Fuel cell hybrid vehicle energy scheduling method and system

Abstract

The application discloses a kind of fuel cell hybrid vehicle energy scheduling method and system, method includes: S1, constructs hydrogen fuel hybrid train power system model;S2, obtain the demand power in the process of train operation, utilize the low-pass filter of adjustable cut-off frequency and resolve demand power into low-frequency component and high-frequency component;Build double-layer reinforcement learning intelligent agent, cut-off frequency of low-pass filter is used as the action variable of reinforcement learning intelligent agent and is carried out real-time dynamic optimization adjustment;S3, design multi-objective reward function and system total cost function, simultaneously optimize fuel cell efficiency, fuel cell power fluctuation, hydrogen consumption, lithium battery SOC approaches initial value;By deep deterministic policy gradient algorithm, update double-layer reinforcement learning intelligent agent network parameters, finally output fuel cell and lithium battery real-time power distribution instruction.The application has the advantages such as reducing the fluctuation of hydrogen fuel cell output power and hydrogen consumption.

Description

Technical Field

[0001] This invention mainly relates to the field of vehicle energy management technology, specifically to a fuel cell hybrid vehicle energy scheduling method and system. Background Technology

[0002] The increase in population and the scale of human production activities have damaged the ecological environment and triggered an energy crisis. A novel power system combining fuel cells and hydrogen is considered a potential solution to replace traditional internal combustion engine technology in the rail transit sector, aiming to reduce pollution emissions and promote green transportation. However, hydrogen fuel cells have drawbacks such as relatively soft output characteristics, slow power response, and the inability to recover energy during train braking, meaning that using only hydrogen fuel cells as a power source cannot meet the load's demands. Typically, batteries and supercapacitors are chosen as auxiliary power sources, forming a hybrid power system together with the hydrogen fuel cell. When the train accelerates or maintains a constant speed, the hydrogen fuel cell acts as the primary power source, providing the required power to the load. If there is a large power demand, the auxiliary power source will supplement the peak power exceeding the hydrogen fuel cell's supply capacity. When the train decelerates, the load power demand becomes negative, and the auxiliary power source will recover and store the feedback energy, improving energy utilization and reducing the overall operating cost of the train. Designing a reasonable and reliable energy management strategy to coordinate energy flow between multiple power sources and between power sources and loads is a key research area for fuel cell hybrid trains.

[0003] Current theoretical and experimental research mainly focuses on single aspects such as reducing fuel consumption or improving fuel cell lifespan. Furthermore, energy scheduling methods based on rules and optimization suffer from drawbacks such as reliance on engineering experience, inability to guarantee optimal control, high computational cost, poor real-time performance, and potential for getting trapped in local optima. Therefore, this invention constructs a power system model of a hydrogen fuel cell hybrid train, comprehensively considering fuel cell lifespan and energy efficiency, and proposes an energy scheduling method based on frequency decoupling and reinforcement learning to allocate the optimal fuel cell output power.

[0004] The power system models of hydrogen fuel cell hybrid trains are mostly modeled using formula approximation methods, resulting in low model accuracy.

[0005] Energy scheduling methods based on rules and optimization have drawbacks such as reliance on engineering experience, large computational load, and poor real-time performance. Energy scheduling methods based on reinforcement learning rely on approximate formulas for environment modeling, resulting in poor accuracy, poor policy usability, and often considering only one objective. Summary of the Invention

[0006] In view of the technical problems existing in the prior art, the present invention provides a method and system for energy scheduling of fuel cell hybrid vehicles to reduce the fluctuation of hydrogen fuel cell output power and hydrogen consumption.

[0007] To solve the above-mentioned technical problems, the technical solution proposed by this invention is as follows:

[0008] A method for energy dispatching of a fuel cell hybrid vehicle includes the following steps:

[0009] S1. Construct a hydrogen fuel cell hybrid train power system model, which includes a fuel cell model, a lithium battery model, and a traction motor model;

[0010] S2. Obtain the required power during train operation and decompose the required power into low-frequency and high-frequency components using a low-pass filter with an adjustable cutoff frequency.

[0011] A two-layer reinforcement learning agent is constructed, and the cutoff frequency of the low-pass filter is used as the action variable of the reinforcement learning agent for real-time dynamic optimization and adjustment.

[0012] The two-layer reinforcement learning agent includes an outer agent and an inner agent; the outer agent is used to extract global features of the required power and output an initial normalized cutoff frequency value; the inner agent receives the initial normalized cutoff frequency value, optimizes the power allocation strategy in real time, and uses the low-frequency component as the output power of the fuel cell and the high-frequency component as the output power of the lithium battery.

[0013] S3. Design a multi-objective reward function and a total system cost function to simultaneously optimize fuel cell efficiency, fuel cell power fluctuation, hydrogen consumption, and lithium battery SOC to approach the initial value.

[0014] The parameters of the two-layer reinforcement learning agent network are updated by a deep deterministic policy gradient algorithm, and finally the real-time power allocation command of the fuel cell and lithium battery is output.

[0015] Preferably, in step S3, when braking power is detected: if the lithium battery SOC ≥ safety threshold, the regenerative braking energy is allocated to the braking resistor consumption; otherwise, the energy is recovered by the lithium battery.

[0016] Preferably, in step S3, the reward function r Total system cost cost total They are respectively:

[0017]

[0018]

[0019] In the formula, For fuel cell efficiency, This represents the current remaining charge of the power battery. The standard deviation of fuel cell output power. This represents the fuel consumption between two adjacent time steps. These are the corresponding weighting coefficients; The cost of starting a fuel cell once, For the number of fuel cell starts, and To account for the operating costs of fuel cells and power batteries, and This refers to the output power of fuel cells and power batteries.

[0020] Preferably, in the fuel cell model, the output voltage of a single fuel cell is as shown in the following equation:

[0021] ;

[0022] In the formula, The output voltage of a single PEMFC; Nernst voltage; This is due to voltage loss caused by the activation polarization effect; Voltage loss caused by Ohmic polarization effect; This is the voltage loss caused by concentration polarization.

[0023] Preferably, in the lithium battery model, the output voltage expression during lithium battery charging and discharging is as follows:

[0024] Discharge:

[0025]

[0026] Charge:

[0027]

[0028] In the formula, This is the battery output voltage. To maintain a constant output voltage for the battery, It is the polarization constant. For battery capacity, This represents the actual battery capacity. For the amplitude of the exponential region, It is the reciprocal of the amplitude in the exponential region. For internal resistance, Battery current, Filter current for the battery;

[0029] Current state of charge of the battery The expression is:

[0030]

[0031] In the formula, It is the amount of remaining charge in the battery at the initial moment; The next moment The state of charge of the battery; It is the total charge at the current moment; It is the total charge at the next moment; This represents the amount of charge that the battery receives during charging. This represents the amount of charge released by the battery under discharge conditions.

[0032] Preferably, in the traction motor model, the required torque and speed of the traction motor are obtained, and then the required power of the traction motor is obtained.

[0033] The present invention also discloses a computer program product, comprising a computer program that, when executed by a processor, performs the steps of the method described above.

[0034] The present invention further discloses a computer-readable storage medium having a computer program stored thereon, the computer program executing the steps of the method described above when run by a processor.

[0035] The present invention also discloses an energy dispatching system for a fuel cell hybrid vehicle, including a memory and a processor connected to each other, wherein the memory stores a computer program, and the computer program executes the steps of the method described above when run by the processor.

[0036] Compared with the prior art, the advantages of the present invention are as follows:

[0037] This invention designs a "global-instantaneous" dual-agent architecture: the outer layer solves the initial value convergence problem, and the inner layer realizes dynamic power allocation, breaking through the feature extraction limitations of traditional single-agent reinforcement learning; a dynamic frequency decoupling mechanism: the filter cutoff frequency fc is used as a learnable action variable to replace the fixed threshold, so that the power allocation adapts to changes in operating conditions; a multi-objective reward function design: for the first time, it integrates multi-dimensional objectives such as fuel cell life (fluctuation suppression), energy consumption (hydrogen consumption), and battery state (SOC), solving the drawbacks of single-objective optimization in the industry.

[0038] This invention constructs a hybrid power system simulation model based on the mechanism model of fuel cells and power batteries, improves the accuracy of the model, and thus improves the usability of energy management strategies. It also enables the interaction between the environment and the agent's strategies to support the cyclical training of reinforcement learning.

[0039] This invention constructs a hybrid system environment model based on a mechanism model and trains it with a reinforcement learning agent. By designing a reward function, it reduces the energy consumption of the fuel cell. At the same time, it introduces a filter to decouple the required power and allocate the low-frequency part to the fuel cell output, thereby improving its service life and reducing the fluctuation of the hydrogen fuel cell output power and hydrogen consumption. Attached Figure Description

[0040] Figure 1 This is a schematic diagram of the fuel cell hybrid power energy dispatching system of the present invention.

[0041] Figure 2 This is a diagram of the hybrid power system architecture of the present invention.

[0042] Figure 3 (a) is a graph showing the output characteristics of the fuel cell of the present invention; (b) is a graph showing the current-voltage characteristic; and (c) is a graph showing the output power as a function of current.

[0043] Figure 4 (a) is a schematic diagram of the effect of temperature on the output characteristics of the fuel cell in this invention; (b) is a schematic diagram of the effect of temperature on the output voltage; and (c) is a schematic diagram of the effect of temperature on the output efficiency.

[0044] Figure 5 Here are the discharge curves of the lithium battery in this invention; (a) is the discharge curve at the nominal current; (b) is the discharge curve at discharge currents of 100A, 500A, and 1000A.

[0045] Figure 6 This is a schematic diagram of the traction motor calculation model in this invention.

[0046] Figure 7 This is a schematic diagram of the demand frequency decoupling method in this invention.

[0047] Figure 8 This is a schematic diagram of the frequency decoupling power allocation method in this invention.

[0048] Figure 9 This is a training architecture diagram of the DDPG reinforcement learning algorithm in this invention. Detailed Implementation

[0049] The present invention will be further described below with reference to the accompanying drawings and specific embodiments.

[0050] like Figure 1 As shown, the energy scheduling method for fuel cell hybrid vehicles provided in this embodiment of the invention includes the following steps:

[0051] S1. Construct a hydrogen fuel cell hybrid train power system model, which includes a fuel cell model, a lithium battery model, and a traction motor model.

[0052] like Figure 2 As shown, the power source characteristics of hydrogen fuel cell hybrid trains are analyzed, and a corresponding vehicle power system model is established, specifically including:

[0053] Hydrogen fuel cell model: The focus is on analyzing the response characteristics of the fuel cell. The output characteristics of the hydrogen fuel cell are affected by factors such as temperature, gas pressure, and load current. The output voltage of a single fuel cell is shown in equation (1).

[0054] (1)

[0055] In the formula, This refers to the output voltage of a single proton exchange membrane fuel cell. This is the Nernst voltage, which is the theoretical maximum voltage. This is due to voltage loss caused by the activation polarization effect; Voltage loss caused by Ohmic polarization effect; This is the voltage loss caused by concentration polarization.

[0056] Nernst voltage The expression is shown in equation (2):

[0057] (2)

[0058] In the formula, The internal temperature of a hydrogen fuel cell during operation, in units of ; This is Faraday's constant, with a value of 96485; The universal ideal gas constant is 8.314. This is the partial pressure of hydrogen gas; This is the partial pressure of oxygen.

[0059] Voltage loss due to activation polarization effect As shown in equation (3):

[0060] (3)

[0061] In the formula, The conversion factor has a value of 0.5. This represents the actual current density, in units of... ; Exchange current density, unit: The empirical value is .

[0062] Voltage loss due to Ohmic polarization effect As shown in equation (4):

[0063] (4)

[0064] In the formula, This refers to the output current of the hydrogen fuel cell. This is the equivalent resistance of the proton exchange membrane; is a membrane resistance parameter used to characterize the resistance encountered by protons as they pass through the exchange membrane.

[0065] Voltage loss caused by concentration polarization effect As shown in equation (5):

[0066] (5)

[0067] In the formula, It is a constant, determined by the operating state of the hydrogen fuel cell; The maximum current density; This is the ratio of the actual current to the membrane area during hydrogen fuel cell operation, used to characterize the load factor of the hydrogen fuel cell, and is measured in units of... .

[0068] The output characteristic curve of a proton exchange membrane fuel cell is shown below. Figure 3 As shown, when the output power gradually increases from zero, the output voltage drops rapidly, indicating that the system is in the activation polarization stage. Figure 3 As shown in (a); when the output power increases to a certain value, the output voltage decreases linearly with the current, at which point it is in the ohmic polarization stage, as shown in (a). Figure 3 As shown in (b).

[0069] The effect of temperature on the output voltage and efficiency of fuel cells, such as Figure 4 As shown, within a certain temperature range, both the output voltage and efficiency increase with increasing temperature. Figure 4 As shown in (a) and (b) in the figure.

[0070] Lithium battery model: The expression for the output voltage during charging and discharging of a lithium battery is as follows:

[0071] The expression for the output voltage during discharge is:

[0072] (6)

[0073] The expression for the output voltage during charging is:

[0074] (7)

[0075] In the formula, This refers to the battery output voltage (V). The constant output voltage (V) of the battery. is the polarization constant (V / Ah). Battery capacity (Ah). This represents the actual battery capacity (Ah). V represents the amplitude in the exponential region. It is the reciprocal of the amplitude in the exponential region ( / Ah). The internal resistance is (Ω). Battery current (A). The battery filter current (A).

[0076] The SOC estimation method used in this invention is the charge integration method. Given the remaining battery charge at the previous moment, it integrates the charging and discharging currents of the battery over a period of time to obtain the current state of charge (SOC). The calculation method is as follows:

[0077] (8)

[0078] In the formula, It is the initial moment The state of charge of the battery; The next moment The state of charge of the battery; It is the total charge at the current moment; It is the total charge at the next moment; These represent the amount of charge that the battery receives and discharges under charging and discharging conditions, respectively.

[0079] The discharge curve of the hydrogen fuel cell hybrid train's power battery at a rated current (nominal current) of 155.217A is shown below. Figure 5 As shown in (a); the discharge curves for discharge currents of 100A, 500A, and 1000A are as follows. Figure 5 As shown in (b).

[0080] The discharge curve can be divided into three regions, the first region (e.g.) Figure 5 In the yellow area (a), also known as the exponential region, the terminal voltage drops rapidly. If the discharge current increases, the voltage drops even faster. The second area is also called the rated operating area (e.g., the rated operating area). Figure 5 In the middle (a) gray area, also known as the nominal area), the voltage drop slows down, and the smaller the discharge rate, the longer the rated operating area lasts; the third area (such as...) Figure 5 In the blank area A of (a), the voltage drops rapidly when the battery is almost fully discharged.

[0081] Traction motor model: Resistance during train operation Including air resistance and friction The calculation method is as follows:

[0082] (9)

[0083] In the formula, This refers to the air drag coefficient; Let the air density be denoted as . ; The train's frontal area; This refers to the actual speed of the locomotive train. The coefficient of sliding friction; m For the overall quality of the train; Let be the acceleration due to gravity, and take . .

[0084] By performing proportional-derivative control on the difference between the actual train speed and the reference speed, the traction force required for the actual train speed to track the reference speed is obtained. As shown in equation (10), the total traction force required by the train is... For resistance The sum of the traction required to track the reference speed:

[0085] (10)

[0086] Using the total traction force required by the train obtained above and equation (11), the required torque of the traction motor can be calculated. :

[0087] (11)

[0088] In the formula, The radius of the train wheels; The transmission ratio is defined as the ratio of the speed of the main wheel to the speed of the driven wheel. In trains, it is also defined as the ratio of the wheel torque to the traction motor torque. This refers to the efficiency of the gearbox.

[0089] traction motor speed With the actual speed of the train The relationship is shown in equation (12):

[0090] (12)

[0091] In the formula, The speed of the traction motor, in units of ; The reduction ratio, also known as the transmission ratio of a speed reduction device, refers to the ratio of the instantaneous input speed to the output speed in the speed reduction device.

[0092] Substituting the required torque and speed of the traction motor obtained above into equation (13), the required power of the traction motor can be calculated. :

[0093] (13)

[0094] Since the voltage of the DC bus in the system is known, the required current of the traction motor can be calculated.

[0095] In summary, the traction motor calculation model is as follows: Figure 6 As shown.

[0096] S2, such as Figure 1 As shown, the energy scheduling of hydrogen fuel cell hybrid trains adopts a hierarchical structure, mainly composed of calculation modules such as operating speed planning, demand power calculation, and energy scheduling unit. Figure 1 In the process, the speed planning module calculates the global speed curve based on information such as the upper-level timetable, section speed limits, and train parameters; the power demand calculation module calculates the power demand of the route based on the global speed curve and the train power balance equation, and further transmits it to the energy scheduling module for power allocation. The goal is to reduce the fluctuation of fuel cell output power, improve its service life, and reduce the hydrogen consumption of the route. The final power allocation curve is sent to the real-time controller for final control implementation.

[0097] The frequency-decoupling-based energy management strategy, based on the characteristics of hydrogen fuel cells and lithium batteries, divides the required power into low-frequency and high-frequency components. The separated low-frequency component is used as the output power of the hydrogen fuel cell, and the high-frequency component is used as the output power of the lithium battery. Figure 7 As shown. Specifically, the required power is divided into low-frequency and high-frequency components by a second-order low-frequency filter with an adjustable normalized cutoff frequency.

[0098] Construct a two-layer reinforcement learning agent to normalize the cutoff frequency of the low-pass filter. The action variables of the reinforcement learning agent are dynamically optimized and adjusted in real time. The calculation formula is shown below, and the value ranges from 0 to 1:

[0099] (14)

[0100] like Figure 9 As shown, the two-layer reinforcement learning agent consists of an inner agent and an outer agent, which are responsible for learning the instantaneous and global features of the vehicle's power demand, respectively. The outer agent is responsible for predicting the frequency decoupling threshold and feeding the result of frequency decoupling as an initial value to the inner agent, which can effectively solve the problems of initial value convergence and global feature extraction. The inner agent uses the initial normalized cutoff frequency value output by the outer agent as a starting point to optimize the power allocation strategy in real time, using the low-frequency component as the fuel cell output power and the high-frequency component as the lithium battery output power.

[0101] S3. Design a multi-objective reward function and a total system cost function to simultaneously optimize fuel cell efficiency, fuel cell power fluctuation, hydrogen consumption, and lithium battery SOC approaching the initial value; update the agent network parameters using the DDPG algorithm (Deep Deterministic Policy Gradient).

[0102] Specifically, the DDPG algorithm uses the target network to update the parameters of the Actor and Critic networks. Optimizing the Critic network aims to minimize its error; training the Critic network involves finding its optimal parameters. The process of finding the optimal solution. The process of training the Actor network is essentially finding the Actor network parameters. The process of finding the optimal solution involves optimizing the Actor network parameters using stochastic gradient descent.

[0103] In actual training, the method of this invention first initializes the state variables and resets the train's operating conditions in each large loop. After the simulation starts, in each small loop, actions are first obtained from the Actor network. a t ,by a t The average value and the variance parameter var form a normal distribution. An action is randomly selected from the normal distribution as the normalized cutoff frequency of the second-order low-pass filter. .

[0104] This filter is used to filter the demand power; the low-frequency component represents the output power allocated to the fuel cell by the energy management system. The simulation pause time and the output power allocated to the fuel cell are written into the simulation model. Hydrogen consumption, lithium battery SOC, actual fuel cell output power, fuel cell output efficiency, and total system cost are obtained from the model. The fuel cell output efficiency and total system cost are then used as the basis for further analysis. cost total The reward function includes lithium battery SOC, fuel cell output standard deviation, and hydrogen consumption. r The components are calculated as follows:

[0105] (15)

[0106] (16)

[0107] In the formula, For fuel cell efficiency, This represents the current remaining charge of the power battery. The standard deviation of fuel cell output power. This represents the fuel consumption between two adjacent time steps. These are the weighting coefficients for each item; The cost of starting a fuel cell once, For the number of fuel cell starts, and These are the operating costs of fuel cells and power batteries, respectively. and These are the output power of the fuel cell and the power battery, respectively. The higher the output efficiency of the fuel cell, the lower the total system cost, the closer the lithium battery SOC is to the initial value, the smaller the standard deviation of the fuel cell output power, and the lower the hydrogen consumption, the greater the reward value obtained by the intelligent agent, which is more conducive to achieving the system optimization goal.

[0108] This invention uses normalized cutoff frequency, fuel cell efficiency, and lithium battery SOC as state variables. s t When the system simulation pauses for one second, the state variables obtained from the simulation model will constitute the state space for the next moment. s t+1 ; the quadruple ( s t , a t , r t , s t+1 The data is stored in the DDPG experience replay buffer pool, which has a capacity of 10,000. When the buffer pool is full, the normal distribution variance parameter is reduced to decrease the agent's exploration rate. Multiple sets of data are then selected from the buffer pool to train the agent and update the network parameters. s t The current state instructions for the intelligent agent include the normalized cutoff frequency, fuel cell efficiency, and lithium battery SOC. s t+1 This is the state instruction for the agent in the next moment; a t This is the action instruction for the current agent, specifically adjusting the normalization cutoff frequency; r t This is the multi-objective reward function for the current agent.

[0109] The learning rate for both the Actor network and the Critic network is set to 1×10. -3 The parameters of the Actor_target and Critic_target networks are updated with a smoothing rate of 0.01; the Q-target value is updated using a discount factor of 0.9 according to the Bellman equation. The training architecture of the reinforcement learning algorithm is as follows: Figure 9 As shown.

[0110] To prevent overcharging of the lithium battery during train braking, the model of this invention also incorporates a braking resistor. The specific power distribution method is as follows: Figure 8 As shown, the power required is subtracted from the output power of the lithium battery and the output power of the fuel cell, and the remaining power is the power consumed by the braking resistor, thereby preventing the lithium battery from being overcharged.

[0111] This invention designs a "global-instantaneous" two-layer intelligent agent architecture. The outer intelligent agent solves the initial value convergence problem, while the inner intelligent agent realizes dynamic power allocation, breaking through the feature extraction limitations of traditional single-agent reinforcement learning. Through a dynamic frequency decoupling mechanism, the filter cutoff frequency is used as a learnable action variable to replace the fixed threshold, enabling power allocation to adapt to changes in operating conditions. A multi-objective reward function is designed, which for the first time integrates multiple dimensions such as fuel cell lifespan (fluctuation suppression), energy consumption (hydrogen consumption), and battery state of charge (SOC), solving the drawbacks of single-objective optimization in the industry.

[0112] This invention constructs a hybrid power system simulation model based on the mechanism model of fuel cells and power batteries, improves the accuracy of the model, and thus improves the usability of energy management strategies. It also enables the interaction between the environment and the agent's strategies to support the cyclical training of reinforcement learning.

[0113] This invention constructs a hybrid system environment model based on a mechanism model and trains it with a reinforcement learning agent. By designing a reward function, it reduces the energy consumption of the fuel cell. At the same time, it introduces a filter to decouple the required power and allocate the low-frequency part to the fuel cell output, thereby improving its service life and reducing the fluctuation of the hydrogen fuel cell output power and hydrogen consumption.

[0114] This invention presents an energy scheduling method based on frequency decoupling and reinforcement learning that simultaneously considers fuel cell lifespan and energy efficiency. The method is trained using operating data from real vehicle routes, which effectively reduces hydrogen consumption under fixed operating conditions, while also reducing fluctuations in fuel cell output power and improving lifespan.

[0115] This invention presents an energy scheduling method based on frequency decoupling and reinforcement learning. By training the energy scheduling problem using reinforcement learning, and through decoupling the demand power and continuous trial and error updates, an energy scheduling strategy, namely the output power allocation strategy, is obtained that can both improve fuel cell lifespan and reduce fuel consumption. Simulation verification using real-vehicle data from a mainline line validates the effectiveness of this energy scheduling method, laying the foundation for its engineering application.

[0116] This invention also discloses a computer program product, comprising a computer program that, when run by a processor, performs the steps of the method described above. This invention further discloses a computer-readable storage medium storing a computer program that, when run by a processor, performs the steps of the method described above. This invention also discloses a fuel cell hybrid vehicle energy dispatching system, comprising an interconnected memory and a processor, wherein the memory stores a computer program that, when run by a processor, performs the steps of the method described above. The products, media, and systems of this invention, corresponding to the methods described above, also possess the advantages described above.

[0117] The present invention can implement all or part of the processes in the methods of the above embodiments, or it can be implemented by hardware related to computer program instructions. The computer program can be stored in a computer-readable storage medium. When the computer program is executed by a processor, it can implement the steps of the above method embodiments. The computer program includes computer program code, which can be in the form of source code, object code, executable file, or some intermediate form. The computer-readable storage medium includes: any entity or device capable of carrying computer program code, recording media, USB flash drive, portable hard drive, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc. The memory is used to store computer programs and / or modules. The processor implements various functions by running or executing the computer programs and / or modules stored in the memory, and by calling data stored in the memory. The memory may include high-speed random access memory, as well as non-volatile memory, such as hard disks, RAM, plug-in hard disks, smart media cards (SMC), secure digital (SD) cards, flash cards, at least one disk storage device, flash memory device, or other volatile solid-state storage devices.

[0118] The above are merely preferred embodiments of the present invention. The scope of protection of the present invention is not limited to the above embodiments. All technical solutions falling within the scope of the present invention's concept are within the scope of protection of the present invention. It should be noted that for those skilled in the art, any improvements and modifications made without departing from the principles of the present invention should be considered within the scope of protection of the present invention.

Claims

1. A method for energy dispatching of fuel cell hybrid vehicles, characterized in that, Includes the following steps: S1. Construct a hydrogen fuel cell hybrid train power system model, which includes a fuel cell model, a lithium battery model, and a traction motor model; S2. Obtain the required power during train operation and decompose the required power into low-frequency and high-frequency components using a low-pass filter with an adjustable cutoff frequency. A two-layer reinforcement learning agent is constructed, and the cutoff frequency of the low-pass filter is used as the action variable of the reinforcement learning agent for real-time dynamic optimization and adjustment. The two-layer reinforcement learning agent includes an outer agent and an inner agent; the outer agent is used to extract global features of the required power and output an initial normalized cutoff frequency value; the inner agent receives the initial normalized cutoff frequency value, optimizes the power allocation strategy in real time, and uses the low-frequency component as the output power of the fuel cell and the high-frequency component as the output power of the lithium battery. S3. Design a multi-objective reward function and a total system cost function to simultaneously optimize fuel cell efficiency, fuel cell power fluctuation, hydrogen consumption, and lithium battery SOC to approach the initial value. The parameters of the two-layer reinforcement learning agent network are updated by a deep deterministic policy gradient algorithm, and finally the real-time power allocation command of the fuel cell and lithium battery is output.

2. The energy dispatching method for fuel cell hybrid vehicles according to claim 1, characterized in that, In step S3, when braking power is detected: if the lithium battery SOC ≥ safety threshold, the regenerative braking energy is allocated to the braking resistor consumption; otherwise, the energy is recovered by the lithium battery.

3. The energy dispatching method for fuel cell hybrid vehicles according to claim 1 or 2, characterized in that, In step S3, the reward function r Total system cost cost total They are respectively: In the formula, For fuel cell efficiency, This represents the current remaining charge of the power battery. The standard deviation of fuel cell output power. This represents the fuel consumption between two adjacent time steps. These are the corresponding weighting coefficients; The cost of starting a fuel cell once, For the number of fuel cell starts, and For the operating costs of fuel cells and power batteries, and This refers to the output power of fuel cells and power batteries.

4. The energy dispatching method for fuel cell hybrid vehicles according to claim 1 or 2, characterized in that, In the fuel cell model, the output voltage of a single fuel cell is shown in the following equation: ； In the formula, The output voltage of a single PEMFC; This is the Nernst voltage; This is due to voltage loss caused by the activation polarization effect; Voltage loss caused by ohmic polarization effect; This is the voltage loss caused by concentration polarization.

5. The energy dispatching method for fuel cell hybrid vehicles according to claim 1 or 2, characterized in that, In the traction motor model, the required torque and speed of the traction motor are obtained, and then the required power of the traction motor is obtained.

6. A computer program product, comprising a computer program, characterized in that, The computer program is executed by the processor to perform the steps of the method as described in any one of claims 1-5.

7. A computer-readable storage medium having a computer program stored thereon, characterized in that, The computer program, when run by a processor, performs the steps of the method as described in any one of claims 1-5.

8. A fuel cell hybrid vehicle energy dispatching system, comprising a memory and a processor interconnected, wherein the memory stores a computer program, characterized in that, The computer program, when run by a processor, performs the steps of the method as described in any one of claims 1-5.

Images