0105] Step 5: Real-time energy allocation decision-making and online iterative optimization;

[0106] The optimal strategy model after training convergence is deployed to the vehicle controller of the fuel cell vehicle. During actual vehicle operation, the VCU collects intelligent network information, vehicle operating conditions and power system status parameters in real time. The predicted vehicle speed is obtained through the speed prediction model and input to the DDPG network to output the optimal output power command of the fuel cell in real time. The VCU completes the power allocation between the fuel cell and the lithium-ion battery according to the command, realizing adaptive energy management.

[0107] In addition, the vehicle terminal continuously collects intelligent connected operating condition data, power system operation data, and energy distribution decision data of the actual vehicle operation, and uploads the data to the cloud regularly. The cloud uses the new dataset to conduct offline iterative training of the attenuation adaptive DDPG strategy, optimizes the model parameters, and then distributes the updated strategy model to the vehicle, so that the strategy can continuously adapt to the dynamic changes of intelligent connected operating conditions and improve the economy and durability of the whole vehicle throughout its life cycle.

[0108] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A degradation-adaptive energy management method for fuel cell vehicles integrating intelligent connectivity, characterized in that, Includes the following steps: Step 1: Vehicle speed prediction model construction; The vehicle collects driving information through intelligent connected vehicle wireless communication technology, and builds a vehicle speed prediction model based on long short-term memory neural network. The vehicle speed prediction model is used to predict the future driving speed and then calculate the vehicle's power demand. Step 2: Decaying adaptive health state estimation; The health status of fuel cells and lithium-ion batteries is estimated. The cumulative degradation of fuel cells is estimated based on an empirical model of fuel cell efficiency degradation, and their health status is defined. At the same time, the total aging degradation rate of lithium-ion batteries is estimated based on calendar aging model and cycle aging model. Step 3: Building the Deep Reinforcement Learning DDPG Network Architecture; An energy management strategy model is constructed based on the DDPG algorithm, and the health state of the fuel cell is introduced as a state variable into the state space of the model. The state space also includes the vehicle speed, the real-time state of charge of the lithium-ion battery, and the real-time power change rate of the fuel cell. Step 4: Constructing the multi-objective reward function; Construct a multi-objective weighted reward function that includes equivalent hydrogen consumption cost, fuel cell stack degradation cost, lithium-ion battery degradation cost, and lithium-ion battery state of charge deviation cost; Step 5: Energy allocation and iterative optimization; The optimal strategy model after training convergence is deployed to the vehicle controller, which collects vehicle information in real time and inputs the vehicle speed prediction model and the optimal strategy model to output the optimal output power command of the fuel cell to complete the power allocation; and uses the cloud to perform offline iterative training and update of the optimal strategy model. In step 2, based on the empirical model of fuel cell efficiency degradation, the cumulative degradation of the fuel cell is... The health state of a fuel cell is determined by its operating conditions, as shown in formula (5). As in formula (6): （5） （6） The polarization curve of the fuel cell voltage is described by the following empirical formula equation, see formula (7): （7） In the formula, For acceleration coefficient, , , and The attenuation rate is characterized under typical operating conditions such as dynamic loading, start-stop process, idling operation, and high power output. , , and These are the number of dynamic loading cycles, the number of start-stop cycles, the idling time, and the high-power output time, respectively. The upper limit of degradation defined for fuel cells from initial use to the end of their lifespan; For Nernst voltage, Activation voltage, For ohmic voltage, Concentration voltage, This represents the total output voltage of the fuel cell stack. This refers to the number of individual cells in a fuel cell stack. The operating voltage of a single fuel cell unit; As the health of the fuel cell declines, the internal resistance of the fuel cell increases. Will follow The changes are shown in formula (8): （8） In the formula, The initial internal resistance, This represents the maximum increase in internal resistance caused by fuel cell degradation; Substituting the polarization curve, we obtain The voltage expression under the influence is shown in formula (9): （9） As the health of the fuel cell declines, its maximum output power, efficiency and variable load capacity will decrease, and the hydrogen consumption rate will increase. The hydrogen consumption rate is shown in formula (10), the fuel cell output power is shown in formula (11), and the adaptive calculation of the fuel cell system efficiency after degradation is shown in formula (12). （10） （11） （12） In the formula, This represents the theoretical hydrogen consumption mass rate. Here is the molar mass of hydrogen. This refers to the current of a lithium-ion battery. It is Faraday's constant. For real-time power output of fuel cells, This refers to the output voltage of a fuel cell under specific current and health conditions. Because of the low calorific value of hydrogen, To improve the working efficiency of the fuel cell system; In step 4, the reward function is defined. for: （15） in, This is the equivalent hydrogen consumption cost; To account for the degradation cost of fuel cell stacks, the degradation characteristics under low power, high power, and power-variable loads are considered. , , and These are the degradation coefficients for low power, high power, and power-variable loads, respectively. , These refer to the operating times of the fuel cell at low and high power, respectively. The cost of lithium-ion battery degradation. , The charge / discharge rate of lithium-ion batteries. and These are the curve fitting coefficients; This is the reference SOC value for lithium-ion batteries. The deviation of SOC from the reference value is used to constrain SOC to remain stable within a reasonable range; As a weighting factor for hydrogen consumption, As a weighting factor for fuel cell degradation, As a weighting factor for lithium-ion battery degradation, This is the weighting factor for battery SOC fluctuations.

2. The energy management method for adaptive fuel cell vehicles with integrated intelligent connectivity as described in claim 1, characterized in that, In step 1, the collected driving information is preprocessed and denoised: the singular spectrum analysis method is used to eliminate random noise from the sensor and communication, and outliers are removed; and the speed prediction time step and prediction steps are determined according to the vehicle speed and road conditions.

3. The energy management method for adaptive fuel cell vehicles with integrated intelligent connectivity as described in claim 1, characterized in that, In step 2, the empirical model "four-factor model" is selected as the fuel cell efficiency degradation model, and the expression corresponding to this empirical model is shown in formula (1): （1） In the formula, This represents the degradation rate of the fuel cell.

4. The energy management method for adaptive fuel cell vehicles with integrated intelligent connectivity as described in claim 3, characterized in that, In step 2, the calendar aging model and the cyclic aging model are as follows: （2） （3） （4） In the formula, This refers to the calendar aging degradation rate of lithium-ion batteries. Thermodynamic temperature; This refers to the state of charge of a lithium-ion battery. The rate of degradation due to cycle aging of lithium-ion batteries; This is the current factor for lithium-ion batteries, with a value of 1. This represents the total aging degradation rate of lithium-ion batteries.

5. The energy management method for adaptive fuel cell vehicles with integrated intelligent connectivity as described in claim 4, characterized in that, In step 3, the state space Defined as: （13） in, For the total vehicle speed, For the health status of the power system of fuel cell vehicles, This refers to the real-time state of charge of a lithium-ion battery. This represents the real-time power change rate of the fuel cell. Action space Select the real-time output power of the fuel cell As an action variable, it forms part of a continuous action space and is defined as follows: （14） in, This refers to the idle power of the fuel cell. The rated maximum power of the fuel cell, the output power of the lithium-ion battery Power demand of the whole vehicle in real time Real-time output power of fuel cells The difference is determined, that is ,in, For fuel cell DC / DC converter efficiency.

Images