CN111044926A - Method for predicting service life of proton exchange membrane fuel cell - Google Patents
Method for predicting service life of proton exchange membrane fuel cell Download PDFInfo
- Publication number
- CN111044926A CN111044926A CN201911291950.3A CN201911291950A CN111044926A CN 111044926 A CN111044926 A CN 111044926A CN 201911291950 A CN201911291950 A CN 201911291950A CN 111044926 A CN111044926 A CN 111044926A
- Authority
- CN
- China
- Prior art keywords
- particle
- fuel cell
- weight
- state
- particles
- Prior art date
- Legal status (The legal status 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 status listed.)
- Pending
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]
- G01R31/392—Determining battery ageing or deterioration, e.g. state of health
-
- 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]
- G01R31/378—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
Abstract
The invention relates to a method for predicting the service life of a proton exchange membrane fuel cell. The method for predicting the service life of the fuel cell based on the particle filter framework comprises a learning stage and a prediction stage: in the learning stage, the steps of initialization, updating particle state, updating particle weight and resampling are sequentially carried out until t is reachedpTime of day; wherein, the particles refer to possible values of the voltage of the fuel cell, tpThe particle weight is not updated from the moment, and resampling is not performed; in the prediction phase, from tpAnd starting to update the particle state at any moment, obtaining the voltage value of the fuel cell according to the particle state, judging whether the voltage value reaches a cell failure threshold value, obtaining the remaining service life of the fuel cell if the voltage value reaches the cell failure threshold value, and otherwise, updating the particle state in an iterative manner. The invention adopts a statistical method to update the particle number of each iterationAnd the aging parameter can accurately predict the aging trend of the fuel cell, so that the residual service life of the PEMFC can be calculated.
Description
Technical Field
The invention belongs to the technical field of life prediction of fuel cells, and particularly relates to a method for predicting the life of a proton exchange membrane fuel cell.
Background
Proton Exchange Membrane Fuel Cells (PEMFCs) are one of the most promising new power generation devices, and have been widely used in the fields of military, transportation, cogeneration, and the like due to their characteristics of high power density, environmental friendliness, light weight, and abundant resources.
Prognosis and Health Management (PHM) involves two aspects: prediction refers to the predictive diagnosis of the future health status of a component or the overall system based on the current or historical status of the system, including determining the length of time or Remaining Useful Life (RUL) of the component or system in normal operation; health management refers to the ability to make appropriate decisions about maintenance activities based on diagnostic prognostic information, available maintenance resources, and usage requirements. The PHM utilizes the actual monitoring data to obtain relevant indexes and trends reflecting the health condition of the PEMFC, and after the remaining service life of the battery is obtained, the power supply required by the system can be adjusted, so that the battery is more durable. The current battery life prediction technology is mainly divided into two types: the first category is model-based prediction methods that use physical or mathematical models of the system to make the prediction. The second category is based on data-driven prediction methods, using pattern recognition and machine learning to predict based on test or sensor data.
The model driving method realizes the prediction of the residual service life by relying on the load condition, the material property, the degradation mechanism and the failure mechanism of the fuel cell. The model driving method mainly comprises particle filtering, Kalman filtering, a degradation mechanism model and an experience degradation model. PEMFC systems are complex systems with multiple physics, multiple scales, and high uncertainties. The degradation mechanism is not completely understood, and it is therefore difficult to obtain an accurate analytical model to describe the degradation of the fuel cell system, particularly in noisy or uncertain environments. The data driving method mainly comprises an echo state network, an overrun learning machine, a self-adaptive neural-fuzzy inference system, a correlation vector machine and a Gaussian process state space model.
In the current related art solution, there is only a single battery life prediction solution, and there is no solution for performing life prediction at a stack level. In addition, the current scheme does not consider the influence of dynamic load, and some schemes improve the effect, but are not clearly proved. As the existing prediction method of the residual service life of the PEMFC based on the unscented Kalman filtering, a physical model of the aging of the fuel cell and a relation between the state of the cell and the aging rate are established. The method does not take into account the effect of dynamic loading and has not been validated by experimental studies. A PEMFC aging prediction algorithm based on an echo state network. Inputting the pile voltage data after short-time Fourier transform preprocessing into a network, replacing the neural network of the original hidden layer with a neuron pool, training network parameters by adopting the existing voltage degradation data of the PEMFC, and predicting the pile voltage of the PEMFC by using an iterative structure so as to estimate the residual service life of the PEMFC. Although this method has improved accuracy, it has not been clearly demonstrated. Although some progress has been made in the life prediction technology of fuel cells, more research is needed to fully understand the aging process of the fuel cell and predict the remaining service life of the fuel cell within a smaller error range.
Disclosure of Invention
The invention provides a method for predicting the service life of a proton exchange membrane fuel cell, which predicts the residual service life of the fuel cell based on a particle filtering method. The scheme adopts a statistical method to update the particle number and the aging parameter of each iteration, and can accurately predict the aging trend of the fuel cell, so that the residual service life of the PEMFC is calculated.
The technical scheme adopted by the invention is as follows:
a proton exchange membrane fuel cell life prediction method is characterized in that the fuel cell life prediction is carried out based on a particle filter framework, and comprises a learning stage and a prediction stage:
in the learning stage, initialization, updating of particle state,Updating the particle weight and resampling until t is reachedpTime of day; wherein, the particles refer to possible values of the voltage of the fuel cell, tpThe particle weight is not updated from the moment, and resampling is not performed;
in the prediction phase, from tpAnd starting to update the particle state at any moment, obtaining the voltage value of the fuel cell according to the particle state, judging whether the voltage value reaches a cell failure threshold value, obtaining the remaining service life of the fuel cell if the voltage value reaches the cell failure threshold value, and otherwise, updating the particle state in an iterative manner.
Further, firstly, the voltage data of the fuel cell obtained through test measurement is subjected to denoising processing by utilizing a nuclear smoothing technology, and then a learning stage is carried out.
Further, the initializing includes: the number of particle state updates, k, is 0, and the set of particles is sampled according to the known prior probability density of the particle state.
Further, the particle states are updated according to a state transition equation, which is an equation describing the update of the particle voltage values, as follows:
xk＝-β·(tk-tk-1)+xk-1
wherein xkRepresenting the state of the particle after k updates, where xk-1Denotes the state of the particle after k-1 updates, tkDenotes the time, t, at which the particle state is updated the kth timek-1Indicating the time when the particle state was updated the (k-1) th time, β was initialized based on the distribution of β in the history.
Further, the particle weight is a weight occupied by the particle in the reflected battery voltage value, and the updated particle weight is a weight increased to reflect the true battery voltage value.
Further, the method for updating the particle weight is as follows: calculating the weight of the particles according to the prediction error of the particle state, wherein the weight of the particles with small errors is large, and the weight of the particles with large errors is small; wherein the prediction error is calculated from a difference between the measured value and the predicted value of the battery voltage.
Further, after resampling, the total number of particles remains unchanged, the particles with large weight are divided into a plurality of particles, the particles with small weight are discarded, and the weight of each particle after resampling is the same.
Further, the prediction phase updates the particle states according to the state transition equations in the same manner as the learning phase.
The invention has the following beneficial effects:
the invention provides a method for predicting the residual life of a fuel cell based on a particle filter framework. Experimental results show that the scheme can effectively estimate the remaining service life of the fuel cell with prediction error below 20%, thereby reducing the possibility of failure of the system using the PEMFC before the predicted time.
Drawings
FIG. 1 is a flow chart of the steps of the particle filter framework in an embodiment.
FIG. 2 is a graph of the results of an experiment in accordance with the protocol of the present invention.
Detailed Description
In order to make the aforementioned objects, features and advantages of the present invention comprehensible, the present invention shall be described in further detail with reference to the following detailed description and accompanying drawings.
The main content of the invention comprises:
1) and establishing a state equation (namely a state transition equation) of a voltage attenuation model of the fuel cell under the static action.
2) The initial state parameters of the model are determined, as well as the initial voltage distribution range (a number of particles corresponding to the distribution are selected based on the prior probability density distribution).
3) And denoising the data by utilizing a kernel smoothing technology.
Firstly, denoising battery voltage data obtained through test measurement by utilizing a nuclear smoothing technology.
The kernel smoothing technique is a more efficient data smoothing method, and is a weighted moving average, and the weight depends on k (t). Considering the signal u (t), expressed as n data, the data are smoothed and then represented as f (t)j) Represents j ═ 1,2, 3.
Wherein the content of the first and second substances,
wherein K (t) represents a Gaussian kernel function, tjIndicates the current time, tiAnd h represents the bandwidth at any moment, and controls the local action range of the Gaussian kernel function.
And then, predicting the service life of the fuel cell based on a particle filtering method by using the data subjected to denoising processing. Fig. 1 is a particle filter framework, which mainly includes a learning phase and a prediction phase.
1. The specific algorithm steps in the learning phase are as follows:
(1) initialization
k is 0, according to a known prior probability density p (x)0) A set of particles is sampled.
Where, particles are possible values of the voltage of the battery, k represents the number of times the state of the particles is updated, X representing the state of the particles in fig. 1 represents the voltage, and W represents the weight of the particles. The prior probability density is the prior probability density of the particle state and is obtained empirically. The sampling is to select a plurality of particles according to the prior probability density distribution, and 4000 particles are selected in this embodiment.
(2) Updating particle states
The particle state refers to the voltage value represented by the particle at that moment. The purpose of updating the particle state is to make the voltage prediction of the cell more accurate.
And updating the particle state according to the state transition equation. The state transition equation is an equation used to describe the update of the particle voltage value, as follows:
xk＝-β·(tk-tk-1)+xk-1
wherein xkRepresenting the state of the particle after k updates, where xk-1Denotes the state of the particle after k-1 updates, tkDenotes the time, t, at which the particle state is updated the kth timek-1Is shown asThe time when the particle state is updated k-1 times, β, needs to be initialized according to the distribution range of β in the history data.
(3) Updating weights
The particle weight is a weight that the particle occupies in reflecting the battery voltage value. The purpose of updating the particle weight is to increase the particle weight which can reflect the true value of the battery voltage, so that the battery voltage prediction result is more accurate.
The method for updating the particle weight comprises the following steps: and calculating the weight of the particles according to the prediction error of the particle state, wherein the weight of the particles with small errors is large, and the weight of the particles with large errors is small. Wherein the prediction error is based on the battery voltage measurement ZkAnd calculating the difference value between the predicted value and the reference value.
In fig. 1, "weight degradation" means that most of the particle weights degrade to 0, and the number of particles that can indicate the battery state is small.
(4) Resampling
Resampling means adding high-weight particles and deleting low-weight particles in order to suppress weight degradation.
After resampling, the total number of the particles is kept unchanged, the particles with large weight are divided into a plurality of particles, the particles with small weight are discarded, and the weight of each particle after resampling is the same.
2. The algorithm steps of the prediction phase are as follows:
(1) estimating particle states
As shown in fig. 1, from tpBeginning to predict the remaining useful life of the fuel cell, tpThat is, from this moment, the particle weight is not updated any more, and resampling is not performed any more, only according to tpThe particle at that time updates the particle state. In this example tp＝600h。
(2) Updating particle states
And updating the particle state according to the state transition equation by adopting the same method as the learning stage.
(3) Threshold determination
And obtaining the voltage value of the PEMFC according to the particle state, judging whether the voltage value reaches a preset battery voltage failure threshold value, if so, obtaining the residual service life of the fuel battery (the time interval of the battery failure moment is predicted for the beginning), and otherwise, iteratively updating the particle state. The battery voltage failure threshold may be set based on practical experience.
The above embodiments are only intended to illustrate the technical solution of the present invention and not to limit the same, and a person skilled in the art can modify the technical solution of the present invention or substitute the same without departing from the principle and scope of the present invention, and the scope of the present invention should be determined by the claims.
Claims (8)
1. A proton exchange membrane fuel cell life prediction method is characterized in that the fuel cell life prediction is carried out based on a particle filter framework, and comprises a learning stage and a prediction stage:
in the learning stage, the steps of initialization, updating particle state, updating particle weight and resampling are sequentially carried out until t is reachedpTime of day; wherein, the particles refer to possible values of the voltage of the fuel cell, tpThe particle weight is not updated from the moment, and resampling is not performed;
in the prediction phase, from tpAnd starting to update the particle state at any moment, obtaining the voltage value of the fuel cell according to the particle state, judging whether the voltage value reaches a cell failure threshold value, obtaining the remaining service life of the fuel cell if the voltage value reaches the cell failure threshold value, and otherwise, updating the particle state in an iterative manner.
2. The method of claim 1, wherein the voltage data of the fuel cell obtained by the test measurement is denoised by a nuclear smoothing technique, and then the learning stage is performed.
3. The method of claim 1, wherein said initializing comprises: the number of particle state updates, k, is 0, and the set of particles is sampled according to the known prior probability density of the particle state.
4. The method of claim 1, wherein the particle states are updated according to a state transition equation, the state transition equation being an equation describing the update of the particle voltage values, as follows:
xk＝-β·(tk-tk-1)+xk-1
wherein xkRepresenting the state of the particle after k updates, where xk-1Denotes the state of the particle after k-1 updates, tkDenotes the time, t, at which the particle state is updated the kth timek-1Indicating the time when the particle state was updated the (k-1) th time, β was initialized based on the distribution of β in the history.
5. The method of claim 1, wherein the particle weight is a weight of the particle in the battery voltage, and updating the particle weight is increasing the weight of the particle to reflect the actual battery voltage.
6. The method of claim 5, wherein the method of updating the particle weights is: calculating the weight of the particles according to the prediction error of the particle state, wherein the weight of the particles with small errors is large, and the weight of the particles with large errors is small; wherein the prediction error is calculated from a difference between the measured value and the predicted value of the battery voltage.
7. The method of claim 1, wherein after resampling, the total number of particles remains unchanged, the particles with higher weight are divided into a plurality of particles, the particles with lower weight are discarded, and the weight of each particle after resampling is the same.
8. The method of claim 1, wherein the prediction phase updates the particle states according to state transition equations in the same manner as the learning phase.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911291950.3A CN111044926A (en) | 2019-12-16 | 2019-12-16 | Method for predicting service life of proton exchange membrane fuel cell |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911291950.3A CN111044926A (en) | 2019-12-16 | 2019-12-16 | Method for predicting service life of proton exchange membrane fuel cell |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN111044926A true CN111044926A (en) | 2020-04-21 |
Family
ID=70236514
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201911291950.3A Pending CN111044926A (en) | 2019-12-16 | 2019-12-16 | Method for predicting service life of proton exchange membrane fuel cell |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111044926A (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111680848A (en) * | 2020-07-27 | 2020-09-18 | 中南大学 | Battery life prediction method based on prediction model fusion and storage medium |
| CN114551945A (en) * | 2020-11-27 | 2022-05-27 | 中国科学院大连化学物理研究所 | Automatic optimized fuel cell life prediction method |
| CN114566686A (en) * | 2020-11-27 | 2022-05-31 | 中国科学院大连化学物理研究所 | Method for estimating state and predicting service life of fuel cell |
| CN116502541A (en) * | 2023-05-18 | 2023-07-28 | 淮阴工学院 | Method for predicting service life of proton exchange membrane fuel cell |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101132595A (en) * | 2007-09-29 | 2008-02-27 | 清华大学 | Energy management method for wireless network measurement |
| CN102542155A (en) * | 2011-12-05 | 2012-07-04 | 北京航空航天大学 | Particle filter residual life forecasting method based on accelerated degradation data |
| CN106845866A (en) * | 2017-02-27 | 2017-06-13 | 四川大学 | Equipment method for predicting residual useful life based on improved particle filter algorithm |
| CN107918103A (en) * | 2018-01-05 | 2018-04-17 | 广西大学 | A kind of lithium ion battery residual life Forecasting Methodology based on grey particle filter |
| CN109543317A (en) * | 2018-04-28 | 2019-03-29 | 北京航空航天大学 | A kind of method and device of PEMFC remaining life prediction |
| CN109633474A (en) * | 2018-11-14 | 2019-04-16 | 江苏大学 | A kind of lithium ion battery residual life prediction technique |
-
2019
- 2019-12-16 CN CN201911291950.3A patent/CN111044926A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101132595A (en) * | 2007-09-29 | 2008-02-27 | 清华大学 | Energy management method for wireless network measurement |
| CN102542155A (en) * | 2011-12-05 | 2012-07-04 | 北京航空航天大学 | Particle filter residual life forecasting method based on accelerated degradation data |
| CN106845866A (en) * | 2017-02-27 | 2017-06-13 | 四川大学 | Equipment method for predicting residual useful life based on improved particle filter algorithm |
| CN107918103A (en) * | 2018-01-05 | 2018-04-17 | 广西大学 | A kind of lithium ion battery residual life Forecasting Methodology based on grey particle filter |
| CN109543317A (en) * | 2018-04-28 | 2019-03-29 | 北京航空航天大学 | A kind of method and device of PEMFC remaining life prediction |
| CN109633474A (en) * | 2018-11-14 | 2019-04-16 | 江苏大学 | A kind of lithium ion battery residual life prediction technique |
Non-Patent Citations (1)
| Title |
|---|
| 李奇 等: "质子交换膜燃料电池剩余使用寿命预测方法综述及展望", 《中国电机工程学报》 * |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111680848A (en) * | 2020-07-27 | 2020-09-18 | 中南大学 | Battery life prediction method based on prediction model fusion and storage medium |
| CN114551945A (en) * | 2020-11-27 | 2022-05-27 | 中国科学院大连化学物理研究所 | Automatic optimized fuel cell life prediction method |
| CN114566686A (en) * | 2020-11-27 | 2022-05-31 | 中国科学院大连化学物理研究所 | Method for estimating state and predicting service life of fuel cell |
| CN114566686B (en) * | 2020-11-27 | 2023-11-14 | 中国科学院大连化学物理研究所 | Method for evaluating state and predicting service life of fuel cell |
| CN114551945B (en) * | 2020-11-27 | 2023-11-14 | 中国科学院大连化学物理研究所 | Automatic optimization fuel cell life prediction method |
| CN116502541A (en) * | 2023-05-18 | 2023-07-28 | 淮阴工学院 | Method for predicting service life of proton exchange membrane fuel cell |
| CN116502541B (en) * | 2023-05-18 | 2023-12-22 | 淮阴工学院 | Method for predicting service life of proton exchange membrane fuel cell |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111044926A (en) | Method for predicting service life of proton exchange membrane fuel cell | |
| CN112036084B (en) | Similar product life migration screening method and system | |
| Xie et al. | Prognostic for fuel cell based on particle filter and recurrent neural network fusion structure | |
| Fan et al. | A novel machine learning method based approach for Li-ion battery prognostic and health management | |
| CN112990556A (en) | User power consumption prediction method based on Prophet-LSTM model | |
| CN111339712A (en) | Method for predicting residual life of proton exchange membrane fuel cell | |
| EP3380948B1 (en) | Environmental monitoring systems, methods and media | |
| Zhao et al. | Bearing health condition prediction using deep belief network | |
| CN113205233B (en) | Lithium battery life prediction method based on wolf algorithm and multi-core support vector regression | |
| CN114218872B (en) | DBN-LSTM semi-supervised joint model-based residual service life prediction method | |
| CN108961460B (en) | Fault prediction method and device based on sparse ESGP (Enterprise service gateway) and multi-objective optimization | |
| CN111523727B (en) | Method for predicting remaining life of battery by considering recovery effect based on uncertain process | |
| Zhao et al. | On-line least squares support vector machine algorithm in gas prediction | |
| CN115271186B (en) | Reservoir water level prediction and early warning method based on delay factor and PSO RNN Attention model | |
| CN115513951A (en) | Power load prediction method and system based on concept drift detection | |
| CN112734002A (en) | Service life prediction method based on data layer and model layer joint transfer learning | |
| CN112100574A (en) | Resampling-based AAKR model uncertainty calculation method and system | |
| CN116245373A (en) | Power data quality rule self-adaption method based on deep learning | |
| CN114994555A (en) | Fuel cell residual life prediction method and device based on hybrid model | |
| CN116930609A (en) | Electric energy metering error analysis method based on ResNet-LSTM model | |
| CN116960962A (en) | Mid-long term area load prediction method for cross-area data fusion | |
| CN116861256A (en) | Furnace temperature prediction method, system, equipment and medium for solid waste incineration process | |
| CN112257327A (en) | Submarine pipeline residual life prediction method based on wavelet transform denoising and stacking self-coding feature selection | |
| Zhang et al. | Remaining useful life prediction of lithium-ion batteries based on TCN-DCN fusion model combined with IRRS filtering | |
| CN114169416B (en) | Short-term load prediction method based on migration learning under small sample set |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20200421 |
|
| RJ01 | Rejection of invention patent application after publication |