Power distribution method and system based on fuel cell multi-stack health state difference

Abstract

The application discloses a kind of power distribution method and system based on fuel cell multi-stack health state difference, belongs to fuel cell application technical field.The specific method includes monitoring the operating parameters of multiple fuel cell stacks by sensor;The health state of each fuel cell stack and the health state of lithium battery are obtained;Hydrogen consumption model, fuel cell SOH loss model, multi-stack SOH difference equivalent loss model, lithium battery SOH loss model and lithium battery initial-final SOC loss model are established;Equivalent hydrogen loss is calculated by loss conversion, and a target function model is constructed;The target function is optimized using reinforcement learning algorithm, and the power output of fuel cell and lithium battery is distributed according to the current demand power and health state by energy management system, and the parameter identification module and model parameters are updated regularly, to realize dynamic optimization.The application can identify the life difference between multiple stacks, balance the life difference to achieve multi-stack performance balance and improve the overall life of multi-stack system.

Description

TECHNICAL FIELD

[0001] The present application relates to a power distribution method and system, in particular to a power distribution method based on the health state difference of multiple fuel cell stacks and a multiple fuel cell stack system, belonging to the technical field of fuel cell applications. BACKGROUND

[0002] Currently, the large-scale use of fossil fuels has brought about the problems of environmental pollution and energy shortage, which has led to the large-scale development and use of renewable energy. However, the intermittency and volatility of renewable energy production capacity have caused differences in time and space between the power generation side and the load side. In order to solve this problem, hydrogen energy is a long-term energy storage method that can realize cross-day, month and season, that is, green hydrogen is prepared by electrolyzing water when renewable energy is abundant, and fuel cells are used to generate electricity when renewable energy generation is insufficient, electricity demand is high and difficult to meet.

[0003] Fuel cells have limitations such as limited power of a single stack, reduced mass transfer and heat transfer uniformity of high-power stacks, and high cost of auxiliary components, so multiple fuel cell stacks (MSFC) are often used in parallel to achieve higher power output. MSFC can be designed to operate independently between stacks through its modular design, and through reasonable control methods, it can isolate faulty stacks in time without affecting the normal operation of the entire system, and even replacing faulty stacks will not affect the overall electrical connection, thereby enhancing the reliability and maintainability of the power generation system. In addition, when the faulty stack group recovers its health state from adverse conditions such as flooding and membrane drying, it can be added to the power generation system again through control strategies to improve the fault tolerance of the system. However, the introduction of multiple fuel cells can easily lead to slow dynamic response of fuel cells and differences in operating performance between single stacks, which brings new problems to the control and management of multiple stack fuel cell systems. Slow dynamic response is an inherent characteristic of fuel cells, and fast power response can also cause rapid degradation of fuel cell life. In order to solve this problem, energy storage systems such as lithium batteries or supercapacitors need to be added to the fuel cell power generation system to smooth power fluctuations; at the same time, it is necessary to solve the problem of how to distribute power between multiple fuel cell stacks and energy storage systems according to the power generation characteristics of different power sources to achieve stable fuel cell output power, maintain the state of charge of the energy storage system, and the economy of the system power generation.

[0004] The technical solution of the current application focuses on optimizing the hydrogen consumption, life consumption and equivalent hydrogen consumption of the difference between the start and end SOC of the lithium battery of the multi-stack system to improve the overall economy, but ignores the difference in the health state between the stacks. In the MSFC power generation system, some fuel cells have excellent output performance and can achieve higher output power at the same output voltage, and have a longer remaining service life. While some fuel cells have lower output power and shorter remaining service life. According to the "barrel theory", the remaining service life of the entire MSFC depends on the fuel cell with the lowest remaining service life. Therefore, in order to improve the overall life of the multi-stack system, a power distribution algorithm that can identify the life difference between the multi-stacks, balance the life difference, achieve performance balance of the multi-stacks and improve the overall life of the multi-stack system needs to be developed. SUMMARY

[0005] The present application is to solve the problem of short service life of multi-stack system, and further proposes a power distribution method and system based on the health state difference of fuel cell multi-stack.

[0006] In order to achieve the above purpose, the present application adopts the following technical scheme:

[0007] A power distribution method based on the health state difference of fuel cell multi-stack, the power distribution method based on the health state difference of fuel cell multi-stack is realized by the following steps:

[0008] S1: Monitor the operating parameters of the plurality of fuel cell stacks through a sensor, the operating parameters including stack temperature, anode and cathode inlet gas pressure, anode and cathode flow channel humidity, and anode and cathode gas partial pressure;

[0009] S2: Based on the monitoring data obtained in S1, obtain the health state of each fuel cell stack through a parameter identification module;

[0010] S3: Obtain the health state of the lithium battery according to the state of charge of the lithium battery;

[0011] S4: Based on the health state of the fuel cell stack and the health state of the lithium battery obtained in S2 and S3, establish a hydrogen consumption model, a fuel cell SOH consumption model, a multi-stack SOH difference equivalent consumption model, a lithium battery SOH consumption model and a lithium battery start-end SOC consumption model;

[0012] S5: Convert the consumption calculated by the model established in S4 into equivalent hydrogen consumption to construct a target function model;

[0013] S6: Use the reinforcement learning algorithm Proximal Policy Optimization to optimize the target function, train the model based on historical operation data, and obtain the optimized power distribution strategy;

[0014]

[0014] S7: The optimized power distribution strategy obtained in S6 is used by the energy management system to distribute the power output of the fuel cell and lithium battery according to the current demand power and state of health, output the reference current of DC / DC, and realize power distribution;

[0015] S8: The parameter identification module and model parameters are updated regularly to realize dynamic optimization.

[0016] Further, the state of health of the fuel cell stack is calculated by the following formula:

[0017]

[0018] wherein, is the voltage when the fuel cell stack is not in use, is the current operating voltage, is the current operating rated voltage, is the rated voltage obtained by polarization curve identification; is the output voltage of the PEMFC, E is the open circuit voltage of the PEMFC, is the ohmic loss voltage, is the activation loss voltage, is the concentration difference loss voltage.

[0019] Further, the hydrogen consumption model is calculated by the following formula:

[0020]

[0021] wherein, is the number of single cells in the fuel cell stack, is the molar mass of hydrogen, F is the Faraday constant, is the excess ratio of hydrogen, is the consumption rate of hydrogen; is the power during the operation of the fuel cell.

[0022] Further, the fuel cell SOH loss model is divided into start-stop, idle, low-power constant load, medium-power constant load, high-power constant load, rapid load change, and slow load change according to seven working conditions, and is calculated by the following formula:

[0023] The voltage loss at each moment is:

[0024]

[0025] wherein, is a coefficient that changes with the running time of the fuel cell, A, B, C, D, E, F, and G are coefficients corresponding to the seven working conditions, is the running time, SOH loss of the power change, number of times;

[0026] SOH loss of the power change,

[0027]

[0028] wherein, price coefficient of the fuel cell stack, rated voltage of the stack operation,

[0029] SOH loss of the power change,

[0030]

[0031] Further, the SOH difference equivalent loss model of the multiple stacks is calculated by the following formula:

[0032]

[0033] wherein, the shortest life stack in the multiple stack, price coefficient related to the fuel cell life loss, n is the number of stacks in the multiple stack.

[0034] Further, the SOH loss model of the lithium battery is calculated by the following formula:

[0035] The lithium battery adopts the common second-order RC equivalent circuit model, and the life loss is:

[0036]

[0037] wherein, current of the lithium battery operation, time constant, cycle number related to the battery temperature, battery capacity.

[0038] SOH equivalent loss of the lithium battery,

[0039]

[0040] Further, the lithium battery start-end SOC loss model is calculated by the following formula:

[0041]

[0042] wherein, coefficient related to the electricity price and the hydrogen price, state of charge at the beginning, state of charge at the end.

[0043] Further, the target function is calculated by the following formula:

[0044]

[0045] wherein, is a weight coefficient related to the price of the fuel cell, is a weight coefficient related to the difference in the state of health of the fuel cell, is a weight coefficient related to the loss of the lithium battery life, is a weight coefficient related to the loss of the lithium battery SOC.

[0046] Further, the state of the reinforcement learning algorithm Proximal Policy Optimization is the current demand power and the state of health, the action is the power distribution strategy of the fuel cell and the lithium battery, and the reward is the output of the target function.

[0047] Further, the energy management system outputs the reference current of the DC / DC according to the optimized power distribution strategy, and realizes dynamic optimization and maximization of the system life by periodically updating the parameter identification module and the model parameters.

[0048] A fuel cell multi-stack system for implementing the power distribution method based on the difference in the state of health of the fuel cell multi-stack, the fuel cell multi-stack system comprising:

[0049] a plurality of fuel cell stacks connected in parallel, the fuel cell being connected to a DC bus through a boost DC / DC controller;

[0050] a lithium battery connected to the DC bus through a bidirectional DC / DC;

[0051] a plurality of fuel cell system sensors for monitoring the operating parameters of the stacks;

[0052] a plurality of fuel cell parameter identification modules for obtaining the state of health of the stacks;

[0053] a lithium battery parameter identification module for obtaining the state of health of the lithium battery;

[0054] an energy management system for executing the power distribution strategy.

[0055] Further, the system delays the overall life of the multi-stack system by dynamically optimizing the power distribution strategy, and reduces the consumption of hydrogen, the loss of the lithium battery, and the loss of electric energy.

[0056] The present application has the following advantages:

[0057] 1. This invention can achieve lower hydrogen consumption, single-stall life loss and lithium battery life loss in multi-stall fuel cell systems, and optimize the differences in health status between stacks, thereby improving the remaining service life of the entire multi-stall system.

[0058] 2. This invention can achieve the maximum power distribution effect across the entire lifespan of a multi-stack fuel cell system while ensuring that the sum of hydrogen consumption, fuel cell stack lifespan loss, lithium battery lifespan loss, and power loss is relatively small. Attached Figure Description

[0059] Figure 1 This is a schematic diagram of the architecture of a fuel cell multi-stack system according to one embodiment of the present invention, which implements a power allocation method based on the differences in health status of multiple fuel cell stacks.

[0060] Figure 2 This is a schematic flowchart of the power allocation method based on the differences in health status among multiple fuel cell stacks according to the present invention.

[0061] Figure 3 This is a schematic diagram comparing the changes in SOH (State of Health) of fuel cells in multiple stack systems. Detailed Implementation

[0062] Specific implementation method one: Combining Figures 1-3 This implementation method is described as follows: Figure 1 As shown in this embodiment, the fuel cell multi-stack system implementing a power allocation method based on the differences in the health states of multiple fuel cell stacks includes: multiple fuel cell stacks connected in parallel, with the fuel cells connected to a DC bus via a boost DC / DC controller; lithium batteries connected to the DC bus via a bidirectional DC / DC converter; multiple fuel cell system sensors for monitoring stack operating parameters; multiple fuel cell parameter identification modules for acquiring stack health states; a lithium battery parameter identification module for acquiring lithium battery health states; and an energy management system for executing power allocation strategies. The system extends the overall lifespan of the multi-stack system and reduces hydrogen consumption, lithium battery wear, and electrical energy loss by dynamically optimizing the power allocation strategy.

[0063] Specifically, three fuel cells are connected in parallel to form a multi-stack system. The fuel cells are connected to the DC bus via a boost DC / DC controller, and the lithium batteries are connected to the DC bus via a bidirectional DC / DC converter. Sensors in the fuel cell system monitor the stack temperature. Anode and cathode intake pressure and Anode and cathode flow channel humidity and Partial pressure of anode and cathode gases and , sensor information is transmitted to the parameter identification module to obtain the health state SOH of each fuel cell stack. The lithium battery parameter identification module obtains the health state of the lithium battery according to the state of charge SOC . Finally, the energy management system allocates the reference current of the DC / DC and the current demand power according to the SOH of the fuel cell and the and of the lithium battery.

[0064] As Figure 2 shown, the power distribution method based on the health state difference of multiple fuel cell stacks is realized by the following steps:

[0065] S1: Monitor the operating parameters of multiple fuel cell stacks through sensors, including stack temperature, anode and cathode inlet pressure, anode and cathode flow channel humidity, and anode and cathode gas partial pressure;

[0066] S2: Based on the monitoring data obtained in S1, obtain the health state of each fuel cell stack through the parameter identification module;

[0067] The fuel cell polarization curve model can reflect the evolution law of the output performance of the fuel cell with the change of the health state, but since there are numerous nonlinear parameters in the polarization curve model, identification is required:

[0068] (1)

[0069] In formula (1), is the output voltage of PEMFC, E is the open circuit voltage of PEMFC, is the ohmic loss voltage, is the activation loss voltage, is the concentration loss voltage. These variables are related to the stack temperature , anode and cathode inlet pressure and , anode and cathode flow channel humidity and , anode and cathode gas partial pressure and .

[0070] The health state SOH of the fuel cell stack is calculated by the following formula:

[0071] (2)

[0072] In formula (2), is the voltage of the fuel cell stack when not in use, the voltage for the current operation, the rated voltage for the current operation, the rated voltage obtained by polarization curve identification.

[0073] Specifically, based on the stack temperature measured by the stack sensor , the anode and cathode inlet gas pressures and , the anode and cathode flow channel humidities and , the anode and cathode gas partial pressures and , the SOH of the stack is obtained according to formulas (1) and (2).

[0074] S3: obtaining the health state of the lithium battery according to the state of charge of the lithium battery;

[0075] S4: based on the health state of the fuel cell stack and the health state of the lithium battery obtained by S2 and S3, establishing a hydrogen consumption model, a fuel cell SOH loss model, a multi-stack SOH difference equivalent loss model, a lithium battery SOH loss model, and a lithium battery initial-final SOC loss model;

[0076] Hydrogen consumption model considering the health state difference of multiple stacks:

[0077] The hydrogen consumption of the fuel cell stack is:

[0078] (3)

[0079] In formula (3), is the number of single cells in the fuel cell stack, is the molar mass of hydrogen, and F is the Faraday constant, is the excess ratio of hydrogen, is the hydrogen consumption rate; is the power during the operation of the fuel cell.

[0080] From formula (2), we have:

[0081] (4)

[0082] Combining formula (3), the hydrogen consumption model calculation formula is as follows:

[0083] (5)

[0084] As can be seen from formula (5), in a multi-stack system, the health difference between multiple stacks will cause the hydrogen consumption curves of multiple stacks to be different. Therefore, the energy management system needs to fully consider the system efficiency difference caused by the difference in hydrogen consumption curves.

[0085] Fuel cell SOH loss model:

[0086] According to the load demand, seven working conditions are divided: start-stop, idle, low-power constant load, medium-power constant load, high-power constant load, rapid load change and slow load change. The voltage loss at each moment is:

[0087] (6)

[0088] In formula (6), is a coefficient that changes with the running time of the fuel cell, A, B, C, D, E, F, and G are coefficients corresponding to the seven working conditions, is the running time, is the power change, is the number of times.

[0089] The SOH loss of the stack is:

[0090] (7)

[0091] wherein, is the price coefficient of the fuel cell stack, is the rated voltage of the stack operation,

[0092] The SOH change of the stack is:

[0093] (8)

[0094] Multi-stack SOH difference equivalent loss model:

[0095] The multi-stack SOH difference equivalent loss model is calculated by the following formula:

[0096] (9)

[0097] wherein, is the shortest life stack in the multi-stack electric cone, is the price coefficient related to the life loss of the fuel cell, and n is the number of stacks in the multi-stack electric cone.

[0098] Since the life of the multi-stack depends on the life of the shortest life stack , the sum of the difference between the SOH of the three stacks can be taken as the loss.

[0099]

[0100] Lithium battery SOH loss model:

[0101] The life loss of the lithium battery is:

[0102] (10)

[0103] wherein, is the current of lithium battery operation, is the time constant, is the cycle number related to the battery temperature, is the battery capacity.

[0104] The equivalent loss of lithium battery SOH is:

[0105] (11)

[0106] The lithium battery start-end SOC loss model is:

[0107] During the operation of the lithium battery, the SOC at the beginning and end is different in most states, which means that the lithium battery produces power loss or supplement, therefore, this part also needs to be considered:

[0108] (12)

[0109] wherein, is the coefficient related to the electricity price and hydrogen price, is the SOC value at the beginning, is the SOC value at the end.

[0110] S5: Convert the loss calculated by the model established in S4 into equivalent hydrogen loss, and construct a target function model;

[0111] The target function model is:

[0112] The target function of the power distribution method considering the health difference of multiple stacks is:

[0113] (13)

[0114] wherein, is the weight coefficient related to the fuel cell price, is the weight coefficient related to the health state difference of fuel cell, is the weight coefficient related to the lithium battery life loss, is the weight coefficient related to the lithium battery SOC loss.

[0115] That is, according to the The hydrogen loss model, life loss model and equivalent loss model of multiple stack SOH difference, as well as the lithium battery SOH loss model and the lithium battery start-end SOC loss model are established, the losses calculated by the five models are multiplied by the corresponding coefficients and added together to obtain the target function model.

[0116] S6: The objective function is optimized using the reinforcement learning algorithm Proximal Policy Optimization. The model is trained based on historical running data to obtain the optimized power allocation strategy.

[0117] The reinforcement learning algorithm Proximal Policy Optimization's state is the current power demand and health status, the action is the power allocation strategy for the fuel cell and lithium battery, and the reward is the output of the objective function.

[0118] This implementation uses the reinforcement learning algorithm Proximal Policy Optimization (PPO), trained on historical operational data from multiple reactors, and then the optimized algorithm parameters are retained before actual operation on a multi-reactor energy management platform. The objective function model is updated periodically based on the parameter identification results. The algorithm's state and action are as follows:

[0119] (14)

[0120] (15)

[0121] S7: The optimized power allocation strategy obtained from S6 is used by the energy management system to allocate the power output of the fuel cell and lithium battery according to the current power demand and health status, and outputs a DC / DC reference current to achieve power allocation;

[0122] That is, the output of the objective function is used as the reward for the Proximal Policy Optimization (PPO) algorithm, and the state is... action is Using historical demand power data from multiple reactors as the training set, convergent parameters are obtained, and the optimized algorithm is run on the energy management platform.

[0123] S8: Regularly update the parameter identification module and model parameters to achieve dynamic optimization; regularly identify and update the SOH of each fuel cell stack based on sensor data parameters.

[0124] like Figure 3 As shown on the left (a), a multi-stack system is composed of three 60kW fuel cells connected in parallel. When the multi-stack power allocation method of this invention, which considers differences in health status, is adopted, the SOH changes of the three fuel cell stacks are similar, which helps to extend the life of this multi-stack system. Figure 3 As shown on the right (b), when the differences in health status between multiple reactors are not considered, the SOH of a certain reactor in the entire multi-reactor system decays significantly faster, which reduces the lifetime of the entire multi-reactor system.

[0125] Working principle:

[0126] Firstly, a hydrogen consumption model considering the difference of health states of multiple stacks is established according to the information of sensors and the identified health states of fuel cells, and then the required power is divided into seven types: start-stop, idle, low-power constant load, medium-power constant load, high-power constant load, rapid load change and slow load change, so as to establish a fuel cell single-stack SOH loss model. In order to balance the SOH difference between multiple stacks, a multi-stack SOH difference equivalent loss model is established. At the same time, a lithium battery life loss model and a start-to-end electrical energy consumption model are established, and according to the prices of lithium batteries, hydrogen and fuel cells, the losses calculated by the above models are converted into equivalent hydrogen losses, thereby establishing the objective function and constraints; then, the Proximal Policy Optimization (PPO) algorithm is used to train the historical operation data of the multi-stack system as the training set, and the trained model is applied to the energy management system of the multi-stack system; the energy management system outputs the reference current of the DC / DC according to the current health states of the fuel cells and lithium batteries and the power demand; finally, after a period of operation, the parameter identification module updates the corresponding parameters, and the hydrogen consumption model, the fuel cell SOH loss model, the lithium battery life loss model and the start-to-end electrical energy consumption model are updated accordingly.

[0127] The above is only a preferred embodiment of the present application, and does not limit the present application in any form. Although the present application has been disclosed as above with a preferred embodiment, it is not intended to limit the present application. Any skilled person in the art can make some changes or modifications to the above disclosed technical content without departing from the scope of the technical solution of the present application, and can make equivalent embodiments with equivalent changes. However, any simple modification, equivalent replacement and improvement of the above embodiments within the scope of the technical solution of the present application, according to the technical essence of the present application, within the spirit and principles of the present application, are still within the protection scope of the present application.

Claims

1. A power allocation method based on the differences in health states of multiple fuel cell stacks, characterized in that: The power allocation method based on the differences in health status among multiple fuel cell stacks is implemented through the following steps: S1: Monitor the operating parameters of multiple fuel cell stacks through sensors, including stack temperature, anode and cathode inlet pressure, anode and cathode flow channel humidity, and anode and cathode gas partial pressure; S2: Based on the monitoring data obtained in S1, the health status of each fuel cell stack is obtained through the parameter identification module; The health status of the fuel cell stack is calculated using the following formula: in, This is the voltage of the fuel cell stack when it is not in use. The current operating voltage. The rated voltage for current operation. The rated voltage obtained through polarization curve identification; E is the output voltage of the PEMFC, and E is the open-circuit voltage of the PEMFC. This is the ohmic loss voltage. To activate the loss voltage, This is the concentration loss voltage; S3: Obtain the health status of the lithium battery based on its state of charge. S4: Based on the health status of the fuel cell stack and the lithium battery obtained in S2 and S3, establish hydrogen consumption model, fuel cell SOH loss model, multi-stack SOH difference equivalent loss model, lithium battery SOH loss model and lithium battery initial and final SOC loss model. The hydrogen consumption model is calculated using the following formula: in, This refers to the number of individual cells in a fuel cell stack. Let F be the molar mass of hydrogen gas, and F be Faraday's constant. The excess ratio of hydrogen gas. This refers to the power generated during the operation of the fuel cell; The fuel cell SOH loss model is divided into seven operating conditions: start-up / shutdown, idling, low-power constant load, medium-power constant load, high-power constant load, rapid load change, and slow load change. The model is calculated using the following formula: The voltage loss at each moment is: in, Here, A, B, C, D, E, F, and G are coefficients that vary with the fuel cell's operating time, corresponding to the seven operating conditions. For runtime, For power change, For the number of times, The SOH loss of the fuel cell stack is: in, This is the price coefficient for fuel cell stacks. This is the rated voltage for the fuel cell stack operation. The SOH change of the fuel cell stack is as follows: ； The multi-stacking SOH differential equivalent loss model is calculated using the following formula: in, It has the shortest lifespan among multiple electric cone stacks. The price coefficient is related to the lifespan loss of the fuel cell, and n is the number of fuel cell stacks in the multi-stack cone. The lithium battery SOH loss model is calculated using the following formula: Lithium batteries use a common second-order RC equivalent circuit model, and their lifespan loss is as follows: in, This is the operating current of the lithium battery. It is a time constant. The number of cycles related to battery temperature. For battery capacity, The equivalent loss of SOH in a lithium battery is: ； The lithium battery's initial and final SOC loss model is calculated using the following formula: in, The coefficients related to electricity and hydrogen prices, The initial state of charge, The state of charge at the end; S5: Convert the losses calculated by the model established in S4 into equivalent hydrogen losses and construct the objective function model; The objective function is calculated using the following formula: in, The weighting coefficients related to the price of fuel cells, Weighting coefficients related to differences in fuel cell health status. The weighting coefficients related to lithium battery life degradation Weighting coefficients related to lithium battery SOC loss; S6: The objective function is optimized using the reinforcement learning algorithm Proximal Policy Optimization. The model is trained based on historical running data to obtain the optimized power allocation strategy. S7: The optimized power allocation strategy obtained from S6 is used by the energy management system to allocate the power output of the fuel cell and lithium battery according to the current power demand and health status, and outputs a DC / DC reference current to achieve power allocation; S8: Regularly update the parameter identification module and model parameters to achieve dynamic optimization.

2. The power allocation method based on the health status differences of multiple fuel cell stacks according to claim 1, characterized in that: The reinforcement learning algorithm Proximal Policy Optimization's state is the current power demand and health status, the action is the power allocation strategy for the fuel cell and lithium battery, and the reward is the output of the objective function.

3. A fuel cell multi-stack system implementing the power allocation method based on the health status differences of multiple fuel cell stacks as described in any one of claims 1 to 2, characterized in that: The fuel cell multi-stack system includes: Multiple fuel cell stacks are connected in parallel, and the fuel cells are connected to the DC bus via a boost DC / DC controller; The lithium battery is connected to the DC bus via a bidirectional DC / DC converter. Multiple fuel cell system sensors are used to monitor stack operating parameters; Multiple fuel cell parameter identification modules are used to obtain the health status of the fuel cell stack; A lithium battery parameter identification module is used to obtain the health status of the lithium battery; An energy management system is used to execute power allocation strategies.

Images