CN117214722A - Battery health degree prediction method and device, electronic equipment and storage medium - Google Patents
Battery health degree prediction method and device, electronic equipment and storage medium Download PDFInfo
- Publication number
- CN117214722A CN117214722A CN202311192098.0A CN202311192098A CN117214722A CN 117214722 A CN117214722 A CN 117214722A CN 202311192098 A CN202311192098 A CN 202311192098A CN 117214722 A CN117214722 A CN 117214722A
- Authority
- CN
- China
- Prior art keywords
- battery
- module
- time sequence
- charging
- battery health
- 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
- 238000000034 method Methods 0.000 title claims abstract description 90
- 230000036541 health Effects 0.000 title claims abstract description 87
- 238000000354 decomposition reaction Methods 0.000 claims abstract description 79
- 239000000523 sample Substances 0.000 claims description 40
- 230000008569 process Effects 0.000 claims description 35
- 238000004590 computer program Methods 0.000 claims description 16
- 230000015654 memory Effects 0.000 claims description 9
- 230000008859 change Effects 0.000 abstract description 7
- 238000005516 engineering process Methods 0.000 abstract description 5
- 230000007246 mechanism Effects 0.000 abstract description 4
- 238000012549 training Methods 0.000 description 11
- 230000004927 fusion Effects 0.000 description 10
- WHXSMMKQMYFTQS-UHFFFAOYSA-N Lithium Chemical compound [Li] WHXSMMKQMYFTQS-UHFFFAOYSA-N 0.000 description 9
- 230000000694 effects Effects 0.000 description 9
- 229910052744 lithium Inorganic materials 0.000 description 9
- 238000012545 processing Methods 0.000 description 7
- 238000010586 diagram Methods 0.000 description 6
- 238000004891 communication Methods 0.000 description 5
- 238000004146 energy storage Methods 0.000 description 5
- 238000012360 testing method Methods 0.000 description 4
- 230000009286 beneficial effect Effects 0.000 description 3
- 230000002159 abnormal effect Effects 0.000 description 2
- 238000013480 data collection Methods 0.000 description 2
- 238000012417 linear regression Methods 0.000 description 2
- 238000010801 machine learning Methods 0.000 description 2
- 238000010606 normalization Methods 0.000 description 2
- 230000003287 optical effect Effects 0.000 description 2
- 238000013450 outlier detection Methods 0.000 description 2
- 238000007781 pre-processing Methods 0.000 description 2
- 230000003044 adaptive effect Effects 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 238000007599 discharging Methods 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 238000012423 maintenance Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 238000011112 process operation Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000006403 short-term memory Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Landscapes
- Charge And Discharge Circuits For Batteries Or The Like (AREA)
Abstract
The invention provides a battery health degree prediction method, a device, electronic equipment and a storage medium, wherein the method comprises the following steps: acquiring charging characteristic data of a target battery in a historical time period; predicting the future battery health of the target battery based on the charging characteristic data of the target battery in the historical time period and a converter model with fused time sequence decomposition; the converter model is trained based on sample charging characteristic data with battery health labels. According to the battery health degree prediction method, the device, the electronic equipment and the storage medium, the change trend of the battery health state is better revealed by introducing the time sequence decomposition technology, and the nonlinear relation and long-distance dependence of battery data are captured by fully utilizing the self-attention mechanism of the transducer model, so that the accuracy of a prediction result is improved.
Description
Technical Field
The present invention relates to the field of battery technologies, and in particular, to a method and apparatus for predicting battery health, an electronic device, and a storage medium.
Background
Currently, the use of lithium batteries in various fields, such as electric vehicles, energy storage, wearable devices, and the like, is in an increasing trend. However, the continuous attenuation of the capacity of a lithium battery during the use process is a puzzled problem, and monitoring and predicting the health of the lithium battery at any time is a main path for solving the problem.
Currently, the prediction method for the health state of a lithium battery is roughly divided into two directions, namely, feature-based prediction and data-driven prediction. The prediction based on the characteristics has the problems of low precision, poor robustness and the like; however, the data-driven prediction method represented by machine learning is a research hotspot in the field at present due to flexibility and accuracy. At present, a linear regression or Long Short-Term Memory (LSTM) model is mainly used for predicting the health state of a battery by a machine learning model, but the linear regression model has some problems, such as faster prediction, stronger model interpretation and weak fitting capability; LSTM models are relatively strong in fitting ability, but when processing long sequences, it is difficult to capture long-range dependencies and prediction efficiency is low due to their parallel limitations.
Disclosure of Invention
Aiming at the problems existing in the prior art, the invention provides a battery health degree prediction method, a device, electronic equipment and a storage medium.
In a first aspect, the present invention provides a method for predicting the health of a battery, including:
acquiring charging characteristic data of a target battery in a historical time period;
predicting the future battery health of the target battery based on the charging characteristic data of the target battery in the historical time period and a converter model with fused time sequence decomposition;
the converter model is trained based on sample charging characteristic data with battery health labels.
In some embodiments, the transducer model includes an encoder module and a decoder module;
each encoder comprises a first self-attention module and a first time sequence decomposition module, wherein the first time sequence decomposition module is used for decomposing a first trend item for the characteristics output by the first self-attention module, and the first trend item is used as the output of the encoder;
each decoder comprises a second self-attention module, a second time sequence decomposition module, a third self-attention module and a third time sequence decomposition module which are sequentially connected, wherein the second time sequence decomposition module is used for decomposing the characteristics output by the second self-attention module into a second trend item and a second fluctuation item, outputting the second fluctuation item to the third self-attention module, and the third time sequence decomposition module is used for decomposing the characteristics output by the third self-attention module into a third trend item and a third fluctuation item, and the third trend item, the third fluctuation item and the second trend item are spliced and then are used as the output of the decoder;
the transducer model further comprises a fourth time sequence decomposition module, wherein the fourth time sequence decomposition module is used for performing time sequence decomposition on the input data of the encoder module and outputting the obtained fluctuation item data to the decoder module.
In some embodiments, the encoder module comprises 4 of the encoders and the decoder module comprises 2 of the decoders.
In some embodiments, the charging profile includes at least one of the following profile during each active charging:
the method comprises the steps of data acquisition time, the highest voltage of each single battery at each data acquisition time, the lowest voltage of each single battery at each data acquisition time, the highest temperature of each temperature probe at each data acquisition time, the lowest temperature of each temperature probe at each data acquisition time, the time period for charging the residual electric quantity from 40% to 90% and the residual electric quantity at the charging starting time.
In some embodiments, the effective charging process is a charging process that includes a process in which the remaining charge of the battery is charged from 40% to 90%.
In some embodiments, the battery health corresponding to the sample charging characteristic data is characterized by an amount of charge in the process of charging from 40% to 90% based on a remaining amount of the battery.
In a second aspect, the present invention also provides a battery health degree prediction apparatus, including:
the acquisition module is used for acquiring charging characteristic data of the target battery in a historical time period;
the prediction module is used for predicting the future battery health of the target battery based on the charging characteristic data of the target battery in the historical time period and a converter model with fused time sequence decomposition;
the converter model is trained based on sample charging characteristic data with battery health labels.
In some embodiments, the transducer model includes an encoder module and a decoder module;
each encoder comprises a first self-attention module and a first time sequence decomposition module, wherein the first time sequence decomposition module is used for decomposing a first trend item for the characteristics output by the first self-attention module, and the first trend item is used as the output of the encoder;
each decoder comprises a second self-attention module, a second time sequence decomposition module, a third self-attention module and a third time sequence decomposition module which are sequentially connected, wherein the second time sequence decomposition module is used for decomposing the characteristics output by the second self-attention module into a second trend item and a second fluctuation item, outputting the second fluctuation item to the third self-attention module, and the third time sequence decomposition module is used for decomposing the characteristics output by the third self-attention module into a third trend item and a third fluctuation item, and the third trend item, the third fluctuation item and the second trend item are spliced and then are used as the output of the decoder;
the transducer model further comprises a fourth time sequence decomposition module, wherein the fourth time sequence decomposition module is used for performing time sequence decomposition on the input data of the encoder module and outputting the obtained fluctuation item data to the decoder module.
In some embodiments, the encoder module comprises 4 of the encoders and the decoder module comprises 2 of the decoders.
In some embodiments, the charging profile includes at least one of the following profile during each active charging:
the method comprises the steps of data acquisition time, the highest voltage of each single battery at each data acquisition time, the lowest voltage of each single battery at each data acquisition time, the highest temperature of each temperature probe at each data acquisition time, the lowest temperature of each temperature probe at each data acquisition time, the time period for charging the residual electric quantity from 40% to 90% and the residual electric quantity at the charging starting time.
In some embodiments, the effective charging process is a charging process that includes a process in which the remaining charge of the battery is charged from 40% to 90%.
In some embodiments, the battery health corresponding to the sample charging characteristic data is characterized by an amount of charge in the process of charging from 40% to 90% based on a remaining amount of the battery.
In a third aspect, the present invention also provides an electronic device, including a memory, a processor, and a computer program stored on the memory and running on the processor, the processor implementing the battery health prediction method according to the first aspect as described above when executing the program.
In a fourth aspect, the present invention also provides a non-transitory computer readable storage medium having stored thereon a computer program which, when executed by a processor, implements the battery health prediction method according to the first aspect as described above.
In a fifth aspect, the present invention also provides a computer program product comprising a computer program which, when executed by a processor, implements the battery health prediction method according to the first aspect as described above.
According to the battery health degree prediction method, the device, the electronic equipment and the storage medium, the change trend of the battery health state is better revealed by introducing the time sequence decomposition technology, and the nonlinear relation and long-distance dependence of battery data are captured by fully utilizing the self-attention mechanism of the transducer model, so that the accuracy of a prediction result is improved.
Drawings
In order to more clearly illustrate the invention or the technical solutions in the related art, the following description will briefly explain the drawings used in the embodiments or the related art description, and it is obvious that the drawings in the following description are some embodiments of the invention, and other drawings can be obtained according to the drawings without inventive effort for those skilled in the art.
FIG. 1 is a schematic flow chart of a battery health prediction method according to the present invention;
FIG. 2 is an exemplary diagram of a fusion time-series decomposed transducer model provided by the present invention;
fig. 3 is a flowchart for predicting the state of health of a lithium battery according to the present invention;
FIG. 4 is a graph showing the comparison of the effect of the fusion time series decomposed transducer model and the LSTM model on the test set;
fig. 5 is a schematic structural diagram of a battery health prediction device according to the present invention;
fig. 6 is a schematic structural diagram of an electronic device provided by the present invention.
Detailed Description
The term "and/or" in the present invention describes an association relationship of association objects, which means that three relationships may exist, for example, a and/or B may mean: a exists alone, A and B exist together, and B exists alone. The character "/" generally indicates that the context-dependent object is an "or" relationship.
The term "plurality" in the present invention means two or more, and other adjectives are similar thereto.
The terms "first," "second," and the like, herein, are used for distinguishing between similar objects and not necessarily for describing a particular sequential or chronological order. It is to be understood that the terms so used are interchangeable under appropriate circumstances such that the embodiments of the invention are capable of operation in sequences other than those illustrated or otherwise described herein, and that the "first" and "second" distinguishing between objects generally are not limited in number to the extent that the first object may, for example, be one or more.
For the purpose of making the objects, technical solutions and advantages of the present invention more apparent, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings, and it is apparent that the described embodiments are some embodiments of the present invention, not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
Fig. 1 is a schematic flow chart of a battery health prediction method provided by the invention, as shown in fig. 1, the method comprises the following steps:
step 100, acquiring charging characteristic data of a target battery in a historical time period.
And step 101, predicting the future battery health of the target battery based on the charging characteristic data of the target battery in the historical time period and a converter model of fusion time sequence decomposition.
The transducer model is trained based on sample charging characteristic data with a battery health label.
Specifically, the main execution body of each step in the method may be a battery health degree prediction device, and the device may be implemented by software and/or hardware, and the device may be integrated in an electronic device, where the electronic device may be a terminal device (such as a smart phone, a personal computer, etc.), a server (such as a local server or a cloud server, a server cluster, etc.), a processor, a chip, etc.
The battery in the present invention may be a lithium battery or other type of battery, may be a battery for an electric automobile or a battery for other fields, and the present invention is not limited thereto. The battery in the present invention may be constituted by a plurality of unit cells.
According to the invention, by adopting the transducer model, the self-attention mechanism and the parallel processing capability of the transducer model are fully utilized, and the nonlinear relation and long-distance dependence in the battery performance data can be captured, so that the accuracy, the efficiency and the generalization capability of the prediction model are improved. On the basis, the invention also introduces a time sequence decomposition (time sequence decomposition for short) technology, and the change trend under different time scales is revealed by decomposing the battery health state change data into trend items and fluctuation items, so that a more stable basis is provided for accurate prediction.
Before the health degree prediction is performed on the target battery (i.e. the battery requiring the prediction of the health degree of the battery), the charging data of a large number of batteries collected in a history can be preprocessed to form a training data set, the training data set comprises sample charging characteristic data of different histories and battery health degree labels corresponding to the charging characteristic data of each sample, the training data set can be used for training a fusion time sequence decomposed transducer model, and the training process can refer to the traditional transducer model and is not repeated herein.
The trained converter model can be used for predicting the battery health (or battery health state) of the target battery, and the battery health predicting device can correspondingly acquire the charging characteristic data of the target battery in a historical time period according to the sample charging characteristic data used in the process of training the converter model, and then predict the future battery health of the target battery based on the charging characteristic data of the target battery in the historical time period and the trained converter model. For example, the battery health of a battery in the next 3 months can be predicted using a trained transducer model based on the charging characteristics data of that battery over the last 1 year.
In some embodiments, the charging profile includes at least one of the following profile during each active charging:
(1) And (5) data acquisition time. I.e. the moment of each data acquisition, for example, charging data may be acquired every 20 s.
(2) The highest voltage of each single battery at each data acquisition time. The highest voltage of each single battery at a certain data acquisition time refers to the maximum value among the voltage values of each single battery acquired at the data acquisition time.
(3) The lowest voltage of each single battery at each data acquisition time. The minimum voltage of each single battery at a certain data acquisition time refers to the minimum value among the voltage values of each single battery acquired at the data acquisition time.
(4) The highest temperature of each temperature probe at each data acquisition time. The maximum temperature of each temperature probe at a certain data acquisition time is the maximum value among the temperature values of each temperature probe acquired at the data acquisition time.
(5) The lowest temperature of each temperature probe at each data acquisition time. The minimum temperature of each temperature probe at a certain data acquisition time is the minimum value among the temperature values of each temperature probe acquired at the data acquisition time.
(6) The remaining charge is charged from 40% to 90% for the duration. The remaining Charge is the State of Charge (SOC).
(7) The remaining capacity at the charge start time.
In some embodiments, the extraction of the charging characteristic data may be performed by:
firstly, collecting battery charge and discharge process operation data, extracting characteristics capable of reflecting battery health state change from the data, and comprising the following steps: data time, temperature values detected by each temperature probe of the chargeable energy storage subsystem, chargeable energy storage device current, each single battery voltage, battery SOC and the like.
Then, preprocessing the collected data, including filling the missing value by interpolation, processing the abnormal value by outlier detection, and the like, and further extracting useful features to finally obtain charging feature data, including: the data acquisition time, the highest voltage of each single battery, the lowest voltage of each single battery, the highest temperature of each temperature probe, the lowest temperature of each temperature probe, the time period for charging the SOC from 40% to 90%, the charge start SOC and the like.
In some embodiments, the effective charging process is a charging process that includes a process in which the remaining charge of the battery is charged from 40% to 90%. That is, only the process that the residual capacity of the battery is completely charged from 40% to 90% is the charging process considered to be an effective charging process, otherwise, the charging process considered to be an ineffective charging process is the data considered to be ineffective, and the data of the ineffective charging process are not used for model training and battery health prediction.
In some embodiments, the data collection time in the above charging characteristic data is a data collection time in a period from 40% to 90% of the remaining power of the battery.
In some embodiments, the battery health corresponding to the sample charging characteristic data is characterized by the amount of charge charged during a 40% to 90% charge based on the remaining amount of the battery. For example, the electric quantity charged in the process of charging the battery SOC from 40% to 90% in the process of effective charging can be calculated by an ampere-hour integral (i.e. the integral of charging current with respect to time) method according to the data of charging current and charging time in the process of effective charging, and the electric quantity is used as the battery health label of the charging characteristic data corresponding to the effective charging process.
A conventional measure of the health of a battery is SOH (state of health of the battery), i.e. the ratio of the current capacity of the battery to the initial capacity. According to the invention, the accumulated charge quantity S (Ah) from 40% to 90% of the battery SOC is measured, and the attenuation of the battery health state is represented by the attenuation of S, so that the battery health state can be more accurately quantified, and a basis is provided for making an optimal strategy in aspects of battery use, maintenance, and the like.
According to the battery health degree prediction method provided by the invention, the change trend of the battery health state is better revealed by introducing a time sequence decomposition technology, and the nonlinear relation and long-distance dependence of battery data are captured by fully utilizing the self-attention mechanism of the transducer model, so that the accuracy of a prediction result is improved.
In some embodiments, the transducer model includes an encoder module and a decoder module;
the encoder comprises an encoder module, a first self-attention module and a first time sequence decomposition module, wherein the first time sequence decomposition module is used for decomposing a first trend item for the characteristics output by the first self-attention module, and the first trend item is used as the output of the encoder;
in the decoder modules, each decoder comprises a second self-attention module, a second time sequence decomposition module, a third self-attention module and a third time sequence decomposition module which are sequentially connected, wherein the second time sequence decomposition module is used for decomposing the characteristics output by the second self-attention module into a second trend item and a second fluctuation item, outputting the second fluctuation item to the third self-attention module, and the third time sequence decomposition module is used for decomposing the characteristics output by the third self-attention module into a third trend item and a third fluctuation item, and the third trend item, the third fluctuation item and the second trend item are spliced to be used as the output of the decoder;
the transducer model further comprises a fourth time sequence decomposition module, wherein the fourth time sequence decomposition module is used for performing time sequence decomposition on the input data of the encoder module and outputting the obtained fluctuation item data to the decoder module.
In some embodiments, the encoder module may include 4 encoders and the decoder module may include 2 decoders to obtain the best balance between model size and prediction accuracy.
FIG. 2 is a diagram of an exemplary fusion time sequence decomposition converter model according to the present invention, and as shown in FIG. 2, the converter model mainly improves the structure of an encoder and a decoder based on a conventional converter model, in the encoder, the original characteristics output by the self-attention module are not directly input to the next encoder or decoder module, but are improved to be input to the time sequence decomposition module connected with the encoder or decoder module, and after the time sequence decomposition module decomposes the trend item, the trend item is input to the next encoder or decoder module; in the decoder, similarly, the original characteristic output by the 1 st self-attention module (masked multihead attention) is not directly input to the 2 nd self-attention module (encoder-decoder multihead attention), the characteristic output by the 2 nd self-attention module is not directly input to the next decoder or the subsequent prediction module (linear layer+softmax layer), the characteristic output by the 1 st self-attention module is input to the time sequence decomposition module connected with the decoder, the trend item and the fluctuation item are decomposed by the time sequence decomposition module, the fluctuation item is input to the 2 nd self-attention module, then the characteristic output by the 2 nd self-attention module is input to the time sequence decomposition module connected with the decoder, the trend item and the fluctuation item are decomposed by the time sequence decomposition module, the trend item and the fluctuation item are subjected to splicing processing together with the trend item decomposed by the previous time sequence decomposition module, and the characteristic after splicing processing is input to the next decoder or the subsequent prediction module.
In addition, unlike the conventional transducer model, the input data of the decoder module of the fused time-series resolved transducer model further includes fluctuation item data obtained by time-series resolving the input data of the encoder module, which is input as Q (query) to the self-attention module of the decoder. Namely, input data of the encoder module is input into the encoder module for feature encoding on one hand, and is also input into a time sequence decomposition module for time sequence decomposition on the other hand, and fluctuation item data output by the time sequence decomposition module is input into the decoder module.
For the structures of other parts of the model, the adaptive setting can be performed by referring to a traditional transducer model, which is not described herein.
During model training, the input data of the encoder module is sample charging characteristic data (corresponding to the encoder input set in fig. 2), and the input data of the decoder module correspondingly comprises the time and the battery health label (corresponding to the decoder input set in fig. 2) corresponding to the sample charging characteristic data.
In model prediction, the input data of the encoder module is the historical charging characteristic data of the target battery (corresponding to the encoder input set in fig. 2), and the input data of the decoder module includes the target time to be predicted (corresponding to the decoder input set in fig. 2).
In fig. 2, an Embedding (Embedding) module connected to the encoder input set is used for performing embedded representation on time in the encoder input set, and an Embedding feature corresponding to the time data is spliced with other input data in the encoder input set and then input into the encoder module; the embedding module connected with the decoder input set is used for carrying out embedding representation on time in the decoder input set, and the embedding feature corresponding to the time data is spliced with other input data in the decoder input set and then is input into the decoder module.
The time sequence decomposition module in the invention can be realized by using any model or algorithm capable of performing time sequence decomposition, and the invention is not limited by the specific implementation mode. The time sequence decomposition module is mainly used for decomposing a trend term (trend) and a fluctuation term (or season term, seal) for an input sequence, wherein the trend term refers to a mean value, and the fluctuation term refers to the difference between a sequence value and the mean value.
The above method provided by the invention is illustrated by way of example of a specific application scenario.
Fig. 3 is a flowchart for predicting the state of health of a lithium battery according to the present invention, as shown in fig. 3, the main process includes:
step 1, collecting operation data of a lithium battery charging and discharging process, extracting characteristics capable of reflecting battery health state change from the data, and comprising the following steps: data time, temperature values detected by each temperature probe of the chargeable energy storage subsystem, chargeable energy storage device current, each single battery voltage, and battery SOC (state of charge, i.e., residual electric quantity).
Step 2, preprocessing the acquired data, namely filling the missing value by an interpolation method, processing the abnormal value by an outlier detection method and the like, and further extracting useful characteristics, wherein the final characteristics comprise: data time, highest voltage of each single battery, lowest voltage of each single battery, highest temperature of each temperature probe, lowest temperature of each temperature probe, time taken for charging the SOC from 40% to 90%, and charge start SOC.
And 3, calculating the electric quantity charged in the charging process of 40-90% of each battery SOC according to the definition of the charging current, the charging time and the lithium battery health state, so as to represent the health state (namely the battery health degree) of the battery.
And 4, dividing the data set into a training set and a testing set according to a ratio of 7:3, and carrying out maximum-minimum value normalization processing on the data. The max-min normalization process refers to normalizing the range of values of each feature to [0,1 ] by scaling]Or [ -1,1]Between them. For each dimension of feature x i (i=1, 2,.) the eigenvalue x of the kth sample ik (k=1, 2,) normalized value x i ` k The method comprises the following steps:
wherein min (x i ) And max (x) i ) Features x respectively i Minimum and maximum values over all samples.
And 5, establishing a time sequence decomposition model, wherein the time sequence decomposition model is used for decomposing the input sequence into a trend term (mean value) and a fluctuation term (difference between time sequence values and the mean value).
And 6, establishing a transducer model, and integrating the time sequence decomposition model into a transducer network as a module. The final model comprises four layers of encoders and two layers of decoders, each layer of encoder comprises a self-attention module and a time sequence decomposition module, and each layer of decoder comprises two self-attention modules and two time sequence decomposition modules, and the specific model is shown in fig. 2.
And 7, inputting a training set into a transducer network model as shown in fig. 2 for training, wherein the learning rate is self-adaptively adjusted from 0.0003 according to the fitting condition.
And 8, after a trained model is obtained, inputting a test set to obtain a group of battery health state predicted values, and calculating the deviation between the predicted values and the true values through RMSE (root mean square error) and MAPE (average percentage error) to judge the model predicted effect. Fig. 4 is a graph showing the comparison of the effects of the fusion time series decomposed transducer model and the LSTM model on the test set, and as can be seen from the effect of fig. 4, in the process of 40 times of charging, the charging electric quantity (line 2) predicted by using the transducer model is very close to the actual charging electric quantity (line 3), while the charging electric quantity (line 1) predicted by using the conventional LSTM model is far from the line 3, so that the prediction effect of the fusion time series decomposed transducer model is obviously improved.
The battery health degree prediction device provided by the invention is described below, and the battery health degree prediction device described below and the battery health degree prediction method described above can be referred to correspondingly.
Fig. 5 is a schematic structural diagram of a battery health prediction apparatus according to the present invention, as shown in fig. 5, the apparatus includes:
an acquisition module 500, configured to acquire charging characteristic data of a target battery in a historical period;
the prediction module 510 is configured to predict a future battery health of the target battery based on the charging characteristic data of the target battery in the historical time period and the transform model with the fused time sequence decomposition;
the transducer model is trained based on sample charging characteristic data with a battery health label.
In some embodiments, the transducer model includes an encoder module and a decoder module;
the encoder comprises an encoder module, a first self-attention module and a first time sequence decomposition module, wherein the first time sequence decomposition module is used for decomposing a first trend item for the characteristics output by the first self-attention module, and the first trend item is used as the output of the encoder;
in the decoder modules, each decoder comprises a second self-attention module, a second time sequence decomposition module, a third self-attention module and a third time sequence decomposition module which are sequentially connected, wherein the second time sequence decomposition module is used for decomposing the characteristics output by the second self-attention module into a second trend item and a second fluctuation item, outputting the second fluctuation item to the third self-attention module, and the third time sequence decomposition module is used for decomposing the characteristics output by the third self-attention module into a third trend item and a third fluctuation item, and the third trend item, the third fluctuation item and the second trend item are spliced to be used as the output of the decoder;
the transducer model further comprises a fourth time sequence decomposition module, wherein the fourth time sequence decomposition module is used for performing time sequence decomposition on the input data of the encoder module and outputting the obtained fluctuation item data to the decoder module.
In some embodiments, the encoder module includes 4 encoders and the decoder module includes 2 decoders.
In some embodiments, the charging profile includes at least one of the following profile during each active charging:
the method comprises the steps of data acquisition time, the highest voltage of each single battery at each data acquisition time, the lowest voltage of each single battery at each data acquisition time, the highest temperature of each temperature probe at each data acquisition time, the lowest temperature of each temperature probe at each data acquisition time, the time period for charging the residual electric quantity from 40% to 90% and the residual electric quantity at the charging starting time.
In some embodiments, the effective charging process is a charging process that includes a process in which the remaining charge of the battery is charged from 40% to 90%.
In some embodiments, the battery health corresponding to the sample charging characteristic data is characterized by the amount of charge charged during a 40% to 90% charge based on the remaining amount of the battery.
It should be noted that, the device provided by the present invention can implement all the method steps implemented by the method embodiment and achieve the same technical effects, and the parts and beneficial effects that are the same as those of the method embodiment in the present embodiment are not described in detail herein.
Fig. 6 is a schematic structural diagram of an electronic device according to the present invention, as shown in fig. 6, the electronic device may include: processor 610, communication interface (Communications Interface) 620, memory 630, and communication bus 640, wherein processor 610, communication interface 620, and memory 630 communicate with each other via communication bus 640. The processor 610 may invoke logic instructions in the memory 630 to perform any of the battery health prediction methods provided in the embodiments described above, such as: acquiring charging characteristic data of a target battery in a historical time period; predicting the future battery health of the target battery based on the charging characteristic data of the target battery in the historical time period and a converter model of fusion time sequence decomposition; the transducer model is trained based on sample charging characteristic data with a battery health label.
Further, the logic instructions in the memory 630 may be implemented in the form of software functional units and stored in a computer-readable storage medium when sold or used as a stand-alone product. Based on this understanding, the technical solution of the present invention may be embodied essentially or in a part contributing to the prior art or in a part of the technical solution, in the form of a software product stored in a storage medium, comprising several instructions for causing a computer device (which may be a personal computer, a server, a network device, etc.) to perform all or part of the steps of the method according to the embodiments of the present invention. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read-Only Memory (ROM), a random access Memory (RAM, random Access Memory), a magnetic disk, or an optical disk, or other various media capable of storing program codes.
It should be noted that, the electronic device provided by the present invention can implement all the method steps implemented by the method embodiments and achieve the same technical effects, and the details and beneficial effects of the same parts and advantages as those of the method embodiments in the present embodiment are not described in detail.
In another aspect, the present invention also provides a non-transitory computer readable storage medium, on which a computer program is stored, which when executed by a processor, can implement any of the battery health prediction methods provided in the above embodiments, for example: acquiring charging characteristic data of a target battery in a historical time period; predicting the future battery health of the target battery based on the charging characteristic data of the target battery in the historical time period and a converter model of fusion time sequence decomposition; the transducer model is trained based on sample charging characteristic data with a battery health label.
It should be noted that, the non-transitory computer readable storage medium provided by the present invention can implement all the method steps implemented by the method embodiments and achieve the same technical effects, and detailed descriptions of the same parts and beneficial effects as those of the method embodiments in this embodiment are omitted.
In yet another aspect, the present invention also provides a computer program product comprising a computer program, the computer program being storable on a non-transitory computer readable storage medium, the computer program, when executed by a processor, being capable of performing any of the battery health prediction methods provided in the above embodiments, for example: acquiring charging characteristic data of a target battery in a historical time period; predicting the future battery health of the target battery based on the charging characteristic data of the target battery in the historical time period and a converter model of fusion time sequence decomposition; the transducer model is trained based on sample charging characteristic data with a battery health label.
It should be noted that, the computer program product provided by the present invention can implement all the method steps implemented by the method embodiments and achieve the same technical effects, and the details of the same parts and the advantages as those of the method embodiments in the present embodiment are not described herein.
The apparatus embodiments described above are merely illustrative, wherein the modules illustrated as separate components may or may not be physically separate, and the components shown as modules may or may not be physical units, may be located in one place, or may be distributed over a plurality of network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment. Those of ordinary skill in the art will understand and implement the present invention without undue burden.
From the above description of the embodiments, it will be apparent to those skilled in the art that the embodiments may be implemented by means of software plus necessary general hardware platforms, or of course may be implemented by means of hardware. Based on this understanding, the foregoing technical solution may be embodied essentially or in a part contributing to the prior art in the form of a software product, which may be stored in a computer readable storage medium, such as ROM/RAM, a magnetic disk, an optical disk, etc., including several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute the method described in the respective embodiments or some parts of the embodiments.
Finally, it should be noted that: the above embodiments are only for illustrating the technical solution of the present invention, and are not limiting; although the invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical scheme described in the foregoing embodiments can be modified or some technical features thereof can be replaced by equivalents; such modifications and substitutions do not depart from the spirit and scope of the technical solutions of the embodiments of the present invention.
Claims (10)
1. A battery health prediction method, comprising:
acquiring charging characteristic data of a target battery in a historical time period;
predicting the future battery health of the target battery based on the charging characteristic data of the target battery in the historical time period and a converter model with fused time sequence decomposition;
the converter model is trained based on sample charging characteristic data with battery health labels.
2. The battery health prediction method of claim 1, wherein the transducer model comprises an encoder module and a decoder module;
each encoder comprises a first self-attention module and a first time sequence decomposition module, wherein the first time sequence decomposition module is used for decomposing a first trend item for the characteristics output by the first self-attention module, and the first trend item is used as the output of the encoder;
each decoder comprises a second self-attention module, a second time sequence decomposition module, a third self-attention module and a third time sequence decomposition module which are sequentially connected, wherein the second time sequence decomposition module is used for decomposing the characteristics output by the second self-attention module into a second trend item and a second fluctuation item, outputting the second fluctuation item to the third self-attention module, and the third time sequence decomposition module is used for decomposing the characteristics output by the third self-attention module into a third trend item and a third fluctuation item, and the third trend item, the third fluctuation item and the second trend item are spliced and then are used as the output of the decoder;
the transducer model further comprises a fourth time sequence decomposition module, wherein the fourth time sequence decomposition module is used for performing time sequence decomposition on the input data of the encoder module and outputting the obtained fluctuation item data to the decoder module.
3. The battery health prediction method of claim 2, wherein the encoder module includes 4 of the encoders and the decoder module includes 2 of the decoders.
4. A battery health prediction method according to any one of claims 1 to 3, wherein the charging characteristic data includes at least one of the following characteristic data during each active charging:
the method comprises the steps of data acquisition time, the highest voltage of each single battery at each data acquisition time, the lowest voltage of each single battery at each data acquisition time, the highest temperature of each temperature probe at each data acquisition time, the lowest temperature of each temperature probe at each data acquisition time, the time period for charging the residual electric quantity from 40% to 90% and the residual electric quantity at the charging starting time.
5. The method according to claim 4, wherein the effective charging process is a charging process including a process of charging the remaining capacity of the battery from 40% to 90%.
6. A method according to any one of claims 1 to 3, wherein the battery health corresponding to the sample charge characteristic data is characterized by the amount of charge charged during the period from 40% charge to 90% charge based on the remaining amount of the battery.
7. A battery health degree prediction apparatus, comprising:
the acquisition module is used for acquiring charging characteristic data of the target battery in a historical time period;
the prediction module is used for predicting the future battery health of the target battery based on the charging characteristic data of the target battery in the historical time period and a converter model with fused time sequence decomposition;
the converter model is trained based on sample charging characteristic data with battery health labels.
8. An electronic device comprising a memory, a processor and a computer program stored on the memory and running on the processor, wherein the processor implements the battery health prediction method of any one of claims 1 to 6 when the program is executed by the processor.
9. A non-transitory computer readable storage medium having stored thereon a computer program, wherein the computer program when executed by a processor implements the battery health prediction method according to any of claims 1 to 6.
10. A computer program product comprising a computer program, characterized in that the computer program, when executed by a processor, implements the battery health prediction method according to any one of claims 1 to 6.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311192098.0A CN117214722A (en) | 2023-09-14 | 2023-09-14 | Battery health degree prediction method and device, electronic equipment and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311192098.0A CN117214722A (en) | 2023-09-14 | 2023-09-14 | Battery health degree prediction method and device, electronic equipment and storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN117214722A true CN117214722A (en) | 2023-12-12 |
Family
ID=89040242
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311192098.0A Pending CN117214722A (en) | 2023-09-14 | 2023-09-14 | Battery health degree prediction method and device, electronic equipment and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117214722A (en) |
-
2023
- 2023-09-14 CN CN202311192098.0A patent/CN117214722A/en active Pending
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Li et al. | Lithium-ion battery capacity estimation—A pruned convolutional neural network approach assisted with transfer learning | |
| Kim et al. | Data-driven state of health estimation of li-ion batteries with RPT-reduced experimental data | |
| CN113022378B (en) | Temperature consistency prediction method, temperature consistency prediction device, prediction equipment and storage medium | |
| Zhou et al. | Battery health prognosis using improved temporal convolutional network modeling | |
| CN114372417A (en) | Electric vehicle battery health state and remaining life evaluation method based on charging network | |
| CN109001640B (en) | Data processing method and device for power battery | |
| CN116381517A (en) | Lithium battery remaining life probability prediction method based on time convolution attention mechanism | |
| Su et al. | Prediction of remaining useful life for lithium-ion battery with multiple health indicators | |
| CN114035098A (en) | Lithium battery health state prediction method integrating future working condition information and historical state information | |
| CN111983474A (en) | Lithium ion battery life prediction method and system based on capacity decline model | |
| CN116184210A (en) | Battery abnormality detection method, device, system and electronic device | |
| CN116794543A (en) | Lithium battery performance prediction method based on GBLS boost multitask learning model | |
| Mazzi et al. | Lithium-ion battery state of health estimation using a hybrid model based on a convolutional neural network and bidirectional gated recurrent unit | |
| CN117169761A (en) | Battery state evaluation method, apparatus, device, storage medium, and program product | |
| Wang et al. | Feature disentanglement and tendency retainment with domain adaptation for lithium-ion battery capacity estimation | |
| CN118033432A (en) | Battery state of charge estimation method and device and computer equipment | |
| CN114879070A (en) | Battery state evaluation method and related equipment | |
| Wang et al. | Joint prediction of Li-ion battery state of charge and state of health based on the DRSN-CW-LSTM model | |
| CN113297129B (en) | SOC calculation method and system of energy storage system | |
| Song et al. | Capacity estimation method of lithium-ion batteries based on deep convolution neural network | |
| CN116819364A (en) | Battery health determination method, application method, electronic device and storage medium | |
| CN116908727A (en) | Multi-feature combination-based energy storage power station lithium battery capacity estimation method | |
| CN117214722A (en) | Battery health degree prediction method and device, electronic equipment and storage medium | |
| Lin et al. | Battery Health Prognosis Based on Sliding Window Sampling of Charging Curves and Independently Recurrent Neural Network | |
| Xia et al. | Technologies for Energy Storage Power Stations Safety Operation: Battery State Evaluation Survey and a Critical Analysis |
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 |