WO2023000985A1 - 燃料电池的剩余寿命预测方法及装置 - Google Patents

燃料电池的剩余寿命预测方法及装置 Download PDF

Info

Publication number
WO2023000985A1
WO2023000985A1 PCT/CN2022/104187 CN2022104187W WO2023000985A1 WO 2023000985 A1 WO2023000985 A1 WO 2023000985A1 CN 2022104187 W CN2022104187 W CN 2022104187W WO 2023000985 A1 WO2023000985 A1 WO 2023000985A1
Authority
WO
WIPO (PCT)
Prior art keywords
fuel cell
voltage
behavior
remaining life
predicting
Prior art date
Application number
PCT/CN2022/104187
Other languages
English (en)
French (fr)
Inventor
王恺
赵洪辉
王宇鹏
都京
魏凯
马秋玉
刘岩
汝春宇
吕文博
Original Assignee
中国第一汽车股份有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 中国第一汽车股份有限公司 filed Critical 中国第一汽车股份有限公司
Publication of WO2023000985A1 publication Critical patent/WO2023000985A1/zh

Links

Images

Classifications

    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
    • G01R31/00—Arrangements for testing electric properties; Arrangements for locating electric faults; Arrangements for electrical testing characterised by what is being tested not provided for elsewhere
    • G01R31/36—Arrangements for testing, measuring or monitoring the electrical condition of accumulators or electric batteries, e.g. capacity or state of charge [SoC]
    • G01R31/378—Arrangements for testing, measuring or monitoring the electrical condition of accumulators or electric batteries, e.g. capacity or state of charge [SoC] specially adapted for the type of battery or accumulator
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
    • G01R31/00—Arrangements for testing electric properties; Arrangements for locating electric faults; Arrangements for electrical testing characterised by what is being tested not provided for elsewhere
    • G01R31/36—Arrangements for testing, measuring or monitoring the electrical condition of accumulators or electric batteries, e.g. capacity or state of charge [SoC]
    • G01R31/367—Software therefor, e.g. for battery testing using modelling or look-up tables
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
    • G01R31/00—Arrangements for testing electric properties; Arrangements for locating electric faults; Arrangements for electrical testing characterised by what is being tested not provided for elsewhere
    • G01R31/36—Arrangements for testing, measuring or monitoring the electrical condition of accumulators or electric batteries, e.g. capacity or state of charge [SoC]
    • G01R31/392—Determining battery ageing or deterioration, e.g. state of health
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02E—REDUCTION OF GREENHOUSE GAS [GHG] EMISSIONS, RELATED TO ENERGY GENERATION, TRANSMISSION OR DISTRIBUTION
    • Y02E60/00—Enabling technologies; Technologies with a potential or indirect contribution to GHG emissions mitigation
    • Y02E60/30—Hydrogen technology
    • Y02E60/50—Fuel cells

Definitions

  • the present application relates to the technical field of fuel cells, for example, to a method and device for predicting the remaining life of a fuel cell.
  • Fuel cells have the advantages of less greenhouse gas emissions and high energy conversion efficiency. However, the vehicle-mounted fuel cell system is frequently started and stopped, and the working conditions are complicated, which seriously affects the service life of the fuel cell system. Therefore, it is of great significance to effectively predict the remaining service life of fuel cells to prolong their service life and promote the commercial development of fuel cells.
  • the present application provides a method and device for predicting the remaining life of a fuel cell, so as to realize accurate prediction of the remaining life of the fuel cell under driving conditions.
  • the present application provides a method for predicting the remaining life of a fuel cell, the fuel cell is composed of a plurality of monomers, and the method for predicting the remaining life of the fuel cell includes:
  • parameter data of the fuel cell during operating conditions at least includes voltage, current, cell voltage, air intake flow rate, air intake pressure and cooling water inlet temperature;
  • the acquisition of parameter data of the fuel cell during operating conditions includes:
  • the acquiring the fusion health indicator according to the parameter data includes:
  • the obtaining the standard voltage V ek of the fuel cell at different times during the operating conditions according to the parameter data includes:
  • V ek af(I)+bf(p)+cf(W)+df(T in ) of the fuel cell at different times during the operating conditions; wherein, I is the current, p is the air intake pressure, W is the air intake flow rate, T in is the cooling water inlet temperature, a, b, c and d are coefficients, f (I), f (p), f (W) and f(T in ) refer to functions with respect to I, p, W and T in , respectively.
  • the obtaining cell voltage fluctuation rate C v according to the cell voltage and the number of cells in the fuel cell includes:
  • V i is the cell voltage of the ith cell of the fuel cell, is the average voltage of all the cells in the fuel cell, N is the number of cells in the fuel cell, 1 ⁇ i ⁇ N.
  • the obtaining the health state value of the fuel cell according to the fusion health index includes:
  • the method further includes:
  • Each behavior of the fuel cell during the operating conditions is digitally marked; wherein, the marked data of the behavior can be determined according to the impact level of the behavior on the voltage decay of the fuel cell Setting; different data represent different levels of influence.
  • the training of the long-short-term memory neural network model according to time, the health state value and the behavior includes:
  • the long-short-term memory neural network model is trained so that the trained long-short-term memory neural network model simulates the operating conditions of the fuel cell.
  • the predicting the remaining life of the fuel cell according to the long-short-term memory neural network model includes:
  • the operating time of the fuel cell when running to the failure threshold is determined according to the long short-term memory neural network model, and the remaining life of the fuel cell is determined according to the operating time.
  • the application provides a fuel cell remaining life prediction device, including:
  • a parameter acquisition module configured to acquire parameter data of the fuel cell during the operating conditions; wherein the parameter data at least includes voltage, current, cell voltage, air intake flow rate, air intake pressure and cooling water inlet temperature ;
  • the index fitting module is configured to obtain the fusion health index according to the parameter data; and obtain the health state value of the fuel cell according to the fusion health index;
  • a behavior acquisition module configured to acquire the behavior of the fuel cell during the operating conditions; wherein the behavior at least includes start, stop, load change and emergency stop;
  • a model training module configured to train a long-short-term memory neural network model according to time, the health state value and the behavior;
  • the life prediction module is configured to predict the remaining life of the fuel cell according to the long short-term memory neural network model.
  • the present application provides a computer device, including: a memory and a processor, the memory stores a computer program, and the processor is configured to implement the above method for predicting the remaining life of a fuel cell when executing the computer program.
  • the present application provides a computer-readable storage medium, which stores at least one program instruction, and the at least one program instruction is used to execute the above method for predicting the remaining service life of a fuel cell.
  • Fig. 1 is a flowchart of a method for predicting the remaining life of a fuel cell provided in an embodiment of the present application
  • FIG. 2 is a flow chart of another method for predicting the remaining life of a fuel cell provided in an embodiment of the present application
  • Fig. 3 is a flow chart of another method for predicting the remaining life of a fuel cell provided in an embodiment of the present application
  • Fig. 4 is a training flowchart of a behavior-based long-short-term memory neural network model provided by the embodiment of the present application;
  • Fig. 5 is the training flowchart of another behavior-based long-short-term memory neural network model provided by the embodiment of the present application.
  • FIG. 6 is a schematic structural diagram of a fuel cell remaining life prediction device provided in an embodiment of the present application.
  • FIG. 7 is a schematic structural diagram of a computer device provided by an embodiment of the present application.
  • Fig. 1 is a flow chart of the first method for predicting the remaining life of a fuel cell provided in an embodiment of the present application, which is suitable for predicting the remaining life of a fuel cell.
  • Fuel cells are composed of multiple monomers, and the remaining life prediction methods of fuel cells include:
  • parameter data of the fuel cell during operating conditions at least includes voltage, current, cell voltage, air intake flow rate, air intake pressure, and cooling water inlet temperature.
  • Voltage refers to the actual voltage value of the fuel cell at the current moment. When the fuel cell is working, it is equivalent to a DC source.
  • the anode is the negative pole of the power supply
  • the cathode is the positive pole of the power supply.
  • the voltage between the anode and the cathode is the voltage in this implementation. The magnitude of the voltage value can determine whether the fuel cell can continue to be used.
  • the current refers to the current of the stack.
  • the fuel cell is composed of multiple monomers. A single cell is a sub-battery unit. Stacking multiple monomers can form a fuel cell stack whose output voltage meets the actual load requirements, referred to as the stack.
  • the cell voltage refers to the voltage value of the cell in the fuel cell at the current moment. The difference in the voltage distribution of the fuel cell cells exists objectively. If the body is reversed, it will seriously affect the durability of the fuel cell.
  • Air intake flow refers to the gas mass flow rate of fuel gas entering the fuel cell from the anode or cathode of the fuel cell.
  • hydrogen flows into the fuel cell from the inlet
  • oxygen flows into the fuel cell from the inlet.
  • the air intake pressure refers to the gas pressure at which the fuel gas enters the fuel cell from the anode or cathode of the fuel cell.
  • the pressure of the reaction gas on both sides of the proton exchange membrane should be kept relatively balanced.
  • the diffusion of fuel into the proton exchange membrane can be controlled to a minimum, and on the other hand, it can protect the proton exchange membrane. effect.
  • the cooling water inlet temperature refers to the temperature of the cooling water at the inlet of the fuel cell cooling device.
  • the fuel cell also generates a large amount of heat energy in the process of generating electric energy, so the cooling device is required to absorb the heat of the fuel cell, for example, it can be the cooling water inlet of the cooling water pump temperature.
  • parameters such as voltage, current, cell voltage, air intake flow rate, air intake pressure, and cooling water inlet temperature can be obtained.
  • the above parameter data are all measured during the operating condition of the fuel cell.
  • the operating condition is the situation where the fuel cell outputs power to the outside. For example, for a vehicle powered by a fuel cell, when the vehicle starts or drives , which is during the operating conditions of the fuel cell.
  • Converged health indicators refer to non-single health indicators, which are comprehensive health indicators established from a variety of different data.
  • the health index established comprehensively from data such as fuel cell voltage, current, cell voltage, air intake flow rate, air intake pressure, and cooling water inlet temperature, a single health index cannot fully reflect the health status of the fuel cell.
  • the health status value refers to the value calculated by the fuel cell based on the fusion health index. Different health status values represent different performances of the fuel cell. For example, the lower the health status value of the fuel cell, the worse the performance of the fuel cell. The health of the fuel cell The higher the status value, the better the performance of the fuel cell.
  • a fusion health index based on the obtained parameter data such as voltage, current, cell voltage, air intake flow, air intake pressure, and cooling water inlet temperature, obtain the value of the parameter data of the fuel cell, and according to the fusion health index can be obtained It corresponds to the state of health value of the fuel cell.
  • Time refers to the time the fuel cell is running.
  • the long short memory neural network model refers to a deep learning method for processing sequence data.
  • the long-short-term memory neural network model is trained by using time, health state value and behavior as input to obtain a training model with accurate targets, and the remaining life time of the fuel cell is predicted according to the long-short-term memory neural network model.
  • the fusion is obtained according to the parameter data Health indicators, and then obtain the health status value of the fuel cell, and obtain the behaviors of the fuel cell such as start, stop, load change and emergency stop during the operating condition, and train the long-term short-term memory neural network model according to time, health status value and behavior, And predict the remaining life of the fuel cell according to the long short-term memory neural network model. Realize the accurate prediction of the remaining life of the fuel cell by fusing health indicators and behaviors under dynamic driving conditions.
  • acquiring the parameter data of the fuel cell during the operating condition may include: acquiring a stable time interval of the voltage of the fuel cell during the operating condition; acquiring parameter data within the stable time interval.
  • the stable time interval refers to the time interval after the vehicle starts for a period of time. Since the output voltage of the fuel cell is unstable when the vehicle is just started, the parameter data corresponding to the stable time interval of the output voltage of the fuel cell is obtained to ensure the reliability and validity of the data.
  • Fig. 2 is a flow chart of another method for predicting the remaining life of a fuel cell provided in the embodiment of the present application. As shown in Fig. 2, on the basis of the above embodiment, obtaining fusion health indicators according to parameter data mainly includes:
  • parameter data of the fuel cell during operating conditions at least include voltage, current, cell voltage, air intake flow rate, air intake pressure, and cooling water inlet temperature.
  • the standard voltage is calculated according to the parameter data of the fuel cell in the stable time interval, when the fuel cell has a stable output voltage and is in good condition.
  • the actual voltage V real of the fuel cell at different times is obtained, and obtained according to the standard voltage V ek and the actual voltage V real at each time
  • the difference ⁇ V may be a positive value, a negative value or zero.
  • a, b, c and d are coefficients, and f(I), f(p), f(W) and f(T in ) refer to current I, air intake pressure p, air intake flow rate W and cooling water Function of inlet temperature T in .
  • the number of cells is the number of single cells in the fuel cell. According to the cell voltage and the number of cells in the fuel cell, the cell voltage fluctuation rate C v at the current moment can be obtained from the cell voltage fluctuation rate C v. The balance of the cell voltage of the fuel cell, the better the balance of the cell voltage, the better the durability of the fuel cell.
  • obtaining the cell voltage fluctuation rate C v according to the cell voltage and the number of cells in the fuel cell includes: obtaining the cell voltage fluctuation rate in the fuel cell Among them, V i is the cell voltage of the i-th cell of the fuel cell, is the average voltage of all monomers in the fuel cell, and N is the number of monomers in the fuel cell.
  • the cell voltage fluctuation rate C v is used to evaluate the cell voltage balance of the fuel cell, for example, the larger the cell voltage fluctuation rate C v is, the worse the cell voltage balance of the fuel cell is.
  • the cell voltage fluctuation rate C v in the fuel cell is obtained, where 1 ⁇ i ⁇ N.
  • the cell voltage fluctuation rate C v is used as the input of the fusion health index HI, and the cell voltage balance is taken into account to ensure a more accurate prediction of the remaining service life of the fuel cell.
  • the cell data n of the lowest cell voltage at different times can be obtained statistically. For example, you can set the minimum cell voltage value, obtain the cell voltage of each cell at different times, and compare it with the minimum cell voltage value, if there are n cells whose cell voltage is less than or equal to the minimum cell voltage at the current moment The body voltage value, then the cell data of the lowest cell voltage is n.
  • the fusion health index is represented by HI, and f( ⁇ V,n,C v ) refers to the functional relationship formed by the difference ⁇ V, the cell voltage fluctuation rate C v and the cell data n of the lowest cell voltage.
  • the fusion health index HI is established according to the difference ⁇ V, the cell voltage fluctuation rate C v and the cell data n of the lowest cell voltage.
  • the fusion health index HI is about the difference ⁇ V, cell voltage fluctuation rate C v and the minimum cell It is a function of the individual data n of the body voltage, and any change in any variable will affect the value of the fusion health index HI.
  • the above S202-S205 is the process of "obtaining a fusion health indicator according to the parameter data". In this way, multiple health indicators are comprehensively considered to realize accurate evaluation of the current health status of the fuel cell.
  • This embodiment details the process of obtaining fusion health indicators based on parameter data, through the difference ⁇ V between the standard voltage and the actual voltage of the fuel cell, the cell voltage fluctuation rate Cv , and the cell data n of the lowest cell voltage
  • Establishing the fusion health index and fusing multiple parameter data such as current I, air intake pressure p, air intake flow rate W, and cooling water inlet temperature T in can help fully reflect the health status of the fuel cell.
  • Fig. 3 is a flowchart of another method for predicting the remaining life of a fuel cell provided by the embodiment of the present application. As shown in Fig. 3, on the basis of the above embodiment, this embodiment obtains the health status of the fuel cell according to the fusion health index To illustrate the value, the remaining life prediction methods of fuel cells include:
  • parameter data of the fuel cell during operating conditions at least includes voltage, current, cell voltage, air intake flow rate, air intake pressure, and cooling water inlet temperature.
  • the health index value HI 0 at the initial moment refers to the health index value when the fuel cell is factory-set.
  • Time t is the time when the fuel cell is put into use for a certain period of time starting from the initial time t 0.
  • the voltage, current, cell voltage, air intake flow rate, and air flow rate of the fuel cell at time t are obtained.
  • the above S306-S308 is the process of "obtaining the health state value of the fuel cell according to the fusion health index". In this way, according to the realization process of the fuel cell health status value, the fuel cell health status value in any time period can be obtained, the current health status of the fuel cell can be accurately predicted, and the real-time monitoring of the fuel cell under driving conditions can be realized.
  • S310 Mark each behavior of the fuel cell in a data-based manner during the operating condition; wherein, the data marked by the behavior can be set according to the level of influence of the behavior on the voltage decay of the fuel cell; different data represent different impact level.
  • Dataization refers to a quantitative mode. For example, when the fuel cell has no behavior during the operating condition, it is represented by binary number 0000, start is represented by binary number 0001, stop is represented by binary number 0010, load change is represented by binary number 0011, failure urgent The disabled binary number 0100 indicates that the digital marking method is not limited in the embodiment of this application.
  • the voltage attenuation of the fuel cell refers to the reduction of the output voltage of the fuel cell, for example, it can be the drop of the voltage under the same power, the imbalance of the cells or the increase of the lowest cell voltage, etc., which will affect the voltage attenuation of the fuel cell.
  • Each behavior of the fuel cell during operating conditions is marked digitally, and the marked data can be set according to the impact level of the behavior on the voltage decay of the fuel cell. For example, the greater the value of the marked data, the The greater the influence of the behavior on the voltage decay of the fuel cell. In this way, the influence of the behavior on the voltage attenuation of the fuel cell can be judged according to the size of the marked data, which is convenient for the fuel cell system to analyze and process according to the data marked for each behavior, and then establish a training model that is more in line with the actual situation, and improve the training model. target accuracy.
  • predicting the remaining life of the fuel cell according to the long-short-term memory neural network model includes: obtaining the failure threshold of the fuel cell according to the long-short-term memory neural network model; The operating time of the fuel cell is determined based on the operating time to determine the remaining life of the fuel cell.
  • the failure threshold refers to the minimum output voltage value of the fuel cell during operating conditions, for example, it may be a factory set value of the fuel cell, for example, it may be 20% of the rated output voltage of the fuel cell. When the output voltage of the fuel cell reaches the failure threshold during operating conditions, the fuel cell cannot continue to be used.
  • the running time refers to the time elapsed from the moment the fuel cell starts to run after leaving the factory, and runs until the output voltage of the fuel cell reaches the failure threshold.
  • the failure threshold of the fuel cell can be obtained, and the fuel cell can be calculated according to the long-short-term memory neural network model to obtain the running time when the fuel cell runs to the failure threshold, and then determine the remaining service life of the fuel cell.
  • the long-short-term memory neural network model is used for time-series data processing, and the failure threshold and running time are accurately obtained to accurately predict the remaining service life of the fuel cell.
  • Fig. 4 is a training flowchart of a behavior-based long-short-term memory neural network model provided by the embodiment of the present application
  • Fig. 5 is a training flowchart of another behavior-based long-short-term memory neural network model provided by the embodiment of the present application , as shown in Figure 4 and Figure 5, according to time, health status value and behavior training long short-term memory neural network model, including:
  • the network parameters and weights in the long-short-term memory neural network model are set, and the weights are updated based on the gradient descent method until a training model with accurate targets is obtained, and the network model training is completed at this time.
  • the trained long-short-term memory neural network model can form a fuel cell model, and its operating conditions and working conditions are close to the actual situation of the fuel cell. Then, according to the long-short-term memory neural network model, the failure threshold of the fuel cell can be obtained, and the fuel cell can be calculated according to the long-term short-term memory neural network model to obtain the running time when the fuel cell runs to the failure threshold, and then the remaining service life of the fuel cell can be accurately determined.
  • the present application also provides a fuel cell remaining life prediction device, as shown in FIG. 6 , which is a structural schematic diagram of a fuel cell remaining life prediction device provided in an embodiment of the present application.
  • the device includes: a parameter acquisition module 601, set to obtain the parameter data of the fuel cell during the operating condition; the parameter data at least include voltage, current, cell voltage, air intake flow, air intake pressure and cooling water inlet temperature; the index fitting module 602, set To obtain the fusion health index according to the parameter data; and obtain the health state value of the fuel cell according to the fusion health index; the behavior acquisition module 603 is set to obtain the behavior of the fuel cell during the operating condition; the behavior includes at least start, stop, load change and failure emergency stop; the model training module 604 is set to train the long-term short-term memory neural network model according to time, health state value and behavior; the life prediction module 605 is set to predict the remaining life of the fuel cell according to the long-term short-term memory neural network model.
  • the parameter data of the fuel cell during the operating condition is obtained by setting the parameter acquisition module; the index fitting module obtains the fusion health index according to the parameter data, and obtains the health state value of the fuel cell according to the fusion health index; behavior acquisition The module obtains the behavior of the fuel cell during the operating conditions; the model training module trains the long-term short-term memory neural network model according to time, health status value and behavior; the life prediction module predicts the remaining life of the fuel cell according to the long-term short-term memory neural network model. Realize the accurate prediction of the remaining service life of the fuel cell based on the fusion of health indicators and behaviors under dynamic driving conditions.
  • the present application also provides a kind of computer equipment, comprises processor 10, memory 11, and realizes the input device 12 and the output device 13 that need to operate computer equipment;
  • the quantity of processor 10 in the computer equipment can be One or more, the processor 10, the memory 11, the input device 12 and the output device 13 may be connected via a bus or in other ways.
  • the memory 11 can be configured to store software programs, computer-executable programs and functional modules, such as program instructions/modules corresponding to the embodiments of the present application.
  • the processor 10 executes various functional applications and data processing of the computer equipment by running the software programs, instructions and modules stored in the memory 11 , that is, realizes the method for predicting the remaining life of the fuel cell in the above embodiment.
  • the memory 11 may include a high-speed random access memory, and may also include a non-volatile memory, such as at least one magnetic disk storage device, flash memory device, or other non-volatile solid-state storage devices.
  • memory 11 may include memory located remotely from processor 10, and these remote memories may be connected to the computer device via a network. Examples of the aforementioned networks include, but are not limited to, the Internet, intranets, local area networks, mobile communication networks, and combinations thereof.
  • the input device 12 can be configured to receive inputted numerical or character information, and generate key signal input related to user settings and function control of the computer device.
  • the output device 13 may include a display device such as a display screen.
  • the embodiment of the present application also provides a storage medium containing computer-executable instructions, and the computer-executable instructions are used to execute a method for predicting the remaining life of a fuel cell when executed by a computer processor.
  • the present application can be realized by software and necessary general hardware, or by hardware.
  • the technical solution of the present application can be embodied in the form of software products in essence, and the computer software products can be stored in computer-readable storage media, such as computer floppy disks, read-only memory (Read-Only Memory, ROM), random access Memory (Random Access Memory, RAM), flash memory (FLASH), hard disk or optical disc, etc., including a plurality of instructions to make a computer device (which can be a personal computer, server, or network device, etc.) execute the method described in the embodiment of the present application Fuel Cell Remaining Life Prediction Method.

Landscapes

  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Fuel Cell (AREA)

Abstract

一种燃料电池的剩余寿命预测方法及装置。燃料电池由多个单体组成，燃料电池的剩余寿命预测方法包括：获取燃料电池在运行工况期间的参数数据；参数数据至少包括电压、电流、单体电压、空气进气流量、空气进气压力和冷却水入口温度(S101)；根据参数数据获取融合健康指标；并根据融合健康指标获取燃料电池的健康状态值(S102)；获取燃料电池在运行工况期间内的行为；行为至少包括启动、停止、变载和故障急停(S103)；根据时间、健康状态值和行为训练长短期记忆神经网络模型；并根据长短期记忆神经网络模型预测燃料电池的剩余寿命(S104)。

Description

燃料电池的剩余寿命预测方法及装置
本申请要求在2021年07月22日提交中国专利局、申请号为202110830331.8的中国专利申请的优先权，该申请的全部内容通过引用结合在本申请中。
技术领域
本申请涉及燃料电池技术领域，例如涉及一种燃料电池的剩余寿命预测方法及装置。
背景技术
燃料电池具备温室气体排放少、能源转换效率高等优点，然而车载燃料电池系统启停频繁、工况复杂，严重影响了燃料电池系统的使用寿命。因此，有效的预测燃料电池的剩余使用寿命，对延长其使用寿命、促进燃料电池商业化发展具有重要的意义。
相关技术基于稳态实验条件下的燃料电池的耐久实验，采用数据驱动的方法对其剩余使用寿命进行预测，未考虑到实际运行工况中启停、变载、故障急停等行为对燃料电池性能衰减的影响，而燃料电池在启停过程中会产生移动的氢氧界面，多次启停后阴极会发生铂(Platinum，Pt)催化剂颗粒的流失与碳载体的腐蚀，变载与故障急停会对质子膜造成物理损伤，瞬时的操作条件变化会造成单体阻值分布不均，加剧催化剂的腐蚀，同时单一健康指标并不能完整反映燃料电池的健康状态值，燃料单体电压长时间处于不均衡状态下会出现一节或几节反极的情况，严重影响燃料电池的耐久性。
发明内容
本申请提供了一种燃料电池的剩余寿命预测方法及装置，以实现在行车工况下对燃料电池的剩余寿命的精准预测。
本申请提供了一种燃料电池的剩余寿命预测方法，所述燃料电池由多个单体组成，所述燃料电池的剩余寿命预测方法包括：
获取所述燃料电池在运行工况期间的参数数据；其中，所述参数数据至少包括电压、电流、单体电压、空气进气流量、空气进气压力和冷却水入口温度；
根据所述参数数据获取融合健康指标；并根据所述融合健康指标获取燃料电池的健康状态值；
获取所述燃料电池在所述运行工况期间内的行为；其中，所述行为至少包 括启动、停止、变载和故障急停；
根据时间、所述健康状态值和所述行为训练长短期记忆神经网络模型；并根据所述长短期记忆神经网络模型预测所述燃料电池的剩余寿命。
一实施例中，所述获取所述燃料电池在运行工况期间的参数数据，包括：
获取所述燃料电池在所述运行工况期间的电压的稳定时间区间；
获取所述稳定时间区间内的所述参数数据。
一实施例中，所述根据所述参数数据获取所述融合健康指标，包括：
根据所述参数数据获取所述燃料电池在所述运行工况期间内不同时刻的标准电压V ek；并计算每个时刻所述燃料电池的标准电压V ek与实际电压V real之间的差值ΔV＝V ek-V real；
根据所述单体电压和所述燃料电池中的单体个数获取单体电压波动率C v；
统计所述燃料电池在所述不同时刻的最低单体电压的单体数据n；
根据所述燃料电池的标准电压与实际电压之间的差值ΔV、所述单体电压波动率C v和所述最低单体电压的单体数据n建立所述融合健康指标HI＝f(ΔV,n,C v)。
一实施例中，所述根据所述参数数据获取所述燃料电池在所述运行工况期间内不同时刻的标准电压V ek，包括：
获取所述燃料电池在所述运行工况期间内不同时刻的标准电压V ek＝af(I)+bf(p)+cf(W)+df(T in)；其中，I为所述电流，p为所述空气进气压力，W为所述空气进气流量，T in为所述冷却水入口温度，a、b、c和d均为系数，f(I)、f(p)、f(W)和f(T in)分别指关于I、p、W和T in的函数。
一实施例中，所述根据所述单体电压和所述燃料电池中的单体个数获取单体电压波动率C v，包括：
获取所述燃料电池中的单体电压波动率
Figure PCTCN2022104187-appb-000001
其中，V i为所述燃料电池的第i个单体的单体电压，
Figure PCTCN2022104187-appb-000002
为所述燃料电池中所有单体的单体电压均值，N为所述燃料电池中的单体个数，1≤i≤N。
一实施例中，所述根据所述融合健康指标获取所述燃料电池的健康状态值，包括：
根据所述融合健康指标获取所述燃料电池的初始时刻的健康指标值HI 0；
根据所述融合健康指标获取所述燃料电池从初始时刻开始运行至t时刻的健康指标值HI t；
获取所述燃料电池的健康状态值ΔHI＝HI 0-HI t。
一实施例中，在所述获取所述燃料电池在所述运行工况期间内的行为之后，还包括：
将所述燃料电池在所述运行工况期间的每个行为采用数据化的方式进行标记；其中，所述行为被标记的数据可根据所述行为对所述燃料电池的电压衰减的影响等级进行设定；不同数据表示不同的影响等级。
一实施例中，所述根据时间、所述健康状态值和所述行为训练长短期记忆神经网络模型，包括：
将所述健康状态值的时间序列以及不同时刻的行为输入至所述长短期记忆神经网络模型；
对所述长短期记忆神经网络模型进行训练使训练后的长短期记忆神经网络模型模拟所述燃料电池的运行工况。
一实施例中，所述根据所述长短期记忆神经网络模型预测所述燃料电池的剩余寿命，包括：
根据所述长短期记忆神经网络模型获取所述燃料电池的失效阈值；
根据所述长短期记忆神经网络模型确定运行至所述失效阈值时所述燃料电池的运行时长，并根据所述运行时长确定所述燃料电池的剩余寿命。
本申请提供了一种燃料电池的剩余寿命预测装置，包括：
参数获取模块，设置为获取燃料电池在所述运行工况期间的参数数据；其中，所述参数数据至少包括电压、电流、单体电压、空气进气流量、空气进气压力和冷却水入口温度；
指标拟合模块，设置为根据所述参数数据获取融合健康指标；并根据所述融合健康指标获取燃料电池的健康状态值；
行为获取模块，设置为获取所述燃料电池在所述运行工况期间内的行为；其中，所述行为至少包括启动、停止、变载和故障急停；
模型训练模块，设置为根据时间、所述健康状态值和所述行为训练长短期记忆神经网络模型；
寿命预测模块，设置为根据所述长短期记忆神经网络模型预测所述燃料电池的剩余寿命。
本申请提供了一种计算机设备，包括：存储器和处理器，所述存储器存储有计算机程序，所述处理器设置为执行所述计算机程序时实现上述的燃料电池 的剩余寿命预测方法。
本申请提供了一种计算机可读存储介质，存储有至少一个程序指令，所述至少一个程序指令用于执行上述的燃料电池的剩余寿命预测方法。
附图说明
图1为本申请实施例提供的一种燃料电池的剩余寿命预测方法流程图；
图2为本申请实施例提供的另一种燃料电池的剩余寿命预测方法流程图；
图3为本申请实施例提供的又一种燃料电池的剩余寿命预测方法流程图；
图4为本申请实施例提供的一种基于行为的长短期记忆神经网络模型的训练流程图；
图5为本申请实施例提供的另一种基于行为的长短期记忆神经网络模型的训练流程图；
图6为本申请实施例提供的一种燃料电池的剩余寿命预测装置的结构示意图；
图7为本申请实施例提供的一种计算机设备的结构示意图。
具体实施方式
以下将结合本申请实施例中的附图，通过具体实施方式，描述本申请的技术方案。所描述的实施例是本申请的一部分实施例，
图1为本申请实施例提供的第一种燃料电池的剩余寿命预测方法流程图，适用于燃料电池的剩余寿命预测。燃料电池由多个单体组成，燃料电池的剩余寿命预测方法包括：
S101、获取燃料电池在运行工况期间的参数数据；参数数据至少包括电压、电流、单体电压、空气进气流量、空气进气压力和冷却水入口温度。
电压指燃料电池当前时刻的实际电压值，燃料电池工作时相当于一直流源，其阳极即为电源负极，阴极即为电源正极，阳极和阴极之间的电压即为本实施中的电压，根据电压值的大小可以确定燃料电池能否继续使用。
电流指电堆电流，燃料电池由多个单体组成，单体即为一个子电池单元，将多个单体层叠就能构成输出电压满足实际负载需要的燃料电池堆，简称电堆。
单体电压指燃料电池中单体当前时刻的电压值，燃料电池单体电压分布的差异性是客观存在的，当燃料电池单体电压长时间处于不均衡状态下会出现一节或几节单体出现反极，将严重影响燃料电池的耐久性。
空气进气流量是指燃料气体从燃料电池的阳极或阴极进入燃料电池的气体质量流量，例如在燃料电池阳极，氢气从进口流入燃料电池，在燃料电池的阴极，氧气从进口流入燃料电池。
空气进气压力是指燃料气体从燃料电池的阳极或阴极进入燃料电池的气体气压。例如质子交换膜电池工作时，在质子交换膜两侧的反应气体气压要保持相对的平衡，一方面能够将燃料进入质子交换膜的扩散控制在最低限度，另一方面起到保护质子交换膜的作用。
冷却水入口温度是指燃料电池冷却装置入口处的冷却水的温度，燃料电池在产生电能的过程中也会产生大量热能，因此需要冷却装置吸收燃料电池的热量，例如可以是冷却水泵冷却水入口温度。
当燃料电池在运行工况期间，可以获取到电压、电流、单体电压、空气进气流量、空气进气压力和冷却水入口温度等参数数据。
上述参数数据均是在燃料电池的运行工况期间测得的，运行工况为燃料电池向外输出功率的情况，示例性的，对于以燃料电池为动力来源的车辆，当车辆启动后或行驶中，即为燃料电池的运行工况期间。
S102、根据参数数据获取融合健康指标；并根据融合健康指标获取燃料电池的健康状态值。
融合健康指标指非单一健康指标，其为由多种不同数据建立的综合健康指标。例如由燃料电池电压、电流、单体电压、空气进气流量、空气进气压力和冷却水入口温度等数据综合建立的健康指标，单一健康指标不能完整反应燃料电池的健康状态。
健康状态值指燃料电池根据融合健康指标计算得出的值，不同的健康状态值表征燃料电池的性能不同，例如燃料电池的健康状态值越低，说明燃料电池的性能越差，燃料电池的健康状态值越高，说明燃料电池的性能越好。
根据获取到的电压、电流、单体电压、空气进气流量、空气进气压力和冷却水入口温度等参数数据建立融合健康指标，获取燃料电池的参数数据的值，并根据融合健康指标可以得到其对应的燃料电池的健康状态值。
S103、获取燃料电池在运行工况期间内的行为；行为至少包括启动、停止、变载和故障急停。
燃料电池在实际工作中，会存在频繁的启动、停止、变载或故障急停等行为，燃料电池在运行工况期间内的行为对燃料电池衰减有不同程度的影响，进而影响燃料电池的使用寿命。
S104、根据时间、健康状态值和行为训练长短期记忆神经网络模型；并根据长短期记忆神经网络模型预测燃料电池的剩余寿命。
时间指燃料电池运行的时间。
长短记忆神经网络模型指一种深度学习方法，用于处理序列数据。
采用时间、健康状态值和行为作为输入对长短记忆神经网络模型进行训练，得到目标精确的训练模型，根据长短期记忆神经网络模型预测燃料电池的剩余寿命时间。
本申请实施例所提供的技术方案，通过获取燃料电池在运行工况期间电压、电流、单体电压、空气进气流量、空气进气压力和冷却水入口温度等参数数据，根据参数数据得到融合健康指标，进而得到燃料电池的健康状态值，获取燃料电池在运行工况期间内启动、停止、变载和故障急停等行为，根据时间、健康状态值和行为训练长短期记忆神经网络模型，并根据长短期记忆神经网络模型预测燃料电池的剩余寿命。实现在行车动态工况下，通过融合健康指标和行为对燃料电池的剩余寿命的精准预测。
一实施例中，获取燃料电池在运行工况期间的参数数据，可以包括：获取燃料电池在运行工况期间的电压的稳定时间区间；获取稳定时间区间内的参数数据。
稳定时间区间指车辆启动一段时间后的时间区间。由于车辆刚刚启动时燃料电池输出的电压是不稳定的，获取燃料电池输出的电压的稳定时间区间内对应的参数数据，保证数据的可靠性和有效性。
图2为本申请实施例提供的另一种燃料电池的剩余寿命预测方法流程图，如图2所示，在上述实施例的基础上，根据参数数据获取融合健康指标主要包括：
S201、获取燃料电池在运行工况期间的参数数据；参数数据至少包括电压、电流、单体电压、空气进气流量、空气进气压力和冷却水入口温度。
S202、根据参数数据获取燃料电池在运行工况期间内不同时刻的标准电压V ek；并计算每个时刻燃料电池的标准电压V ek与实际电压V real之间的差值ΔV＝V ek-V real。
标准电压是根据燃料电池在稳定时间区间内的参数数据计算得到的，此时燃料电池具有稳定的输出电压且状态良好。
根据燃料电池在稳定时间区间内的参数数据计算得到的不同时刻的标准电 压V ek，获取不同时刻燃料电池的实际电压V real，并根据每个时刻的标准电压V ek和实际电压V real得出差值ΔV，即ΔV＝V ek-V real，差值ΔV可能是正值，也可以是负值或零。
一实施例中，根据参数数据获取燃料电池在运行工况期间内不同时刻的标准电压，包括：获取燃料电池在运行工况期间内不同时刻的标准电压V ek＝af(I)+bf(p)+cf(W)+df(T in)；其中，I为电流，p为空气进气压力，W为空气进气流量，T in为冷却水入口温度。
a、b、c和d为系数，f(I)、f(p)、f(W)和f(T in)分别指关于电流I、空气进气压力p、空气进气流量W和冷却水入口温度T in的函数。
根据电流I、空气进气压力p、空气进气流量W和冷却水入口温度T in可以得到燃料电池在不同时刻的标准电压V ek，当其中任何一个变量值发生变化时，都将影响标准电压V ek的值，影响ΔV的值，最终导致融合健康指标HI的值发生变化，将不同的参数数据均考虑在内，保证对燃料电池剩余使用寿命的精确预测。
S203、根据单体电压和燃料电池中的单体个数获取单体电压波动率C v。
单体个数为燃料电池中单个电池的组成个数，根据燃料电池的单体电压和单体个数计算得出当前时刻的单体电压波动率C v，由单体电压波动率C v可知燃料电池单体电压的均衡性，单体电压的均衡性越好则燃料电池的耐久性越好。
一实施例中，根据单体电压和燃料电池中的单体个数获取单体电压波动率C v，包括：获取燃料电池中的单体电压波动率
Figure PCTCN2022104187-appb-000003
其中，V i为燃料电池的第i个单体的单体电压，
Figure PCTCN2022104187-appb-000004
为燃料电池中所有单体的单体电压均值，N为燃料电池中的单体个数。
单体电压波动率C v用于评价燃料电池单体电压均衡性，例如单体电压波动率C v越大，则说明燃料电池的单体电压均衡性越差。
获取燃料电池的第i个单体的单体电压V i和燃料电池中的单体个数N，进而计算得出燃料电池中所有单体的单体电压均值
Figure PCTCN2022104187-appb-000005
并根据公式
Figure PCTCN2022104187-appb-000006
得到燃料电池中的单体电压波动率C v，其中1≤i≤N。将单体电压波动率C v作为融合健康指标HI的输入，将单体电压均衡性考虑在内，保证对燃料电池剩余使用寿命的预测更加精确。
S204、统计燃料电池在不同时刻的最低单体电压的单体数据n。
根据燃料电池在不同时刻的不同单体的单体电压可统计得出不同时刻的最 低单体电压的单体数据n。例如可以设定最低单体电压值，获取每个单体在不同时刻的单体电压，并与最低单体电压值进行比较，若当前时刻存在n个单体的单体电压小于或者等于最低单体电压值，则最低单体电压的单体数据为n。
S205、根据燃料电池的标准电压与实际电压之间的差值ΔV、单体电压波动率C v和最低单体电压的单体数据n建立融合健康指标HI＝f(ΔV,n,C v)。
融合健康指标用HI表示，f(ΔV,n,C v)指由差值ΔV、单体电压波动率C v和最低单体电压的单体数据n构成的函数关系式。
根据差值ΔV、单体电压波动率C v和最低单体电压的单体数据n建立融合健康指标HI，换言之，融合健康指标HI是关于差值ΔV、单体电压波动率C v和最低单体电压的单体数据n的函数，任何一个变量发生变化都将影响融合健康指标HI的值。
上述S202～S205即为“根据参数数据获取融合健康指标”的过程，如此将多个健康指标综合考虑，实现准确评价燃料电池当前的健康状态。
S206、根据融合健康指标获取燃料电池的健康状态值。
S207、获取燃料电池在运行工况期间内的行为；行为至少包括启动、停止、变载和故障急停。
S208、根据时间、健康状态值和行为训练长短期记忆神经网络模型；并根据长短期记忆神经网络模型预测燃料电池的剩余寿命。
本实施例对根据参数数据获取融合健康指标的过程进行详述，通过燃料电池的标准电压与实际电压之间的差值ΔV、单体电压波动率C v和最低单体电压的单体数据n建立所述融合健康指标，融合电流I、空气进气压力p、空气进气流量W和冷却水入口温度T in等多个参数数据，可利于完整体现燃料电池的健康状态。
图3为本申请实施例提供的又一种燃料电池的剩余寿命预测方法流程图，如图3所示，在上述实施例的基础上，本实施例对根据融合健康指标获取燃料电池的健康状态值进行说明，燃料电池的剩余寿命预测方法包括：
S301、获取燃料电池在运行工况期间的参数数据；参数数据至少包括电压、电流、单体电压、空气进气流量、空气进气压力和冷却水入口温度。
S302、根据参数数据获取燃料电池在运行工况期间内不同时刻的标准电压V ek；并计算每个时刻燃料电池的标准电压V ek与实际电压V real之间的差值ΔV＝V ek-V real。
S303、根据单体电压和燃料电池中的单体个数获取单体电压波动率C v。
S304、统计燃料电池不同时刻的最低单体电压的单体数据n。
S305、根据燃料电池的标准电压与实际电压之间的差值ΔV、单体电压波动率C v和最低单体电压的单体数据n建立融合健康指标HI＝f(ΔV,n,C v)。
S306、获取燃料电池的初始时刻的健康指标值HI 0。
初始时刻的健康指标值HI 0指燃料电池出厂设置时的健康指标值。
获取燃料电池初始时刻t 0时电压、电流、单体电压、空气进气流量、空气进气压力和冷却水入口温度等参数数据，根据上述步骤依次计算得到当前时刻单体电压波动率C v、差值ΔV和最低单体电压的单体数据n，进而根据融合健康指标HI＝f(ΔV,n,C v)得到燃料电池的初始时刻的健康指标值HI 0。
S307、获取燃料电池从初始时刻开始运行至t时刻的健康指标值HI t。
t时刻是燃料电池自初始时刻t 0开始，投入使用一定时间段的时刻，例如，车辆使用后半年或一年，获取燃料电池t时刻时电压、电流、单体电压、空气进气流量、空气进气压力和冷却水入口温度等参数数据，根据上述步骤依次计算得到当前时刻单体电压波动率C v、差值ΔV和最低单体电压的单体数据n，进而根据融合健康指标HI＝f(ΔV,n,C v)得到燃料电池的t时刻的健康指标值HI t。
S308、获取燃料电池的健康状态值ΔHI＝HI 0-HI t。
根据S306和S307得到的燃料电池的初始时刻的健康指标值HI 0和t时刻的健康指标值HI t可以计算得出燃料电池的健康状态值，即ΔHI＝HI 0-HI t。
上述S306～S308即为“根据融合健康指标获取燃料电池的健康状态值”的过程。如此，根据燃料电池的健康状态值实现过程可以得到任何时间段内的燃料电池的健康状态值，精确预测燃料电池当前的健康状态，实现在行车工况下对燃料电池的实时监测。
S309、获取燃料电池在运行工况期间内的行为；行为至少包括启动、停止、变载和故障急停。
继续参考图3所示，一实施例中，获取燃料电池在运行工况期间内的行为之后，还包括：
S310、将燃料电池在运行工况期间的每个行为采用数据化的方式进行标记；其中，行为被标记的数据可根据行为对燃料电池的电压衰减的影响等级进行设定；不同数据表示不同的影响等级。
数据化指一种量化的模式，例如燃料电池在运行工况期间无行为则用二进制数字0000表示，启动用二进制数字0001表示，停止用二进制数字0010表示， 变载用二进制数字0011表示，故障急停用二进制数字0100表示，数据化的标记方式，本申请实施例不进行限定。
燃料电池的电压衰减指燃料电池的输出电压减小，例如可以是相同功率下电压的下降、单体的不均衡性或最低单体电压数的增加等，都将影响到燃料电池的电压衰减。
将燃料电池在运行工况期间的每个行为通过数据化的方式进行标记，被标记的数据可根据行为对燃料电池的电压衰减的影响等级进行设定，例如被标记的数据值越大，则代表行为对燃料电池的电压衰减的影响越大。如此可根据被标记的数据的大小判断行为对燃料电池的电压衰减的影响，便于燃料电池系统根据每个行为被标记的数据进行分析处理，进而建立更符合实际情况的训练模型，提高训练模型的目标精确度。
S311、根据时间、健康状态值和行为训练长短期记忆神经网络模型；并根据长短期记忆神经网络模型预测燃料电池的剩余寿命。
一实施例中，根据长短期记忆神经网络模型预测燃料电池的剩余寿命，包括：根据长短期记忆神经网络模型获取燃料电池的失效阈值；根据长短期记忆神经网络模型确定运行至失效阈值时燃料电池的运行时长，并根据运行时长确定燃料电池的剩余寿命。
失效阈值指燃料电池在运行工况期间最低输出电压值，例如可以是燃料电池的出厂设定值，例如可以为燃料电池额定输出电压的20％。当燃料电池在运行工况期间输出的电压达到失效阈值时，燃料电池将无法继续使用。
运行时长是指燃料电池从出厂后开始启动运行的时刻，运行至燃料电池输出的电压达到失效阈值的时刻所经历的时间。
根据长短期记忆神经网络模型可获取燃料电池的失效阈值，并可根据长短期记忆神经网络模型计算得出燃料电池运行至失效阈值时的运行时长，进而确定燃料电池的剩余使用寿命。采用长短期记忆神经网络模型进行时间序列数据处理，精确获取失效阈值和运行时长，达到对燃料电池剩余使用寿命的精确预测。
图4为本申请实施例提供的一种基于行为的长短期记忆神经网络模型的训练流程图，图5为本申请实施例提供的另一种基于行为的长短期记忆神经网络模型的训练流程图，结合如图4和图5所示，根据时间、健康状态值和行为训练长短期记忆神经网络模型，包括：
S401、将健康状态值的时间序列以及不同时刻的行为输入至长短期记忆神经网络模型。
获取融合健康指标HI计算得出燃料电池的健康状态值以及燃料电池在运行工况期间的不同时刻的多个行为，将两者作为长短期记忆神经网络模型的输入，对长短期记忆神经网络模型进行训练。
S402、对长短期记忆神经网络模型进行训练使其模拟燃料电池的运行工况。
如图5所示，对长短期记忆神经网络模型中的网络参数和权重进行设置，并基于梯度下降法更新权重，直至获得目标精确的训练模型，此时完成网络模型训练。
经过训练后的长短期记忆神经网络模型可以形成燃料电池的模型，其运行工况和工作状态逼近燃料电池实际情况。则根据长短期记忆神经网络模型可获取燃料电池的失效阈值，并可根据长短期记忆神经网络模型计算得出燃料电池运行至失效阈值时的运行时长，进而精准确定燃料电池的剩余使用寿命。
本申请还提供了一种燃料电池的剩余寿命预测装置，如图6所示，图6为本申请实施例提供的一种燃料电池的剩余寿命预测装置的结构示意图，该装置包括：参数获取模块601，设置为获取燃料电池在运行工况期间的参数数据；参数数据至少包括电压、电流、单体电压、空气进气流量、空气进气压力和冷却水入口温度；指标拟合模块602，设置为根据参数数据获取融合健康指标；并根据融合健康指标获取燃料电池的健康状态值；行为获取模块603，设置为获取燃料电池在运行工况期间内的行为；行为至少包括启动、停止、变载和故障急停；模型训练模块604，设置为根据时间、健康状态值和行为训练长短期记忆神经网络模型；寿命预测模块605，设置为根据长短期记忆神经网络模型预测燃料电池的剩余寿命。
本申请实施例中，通过设置参数获取模块获取燃料电池在运行工况期间的参数数据；指标拟合模块根据参数数据获取融合健康指标，并根据融合健康指标获取燃料电池的健康状态值；行为获取模块获取燃料电池在运行工况期间内的行为；模型训练模块根据时间、健康状态值和行为训练长短期记忆神经网络模型；寿命预测模块根据长短期记忆神经网络模型预测燃料电池的剩余寿命。实现在行车动态工况下，基于融合健康指标和行为对燃料电池剩余使用寿命的精准预测。
如图7所示，本申请还提供了一种计算机设备，包括处理器10、存储器11，以及实现操作计算机设备所需的输入装置12和输出装置13；计算机设备中处理器10的数量可以是一个或多个，处理器10、存储器11、输入装置12和输出装 置13可以通过总线或其他方式连接。
存储器11作为一种计算机可读存储介质，可设置为存储软件程序、计算机可执行程序以及功能模块，如与本申请实施例对应的程序指令/模块。处理器10通过运行存储在存储器11中的软件程序、指令以及模块，从而执行计算机设备的多种功能应用以及数据处理，即实现上述实施例的燃料电池的剩余寿命预测方法。
存储器11可以包括高速随机存取存储器，还可以包括非易失性存储器，例如至少一个磁盘存储器件、闪存器件、或其他非易失性固态存储器件。在一些实例中，存储器11可包括相对于处理器10远程设置的存储器，这些远程存储器可以通过网络连接至计算机设备。上述网络的实例包括但不限于互联网、企业内部网、局域网、移动通信网及其组合。
输入装置12可设置为接收输入的数字或字符信息，以及产生与计算机设备的用户设置以及功能控制有关的键信号输入。输出装置13可包括显示屏等显示设备。
本申请实施例还提供一种包含计算机可执行指令的存储介质，计算机可执行指令在由计算机处理器执行时用于执行燃料电池的剩余寿命预测方法。
通过以上关于实施方式的描述可知，本申请可借助软件及必需的通用硬件来实现，也可以通过硬件实现。本申请的技术方案本质上可以以软件产品的形式体现出来，该计算机软件产品可以存储在计算机可读存储介质中，如计算机的软盘、只读存储器(Read-Only Memory，ROM)、随机存取存储器(RandomAccess Memory，RAM)、闪存(FLASH)、硬盘或光盘等，包括多个指令用以使得一台计算机设备(可以是个人计算机，服务器，或者网络设备等)执行本申请实施例所述的燃料电池的剩余寿命预测方法。

Claims (12)

  1. 一种燃料电池的剩余寿命预测方法，所述燃料电池由多个单体组成，所述燃料电池的剩余寿命预测方法包括：
    获取所述燃料电池在运行工况期间的参数数据；其中，所述参数数据至少包括电压、电流、单体电压、空气进气流量、空气进气压力和冷却水入口温度；
    根据所述参数数据获取融合健康指标；并根据所述融合健康指标获取所述燃料电池的健康状态值；
    获取所述燃料电池在所述运行工况期间内的行为；其中，所述行为至少包括启动、停止、变载和故障急停；
    根据时间、所述健康状态值和所述行为训练长短期记忆神经网络模型；并根据所述长短期记忆神经网络模型预测所述燃料电池的剩余寿命。
  2. 根据权利要求1所述的燃料电池的剩余寿命预测方法，其中，所述获取所述燃料电池在运行工况期间的参数数据，包括：
    获取所述燃料电池在所述运行工况期间的电压的稳定时间区间；
    获取所述稳定时间区间内的所述参数数据。
  3. 根据权利要求1所述的燃料电池的剩余寿命预测方法，其中，所述根据所述参数数据获取所述融合健康指标，包括：
    根据所述参数数据获取所述燃料电池在所述运行工况期间内不同时刻的标准电压V ek；并计算每个时刻所述燃料电池的标准电压V ek与实际电压V real之间的差值ΔV＝V ek-V real；
    根据所述单体电压和所述燃料电池中的单体个数获取单体电压波动率C v；
    统计所述燃料电池在所述不同时刻的最低单体电压的单体数据n；
    根据所述燃料电池的标准电压与实际电压之间的差值ΔV、所述单体电压波动率C v和所述最低单体电压的单体数据n建立所述融合健康指标HI＝f(ΔV,n,C v)。
  4. 根据权利要求3所述的燃料电池的剩余寿命预测方法，其中，所述根据所述参数数据获取所述燃料电池在所述运行工况期间内不同时刻的标准电压V ek，包括：
    获取所述燃料电池在所述运行工况期间内不同时刻的标准电压V ek＝af(I)+bf(p)+cf(W)+df(T in)；其中，I为所述电流，p为所述空气进气压力，W为所述空气进气流量，T in为所述冷却水入口温度，a、b、c和d均为系数，f(I)、f(p)、f(W)和f(T in)分别指关于I、p、W和T in的函数。
  5. 根据权利要求3所述的燃料电池的剩余寿命预测方法，其中，所述根据 所述单体电压和所述燃料电池中的单体个数获取单体电压波动率C v，包括：
    获取所述燃料电池中的单体电压波动率
    Figure PCTCN2022104187-appb-100001
    其中，V i为所述燃料电池的第i个单体的单体电压，
    Figure PCTCN2022104187-appb-100002
    为所述燃料电池中所有单体的单体电压均值，N为所述燃料电池中的单体个数，1≤i≤N。
  6. 根据权利要求1所述的燃料电池的剩余寿命预测方法，其中，所述根据所述融合健康指标获取所述燃料电池的健康状态值，包括：
    根据所述融合健康指标获取所述燃料电池的初始时刻的健康指标值HI 0；
    根据所述融合健康指标获取所述燃料电池从所述初始时刻开始运行至t时刻的健康指标值HI t；
    获取所述燃料电池的健康状态值ΔHI＝HI 0-HI t。
  7. 根据权利要求1所述的燃料电池的剩余寿命预测方法，在所述获取所述燃料电池在所述运行工况期间内的行为之后，还包括：
    将所述燃料电池在所述运行工况期间的每个行为采用数据化的方式进行标记；其中，所述行为被标记的数据可根据所述行为对所述燃料电池的电压衰减的影响等级进行设定；不同数据表示不同的影响等级。
  8. 根据权利要求1所述的燃料电池的剩余寿命预测方法，其中，所述根据时间、所述健康状态值和所述行为训练长短期记忆神经网络模型，包括：
    将所述健康状态值的时间序列以及不同时刻的行为输入至所述长短期记忆神经网络模型；
    对所述长短期记忆神经网络模型进行训练使训练后的长短期记忆神经网络模型模拟所述燃料电池的运行工况。
  9. 根据权利要求1所述的燃料电池的剩余寿命预测方法，其中，所述根据所述长短期记忆神经网络模型预测所述燃料电池的剩余寿命，包括：
    根据所述长短期记忆神经网络模型获取所述燃料电池的失效阈值；
    根据所述长短期记忆神经网络模型确定运行至所述失效阈值时所述燃料电池的运行时长，并根据所述运行时长确定所述燃料电池的剩余寿命。
  10. 一种燃料电池的剩余寿命预测装置，包括：
    参数获取模块，设置为获取燃料电池在运行工况期间的参数数据；其中，所述参数数据至少包括电压、电流、单体电压、空气进气流量、空气进气压力和冷却水入口温度；
    指标拟合模块，设置为根据所述参数数据获取融合健康指标；并根据所述融合健康指标获取燃料电池的健康状态值；
    行为获取模块，设置为获取所述燃料电池在所述运行工况期间内的行为；其中，所述行为至少包括启动、停止、变载和故障急停；
    模型训练模块，设置为根据时间、所述健康状态值和所述行为训练长短期记忆神经网络模型；
    寿命预测模块，设置为根据所述长短期记忆神经网络模型预测所述燃料电池的剩余寿命。
  11. 一种计算机设备，包括：存储器和处理器，所述存储器存储有计算机程序，所述处理器设置为执行所述计算机程序时实现如权利要求1-9中任一项所述的燃料电池的剩余寿命预测方法。
  12. 一种计算机可读存储介质，存储有至少一个程序指令，所述至少一个程序指令用于执行如权利要求1-9中任一项所述的燃料电池的剩余寿命预测方法。
PCT/CN2022/104187 2021-07-22 2022-07-06 燃料电池的剩余寿命预测方法及装置 WO2023000985A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202110830331.8A CN113406505A (zh) 2021-07-22 2021-07-22 一种燃料电池的剩余寿命预测方法及装置
CN202110830331.8 2021-07-22

Publications (1)

Publication Number Publication Date
WO2023000985A1 true WO2023000985A1 (zh) 2023-01-26

Family

ID=77687316

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2022/104187 WO2023000985A1 (zh) 2021-07-22 2022-07-06 燃料电池的剩余寿命预测方法及装置

Country Status (2)

Country Link
CN (1) CN113406505A (zh)
WO (1) WO2023000985A1 (zh)

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116068413A (zh) * 2023-03-22 2023-05-05 长安新能源南京研究院有限公司 一种电池电压预测方法、装置、设备及存储介质
CN116502544A (zh) * 2023-06-26 2023-07-28 武汉新威奇科技有限公司 一种基于数据融合的电动螺旋压力机寿命预测方法及系统
CN116995276A (zh) * 2023-09-27 2023-11-03 爱德曼氢能源装备有限公司 燃料电池发电系统冷却方法及系统
CN117350174A (zh) * 2023-12-04 2024-01-05 国网天津市电力公司营销服务中心 预测智能电表剩余寿命的方法、系统、电子设备及介质
CN117491793A (zh) * 2023-12-29 2024-02-02 国网浙江省电力有限公司电力科学研究院 一种氢电耦合系统综合性能测试方法、装置及介质

Families Citing this family (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113406505A (zh) * 2021-07-22 2021-09-17 中国第一汽车股份有限公司 一种燃料电池的剩余寿命预测方法及装置
CN113871661B (zh) * 2021-09-23 2023-05-12 中国第一汽车股份有限公司 一种燃料电池的控制方法及控制装置
CN114142069A (zh) * 2021-11-26 2022-03-04 广东电网有限责任公司广州供电局 一种基于燃料电池健康状态评估的在线监控装置及方法
CN114759227A (zh) * 2022-05-07 2022-07-15 中国第一汽车股份有限公司 燃料电池性能衰减的确定方法以及确定装置
CN115312803A (zh) * 2022-08-31 2022-11-08 佛山仙湖实验室 一种燃料电池系统及其再利用寿命评估方法
CN115453369B (zh) * 2022-09-20 2023-05-05 中国汽车工程研究院股份有限公司 一种燃料电池一致性预测及故障诊断的方法

Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103698709A (zh) * 2013-12-11 2014-04-02 清华大学 一种燃料电池剩余寿命预测方法
EP3503274A1 (fr) * 2017-12-11 2019-06-26 Commissariat à l'Energie Atomique et aux Energies Alternatives Methode pour estimer l'etat de sante d'une pile a combustible a partir de mesures en usage reel
CN110190306A (zh) * 2019-06-04 2019-08-30 昆山知氢信息科技有限公司 一种用于燃料电池系统的在线故障诊断方法
CN110929451A (zh) * 2019-10-24 2020-03-27 同济大学 一种燃料电池单体电压一致性预测方法
CN111103544A (zh) * 2019-12-26 2020-05-05 江苏大学 基于长短时记忆lstm和粒子滤波pf的锂离子电池剩余使用寿命预测方法
CN111880100A (zh) * 2020-08-07 2020-11-03 同济大学 基于自适应扩展卡尔曼滤波的燃料电池剩余寿命预测方法
CN112784216A (zh) * 2021-01-26 2021-05-11 武汉船用电力推进装置研究所（中国船舶重工集团公司第七一二研究所） 一种质子交换膜燃料电池系统的健康度评估方法及系统
CN113406505A (zh) * 2021-07-22 2021-09-17 中国第一汽车股份有限公司 一种燃料电池的剩余寿命预测方法及装置

Family Cites Families (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108417868B (zh) * 2018-03-05 2020-04-14 中国第一汽车股份有限公司 一种车用燃料电池使用寿命加速测试与分析方法
CN109991542B (zh) * 2019-03-27 2021-05-18 东北大学 基于wde优化lstm网络的锂离子电池剩余寿命预测方法
CN110059377B (zh) * 2019-04-02 2022-07-05 西南交通大学 一种基于深度卷积神经网络的燃料电池寿命预测方法
KR102216850B1 (ko) * 2019-12-05 2021-02-18 (주)프라즈마 사이언스 인공지능 기반의 연료 전지 관리 시스템
CN111146478B (zh) * 2019-12-22 2021-02-02 同济大学 一种用于质子交换膜燃料电池堆剩余使用寿命的预测方法
CN111339712B (zh) * 2020-02-24 2023-05-30 电子科技大学 质子交换膜燃料电池剩余寿命预测方法
CN111443294B (zh) * 2020-04-10 2022-09-23 华东理工大学 一种锂离子电池剩余寿命间接预测方法及装置
CN112540317B (zh) * 2020-12-16 2022-12-02 武汉理工大学 基于实车数据的电池健康状态估计与剩余寿命预测方法
CN112986827B (zh) * 2021-04-12 2022-06-03 山东凯格瑞森能源科技有限公司 一种基于深度学习的燃料电池剩余寿命预测方法
CN112926273B (zh) * 2021-04-13 2023-04-18 中国人民解放军火箭军工程大学 一种多元退化设备剩余寿命预测方法

Patent Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103698709A (zh) * 2013-12-11 2014-04-02 清华大学 一种燃料电池剩余寿命预测方法
EP3503274A1 (fr) * 2017-12-11 2019-06-26 Commissariat à l'Energie Atomique et aux Energies Alternatives Methode pour estimer l'etat de sante d'une pile a combustible a partir de mesures en usage reel
CN110190306A (zh) * 2019-06-04 2019-08-30 昆山知氢信息科技有限公司 一种用于燃料电池系统的在线故障诊断方法
CN110929451A (zh) * 2019-10-24 2020-03-27 同济大学 一种燃料电池单体电压一致性预测方法
CN111103544A (zh) * 2019-12-26 2020-05-05 江苏大学 基于长短时记忆lstm和粒子滤波pf的锂离子电池剩余使用寿命预测方法
CN111880100A (zh) * 2020-08-07 2020-11-03 同济大学 基于自适应扩展卡尔曼滤波的燃料电池剩余寿命预测方法
CN112784216A (zh) * 2021-01-26 2021-05-11 武汉船用电力推进装置研究所（中国船舶重工集团公司第七一二研究所） 一种质子交换膜燃料电池系统的健康度评估方法及系统
CN113406505A (zh) * 2021-07-22 2021-09-17 中国第一汽车股份有限公司 一种燃料电池的剩余寿命预测方法及装置

Cited By (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116068413A (zh) * 2023-03-22 2023-05-05 长安新能源南京研究院有限公司 一种电池电压预测方法、装置、设备及存储介质
CN116068413B (zh) * 2023-03-22 2023-08-18 长安新能源南京研究院有限公司 一种电池电压预测方法、装置、设备及存储介质
CN116502544A (zh) * 2023-06-26 2023-07-28 武汉新威奇科技有限公司 一种基于数据融合的电动螺旋压力机寿命预测方法及系统
CN116502544B (zh) * 2023-06-26 2023-09-12 武汉新威奇科技有限公司 一种基于数据融合的电动螺旋压力机寿命预测方法及系统
CN116995276A (zh) * 2023-09-27 2023-11-03 爱德曼氢能源装备有限公司 燃料电池发电系统冷却方法及系统
CN116995276B (zh) * 2023-09-27 2023-12-29 爱德曼氢能源装备有限公司 燃料电池发电系统冷却方法及系统
CN117350174A (zh) * 2023-12-04 2024-01-05 国网天津市电力公司营销服务中心 预测智能电表剩余寿命的方法、系统、电子设备及介质
CN117350174B (zh) * 2023-12-04 2024-04-02 国网天津市电力公司营销服务中心 预测智能电表剩余寿命的方法、系统、电子设备及介质
CN117491793A (zh) * 2023-12-29 2024-02-02 国网浙江省电力有限公司电力科学研究院 一种氢电耦合系统综合性能测试方法、装置及介质
CN117491793B (zh) * 2023-12-29 2024-05-10 国网浙江省电力有限公司电力科学研究院 一种氢电耦合系统综合性能测试方法、装置及介质

Also Published As

Publication number Publication date
CN113406505A (zh) 2021-09-17

Similar Documents

Publication Publication Date Title
WO2023000985A1 (zh) 燃料电池的剩余寿命预测方法及装置
CN107180983B (zh) 一种固体氧化物燃料电池电堆故障诊断方法和系统
CN108761341A (zh) 一种锂离子电池电化学模型参数获取方法
CN111146478A (zh) 一种用于质子交换膜燃料电池堆剩余使用寿命的预测方法
CN108417868A (zh) 一种车用燃料电池使用寿命加速测试与分析方法
CN106532688A (zh) 一种用于评估微电网运行可靠性的方法及系统
CN105203968B (zh) 一种铅酸蓄电池剩余电量的在线测量系统
CN104316879A (zh) 一种铅酸蓄电池组寿命的预测方法
CN115186028A (zh) 碳强度的展示方法、装置、设备、存储介质及程序产品
CN112072643A (zh) 一种基于深度确定性梯度策略的光-蓄系统在线调度方法
CN105487017B (zh) 一种变电站ups用阀控式密封铅酸蓄电池状态估计与预测方法
CN116381541A (zh) 一种储能锂电池系统的健康评估方法及系统
Chen et al. State of health estimation of lithium-ion batteries based on equivalent circuit model and data-driven method
Wang et al. Fuel cell fault forecasting system using grey and extension theories
CN111563826A (zh) 一种基于电动汽车用电行为的电池信息预测系统及方法
CN116184246A (zh) 一种全钒液流电池的soc与soh联合估计方法、装置及终端设备
CN113064075B (zh) 一种基于边缘计算及深度学习的电机寿命预估方法
CN116298947A (zh) 一种蓄电池核容监测装置
CN115422846A (zh) 燃料电池的老化建模方法、装置、设备和介质
Cao et al. A neural network based method for real-time measurement of capacity and SOC of vanadium redox flow battery
CN101071956A (zh) 太阳能光伏系统中二次电池阵列的调控方法及其调控装置
CN112713291A (zh) 一种燃料电池系统及其控制方法
CN104466213A (zh) 一种水冷pemfc空气过量系数控制系统及方法
CN112819350B (zh) 一种公用工程设备的管理方法及系统
CN109927584A (zh) 电动车电池管理方法、计算机设备及存储介质

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 22845150

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE