AU2021102629A4 - Lithium battery thermal runaway multi-level detection and early warning system and judgment method - Google Patents
Lithium battery thermal runaway multi-level detection and early warning system and judgment method Download PDFInfo
- Publication number
- AU2021102629A4 AU2021102629A4 AU2021102629A AU2021102629A AU2021102629A4 AU 2021102629 A4 AU2021102629 A4 AU 2021102629A4 AU 2021102629 A AU2021102629 A AU 2021102629A AU 2021102629 A AU2021102629 A AU 2021102629A AU 2021102629 A4 AU2021102629 A4 AU 2021102629A4
- Authority
- AU
- Australia
- Prior art keywords
- thermal runaway
- battery
- early warning
- prediction
- values
- 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.)
- Ceased
Links
- 238000001514 detection method Methods 0.000 title claims abstract description 28
- WHXSMMKQMYFTQS-UHFFFAOYSA-N Lithium Chemical compound [Li] WHXSMMKQMYFTQS-UHFFFAOYSA-N 0.000 title claims abstract description 26
- 229910052744 lithium Inorganic materials 0.000 title claims abstract description 26
- 238000000034 method Methods 0.000 title abstract description 18
- 238000012549 training Methods 0.000 claims abstract description 29
- 238000013528 artificial neural network Methods 0.000 claims abstract description 24
- 230000008878 coupling Effects 0.000 claims abstract description 21
- 238000010168 coupling process Methods 0.000 claims abstract description 21
- 238000005859 coupling reaction Methods 0.000 claims abstract description 21
- HBBGRARXTFLTSG-UHFFFAOYSA-N Lithium ion Chemical compound [Li+] HBBGRARXTFLTSG-UHFFFAOYSA-N 0.000 claims abstract description 20
- 229910001416 lithium ion Inorganic materials 0.000 claims abstract description 20
- 238000001914 filtration Methods 0.000 claims abstract description 7
- 230000006870 function Effects 0.000 claims description 16
- 230000008859 change Effects 0.000 claims description 5
- 230000006872 improvement Effects 0.000 description 8
- 238000010586 diagram Methods 0.000 description 7
- 230000001413 cellular effect Effects 0.000 description 6
- 238000004146 energy storage Methods 0.000 description 5
- 238000012545 processing Methods 0.000 description 5
- 238000006243 chemical reaction Methods 0.000 description 4
- 230000008569 process Effects 0.000 description 4
- 230000002159 abnormal effect Effects 0.000 description 3
- 238000011161 development Methods 0.000 description 3
- 238000009826 distribution Methods 0.000 description 3
- 238000004880 explosion Methods 0.000 description 3
- 238000012544 monitoring process Methods 0.000 description 3
- 210000002569 neuron Anatomy 0.000 description 3
- 238000003860 storage Methods 0.000 description 3
- 230000004913 activation Effects 0.000 description 2
- 210000004027 cell Anatomy 0.000 description 2
- 230000010485 coping Effects 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 206010000369 Accident Diseases 0.000 description 1
- 239000010405 anode material Substances 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 239000010406 cathode material Substances 0.000 description 1
- 230000001364 causal effect Effects 0.000 description 1
- 238000007418 data mining Methods 0.000 description 1
- 238000000354 decomposition reaction Methods 0.000 description 1
- 230000007812 deficiency Effects 0.000 description 1
- 210000001787 dendrite Anatomy 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000009792 diffusion process Methods 0.000 description 1
- 238000004134 energy conservation Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000001125 extrusion Methods 0.000 description 1
- 230000007774 longterm Effects 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 239000012528 membrane Substances 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 230000035939 shock Effects 0.000 description 1
- 239000002341 toxic gas Substances 0.000 description 1
- 238000012546 transfer Methods 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
- 239000013598 vector Substances 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- 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
-
- 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]
- G01R31/382—Arrangements for monitoring battery or accumulator variables, e.g. SoC
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/044—Recurrent networks, e.g. Hopfield networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Software Systems (AREA)
- Mathematical Physics (AREA)
- Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Computing Systems (AREA)
- Molecular Biology (AREA)
- General Health & Medical Sciences (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Secondary Cells (AREA)
Abstract
The present invention provides a lithium battery thermal runaway multi-level
detection and early warning system and a judgment method. In the method, fault-free
battery characteristic elements are employed to train the thermal runaway prediction
ability of a prediction model formed by a long-short term memory network, a time
5 convolutional network and a GRU neural network; and then the battery characteristic
elements of a to-be-monitored battery in time series are input into the prediction
model to obtain prediction values in a normal state at a certain moment, then the
prediction values are compared with collected true values corresponding to the
moment, and the larger difference are, the larger a battery thermal runaway risk is.
10 Through such operation, respective advantages are combined, and the problem of
overlarge deviation of coefficients trained by respective models is avoided, so that
high-precision battery thermal runaway grading detection is achieved, and speed and
precision of lithium ion battery thermal runaway early warning are improved.
Obtaining a plurality of element
data related to a fault-free lithium battery pack
Filtering data of a fault battery pack at a high
frequency with a Gaussian filter to form a training set
Returning respective loss values of prediction values
obtained by coupling with a TCN model, a GRU model
and an LSTM model respectively, and continuously
conducting training
Fig. 1
Training module Prediction module
Data Gaussian Equal weight
acquisition Filter c otu
Difference percentage grading
Fig. 2
1/3
Description
Obtaining a plurality of element data related to a fault-free lithium battery pack
Filtering data of a fault battery pack at a high frequency with a Gaussian filter to form a training set
Returning respective loss values of prediction values obtained by coupling with a TCN model, a GRU model and an LSTM model respectively, and continuously conducting training
Fig. 1
Training module Prediction module
Data Gaussian Equal weight acquisition Filter c otu
Difference percentage grading
Fig. 2
1/3
The present invention relates to the technical field of battery safety early warning,
in particular to a lithium battery thermal run away multi-level detection and early
warning system and a judgment method.
In recent years, a lithium ion battery energy storage system has been widely
applied to various regions, such as new energy source grid connection, a micro grid
and a smart power grid. A lithium ion battery has wide prospects in a large energy
storage system due to the characteristics of high energy density, high transfer
efficiency and rapid reaction. However, with scale application of the lithium ion
battery energy storage system, the safety problem of the energy storage system
arouses great attention. Discovered from statistical survey on accidents, thermal
runaway of the lithium ion battery caused by its chemical reaction or external
influence has become the main cause of the safety problem. Thermal runaway of the
lithium battery has the characteristics of rapid diffusion rate, large flame strength,
production of a great quantity of toxic gas and the like. Although a fire is put out, a
reignition phenomenon may occur, thereby severely threatening the whole energy
storage system.
Thermal runaway of the battery is caused by the fact that the battery produces
heat, a temperature abnormally rises, and the chain exothermic reaction is caused,
leading to battery burning and explosion. According to triggering conditions, there are
mainly three kinds of reasons that cause thermal runaway of the battery: machinery
abuse, such as mechanical deformation caused by needling, extrusion, heavy weight
shock and the like; electrical abuse which mainly includes faults of electric
components, for example, lithium dendrites are produced in the batteries due to
overcharge and overdischarge and pierce a membrane; and thermal abuse, such as decomposition of anode and cathode materials caused by too high temperature. Thus, accurate prediction for a lithium ion thermal runaway fault would accelerate development of an energy storage technology and has a great significance to promote transformation of new energy source structure in China, guarantee energy safety and achieve the targets of energy conservation and emission reduction. The patent
CN20160282373.1 discloses a battery fault detection method and a battery fault
detection apparatus. In the method, a battery cell in a fault in a battery pack is
recognized by employing a data mining algorithm according to the battery
characteristic elements of the battery pack. A threshold which is set in advance
without using human based experience can improve the accuracy of battery fault
detection. However, the method can only detect the fault on a current battery state,
cannot achieve advanced early warning, and thus easily causes the problem that early
warning and coping are not timely.
In view of this, there is a need to design an improved lithium battery thermal
runaway multi-level detection and early warning system and judgment method to
solve the above problems, or to at least provide a useful alternative to existing
systems.
In order to overcome the deficiencies of the prior art or to provide a useful
alternative embodiments of a lithium battery thermal runaway multi-level detection
and early warning system and a judgment method will be described. By employing
the long-short term memory network, the time convolutional network and the GRU
neural network for prediction respectively and conducting equal weight coupling,
respective advantages are combined, and the problem of overlarge deviation of
coefficients trained by respective models is avoided, so that high-precision battery
thermal runaway grading detection is achieved, and speed and precision of lithium ion
battery thermal runaway early warning are improved. According to a first aspect, there is provided a lithium battery thermal runaway multi-level detection and early warning system which includes the following steps:
Si. collecting fault-free battery characteristic elements, and respectively training prediction abilities of the long-short term memory network, the time convolutional network and the GRU neural network for the battery characteristic elements in time series; S2. collecting the battery characteristic elements of a to-be-monitored battery in time series, inputting the battery characteristic elements into the trained long-short term memory network, time convolutional network and GRU neural network respectively to obtain three groups of prediction values, and conducting equal weight coupling on the three groups of prediction values to obtain values as fault-free prediction values; and S3. continuously collecting the battery characteristic elements of the to-be-monitored battery in time series as true values, comparing the true values with the fault-free prediction values obtained in the step S2, and obtaining a result that the lager differences are, the larger a battery thermal runaway risk is. As further improvement on the present invention, the battery characteristic elements include but are not limited to one or more of a temperature state, a current state, a voltage state and a charge state of the battery. As further improvement on the present invention, in the step Sl, the fault-free battery characteristic elements are obtained by filtering off the battery characteristic elements of thermal runaway with a Gaussian filter. As further improvement on the present invention, the Gaussian filter is a
low-pass filter with a frequency domain width of ±200w and a threshold of1-I/100.
As further improvement on the present invention, in the step Si, training the prediction ability includes the following steps: employing the fault-free battery characteristic elements in time series to train the long-short term memory network, the time convolutional network and the GRU neural network, and conducting coupling to obtain a prediction result, i.e. a change trend value of the battery characteristic elements over time; and then returning respective loss values of three networks, and continuously conducting training. As further improvement on the present invention, the loss values of the long-short term memory network and the GRU neural network are obtained through an MSE loss function; and the loss value of the time convolutional network is obtained through a mean absolute deviation (MAD).
As further improvement on the present invention, the step S3 further includes the following steps: taking a ratio of an absolute value of a difference between each true value and the corresponding fault-free prediction value to the corresponding fault-free prediction value as a thermal runaway early warning coefficient, wherein the larger the thermal runaway early warning coefficient is, the lager the battery thermal runaway risk is. As further improvement on the present invention, the step S3 further includes the following steps: dividing the thermal runaway early warning coefficients into a plurality of grade intervals for grading early warning, wherein the larger an upper limit value of a grade interval in which thermal runaway is located is, the larger the battery thermal runaway risk is. In order to achieve the above purpose of the present invention, the present invention further provides a lithium ion battery thermal runaway grading early warning system which includes: a data collecting module, used for collecting characteristic element data of a battery; a thermal runaway model building module, used for building a long-short term memory network, a time convolutional network and the GRU neural network into a prediction model associated with the battery characteristic elements in parallel, and employing the fault-free battery characteristic elements to train the prediction ability of the prediction model; and a thermal runaway early warning module, used for inputting the battery characteristic elements of a to-be-monitored battery in time series into the prediction model to obtain prediction values at a certain moment, and then comparing the prediction values with collected true values corresponding to the moment to obtain a thermal runaway early warning result. As further improvement on the present invention, the lithium ion battery thermal runaway grading early warning system further includes an equal weight coupling module used for conducting equal weight coupling on the prediction values of the long-short term memory network, the time convolutional network and the GRU neural network.
The present invention has the beneficial effects that: 1. In the lithium battery thermal runaway multi-level detection and early warning system provided by the present invention, by employing the long-short term memory network, the time convolutional network and the GRU neural network for prediction respectively and conducting equal weight coupling, respective advantages are combined, the problem of overlarge deviation of coefficients trained by respective models is avoided, and a training period is shortened, so that high-precision battery thermal runaway grading detection is achieved, speed and precision of lithium ion battery thermal runaway early warning are improved, intelligent progress of new energy source storage safety is promoted, and application of the Fourth Industrial
Revolution in the energy and energy storage industry is accelerated.
2. In the lithium battery thermal runaway multi-level detection and early warning
system provided by the present invention, with historical battery characteristic
elements of a normal battery as a training set, differences between values of the
predicted characteristic elements and the true values of the historical battery
characteristic elements are gradually corrected and reduced; during prediction,
because the model is trained and verified by using normal data, the prediction values
obtained according to real-time monitoring data are all prediction values in a normal
state; then the prediction values are compared with the true values at the moment,
showing that the larger the differences are, the larger the battery thermal runaway risk
is. Through such operation, a working state of the battery can be predicted in advance
in time series, and thus a real sense of thermal runaway early warning is achieved.
3. In the lithium battery thermal runaway multi-level detection and early warning
system provided by the present invention, when the model is trained, loss produced by
the prediction values and the true values in the training set is returned, so that the
training effect is strengthened, and a working workload of the whole model can be
reduced.
Fig. 1 is a flow chart of a model training method of a lithium battery thermal
runaway multi-level detection and early warning system of the present invention.
Fig. 2 is a block diagram of a composition structure of lithium ion battery thermal runaway grading early warning of the present invention.
Fig. 3 is a structural diagram of a long-short term memory network.
Fig. 4 is a structural diagram of a time convolutional network.
Fig. 5 is a structural diagram of a GRU neural network.
Fig. 6 is an image of probability density functions in normal distribution in a
frequency domain.
DETAILED DESCRIPTION OF THE PRESENT INVENTION In order to make the purpose, technical solutions, and advantages of the present invention clearer, the present invention is described in detail below in combination with specific embodiments. It should also be noted herein that in order to avoid obscuring the present invention due to unnecessary details, only the structure and/or processing steps closely related to the solution of the present invention are shown in the specific embodiments, and other details that are not relevant to the present invention are omitted. In addition, it should also be noted that the terms "including", "include" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements not only includes those elements, but also includes other elements that are not explicitly listed, or elements inherent to the process, method, article, or device. In the long-term power supply working process of a lithium ion battery, due to the service life problem of the lithium ion battery or the external environment problem, abnormal temperature may be caused, and severely, a fire accident may be caused. Occurrence of abnormal temperature is not suddenly changed generally, and the abnormal temperature presents a change trend generally and may gradually develop into an uncontrollable thermal runaway accident if not being timely warned and treated. Based on this, the present invention provides a high-precision lithium battery thermal runaway multi-level detection and early warning system. According to historical monitoring parameters of the lithium ion battery, parameters in a normal state at a subsequent moment are accurately predicted and then are compared with parameters monitored in real time; and if differences between the parameters monitored in real time and the predicted parameters at the corresponding moment are too large, it indicates that the battery may have a thermal runaway risk, thereby realizing advanced early warning. As shown in Fig. 1 and Fig. 2, the present invention provides the lithium battery thermal runaway multi-level detection and early warning system which includes the following steps: SI, characteristic elements (historical battery characteristic elements) of a fault-free battery in time series are collected; and the prediction abilities of a long-short term memory network, a time convolutional network and a GRU neural network for the battery characteristic elements in time series are trained (that is, differences between values of the predicted characteristic elements and the true values of the historical battery characteristic elements are gradually corrected and reduced according to the historical battery characteristic elements, one prediction value and one true value which are arbitrarily compared with each other in a training set should be values at a same time point, and thus accurate prediction of a change trend is achieved). Through such operation, historical data of the normal battery is used to train and verify network models, so that the battery characteristic elements in the normal state at a subsequent moment can be predicted in advance according to the historical data; and during prediction, because the models are trained and verified by using normal data, prediction values obtained according to real-time monitoring data are all prediction values in the normal state. S2. The battery characteristic elements of a to-be-monitored battery in time series are collected and input into the trained long-short term memory network, time convolutional network and GRU neural network respectively to obtain three groups of prediction values, and equal weight coupling is conducted on the three groups of prediction values to obtain values as fault-free prediction values. Equal weight coupling conducted by combining the three groups of prediction values may correct a result, reduce the deviation of coefficients obtained by training each model, and shorten a training period, so that a prediction result is more accurate. S3. The battery characteristic elements of the to-be-monitored battery in time series are continuously collected as true values and compared with the fault-free prediction values obtained in the step S2, to obtain a result that the lager differences are, the larger a battery thermal runaway risk is. The step S3 further includes: a ratio of an absolute value of a difference between each true value and the corresponding fault-free prediction value to the corresponding fault-free prediction value is taken as a thermal runaway early warning coefficient, wherein the larger the thermal runaway early warning coefficient is, the lager the battery thermal runaway risk is. The step S3 further includes: the thermal runaway early warning coefficients are divided into a plurality of grade intervals for grading early warning, wherein the larger an upper limit value of a grade interval in which thermal runaway is located is, the larger the battery thermal runaway risk is. For example, according to a scale that a whole interval is 0-100% with a range of each interval being 25%, the thermal runaway early warning coefficients are divided into four grade intervals: [0-25%],
[26-50%], [51-75%] and [76-100%]. When the values of intervals in which the obtain thermal runaway early warning coefficients are located are lager, the thermal runaway risks are larger to conduct early warning to different degrees. The battery characteristic elements include but are not limited to one or more of a temperature state, a current state, a voltage state and a charge state of the battery, and preferably, in the temperature state, the current state, the voltage state and the charge state, the prediction accuracy of the normal state of the battery can be improved through centralized training of various battery characteristic elements. By employing the early warning method of the present invention, because the early warning model is trained based on the characteristic elements of a thermal runaway fault-free battery, battery characteristic element prediction values at some a subsequent moment can be sensitively given according to the historical battery characteristic elements of the normal battery. For example, if the battery at ti-t3 moments normally works, the battery at a t4 moment has a trend of thermal runaway, or the t4 moment is a starting point of development of thermal runaway. Then, the prediction model formed by the above three networks can provide the battery characteristic elements in the normal state at the t4 moment according to the battery characteristic elements at the t 1-t3 moments; an exception at the t4 moment can be accurately judged by comparing the battery characteristic elements with true values monitored at the t4 moment, and early warning is conducted immediately to employ a corresponding coping measure. However, by using the prior art, the starting point of development of thermal runaway at the t4 moment cannot be sensitively judged as a thermal runaway state, and thus advanced early warning cannot be conducted. In the step S, the fault-free battery characteristic elements are obtained by filtering off the battery characteristic elements of thermal runaway with a Gaussian filter, that is, the Gaussian filter filters bizarre points (i.e. thermal runaway battery data) of the collected battery characteristic elements. The Gaussian filter is a low-pass filter with a frequency domain width of±200w and a threshold ofl/100. For example, a mean filter is used for approximation, and a reaction function is as follows:
As shown in Fig. 6, an image of probability density functions in normal distribution in a frequency domain is shown. a2 represents a normal distribution variance, and the larger a is, the wider a graph is, the lower a peak value is; x is an argument, representing that an input time series is converted to a main frequency (i.e. a frequency peak value) in a frequency domain through Fourier; because a thermal runaway battery signal at a high frequency needs to be filtered, the Gaussian filter is employed; and as the main frequency increases, a weight g(x) is lower and approaches zero, and then filtering may be completed. In the step Sl, training the prediction ability includes: the fault-free battery characteristic elements in time series are employed to train the long-short term memory network, the time convolutional network and the GRU neural network, and coupling is conducted to obtain a prediction result, i.e. a change trend value of the battery characteristic elements over time; and then respective loss values of three networks are returned, and training is continuously conducted. Specifically, as shown in Figs. 3-5, Fig. 3 is a structural diagram of a long-short term memory network (LSTM). A structure in an A frame in Fig. 3 represents a cellular state (ct); a structure in a B frame represents a forget gate (ft), wherein C is a sigmoid function, and output ft is 0-1, representing a forget gate probability; a structure in a C frame represents an input gate, wherein a is still a sigmoid function, tanh is an activation function, and the cellular state is upgraded by combining the functions; a structure in a D frame represents cell renewing, wherein an input cellular state is multiplied by a forget gate output, and then a product is added to a product of it and at of the input gate; and a structure in an E frame represents a hidden state ht which is a product by multiplying a hidden state at a last node by an input signal at a current node through the sigmoid function and the cellular state.
The loss function of the LSTM is an MSE loss function, loss(xry) = (x - Y) I wherein xi is a prediction element, yi is an actual element output prediction value, and xi and yi may be returned in a form of a square of a difference.
The long-short term memory network has the following advantages: (1) gradient
vanishing or gradient explosion cannot occur, and connection weights may be
changed at each time step. (2) Functions, such as forget gate, sigmoid and tanh,
participate in internal self-circulation of nerve cells; and as for a system for treating a
time series signal, input of time information of an internal cellular state of the LSTM
is output of a previous neuron, so that the interior of the neuron is self-loop without
manually deciding which information the LSTM should be forgotten or remembered
externally.
Fig. 4 is a structural diagram of a time convolutional network which is a causal
convolutional TCN model. The TCN model includes an input layer, two hidden layers
and an output layer, that is, a value at the t moment of a previous layer only depends
on values at the t moment of a next layer and a previous value. A first bit at each layer
is filled with 0, future data cannot be seen due to strict one way, and the first bit is of a
one-way structure rather than two-way structure and is a strict time series model. The
loss function employs a mean absolute deviation (MAD), as shown in a following
formula:
MAD= oIx- m-).
wherein xi is a true value, and m(x) is a prediction value. The time convolutional network has the following characteristics: (1) With the natural characteristics of the time convolutional network, the TCN can extract different characteristic values and characteristic vectors in the time series signal. (2) The time convolutional network has the parallelism of data processing. When a time series signal is given, the TCN may conduct parallel processing on the time series signal without sequential processing like RNN. (3) The time convolutional network also rarely has the problem of gradient vanishing or gradient explosion (in the signal processing process of a traditional neural network, because each layer has the effect of the activation function, some feature weights may be continuously increased with increase of the amount of layers, while some feature weights may be continuously reduced); and because a convolution kernel is mainly employed for convolution, a fully connected layer can uniformly conduct weighted mapping on the extracted different characteristic values only. (4) A consuming memory is smaller. When the RNN is used, information at each step needs to be stored, which may consume a great quantity of memory; and the TCN shares the convolution kernel in one layer, and memory usage is lower.
Fig. 5 is a structural diagram of a GRU neural network. Compared with the
LSTM, in the GRU, the forget gate and the input gate are synthesized into one single
updating gate. Similarly, the cellular state and the hidden state are mixed and some
other changes are added. A final model is simpler than a standard LSTM model. The
loss function also employs the MSE function. The GRU is one of the LSTM. The forget gate and the input gate are combined into one updating gate, so that many coefficients and computing requirements are reduced. Although the accuracy is not as good as that of the LSTM, computing is rapid, and errors caused by too much computing of the LSTM may be avoided. Thus, the three networks are combined for equal weight coupling, so that the result may be corrected, deviation of coefficients trained by respective models is reduced, the training period can also be shortened, and a prediction result is more accurate. The present invention further provides a lithium ion battery thermal runaway grading early warning system which includes: a data collecting module, used for collecting characteristic element data of a battery; a thermal runaway model building module, used for building a long-short term memory network, a time convolutional network and the GRU neural network into a prediction model associated with the battery characteristic elements in parallel, and employing the fault-free battery characteristic elements to train the prediction ability of the prediction model; and a thermal runaway early warning module, used for inputting the battery characteristic elements of a to-be-monitored battery in time series into the prediction model to obtain prediction values at a certain moment, and then comparing the prediction values with collected true values corresponding to the moment to obtain a thermal runaway early warning result. The lithium ion battery thermal runaway grading early warning system further includes an equal weight coupling module used for conducting equal weight coupling on the prediction values of the long-short term memory network, the time convolutional network and the GRU neural network. Thus, the result may be corrected, the deviation of coefficients trained by respective models is reduced, the training period can also be shortened, and the prediction result is more accurate.
To sum up, for the thermal runaway grading early warning method and early warning system provided by the present invention, by employing the long-short term memory network, the time convolutional network and the GRU neural network for prediction respectively and conducting equal weight coupling, respective advantages are combined, the problem of overlarge deviation of coefficients trained by respective models is avoided, and the training period is shortened, so that high-precision battery thermal runaway grading detection is achieved, and speed and precision of lithium ion battery thermal runaway early warning are improved; and moreover, the battery working state can be predicted in time series in advance, and a real sense of thermal runaway early warning is achieved. It will be understood that the terms "comprise" and "include" and any of their derivatives (eg comprises, comprising, includes, including) as used in this specification is to be taken to be inclusive of features to which the term refers, and is not meant to exclude the presence of any additional features unless otherwise stated or implied The reference to any prior art in this specification is not, and should not be taken as, an acknowledgement or any form of suggestion that such prior art forms part of the common general knowledge. The above embodiments are only used to illustrate the technical solution of the present invention, not to limit the technical solution. Although the present invention is described in detail by referring to the preferred embodiments, those ordinary skilled in the art shall understand that the technical solution of the present invention can be amended, or replaced equivalently without departing from the spirit and the scope of the technical solution of the present invention.
Claims (10)
1. A lithium battery thermal runaway multi-level detection and early warning system, comprising the following steps: SI. collecting fault-free battery characteristic elements, and respectively training prediction abilities of the long-short term memory network, the time convolutional network and the GRU neural network for the battery characteristic elements in time series; S2. collecting the battery characteristic elements of a to-be-monitored battery in time series, inputting the battery characteristic elements into the trained long-short term memory network, time convolutional network and GRU neural network respectively to obtain three groups of prediction values, and conducting equal weight coupling on the three groups of prediction values to obtain values as fault-free prediction values; and S3. continuously collecting the battery characteristic elements of the to-be-monitored battery in time series as true values, comparing the true values with the fault-free prediction values obtained in the step S2, and obtaining a result that the lager differences are, the larger a battery thermal runaway risk is.
2. The lithium battery thermal runaway multi-level detection and early warning system according to claim 1, wherein the battery characteristic elements comprise but are not limited to one or more of a temperature state, a current state, a voltage state and a charge state of the battery.
3. The lithium battery thermal runaway multi-level detection and early warning system according to claim 1, wherein in the step Sl, the fault-free battery characteristic elements are obtained by filtering off the battery characteristic elements of thermal runaway with a Gaussian filter.
4. The lithium battery thermal runaway multi-level detection and early warning system according to claim 3, wherein the Gaussian filter is a low-pass filter with a
frequency domain width of ±200w and a threshold of 1/100.
5. The lithium battery thermal runaway multi-level detection and early warning system according to claim 1, wherein in the step S, training the prediction ability comprises the following steps: employing the fault-free battery characteristic elements in time series to train the long-short term memory network, the time convolutional network and the GRU neural network, and conducting coupling to obtain a prediction result, i.e. a change trend value of the battery characteristic elements over time; and then returning respective loss values of three networks, and continuously conducting training.
6. The lithium battery thermal runaway multi-level detection and early warning system according to claim 5, wherein the loss values of the long-short term memory network and the GRU neural network are obtained through an MSE loss function; and the loss value of the time convolutional network is obtained through a mean absolute deviation (MAD).
7. The lithium battery thermal runaway multi-level detection and early warning system according to claim 1, wherein the step S3 further comprises the following steps: taking a ratio of an absolute value of a difference between each true value and the corresponding fault-free prediction value to the corresponding fault-free prediction value as a thermal runaway early warning coefficient; and the larger the thermal runaway early warning coefficient is, the lager the battery thermal runaway risk is.
8. The lithium battery thermal runaway multi-level detection and early warning system according to claim 7, wherein the step S3 further comprises the following steps: dividing the thermal runaway early warning coefficients into a plurality of grade intervals for grading early warning, wherein the larger an upper limit value of a grade interval in which thermal runaway is located is, the larger the battery thermal runaway risk is.
9. A lithium ion battery thermal runaway grading early warning system, comprising: a data collecting module, used for collecting characteristic element data of a battery; a thermal runaway model building module, used for building a long-short term memory network, a time convolutional network and the GRU neural network into a prediction model associated with the battery characteristic elements in parallel, and employing the fault-free battery characteristic elements to train the prediction ability of the prediction model; and a thermal runaway early warning module, used for inputting the battery characteristic elements of a to-be-monitored battery in time series into the prediction model to obtain prediction values at a certain moment, and then comparing the prediction values with collected true values corresponding to the moment to obtain a thermal runaway early warning result.
10. The lithium ion battery thermal runaway grading early warning system according to claim 9, wherein the lithium ion battery thermal runaway grading early warning system further comprises an equal weight coupling module used for conducting equal weight coupling on the prediction values of the long-short term memory network, the time convolutional network and the GRU neural network.
Obtaining a plurality of element data related to a fault-free lithium battery pack v 2021102629
Filtering data of a fault battery pack at a high frequency with a Gaussian filter to form a training set
v Returning respective loss values of prediction values obtained by coupling with a TCN model, a GRU model and an LSTM model respectively, and continuously conducting training
Fig. 1
Training module Prediction module
> TCN
Data Gaussian > LSTM ^ Equal weight acquisition Filter coupling output
> GRU V Difference percentage grading
Fig. 2
1/3
Applications Claiming Priority (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110435874.X | 2021-04-22 | ||
| CN202110435874.XA CN113344024B (en) | 2021-04-22 | 2021-04-22 | Lithium ion battery thermal runaway grading early warning method and early warning system |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| AU2021102629A4 true AU2021102629A4 (en) | 2021-07-08 |
Family
ID=76662634
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| AU2021102629A Ceased AU2021102629A4 (en) | 2021-04-22 | 2021-05-17 | Lithium battery thermal runaway multi-level detection and early warning system and judgment method |
Country Status (2)
| Country | Link |
|---|---|
| CN (1) | CN113344024B (en) |
| AU (1) | AU2021102629A4 (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113591404A (en) * | 2021-09-29 | 2021-11-02 | 杭州宇谷科技有限公司 | Battery abnormity detection system and method based on deep learning |
| CN113702836A (en) * | 2021-07-23 | 2021-11-26 | 国家电网有限公司西北分部 | Lithium ion battery state-of-charge estimation method based on EMD-GRU |
| CN115034312A (en) * | 2022-06-14 | 2022-09-09 | 燕山大学 | Fault diagnosis method for dual neural network model satellite power supply system |
| CN115032548A (en) * | 2022-05-25 | 2022-09-09 | 广州汽车集团股份有限公司 | Early warning method and system for automobile storage battery |
| CN117346946A (en) * | 2023-11-29 | 2024-01-05 | 宁德时代新能源科技股份有限公司 | Air pressure sampling circuit, method, battery management system and power utilization device |
Families Citing this family (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113962452A (en) * | 2021-10-15 | 2022-01-21 | 佛山科学技术学院 | Thermal runaway spread prediction method and prediction system |
| CN114460481A (en) * | 2022-01-27 | 2022-05-10 | 重庆邮电大学 | Energy storage battery thermal runaway early warning method based on Bi-LSTM and attention mechanism |
| WO2023159638A1 (en) * | 2022-02-28 | 2023-08-31 | 宁德时代新能源科技股份有限公司 | Battery fault detection system, method and apparatus |
| CN115294745B (en) * | 2022-05-23 | 2023-09-29 | 电子科技大学 | Lithium battery thermal runaway layered early warning method based on neural network and data difference |
| CN115184808B (en) * | 2022-07-05 | 2024-06-18 | 东莞新能安科技有限公司 | Battery thermal runaway risk detection method, device, equipment and computer storage medium |
| CN115420401A (en) * | 2022-08-25 | 2022-12-02 | 上海玫克生储能科技有限公司 | Early warning method, early warning system, storage medium and electronic equipment |
| CN117388725B (en) * | 2023-03-13 | 2024-07-05 | 中国石油大学(华东) | Early abnormality early warning method based on expansion force of lithium ion battery |
| CN116430245A (en) * | 2023-06-14 | 2023-07-14 | 威海谱跃光电科技有限公司 | Battery thermal runaway prediction method based on gradient optimization multi-physical information neural network |
| CN116720073A (en) * | 2023-08-10 | 2023-09-08 | 江苏金恒信息科技股份有限公司 | Abnormality detection extraction method and system based on classifier |
| CN117691227B (en) * | 2024-02-04 | 2024-04-26 | 江苏林洋亿纬储能科技有限公司 | Method and system for safety pre-warning of battery energy storage system and computing device |
| CN117895119B (en) * | 2024-03-13 | 2024-07-09 | 天芯电子科技(山东)有限公司 | Lithium battery thermal runaway early warning system and method |
Family Cites Families (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2019017991A1 (en) * | 2017-07-21 | 2019-01-24 | Quantumscape Corporation | Predictive model for estimating battery states |
| CN108416103A (en) * | 2018-02-05 | 2018-08-17 | 武汉大学 | A kind of method for diagnosing faults of electric automobile of series hybrid powder AC/DC convertor |
| CN108519556A (en) * | 2018-04-13 | 2018-09-11 | 重庆邮电大学 | A kind of lithium ion battery SOC prediction techniques based on Recognition with Recurrent Neural Network |
| CN111999648A (en) * | 2020-08-20 | 2020-11-27 | 浙江工业大学 | Lithium battery residual life prediction method based on long-term and short-term memory network |
-
2021
- 2021-04-22 CN CN202110435874.XA patent/CN113344024B/en active Active
- 2021-05-17 AU AU2021102629A patent/AU2021102629A4/en not_active Ceased
Cited By (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113702836A (en) * | 2021-07-23 | 2021-11-26 | 国家电网有限公司西北分部 | Lithium ion battery state-of-charge estimation method based on EMD-GRU |
| CN113702836B (en) * | 2021-07-23 | 2023-08-18 | 国家电网有限公司西北分部 | Lithium ion battery state of charge estimation method based on EMD-GRU |
| CN113591404A (en) * | 2021-09-29 | 2021-11-02 | 杭州宇谷科技有限公司 | Battery abnormity detection system and method based on deep learning |
| CN113591404B (en) * | 2021-09-29 | 2021-12-07 | 杭州宇谷科技有限公司 | Battery abnormity detection system and method based on deep learning |
| CN115032548A (en) * | 2022-05-25 | 2022-09-09 | 广州汽车集团股份有限公司 | Early warning method and system for automobile storage battery |
| CN115034312A (en) * | 2022-06-14 | 2022-09-09 | 燕山大学 | Fault diagnosis method for dual neural network model satellite power supply system |
| CN115034312B (en) * | 2022-06-14 | 2023-01-06 | 燕山大学 | Fault diagnosis method for dual neural network model satellite power system |
| CN117346946A (en) * | 2023-11-29 | 2024-01-05 | 宁德时代新能源科技股份有限公司 | Air pressure sampling circuit, method, battery management system and power utilization device |
| CN117346946B (en) * | 2023-11-29 | 2024-05-03 | 宁德时代新能源科技股份有限公司 | Air pressure sampling circuit, method, battery management system and power utilization device |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113344024B (en) | 2022-04-12 |
| CN113344024A (en) | 2021-09-03 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| AU2021102629A4 (en) | Lithium battery thermal runaway multi-level detection and early warning system and judgment method | |
| CN108847686B (en) | Photovoltaic inverter fault prediction method | |
| CN107656216A (en) | A kind of lead-acid accumulator on-line monitoring is safeguarded and early warning system and performance estimating method | |
| CN108960321B (en) | Battery fault prediction method for large lithium battery energy storage power station | |
| CN104732276A (en) | On-line diagnosing method for faults of metering production facility | |
| CN116626505B (en) | Battery pack consistency anomaly detection method based on Internet of vehicles big data | |
| CN109002781B (en) | Fault prediction method for energy storage converter | |
| Li et al. | The state-of-charge predication of lithium-ion battery energy storage system using data-driven machine learning | |
| Diao et al. | Research on Electric Vehicle Charging Safety Warning Based on A-LSTM Algorithm | |
| Wang et al. | Charging diagnosis of power battery based on adaptive STCKF and BLS for electric vehicles | |
| Xu et al. | Safety warning analysis for power battery packs in electric vehicles with running data | |
| Pang et al. | A Review on the Prediction of Health State and Serving Life of Lithium‐Ion Batteries | |
| Zhang et al. | Lithium battery SOC prediction based on mproved BP eural etwork algorithm | |
| CN102095953B (en) | A kind of performance of accumulator charger online test method | |
| Guo et al. | An unscented kalman filtering method for estimation of state-of-charge of lithium-ion battery | |
| Zhang et al. | Online state-of-health estimation for the lithium-ion battery based on an LSTM neural network with attention mechanism | |
| Zhang et al. | Research for SOC prediction of lithium battery based on GA-ESN | |
| Zhang et al. | Remaining useful life prediction of lithium-ion batteries based on TCN-DCN fusion model combined with IRRS filtering | |
| CN116400224A (en) | Battery remaining service life prediction method based on working temperature correction | |
| Liu et al. | Battery fault diagnosis scheme based on improved RBF neural network | |
| Mehta et al. | A Comprehensive study of Machine Learning Techniques used for estimating State of Charge for Li-ion Battery | |
| LU500227B1 (en) | A system and a method for smart grading and early warning the thermal runaway of a lithium-ion battery | |
| Liu et al. | Internal short circuit fault diagnosis based on 1DVCNN in lithium battery storage system | |
| CN115730525A (en) | Rail transit UPS storage battery health state prediction method | |
| Wang et al. | The Application Research of Deep Learning Based Method for Short-term Wind Speed Forecasting |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| FGI | Letters patent sealed or granted (innovation patent) | ||
| MK22 | Patent ceased section 143a(d), or expired - non payment of renewal fee or expiry |