Fuel cell system energy management strategy combining aging long-term prediction and model online optimization

Abstract

The present application relates to the technical field of lithium battery and fuel cell hybrid power system energy management, in particular to a fuel cell system energy management strategy combining aging long-term prediction and model online optimization, comprising: obtaining a fuel cell and a health state corresponding to the fuel cell; constructing a long-term prediction data set and obtaining a long-term prediction model of the fuel cell health state; constructing an objective function; and obtaining an optimal supply power distribution result of the hybrid power system. The method of the present application can prolong the remaining service life of the fuel cell, improve the durability of the fuel cell, realize the collaborative optimization of the health state of the fuel cell in the hybrid system, and improve the performance of the hybrid power system while ensuring efficient and economical operation of the hybrid power system.

Description

Technical Field

[0001] This invention relates to the field of energy management technology for lithium battery and fuel cell hybrid power systems, and specifically to an energy management strategy for fuel cell systems that combines long-term aging prediction and online model optimization. Background Technology

[0002] With development, the automotive industry has been driven towards energy conservation and cleaner manufacturing. Proton exchange membrane fuel cells (PEMFCs) have demonstrated significant advantages in the new energy vehicle sector due to their high energy conversion efficiency and environmentally friendly characteristics, with only water as a byproduct. Currently, hybrid power systems combining fuel cells and lithium batteries are becoming an important path to technological commercialization. In the lithium-fuel cell hybrid architecture, lithium batteries possess high power density, while fuel cells excel in high energy density; their performances complement each other, but this also places higher demands on system energy management. As a core technology of hybrid vehicles, energy management strategies directly affect the vehicle's hydrogen consumption economy, lithium battery life, fuel cell durability, vehicle power performance, and driving comfort, making it a challenging and cutting-edge research direction in this field. To achieve efficient, safe, and reliable system operation, the energy management system needs to optimize the power distribution between power sources in real time, meeting load requirements while considering multiple objectives such as system efficiency, component lifespan, and operating costs.

[0003] Currently, research on energy management strategies that comprehensively consider fuel cell durability is still insufficient. Existing real-time optimization control typically only considers changes in the state of charge (SOC) of lithium-ion batteries, and a forward-looking health state optimization mechanism capable of mitigating fuel cell aging has not yet been developed.

[0004] Therefore, there is a need to provide an energy management strategy for fuel cell systems that combines long-term aging prediction and online model optimization to address the above problems. Summary of the Invention

[0005] To address the problem that existing real-time optimization control technologies typically only consider changes in the state of charge (SOC) of lithium batteries and have not yet developed a forward-looking health state optimization mechanism to mitigate fuel cell aging, this invention provides a fuel cell system energy management strategy that combines long-term aging prediction and online model optimization to solve the existing problems.

[0006] The first aspect of this invention provides an energy management strategy for a fuel cell system that combines long-term aging prediction and online model optimization, employing the following technical solution: Construct vehicle dynamics model, fuel cell equivalent circuit model and lithium battery equivalent circuit model, and build hybrid power system based on vehicle dynamics model, fuel cell equivalent circuit model and lithium battery equivalent circuit model; The health state of the fuel cell is obtained based on the equivalent circuit model of the fuel cell; the state of charge of the lithium battery is obtained based on the equivalent circuit model of the lithium battery, and the state of charge of the lithium battery is taken as the health state of the lithium battery. Macroscopic reference operating condition data of the battery system within the prediction time domain is obtained based on historical vehicle data. The macroscopic reference operating condition data includes the power, temperature, voltage and current of the fuel cell. A long-term prediction dataset is constructed based on the macroscopic reference operating condition data and the health status of the fuel cell. A decoupled echo state network model is constructed, and a trained target decoupled echo state network model is obtained by training the decoupled echo state network model based on the dataset. The target decoupled echo state network model is used as a long-term prediction model for the health status of the fuel cell. The predicted health status value of the fuel cell is obtained by using the long-term prediction model for the health status of the fuel cell. The following calculations are performed: 1) Power jump aging cost of fuel cell is constructed based on changes in fuel cell power supply; 2) Hydrogen consumption cost is obtained based on real-time fuel cell power supply; 3) Aging jump cost of fuel cell is obtained based on real-time fuel cell health status and health status reference value; 4) Aging cost due to temperature-current aging feedforward is obtained based on real-time fuel cell power supply and real-time operating temperature; 5) State of charge deviation consumption cost of lithium battery is obtained based on real-time lithium battery power supply; 6) Depreciation cost of lithium battery is obtained based on real-time lithium battery power supply; 7) Objective function is constructed based on fuel cell hydrogen consumption cost, power jump aging cost, temperature-current aging feedforward aging cost, aging jump cost, and lithium battery depreciation cost and state of charge deviation consumption cost. Based on preset constraints, the objective function is solved using a model predictive control algorithm until the objective function value is minimized. The real-time power supply of the lithium battery and fuel cell is then used as the optimal power supply allocation result for the hybrid power system.

[0007] A further technical solution of the present invention is that the step of obtaining the health status of the fuel cell based on the equivalent circuit model of the fuel cell is as follows: The first indicator factor characterizing the health status of a fuel cell is obtained based on the equivalent circuit model of the fuel cell. Based on the equivalent circuit model of the fuel cell and the semi-empirical model of the fuel cell output voltage loss, a correlation model combining macroscopic and microscopic aspects of the fuel cell is obtained; based on the correlation model, a second indicator factor characterizing the health status of the fuel cell is constructed. The health status of the fuel cell is obtained based on the first indicator factor and the second indicator factor.

[0008] A further technical solution of the present invention is that the expression of the first index factor is:

[0009]

[0010] In the formula, The primary indicator factor representing the health status of a fuel cell. The normalized ohmic inductance in the equivalent circuit model of a fuel cell The superimposed weighting factors; This represents the normalized ohmic resistance in the equivalent circuit model of a fuel cell. The superimposed weighting factors; The normalized polarization capacitance in the equivalent circuit model of a fuel cell The superimposed weighting factors; This represents the normalized polarization resistance in the equivalent circuit model of a fuel cell. The superimposed weighting factors; This represents the normalized concentration polarization resistance in the equivalent circuit model of a fuel cell. The superimposed weighting factors; This represents the double-layer capacitance of the normalized constant-phase capacitor element in the equivalent circuit model of a fuel cell. The superimposed weighting factors; This represents the exponential parameter of the normalized constant-phase capacitor element in the equivalent circuit model of a fuel cell. The superimposed weighting factors.

[0011] A further technical solution of the present invention is that the expression of the association model is:

[0012] In the formula, For the association model; The effective activation area of ​​the membrane electrode. Indicates the calculation of the modulus; The correlation coefficient of ohmic polarization loss in the correlation model; The correlation coefficient for activation polarization loss in the correlation model; The correlation coefficient for concentration polarization loss in the correlation model; This refers to the ohmic polarization impedance in the equivalent circuit model. The active polarization impedance in the constructed equivalent circuit model; The concentration polarization impedance in the constructed equivalent circuit model; This indicates the total number of individual fuel cells in the fuel cell stack; This indicates the open-circuit voltage at rated pressure and rated temperature. Indicates the load current density; Represents the concentration constant; Represents the Tafel constant; This indicates the temperature of the fuel cell stack.

[0013] A further technical solution of the present invention is that the expression for the second index factor is:

[0014]

[0015] In the formula, This represents the second indicator factor characterizing the health status of a fuel cell. Represents the open-circuit voltage after normalization. The superimposed weighting factors; The correlation coefficient represents the ohmic polarization loss after normalization. The superimposed weighting factors; The correlation coefficient represents the activation polarization loss after normalization. The superimposed weighting factors; The correlation coefficient represents the concentration polarization loss after normalization. The superimposed weighting factors.

[0016] A further technical solution of the present invention is that the expression of the objective function is:

[0017]

[0018] In the formula, This represents the objective function value, which is the total cost consumed by the fuel cell and lithium battery in the hybrid power system; This indicates the cost of hydrogen consumption in a fuel cell; This indicates the depreciation cost of lithium batteries; This indicates the cost incurred due to the deviation in the state of charge of lithium batteries. This indicates the power jump aging cost of the fuel cell; Indicates the cost of fuel cell aging rate; Indicates the cost based on the aging degree of the fuel cell; The weight representing the cost of hydrogen consumption; Indicates the weight of depreciation costs; The weight representing the cost incurred due to the deviation in the state of charge; The real-time weighting of power transition aging costs; This indicates the weight of the feedforward effect on aging costs; The weighting of costs based on the degree of fuel cell aging; This indicates the real-time power supply of the fuel cell; This indicates the real-time power supply of the lithium battery. These are weighting coefficients used to limit fluctuations in the state of charge of lithium batteries. This indicates the real-time health status of the lithium battery. Indicates reference values ​​for health status; This indicates the change in the power supplied by the fuel cell; Indicates the real-time health status of the fuel cell; Indicates a reference value for the health status of the fuel cell; Represents the frequency factor; Indicates activation energy; Represents the ideal gas constant; Indicating fuel cells in Real-time operating temperature at all times; Indicates the rated power of the fuel cell; Indicates the current stress index; This represents the aging rate function of the fuel cell.

[0019] A further technical solution of the present invention is that the expression for the health status reference value of the fuel cell is:

[0020] In the formula, Indicating fuel cells in Reference values ​​for health status at any given time; Indicates the total lifespan of the fuel cell system design; This represents the end-of-life threshold of the fuel cell, which is the last health state prediction value in the sequence of health state prediction values ​​predicted by the long-term prediction model.

[0021] A further technical solution of the present invention is that the expression for the aging rate function of the fuel cell is:

[0022] In the formula, This represents the aging rate function of the fuel cell; Represents the frequency factor; Indicates activation energy; Represents the ideal gas constant; Indicating fuel cells in Real-time operating temperature at all times; Indicates the rated power of the fuel cell; Indicates the current stress index; Indicating fuel cells in Real-time power supply.

[0023] A further technical solution of the present invention is subject to the following constraints:

[0024] In the formula, This indicates the real-time power supply of the fuel cell; This indicates the minimum power output of the fuel cell; This indicates the maximum power output of the fuel cell; This indicates the maximum change in the power supplied by the fuel cell during this period; This indicates the real-time power supply of the fuel cell; This indicates the real-time power supply of the lithium battery. This indicates the depth of charge / discharge calculated based on the capacity decay of the lithium battery within this cycle. This indicates the real-time health status of the lithium battery. This indicates the minimum healthy state of a lithium battery. This indicates the maximum health status of the lithium battery; This indicates the real-time operating temperature of the fuel cell; Indicates the minimum operating temperature of the fuel cell; Indicates the maximum operating temperature of the fuel cell; Indicates the real-time health status of the fuel cell; This represents the end reference value for health status prediction. The end reference value for health status prediction is the health status prediction value obtained by using the long-term health status prediction model of fuel cell when it is equal to the preset lifespan threshold. The terminal constraint relaxation tolerance represents the health status of the fuel cell.

[0025] A further technical solution of the present invention is to construct a state vector group of the battery system in the hybrid power system based on the voltage, current and operating temperature of the fuel cell and the voltage, current and operating temperature of the lithium battery. A dual extended Kalman filter observation model is constructed, and the state vector group of the battery system is input into the dual extended Kalman filter observation model to obtain the health state estimates of the fuel cell and lithium battery. Based on the preset initial health state of lithium batteries and fuel cells, the model predictive control algorithm is used to obtain the predicted health state values ​​of lithium batteries and fuel cells. Based on the predicted health state values ​​and the estimated health state values ​​of fuel cells and lithium batteries, the error vector of the dual extended Kalman filter observation model is obtained. Based on the error vector, the model parameters of the model predictive control algorithm are updated. Based on the objective function, the updated model predictive control algorithm is used to optimize the power allocation results of the fuel cell and lithium battery. The power allocation result of the fuel cell and lithium battery with the minimum objective function is taken as the optimal power allocation result.

[0026] The beneficial effects of this invention are: This invention proposes an aging factor extraction method combining equivalent circuits and semi-empirical models to assess the health status of fuel cells in hybrid power systems, intuitively reflecting the dynamic changes in the aging process of proton exchange membrane fuel cells. The invention trains indicators characterizing the health status of fuel cells during operation and uses neural networks to predict degradation trends over a future period, accurately reflecting the monotonic decline process of fuel cell performance, thus supporting degradation assessment and feedback optimization. Secondly, an objective function is constructed based on the fuel cell's hydrogen consumption cost, power jump aging cost, temperature-current aging feedforward impact aging cost, and aging jump cost, as well as the lithium battery's depreciation cost and state-of-charge deviation consumption cost. A model predictive control algorithm is used to solve the objective function until the real-time power supply of the lithium battery and fuel cell reaches its minimum value, which is then used as the optimal power supply allocation result for the hybrid power system. This invention combines long-term fuel cell lifetime prediction with real-time optimization of the hybrid power system model, dynamically adjusting the power allocation between the fuel cell and the lithium battery based on the degree of fuel cell degradation and the lithium battery's state of charge, achieving adaptive collaborative optimization in energy management for multi-power systems and effectively improving the overall system durability. Attached Figure Description

[0027] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0028] Figure 1 A flowchart illustrating an energy management strategy for a fuel cell system that combines long-term aging prediction and online model optimization according to the present invention; Figure 2 This is a schematic diagram of the architecture of the hybrid power system in an embodiment of the present invention; Figure 3 This is a schematic diagram showing the power distribution results of the hybrid power system using the strategy of this invention; Figure 4 The predicted results of the lithium battery state of charge degradation trend in the power distribution of the hybrid power system in the strategy of this invention; Figure 5 This is a prediction result of the fuel cell state-of-charge degradation trend in the power distribution of the hybrid power system in the strategy of this invention. Detailed Implementation

[0029] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0030] An embodiment of the present invention provides an energy management strategy for a fuel cell system that combines long-term aging prediction and online model optimization, such as... Figure 1 As shown, it includes: S1. Obtain the fuel cell and its corresponding health status; Specifically, such as Figure 2 As shown, a vehicle dynamics model, a fuel cell equivalent circuit model, and a lithium battery equivalent circuit model are constructed. A hybrid power system is then built based on these models. The health state of the fuel cell is obtained from the fuel cell equivalent circuit model. The state of charge of the lithium battery is obtained from the lithium battery equivalent circuit model, and the state of charge of the lithium battery is used as the health state of the lithium battery.

[0031] S11. The steps for constructing a vehicle dynamics model are as follows: (1) in, F drive This indicates the total force that the drive wheels must provide; F roll Rolling resistance refers to the resistance generated by the deformation of the tire in contact with the road surface and the slight plastic deformation of the road surface; F aero Air resistance refers to the pressure resistance generated by a vehicle pushing against the air, and is proportional to the square of the vehicle speed. F grade Gradient resistance refers to the component of gravity along the slope, with positive values ​​for uphill and negative values ​​for downhill. F accel This represents acceleration drag, referring to the inertial force that causes acceleration in the translational and rotating components of the vehicle. The mathematical expressions for the four sub-modules of the model are as follows: (2) in, m Indicates the overall vehicle weight; g This represents the acceleration due to gravity, taken as 9.8 m / s². -2 ; C r This represents the rolling resistance coefficient, with a value ranging from 0.008 to 0.015. θ This represents the road surface inclination angle, measured in rad. For flat roads, the value is cos(θ). θ ≈1; ρ This represents air density, with a value of 1.225 kg·m³. -3 ; C d This represents the air drag coefficient, with a value ranging from 0.008 to 0.015. A Indicates the windward area; v This indicates vehicle speed, measured in m / s. -1 ; δ This represents the rotational mass coefficient, with a value ranging from 1.03 to 1.08. a This represents longitudinal acceleration, with units of m·s². -2 .

[0032] S12. The steps for constructing the equivalent circuit model of the fuel cell and the equivalent circuit model of the lithium battery are as follows: Figure 2 As shown, the equivalent circuit model of a fuel cell includes an ohmic resistor, an ohmic inductor, a polarization capacitor, a polarization resistor, a concentration resistor, and a constant-phase capacitor. The equivalent circuit model of a lithium battery includes a controlled voltage source to characterize the battery's open-circuit voltage; a polarization resistor to reflect the polarization effect of the lithium battery during charging and discharging as the current direction changes; a low-pass filter to filter the input current and suppress high-frequency noise; and a series internal resistance to represent the battery's internal equivalent impedance. A unidirectional converter and a bidirectional converter are selected to connect the fuel cell and lithium battery, respectively. The unidirectional converter consists of a power switch, a freewheeling diode, a filter inductor, and a filter capacitor; the bidirectional converter consists of a power switch, a resonant inductor, a resonant capacitor, a magnetizing inductor, a high-frequency transformer, and a rectifier diode.

[0033] S13. The steps for obtaining the health status of a fuel cell based on its equivalent circuit model are as follows: For example, in one specific embodiment, index factors characterizing the health status are constructed based on the fuel cell equivalent circuit model. : (4) (5) in, The primary indicator factor representing the health status of a fuel cell. The normalized ohmic inductance in the equivalent circuit model of a fuel cell The superimposed weighting factors; This represents the normalized ohmic resistance in the equivalent circuit model of a fuel cell. The superimposed weighting factors; The normalized polarization capacitance in the equivalent circuit model of a fuel cell The superimposed weighting factors; This represents the normalized polarization resistance in the equivalent circuit model of a fuel cell. The superimposed weighting factors; This represents the normalized concentration polarization resistance in the equivalent circuit model of a fuel cell. The superimposed weighting factors; This represents the double-layer capacitance of the normalized constant-phase capacitor element in the equivalent circuit model of a fuel cell. The superimposed weighting factors; This represents the exponential parameter of the normalized constant-phase capacitor element in the equivalent circuit model of a fuel cell. The superimposed weighting factors.

[0034] The semi-empirical model for the output voltage loss of a fuel cell is as follows: (6) in, V fc This is the fuel cell stack voltage. N cell This represents the number of individual fuel cells in the fuel cell stack. E OCV The open-circuit voltage at rated pressure and rated temperature. R eq It is the total resistance. i fc It is the load current density. i limit For limiting current density, T work For the fuel cell stack temperature, A ta It is the Tafel constant. B con It is a concentration constant. i change For varying current density.

[0035] Meanwhile, based on the semi-empirical model and the equivalent circuit model of the fuel cell, the correlation model of fuel cell aging, which combines macroscopic and microscopic aspects, is obtained as follows: (7) in, For the association model; The effective activation area of ​​the membrane electrode. Indicates the calculation of the modulus; The correlation coefficient of ohmic polarization loss in the correlation model; The correlation coefficient for activation polarization loss in the correlation model; The correlation coefficient for concentration polarization loss in the correlation model; This refers to the ohmic polarization impedance in the equivalent circuit model. The active polarization impedance in the constructed equivalent circuit model; The concentration polarization impedance in the constructed equivalent circuit model; This indicates the total number of individual fuel cells in the fuel cell stack; This indicates the open-circuit voltage at rated pressure and rated temperature. Indicates the load current density; Represents the concentration constant; Represents the Tafel constant; This indicates the temperature of the fuel cell stack.

[0036] Then, based on the aging correlation model, the index factors characterizing the health status of the fuel cell are obtained. for: (8) (9) in, This represents the second indicator factor characterizing the health status of a fuel cell. Represents the open-circuit voltage after normalization. The superimposed weighting factors; The correlation coefficient represents the ohmic polarization loss after normalization. The superimposed weighting factors; The correlation coefficient represents the activation polarization loss after normalization. The superimposed weighting factors; The correlation coefficient represents the concentration polarization loss after normalization. The superimposed weighting factors.

[0037] The first indicator factor characterizing the health status of a fuel cell obtained from the equivalent circuit model of the fuel cell. The second indicator factor representing the health status of fuel cells obtained from the aging correlation model. Normalization is performed using a weighted calculation method to obtain the health status of the fuel cell: (10) (11) in, This represents the first index factor after normalization. This represents the second index factor after normalization. This represents the superposition weighting factor of the first indicator factor after normalization. This represents the superposition weighting factor of the second indicator factor after normalization.

[0038] S14. The steps for obtaining the state of charge (SOC) of the lithium battery based on the equivalent circuit model of the lithium battery and using the SOC as the health state of the lithium battery are as follows: For example, in one specific embodiment, the output voltage of the lithium battery equivalent circuit model in the hybrid power system is: (12) in, V bat This refers to the output voltage of the lithium battery. E li The initial output voltage of the lithium battery is characterized using a controlled voltage source. R This refers to the internal resistance of a lithium battery. i It provides the output current for the lithium battery.

[0039] Constructing a lithium battery output voltage loss model: The lithium battery output voltage loss model includes: initial open-circuit voltage, voltage drop caused by polarization, voltage loss due to battery capacity decay, and relaxation effect-related terms. The expression for the lithium battery output voltage loss model is as follows: (13) in, E 0 represents the initial open-circuit constant voltage of the lithium battery; R pole Polarization resistor; i * This is the low-frequency filter current; K v The coefficient is used to control the rate of voltage decay, which is caused by the decrease in lithium battery capacity. Q ca This represents the maximum capacitance of the capacitor. it For discharge capacity, i For lithium battery output current, t The time of output current change; A For exponentially changing voltage; B It is an exponentially changing capacitor.

[0040] In different operating modes of lithium batteries, the polarization resistance in formula (13) R pole for: (14) Here, discharge means to discharge, and charge means to charge.

[0041] Since the internal resistance and open-circuit voltage of a lithium battery change dynamically with the state of charge (SOC), a typical ampere-hour integration method is used to calculate the SOC of the lithium battery to effectively characterize its chargeability and discharge characteristics. The SOC is then used to characterize the health state of the lithium battery. The health state of a lithium battery is defined as follows: (15) in, SOC 0 represents the initial state of charge of the lithium battery.

[0042] At this point, the health status of the fuel cell and lithium battery can be obtained.

[0043] S2. Construct a long-term prediction dataset and obtain a long-term prediction model for the health status of fuel cells; Specifically, macroscopic reference operating condition data of the battery system within the prediction time domain is obtained based on historical vehicle data. The macroscopic reference operating condition data includes the power, temperature, voltage, and current of the fuel cell. A long-term prediction dataset is constructed based on the macroscopic reference operating condition data and the health status of the fuel cell. A decoupled echo state network model is constructed, and the decoupled echo state network model is trained based on the dataset to obtain a trained target decoupled echo state network model. The target decoupled echo state network model is used as the long-term prediction model for the health status of the fuel cell. The predicted health status value of the fuel cell is obtained using the long-term prediction model for the health status of the fuel cell.

[0044] S21. The steps for obtaining macroscopic reference operating condition data of the battery system in the prediction time domain are as follows: For example, in one specific embodiment, this embodiment uses the macroscopic reference operating condition data of the battery system in the prediction time domain obtained based on the vehicle's historical data as the initial macroscopic reference operating condition data. Then, the initial macroscopic reference operating condition data is processed by median filtering to obtain the macroscopic reference operating condition data.

[0045] S22. The steps for constructing a decoupled echo state network model and training the decoupled echo state network model based on the dataset to obtain the trained target decoupled echo state network model are as follows: For example, in one specific embodiment, the decoupled echo state neural network achieves more complex state simulation by dividing the reservoirs in the echo state neural network into four equal parts and applying matrix constraints. The corresponding updated state equations for the sub-reservoirs are as follows: (16) in, They represent The update state of the activation vectors of neurons in the reservoir at any given time; f (·) represents the activation function of a neuron; For the input weight matrix, This is the cyclic weight matrix for the water storage tank; yes The input vector at time t, k ( t -1) is ( t -1) Activation vector at time 1; reservoir leakage rate α ∈(0,1).

[0046] The output state equations for each sub-reservoir are as follows: (17) in, SOH Long ( t ) is the output vector; S out To output the weight matrix; the root mean square error is calculated using ridge regression, thus obtaining... S out Define the output layer weights. S out The training objective is to minimize the root mean square error between the network output and the target value, ensuring that the output approximates the expected value, and this is solved using ridge regression. (18) (19) in, D It is the number of all data points in the training dataset. V out Output matrix for the water storage tank. SOH Long target Output a matrix for the target value; γ It is the regularization parameter. I It is an identity matrix.

[0047] A mechanism to suppress hiding is added to the reservoir, and the reservoir's cycle weight is calculated as follows: (20) in, S R The cyclic weight matrix for a decoupled echo state network structure reservoir; I ∈ It is the identity matrix; in the circular matrix, S i ( i =1,2,3,4) are the cyclic weights of the four sub-reservoirs. Substituting formula (20) into formula (16) yields the expressions for the updated state and the current state of the reservoirs: (twenty one) (twenty two) Among them, t Current status of the entire reservoir ( t It consists of the renewal status of 4 sub-reservoirs. i ( t ); k ( t The input represents the current state of the reservoir at time t; and there are four identical inputs. Car ( t This forms a new 4-column input matrix. Car whole ( t ); S in This is the input matrix for the entire reservoir; S in(i) The input matrix is ​​for the four sub-reservoirs; k ( t -1) is the state matrix of the entire reservoir; k i ( t -1) is the state matrix of the four sub-reservoirs; I It is an identity matrix; consisting of leakage rate parameter matrices for four sub-reservoirs. α i The entire leakage rate parameter matrix of the reservoir α ;Г is defined as a predictive operator, used to calculate t The current state of the sub-reservoir association: (twenty three) Calculate the output of the four sub-reservoirs: (twenty four) in, S out(i) The outputs of the four sub-reservoirs are used as inputs. Based on the output weights, the outputs of the four sub-reservoirs are superimposed to obtain a decoupled echo state network model, which then outputs predicted data for the fuel cell's state of health (SOH). SOH Long ( t ): (25) in, SOH Long 12 ( t ) represents the superimposed input values ​​of sub-reservoir 1 and sub-reservoir 2; SOH Long 34 ( t ) represents the superimposed input value of sub-reservoir 3 and sub-reservoir 4; SOH Long (t )express SOH Long12 ( t )and SOH Long34 ( t The summation value of ) is, in this embodiment, the predicted health status value of the fuel cell; w 1, w 2, w 3 represents sub-reservoir 1, sub-reservoir 3, and the superposition value, respectively. SOH Long12 ( t The weight ratio of ) is determined; thus, the trained target decoupled echo state network model can be obtained.

[0048] S23. The target-decoupled echo state network model is used as the long-term prediction model for the fuel cell health state; the predicted health state value of the fuel cell is obtained using the long-term prediction model for the fuel cell health state. That is, only the power, temperature, voltage, and current of the fuel cell at each moment need to be input into the long-term prediction model for the fuel cell health state to predict the predicted health state value of the fuel cell.

[0049] S3. Construct the objective function; Specifically, the following calculations are performed: 1) Power jump aging cost of fuel cell is constructed based on changes in fuel cell power supply; 2) Hydrogen consumption cost is obtained based on real-time fuel cell power supply; 3) Aging jump cost of fuel cell is obtained based on real-time fuel cell health status and health status reference value; 4) Aging cost due to temperature-current aging feedforward is obtained based on real-time fuel cell power supply and real-time operating temperature; 5) Charge deviation consumption cost of lithium battery is obtained based on real-time lithium battery health status and health status reference value; 6) Depreciation cost of lithium battery is obtained based on real-time lithium battery power supply; 7) Objective function is constructed based on fuel cell hydrogen consumption cost, power jump aging cost, temperature-current aging feedforward aging cost, aging jump cost, lithium battery depreciation cost, and charge deviation consumption cost.

[0050] S31. Based on the hybrid power system in step S1, calculate the wheel power that the drive motor needs to output: (26) in, F drive This represents the total force that the drive wheels must provide in the power system model in step S1; P wheel This indicates the wheel power that the drive motor needs to output; v Indicates vehicle speed.

[0051] In a hybrid power system, which consists of a fuel cell, a lithium battery, and a drive motor, the power balance relationship is as follows: (27) in, P MC_in This indicates the actual electrical power required by the drive motor; P fc This indicates the power supply required for the fuel cell; P bat This indicates the power supply required for the lithium battery; η t Indicates mechanical transmission efficiency; η mc The efficiency of the motor controller.

[0052] S32. The steps for constructing the objective function based on the hydrogen consumption cost, power jump aging cost, temperature-current aging feedforward effect aging cost, and aging jump cost of fuel cells, as well as the depreciation cost and state-of-charge deviation consumption cost of lithium batteries, are as follows: For example, in one specific embodiment, the expression for the objective function is:

[0053]

[0054] In the formula, This represents the objective function value, which is the total cost consumed by the fuel cell and lithium battery in the hybrid power system; This indicates the cost of hydrogen consumption in a fuel cell; This indicates the depreciation cost of lithium batteries; This indicates the cost incurred due to the deviation in the state of charge of lithium batteries. This indicates the power jump aging cost of the fuel cell (which in practice includes load jumps, start-stop, no-load, and high-power load). Indicates the cost of fuel cell aging rate; Indicates the cost based on the aging degree of the fuel cell; The weight representing the cost of hydrogen consumption; Indicates the weight of depreciation costs; The weight representing the cost incurred due to the deviation in the state of charge; The real-time weighting of power transition aging costs; This indicates the weight of the feedforward effect on aging costs; The weighting of costs based on the degree of fuel cell aging; This indicates the real-time power supply of the fuel cell; This indicates the real-time power supply of the lithium battery. These are weighting coefficients used to limit fluctuations in the state of charge of lithium batteries. This indicates the real-time health status of the lithium battery. This represents a reference value for health status; in this embodiment, it is taken as... =[0.4, 0.6]; This indicates the change in the power supplied by the fuel cell; Indicates the real-time health status of the fuel cell; Indicates a reference value for the health status of the fuel cell; Represents the frequency factor; Indicates activation energy; Represents the ideal gas constant; Indicating fuel cells in Real-time operating temperature at all times; Indicates the rated power of the fuel cell; Indicates the current stress index; The aging rate function of a fuel cell is specifically defined as... Real-time power supply of fuel cells and real-time operating temperature The function for calculating the influencing factors on the aging degree of fuel cells under the combined effect.

[0055] In this embodiment, the weighting coefficient used to limit the fluctuation of the state of charge of the lithium battery in the depreciation cost is... Specifically, it is as follows: (30) in, m These are weighting coefficients used to limit fluctuations in the state of charge of lithium batteries. This indicates the real-time health status of the lithium battery. This indicates the minimum healthy state of a lithium battery. This indicates the maximum health status of the lithium battery; μ This represents the SOC balance correction factor, set to 0.4.

[0056] In this embodiment, the fuel cell aging rate cost is constructed based on the Arrheniu equation. The aging rate of the fuel cell is mainly affected by the coupling effect of membrane electrode temperature and load current density. High temperature and high current conditions exponentially accelerate irreversible aging processes such as catalyst agglomeration and membrane degradation. Therefore, in this embodiment, the expression for the fuel cell aging rate cost is as follows: (31) In the formula, This represents the aging rate function of the fuel cell; The frequency factor reflects the intrinsic reactivity of the material and is typically taken as 10. 6 ~1012 h -1 ; The activation energy is typically taken as 50–80 kJ / mol. Let J represent the ideal gas constant, taken as 8.314 J / (mol·K); Indicating fuel cells in Real-time operating temperature at all times; Indicates the rated power of the fuel cell; The current stress index is usually taken as... =[1, 3]; Indicating fuel cells in Real-time power supply.

[0057] In this embodiment, the cost of fuel cell aging is based on a threshold value SOH ranging from 1 to the end-of-life threshold of the fuel cell. EOL The straight reference trajectory represents the cost of deviating from the predicted lifespan, while the deviation trajectory represents the cost of fuel cell aging. In the relationship between the reference trajectory and the deviation trajectory, the ratio of the remaining aging state quantity equals the ratio of operating cycles, i.e.: (32) in, Indicating fuel cells in The health status reference value at any given time is a linearly decreasing line from 1 to SOH. EOL The reference trajectory when SOH( i ) <SOH ref ( i ), indicating that aging is too rapid, requiring a penalty positive value to force load reduction protection; when SOH( i )>SOH ref ( i This indicates that aging is too slow, requiring a positive penalty, while allowing for moderate overload. Indicates the total lifespan of the fuel cell system design; This represents the end-of-life threshold of the fuel cell, which is the last predicted health state value in the sequence of predicted health state values ​​from the long-term prediction model. In this embodiment, the end-of-life threshold is set to 0.8 (i.e., 20% initial performance degradation of the fuel cell). Based on the long-term prediction model of the fuel cell health state and the end-of-life threshold SOH from step S2... EOL The total lifespan of the fuel cell system design is obtained.

[0058] At this point, the objective function can be established.

[0059] S4. Obtain the optimal power supply allocation result of the hybrid power system; Specifically, based on preset constraints, the objective function is solved using a model predictive control algorithm until the objective function value is minimized. The real-time power supply of the lithium battery and fuel cell is then used as the optimal power supply allocation result for the hybrid power system.

[0060] S41. The constraint conditions are: For example, in one specific embodiment, the constraints for determining the power dynamic boundaries of the fuel cell and the lithium battery are as follows: (33) In the formula, This indicates the real-time power supply of the fuel cell; This represents the minimum power supplied by the fuel cell, taken as 5% of the fuel cell's rated power. This represents the maximum power output of the fuel cell, taken as 95% of the fuel cell's rated power. This represents the maximum change in the power supplied by the fuel cell during this cycle, taken as 20 kW / s; This indicates the real-time power supply of the fuel cell; This indicates the real-time power supplied by the lithium battery; positive indicates discharging and negative indicates charging. This indicates the depth of charge / discharge calculated based on the capacity decay of the lithium battery within this cycle. This indicates the real-time health status of the lithium battery. This indicates the minimum healthy state of a lithium battery. This indicates the maximum health state of a lithium battery, taking... =[0.2, 0.3], =[0.8, 0.9], to prevent damage from over-discharge and the risk of overcharging; This indicates the real-time operating temperature of the fuel cell; Indicates the minimum operating temperature of the fuel cell; Indicates the maximum operating temperature of the fuel cell; take T fc_min =[60, 65]℃ and T fc_max =[80, 85]℃, the upper limit of the temperature range to prevent flooding and membrane dehydration; Indicates the real-time health status of the fuel cell; This represents the final reference value for health status prediction. Since short-term optimization cannot overdraw the long-term lifespan budget, the final reference value for health status prediction is the health status prediction value obtained by using the long-term health status prediction model of fuel cell equal to the preset lifespan threshold. The terminal constraint relaxation tolerance, representing the health status of the fuel cell, is taken as 0.01~0.05.

[0061] S42. The steps for solving the objective function using a model predictive control algorithm based on preset constraints until the real-time power supply of the lithium battery and fuel cell is minimized and taken as the optimal power supply allocation result for the hybrid power system are as follows: A state vector group of the battery system in the hybrid power system is constructed based on the voltage, current, and operating temperature of the fuel cell and the voltage, current, and operating temperature of the lithium battery; a dual extended Kalman filter observation model is constructed, and the state vector group of the battery system is input into the dual extended Kalman filter observation model to obtain the estimated health state values ​​corresponding to the fuel cell and lithium battery; based on the preset initial health state of the lithium battery and fuel cell, the model predictive control algorithm is used to obtain the predicted health state values ​​corresponding to the lithium battery and fuel cell; based on the predicted health state values ​​and the estimated health state values ​​corresponding to the fuel cell and lithium battery, the error vector of the dual extended Kalman filter observation model is obtained; based on the error vector, the model parameters of the model predictive control algorithm are updated, and multiple sets of power supply allocation results are preset within the power constraints corresponding to the lithium battery and fuel cell; based on the objective function, and using the updated model predictive control algorithm, the power supply allocation results of the fuel cell and lithium battery are optimized, and the power supply allocation result of the fuel cell and lithium battery at the minimum objective function value is taken as the optimal power allocation result.

[0062] S41. Based on the voltage, current, and operating temperature of the fuel cell and the voltage, current, and operating temperature of the lithium battery, construct the state vector group of the battery system in the hybrid power system; construct a dual extended Kalman filter observation model, and input the state vector group of the battery system into the dual extended Kalman filter observation model to obtain the corresponding health state estimates of the fuel cell and lithium battery. For example, in one specific embodiment, the collected voltage of the fuel cell is... V fc The current of fuel cells I fc The voltage of lithium batteries V bat The current of lithium batteries I bat Input to the constructed double extended Kalman filter observation model.

[0063] The state variables of a lithium battery are x bat =[SOC; Q ] The observed variables of the fuel cell are y bat = V bat The state variables of the fuel cell are: x fc =[SOH; R 0] The observed variables of the fuel cell are y fc = V fc .

[0064] The process model for lithium batteries is as follows: (34) (35) in, Q ( k ) represents the capacity decay amount per step.

[0065] The observation model for lithium batteries is as follows: (36) in, OCV (SOC) is the open-circuit voltage corresponding to SOC; R ohm and R ct It is the total equivalent internal resistance of a first-order RC circuit.

[0066] The final output is: (37) in, Q obs To extend the capacity estimation of the Kalman filter output, i.e., the filtering result; SOC obs The health status estimate of the lithium battery is obtained by... Q obs Calculated.

[0067] The fuel cell process model is as follows: (38) (39) in, R 0( k )for k At any given moment, the internal resistance of the Ohm is constant. R 0( k ) represents the single-step internal resistance increment.

[0068] The fuel cell observation model is as follows: (40) in, V fc This represents the measured terminal voltage of the fuel cell stack. E OCV This refers to the open-circuit voltage of a single cell; A ta Let Tafel slope be the slope. iThis refers to the current density of a single cell; R 0( k ) represents the current ohmic internal resistance; m and n This is the concentration polarization coefficient; N cell This represents the number of individual fuel cell units in the fuel cell stack.

[0069] The final output is: (41) in, R obs To extend the internal resistance of the Kalman filter output, i.e., the filtering result; SOH obs The health status estimate of the fuel cell is obtained by... R obs Calculated.

[0070] At this point, the estimated health status values ​​for fuel cells and lithium batteries can be obtained.

[0071] S42. Based on the preset initial health state of the lithium battery and fuel cell, the model predictive control algorithm is used to obtain the predicted health state values ​​of the lithium battery and fuel cell. Based on the predicted health state values ​​and the estimated health state values ​​of the fuel cell and lithium battery, the error vector of the dual extended Kalman filter observation model is obtained.

[0072] Construct the state vector predicted by the model: (42) in, Z model ( k (SOC) represents the predicted state vector obtained by the model predictive control algorithm. model ( k The state of health (SOH) is the predicted value of the lithium battery obtained by the model predictive control algorithm. model ( k The state of health (SOC) is the predicted value of the fuel cell obtained by the model predictive control algorithm. model ( k ) and SOH model ( k The values ​​are obtained by recursion from the internal state-space model of the model predictive control algorithm, and all belong to the prior prediction values. The initial values ​​of the recursion process of the internal state-space model of the model predictive control algorithm are the preset initial health states of the lithium battery and fuel cell.

[0073] The error of the filter estimation is: (43) in, e ( kThe total error vector directly drives the next update, enabling adaptive updating of model parameters in the model predictive control algorithm. Z obs ( k )=[SOC obs ( k SOH obs ( k )] T The estimated state vector is the actual observed output of the extended Kalman filter in step S41, where SOC obs ( k The SOH is estimated in real time based on the measured current through the filtered output capacity. obs ( k The internal resistance of the filtered output is calculated and all belong to the posterior observations. e SOC ( k ) represents the SOC observation error; e SOH ( k The error vector represents the SOH observation error. This error vector directly drives the next step of adaptive model parameter updates.

[0074] Therefore, based on the objective function, the power allocation results of the fuel cell and lithium battery are optimized using the updated model predictive control algorithm. The power allocation result of the fuel cell and lithium battery with the minimum objective function is taken as the optimal power allocation result. This achieves the optimal power allocation under the constraints of meeting the vehicle drive requirements, the health status of the lithium battery and the fuel cell, so as to minimize the comprehensive operating costs (hydrogen consumption, depreciation, aging) and achieve a dynamic balance between economy, power and durability.

[0075] The invention will be described in detail below with reference to simulation data: A simulation model was built in Matlab / Simulink to verify the effectiveness of the proposed strategy. Specifically, the power allocation results of the hybrid power system using the strategy of this invention are as follows: Figure 3 As shown, the predicted results of the lithium battery state-of-charge degradation trend in the power distribution of the hybrid power system in this invention are as follows: Figure 4 As shown, the prediction results of the fuel cell health degradation trend of the hybrid power system using the strategy of this invention are as follows: Figure 5 As shown, Figures 3 to 5As shown, under drastic load conditions, this strategy strictly limits the fuel cell power to within 280W. The strategy is implemented after 200 hours of fuel cell system operation, achieving long-term prediction and online optimization. Aging comparison at 400 hours shows that the SOH (State of Health) of this strategy drops to 82%. The proposed method can extend the remaining service life of the fuel cell while ensuring efficient and economical operation of the hybrid power system, improve fuel cell durability, achieve synergistic optimization of the fuel cell health status in the hybrid system, and enhance the performance of the hybrid power system.

[0076] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. An energy management strategy for a fuel cell system that combines long-term aging prediction and online model optimization, characterized in that, include: Construct vehicle dynamics model, fuel cell equivalent circuit model and lithium battery equivalent circuit model, and build hybrid power system based on vehicle dynamics model, fuel cell equivalent circuit model and lithium battery equivalent circuit model; The health state of the fuel cell is obtained based on the equivalent circuit model of the fuel cell; the state of charge of the lithium battery is obtained based on the equivalent circuit model of the lithium battery, and the state of charge of the lithium battery is taken as the health state of the lithium battery. Macroscopic reference operating condition data of the battery system within the prediction time domain is obtained based on historical vehicle data. The macroscopic reference operating condition data includes the power, temperature, voltage and current of the fuel cell. A long-term prediction dataset is constructed based on the macroscopic reference operating condition data and the health status of the fuel cell. A decoupled echo state network model is constructed, and a trained target decoupled echo state network model is obtained by training the decoupled echo state network model based on the dataset. The target decoupled echo state network model is used as a long-term prediction model for the health status of the fuel cell. The predicted health status value of the fuel cell is obtained by using the long-term prediction model for the health status of the fuel cell. The following calculations are performed: 1) Power jump aging cost of fuel cell is constructed based on changes in fuel cell power supply; 2) Hydrogen consumption cost is obtained based on real-time fuel cell power supply; 3) Aging jump cost of fuel cell is obtained based on real-time fuel cell health status and health status reference value; 4) Aging cost due to temperature-current aging feedforward is obtained based on real-time fuel cell power supply and real-time operating temperature; 5) Charge deviation consumption cost of lithium battery is obtained based on real-time lithium battery power supply; 6) Depreciation cost of lithium battery is obtained based on real-time lithium battery power supply; 7) Objective function is constructed based on fuel cell hydrogen consumption cost, power jump aging cost, temperature-current aging feedforward aging cost, aging jump cost, lithium battery depreciation cost, and charge deviation consumption cost. Based on preset constraints, the objective function is solved using a model predictive control algorithm until the objective function value is minimized. The real-time power supply of the lithium battery and fuel cell is then used as the optimal power supply allocation result for the hybrid power system.

2. The energy management strategy for a fuel cell system combining long-term aging prediction and online model optimization as described in claim 1, characterized in that, The steps to obtain the health status of a fuel cell based on its equivalent circuit model are as follows: The first indicator factor characterizing the health status of a fuel cell is obtained based on the equivalent circuit model of the fuel cell. Based on the equivalent circuit model of the fuel cell and the semi-empirical model of the fuel cell output voltage loss, a correlation model combining macroscopic and microscopic aspects of the fuel cell is obtained. A second indicator factor representing the health status of fuel cells is constructed based on the correlation model; The health status of the fuel cell is obtained based on the first indicator factor and the second indicator factor.

3. The energy management strategy for a fuel cell system combining long-term aging prediction and online model optimization according to claim 2, characterized in that, The expression for the first indicator factor is: In the formula, The primary indicator factor representing the health status of a fuel cell. The normalized ohmic inductance in the equivalent circuit model of a fuel cell The superimposed weighting factors; This represents the normalized ohmic resistance in the equivalent circuit model of a fuel cell. The superimposed weighting factors; The normalized polarization capacitance in the equivalent circuit model of a fuel cell The superimposed weighting factors; This represents the normalized polarization resistance in the equivalent circuit model of a fuel cell. The superimposed weighting factors; This represents the normalized concentration polarization resistance in the equivalent circuit model of a fuel cell. The superimposed weighting factors; This represents the double-layer capacitance of the normalized constant-phase capacitor element in the equivalent circuit model of a fuel cell. The superimposed weighting factors; This represents the exponential parameter of the normalized constant-phase capacitor element in the equivalent circuit model of a fuel cell. The superimposed weighting factors.

4. The energy management strategy for a fuel cell system combining long-term aging prediction and online model optimization according to claim 2, characterized in that, The expression for the association model is: In the formula, For the association model; The effective activation area of ​​the membrane electrode. Indicates the calculation of the modulus; The correlation coefficient of ohmic polarization loss in the correlation model; The correlation coefficient for activation polarization loss in the correlation model; The correlation coefficient for concentration polarization loss in the correlation model; This refers to the ohmic polarization impedance in the equivalent circuit model. The active polarization impedance in the constructed equivalent circuit model; The concentration polarization impedance in the constructed equivalent circuit model; This indicates the total number of individual fuel cells in the fuel cell stack; This indicates the open-circuit voltage at rated pressure and rated temperature. Indicates the load current density; Represents the concentration constant; Represents the Tafel constant; This indicates the temperature of the fuel cell stack.

5. The energy management strategy for a fuel cell system combining long-term aging prediction and online model optimization according to claim 2, characterized in that, The expression for the second indicator factor is: In the formula, This represents the second indicator factor characterizing the health status of a fuel cell. Represents the open-circuit voltage after normalization. The superimposed weighting factor; The correlation coefficient represents the ohmic polarization loss after normalization. The superimposed weighting factor; The correlation coefficient represents the activation polarization loss after normalization. The superimposed weighting factor; The correlation coefficient represents the concentration polarization loss after normalization. The superimposed weighting factors.

6. The energy management strategy for a fuel cell system combining long-term aging prediction and online model optimization according to claim 1, characterized in that, The expression for the objective function is: In the formula, This represents the objective function value, which is the total cost consumed by the fuel cell and lithium battery in the hybrid power system; This indicates the cost of hydrogen consumption in a fuel cell; This indicates the depreciation cost of lithium batteries; This indicates the cost incurred due to the deviation in the state of charge of lithium batteries. This indicates the power jump aging cost of the fuel cell; Indicates the cost of fuel cell aging rate; Indicates the cost based on the aging degree of the fuel cell; The weight representing the cost of hydrogen consumption; Indicates the weight of depreciation costs; The weight representing the cost incurred due to the deviation in the state of charge; The real-time weighting of power transition aging costs; This indicates the weight of the feedforward effect on aging costs; The weighting of costs based on the degree of fuel cell aging; This indicates the real-time power supply of the fuel cell; This indicates the real-time power supply of the lithium battery. These are weighting coefficients used to limit fluctuations in the state of charge of lithium batteries. This indicates the real-time health status of the lithium battery. Indicates reference values ​​for health status; This indicates the change in the power supplied by the fuel cell; Indicates the real-time health status of the fuel cell; Indicates a reference value for the health status of the fuel cell; Represents the frequency factor; Indicates activation energy; Represents the ideal gas constant; Indicating fuel cells in Real-time operating temperature at all times; Indicates the rated power of the fuel cell; Indicates the current stress index; This represents the aging rate function of the fuel cell.

7. The energy management strategy for a fuel cell system combining long-term aging prediction and online model optimization according to claim 6, characterized in that, The expression for the health status reference value of a fuel cell is: In the formula, Indicating fuel cells in Reference values ​​for health status at any given time; Indicates the total lifespan of the fuel cell system design; This represents the end-of-life threshold of the fuel cell, which is the last health state prediction value in the sequence of health state prediction values ​​predicted by the long-term prediction model.

8. The energy management strategy for a fuel cell system combining long-term aging prediction and online model optimization according to claim 6, characterized in that, The expression for the aging rate function of a fuel cell is: In the formula, The aging rate function of a fuel cell; Represents the frequency factor; Indicates activation energy; Represents the ideal gas constant; Indicating fuel cells in Real-time operating temperature at all times; Indicates the rated power of the fuel cell; Indicates the current stress index; Indicating fuel cells in Real-time power supply.

9. The energy management strategy for a fuel cell system combining long-term aging prediction and online model optimization according to claim 1, characterized in that, The constraints are: In the formula, This indicates the real-time power supply of the fuel cell; This indicates the minimum power output of the fuel cell; This indicates the maximum power output of the fuel cell; This indicates the maximum change in the power supplied by the fuel cell during this period; This indicates the real-time power supply of the fuel cell; This indicates the real-time power supply of the lithium battery. This indicates the depth of charge / discharge calculated based on the capacity decay of the lithium battery within this cycle. This indicates the real-time health status of the lithium battery. This indicates the minimum healthy state of a lithium battery. This indicates the maximum health status of the lithium battery; This indicates the real-time operating temperature of the fuel cell; Indicates the minimum operating temperature of the fuel cell; Indicates the maximum operating temperature of the fuel cell; Indicates the real-time health status of the fuel cell; This represents the end reference value for health status prediction. The end reference value for health status prediction is the health status prediction value obtained by using the long-term health status prediction model of fuel cell when it is equal to the preset lifespan threshold. The terminal constraint relaxation tolerance represents the health status of the fuel cell.

10. The energy management strategy for a fuel cell system combining long-term aging prediction and online model optimization according to claim 1, characterized in that, Also includes: Based on the voltage, current, and operating temperature of the fuel cell, and the voltage, current, and operating temperature of the lithium battery, a state vector group of the battery system in the hybrid power system is constructed. A dual extended Kalman filter observation model is constructed, and the state vector group of the battery system is input into the dual extended Kalman filter observation model to obtain the health state estimates of the fuel cell and lithium battery. Based on the preset initial health state of lithium batteries and fuel cells, the model predictive control algorithm is used to obtain the predicted health state values ​​of lithium batteries and fuel cells. Based on the predicted health state values ​​and the estimated health state values ​​of fuel cells and lithium batteries, the error vector of the dual extended Kalman filter observation model is obtained. Based on the error vector, the model parameters of the model predictive control algorithm are updated. Based on the objective function, the updated model predictive control algorithm is used to optimize the power allocation results of the fuel cell and lithium battery. The power allocation result of the fuel cell and lithium battery with the minimum objective function is taken as the optimal power allocation result.

Images