CN112684346B - Lithium battery health state estimation method based on genetic convolutional neural network - Google Patents
Lithium battery health state estimation method based on genetic convolutional neural network Download PDFInfo
- Publication number
- CN112684346B CN112684346B CN202011434822.2A CN202011434822A CN112684346B CN 112684346 B CN112684346 B CN 112684346B CN 202011434822 A CN202011434822 A CN 202011434822A CN 112684346 B CN112684346 B CN 112684346B
- Authority
- CN
- China
- Prior art keywords
- battery
- cnn
- network structure
- data
- network
- 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.)
- Active
Links
Images
Landscapes
- Secondary Cells (AREA)
Abstract
The invention discloses a lithium battery health state estimation method based on a genetic convolutional neural network, which specifically comprises the following steps: charging and discharging are carried out on different types of lithium batteries under constant current conditions until the end of the service life of the batteries is recorded, so that a constant current charging voltage curve of the lithium batteries is formed; after each charging of the battery, determining the current capacity of the battery as a true value of the CNN model; characterizing a voltage curve by using characteristic points on the recorded voltage curve, and using the characteristic points as input data of a CNN model; initializing a network structure and various parameters; grouping the processed training set data, and training each CNN network; and inputting the processed test set data into a group of CNN network structures, and selecting the network structure with the minimum mean square error between the true value and the predicted value as a final prediction model.
Description
Technical Field
The invention belongs to the technical field of battery management, and relates to a lithium battery health state estimation method based on a genetic convolutional neural network.
Background
Lithium Ion Batteries (LIB) have been widely used in the fields of electric automobiles, electric tools, base station backup power supplies, and the like because of their advantages of high energy density, long life, strong stability, and small influence on the environment. In practical application, as charge and discharge are carried out once, a series of irreversible chemical reactions occur in the battery, which causes the battery to gradually age and is represented by capacity degradation, power loss and the like. Therefore, it is necessary to estimate the state of health of the battery in advance during the use of the battery, and to send out early warning information when the battery life reaches the end, so as to prompt the user or the equipment provider to replace the battery in time.
The degraded (aged) state of the battery is typically described by the battery state of health (StateofHealth, SOH). The SOH can be obtained by directly measuring or indirectly calculating the ratio of the current value of a certain characteristic parameter to the initial value, the direct measurement needs to use an instrument to measure the current SOH of the battery after each charge and discharge of the battery, and in actual life, the disassembly measurement of the battery is impossible anytime and anywhere, so that the method is not applicable.
Disclosure of Invention
The invention aims to provide a lithium battery health state estimation method based on a genetic convolutional neural network, which can rapidly and accurately predict the battery health state on the basis of automatically learning a network structure.
The technical scheme adopted by the invention is that the lithium battery health state estimation method based on the genetic convolutional neural network is implemented according to the following steps:
step 9, grouping the processed training set data, wherein every 5 training sets are in a group, circularly inputting each group of training sets into a CNN network structure, and training each CNN network;
step 10, realizing automatic learning of a network structure by means of population evolution through a genetic algorithm, and obtaining a group of good CNN network structures;
step 11, inputting the processed test set data into a group of CNN network structures, and selecting a network structure with the minimum mean square error between a true value and a predicted value as a final prediction model; when the voltage data of another battery is input again, the residual capacity of the battery can be predicted through the trained model, and the SOH of the battery can be calculated rapidly and accurately.
The specific steps of the treatment in the step 5 are as follows: according to 135 voltage curves, 8 characteristic points are extracted from each voltage curve to form a 135-row 8-column matrix, and 8 data of each row are converted into a 2-row 4-column matrix and are input into a CNN model as images.
In step 6, the data set is divided into 80% training set and 20% test set.
In step 8, a set of CNN network structures is randomly initialized, the set of network structures is regarded as a population, each individual in the population represents a CNN network structure, and the population size and the generation number are initialized.
The step 10 specifically comprises the following steps:
step 10.1, coding network structures, dividing each network structure, namely an individual, into different stages by taking a pooling layer as a boundary, combining each stage into a fixed binary character string by coding, and expressing each network structure by adopting genotypes;
step 10.2, evaluating the fitness value of the individual, inputting the voltage data into the individual for training, and taking MSE (mean square error) between the true value and the predicted value as the fitness value of the individual;
step 10.3: and (3) generating a new network structure through selection, intersection and variation, evaluating the fitness value of the new network structure, maintaining individuals with low fitness values in each iteration, deleting individuals with high fitness values, always maintaining the population size as N until the maximum generation number is reached, ending the iteration, and outputting a group of good CNN network structures.
The beneficial effects of the invention are as follows: according to the method, the SOH estimation model is built through offline data, the SOH value is estimated on line in real time, and the SOH value can be estimated conveniently; the CNN network structure is automatically designed through binary codes, and a genetic algorithm is used at the same time, so that a group of network structures with optimal performance are automatically searched, the search space is enlarged, and the operation efficiency is improved; according to the invention, the optimal health characteristic is selected from the charging voltage curve through the dichotomy sampling, and is converted into the matrix to be used as the input of the model, so that the optimal health characteristic is effectively utilized, and the calculation complexity is reduced.
Drawings
FIG. 1 is a dichotomy sampling diagram of a lithium battery state of health estimation method based on a genetic convolutional neural network of the present invention;
FIG. 2 is a line of data converted image in the method for estimating the state of health of a lithium battery based on a genetic convolutional neural network according to the present invention;
FIG. 3 is a single stage block diagram of a method for estimating the state of health of a lithium battery based on a genetic convolutional neural network according to the present invention;
FIG. 4 is a diagram of network structure coding operation in the method for estimating the state of health of a lithium battery based on a genetic convolutional neural network;
FIG. 5 is a flow chart of an evolutionary network structure in a lithium battery state of health estimation method based on a genetic convolutional neural network;
fig. 6 is a general flowchart of a method for estimating the state of health of a lithium battery based on a genetic convolutional neural network according to the present invention.
Detailed Description
The invention will be described in detail below with reference to the drawings and the detailed description.
As shown in fig. 6, the method for estimating the health state of the lithium battery based on the genetic convolutional neural network is implemented specifically according to the following steps:
step 9, grouping the processed training set data, wherein every 5 training sets are in a group, circularly inputting each group of training sets into a CNN network structure, and training each CNN network;
step 10, realizing automatic learning of a network structure by means of population evolution through a genetic algorithm, and obtaining a group of good CNN network structures;
step 11, inputting the processed test set data into a group of CNN network structures, and selecting a network structure with the minimum mean square error between a true value and a predicted value as a final prediction model; when the voltage data of another battery is input again, the residual capacity of the battery can be predicted through the trained model, and the SOH of the battery can be calculated rapidly and accurately.
The specific steps of the treatment in the step 5 are as follows: according to 135 voltage curves, 8 characteristic points are extracted from each voltage curve to form a 135-row 8-column matrix, and 8 data of each row are converted into a 2-row 4-column matrix and are input into a CNN model as images. The specific transformation mode is shown in figure 2.
In step 6, the data set is divided into 80% training set and 20% test set.
In step 8, a set of CNN network structures is randomly initialized, the set of network structures is regarded as a population, each individual in the population represents a CNN network structure, and the population size and the generation number are initialized.
The step 10 specifically comprises the following steps:
step 10.1, coding network structures, dividing each network structure, namely an individual, into different stages by taking a pooling layer as a boundary, combining each stage into a fixed binary character string by coding, and expressing each network structure by adopting genotypes;
step 10.2, evaluating the fitness value of the individual, inputting the voltage data into the individual for training, and taking MSE (mean square error) between the true value and the predicted value as the fitness value of the individual;
step 10.3: and (3) generating a new network structure through selection, intersection and variation, evaluating the fitness value of the new network structure, maintaining individuals with low fitness values in each iteration, deleting individuals with high fitness values, always maintaining the population size as N until the maximum generation number is reached, ending the iteration, and outputting a group of good CNN network structures.
Stage and node division:
the possible phase structures in a part of a single network are shown in fig. 3, each CNN network structure is composed of S phases randomly, symbol V s，ks Represents the S-th stage (s=1, 2,3 … S), ks represents V s，ks The number of nodes involved in the phase. Each node represents a convolution operation, adjacent stages are connected by pooling operations, and the last stage ends in performing an average pooling operation.
Default node:
two default nodes are set in each stage, and the default input node represents V s,0 Accepting data from a previous stage, performing a convolution operation, and sending an output to the absence of a preambleEach node of the nodes such as V s,1 。V s,Ks+1 Is the default output node, accepts data such as V from all nodes without subsequent nodes s,ks Convolutions are performed on their summaries and the outputs are sent to the pooling layer, the connection between the normal node and the default node does not need to be encoded.
A common node.
The common nodes are unique and orderly in numbers except for default nodes in one stage, each node represents a convolution operation, ks node numbers are ordered from small to large, and the symbol V s,ks Indicating that the s-th phase contains Ks nodes.
Inter-node operation.
After summing the values of all input nodes (the lower numbered nodes connected to it), a convolution is performed, followed by a batch normalization, relu operation.
Coding rules:
using
Bits encode directed edges between phase internal nodes, the first bit representing (V s,1 ,V s,2 ) Whether there is a directed edge connection between nodes, the next two bits representing (V s,1 ,V s,3 )，(V s,2 ,V s,3 ) If there is a directed edge connection between them, and so on until the last bit in the phase, if there is a directed edge connection, then the code is 1, otherwise, it is 0. Adding one bit at the end of encoding indicates skipping the connection, forwarding the input information directly to the output, bypassing the entire block.
Special cases:
if an orphan node exists, the orphan node is ignored, no encoding is engaged, and the default node is not connected with the orphan node. If there is no connection in a stage, i.e. all bits of the stage are 0, the convolution operation is performed only once.
Specific examples are as follows:
(1) Lithium batteries were used as a U.S. aerospace agency repository 18650 rated for 2Ah, and B5, B6 batteries were selected for use.
(2) And (3) circularly charging and discharging the B5 lithium battery under the constant current 4A, obtaining a constant current charging voltage change curve after charging once until the battery is scrapped, and obtaining 135 charging voltage curves after 135 times of circular charging and discharging, wherein only charging voltage data of the B6 lithium battery are recorded.
(3) Before each charging of the battery, the residual capacity before the battery is charged is measured, and after the battery is fully charged (the voltage is not changed any more), the current capacity of the battery is measured, wherein the current battery capacity is equal to the sum of the charging quantity required by the current charging and the residual capacity, the current capacity of the battery is taken as a true value of the CNN model, the output of the CNN model is taken as a predicted value, and the optimal CNN model is trained by continuously reducing the error between the true value and the predicted value.
(4) And (3) taking 8 points of each voltage curve recorded in the step (2) as characteristic points by using a dichotomy, representing one voltage curve by using the 8 characteristic points, and taking the voltage curve as input data of a CNN model, wherein a sampling result is shown in fig. 1, and the specific sampling process is as follows:
in fig. 1, the horizontal axis represents time variation, the vertical axis represents voltage variation, the curve in the figure represents a charging voltage curve, characteristic points are extracted on the charging voltage curve, the midpoint taking operation is performed through the voltage variation of the vertical axis, and then whether the midpoint taking point is at the front half part or the rear half part of the midpoint of the current voltage in the next step is determined through the time ratio of the horizontal axis. The starting voltage is currently known to be 2.8V and the terminating voltage is known to be 4.2V.
Taking terminal voltage 4.2V as a 1 st characteristic point F1;
taking the middle point between the starting voltage and the terminal voltage, the obtained voltage value is 3.5V as the 2 nd characteristic point F2, and 2.8V-3.5V is repartitioned as the first half part, and 3.5V-4.2V is the second half part.
After the second characteristic point is determined, the ratio of the corresponding time of the point on the voltage curve in the whole time is observed, the front half part is 0-1700t, the rear half part is 1700t-12300t, and the time interval of the rear half part is relatively large, so that the midpoint is taken at the voltage rear half part of 3.5V-4.2V, the obtained voltage value of 3.85V is taken as a 3 rd characteristic point F3, 3.5V-3.85V is taken as the front half part, and 3.85V-4.2V is taken as the rear half part.
The front half part of the time interval corresponding to the 3 rd characteristic point is 1700t-8000t, the rear half part is 8000t-14000t, and the time intervals are quite different, so that the middle points are respectively taken in the two sections of voltage intervals.
The specific process of extracting feature points in the first half 1700t-8000t is as follows:
the corresponding voltage of the front half part is 3.5V-3.85V, 3.675V obtained by taking the midpoint is taken as a 4 th characteristic point F4, 3.5V-3.675V is taken as the front half part, and 3.675V-3.85V is taken as the rear half part.
The front half part of the time interval corresponding to the 4 th characteristic point is 1700t-2900t, the rear half part is 2900t-8000t, and the rear half part occupies a relatively large area, so that the midpoint is taken in the voltage of 3.675V-3.85V of the rear half part, the obtained 3.7625V is taken as the 5 th characteristic point F5, 3.675V-3.7625V is taken as the front half part, and 3.7625V-3.85V is taken as the rear half part.
The first half part of the time interval corresponding to the 5 th characteristic point is 2900t-4050t, the second half part is 4050t-6300t, and the second half part occupies a relatively large time, so that the midpoint is taken at 3.7625V-3.85V of the second half part, 3.80625V is taken as the 6 th characteristic point F6, 3.7625V-3.80625V is taken as the first half part, and 3.80625V-3.85V is taken as the second half part.
The specific process of extracting feature points in the second half 8000t-14000t is as follows:
the second half 3.85V-4.2V, taking 4.025V obtained from the middle point as 7 th characteristic point F7, and repartitioning 3.85V-4.025V as the first half and 4.025V-4.2V as the second half.
The first half of the time interval corresponding to 4.025V is 8000t-10300t, the second half is 10300t-12300t, and the first half occupies a relatively large area, so that the midpoint is taken from 3.85V-4.025V in the first half, and 3.9375 is taken as the 8 th characteristic point.
(5) According to 135 voltage curves, 8 characteristic points are extracted from each voltage curve to form a 135-row 8-column matrix, wherein the data of each row corresponds to the current capacity value of a battery, and the 8 data of each row is converted into a 2-row 4-column matrix which is used as the input of a CNN model, and the specific conversion mode is shown in figure 2.
(6) The data set was partitioned, and 135 pieces of data were partitioned into 80% training set and 20% test set.
(7) SOH was calculated. The current state of health of the lithium battery is estimated by calculating the ratio of the current capacity of the battery to the rated capacity of the battery when the battery leaves the factory, and the current SOH of the battery can be calculated by obtaining the current capacity value of the battery because the rated capacity of the battery when the battery leaves the factory is given, and the SOH is estimated by fitting the current capacity of the battery.
The expression for calculating SOH is as follows:
wherein C is current Representing the current capacity of the battery, C new Indicating the rated capacity of the battery as shipped.
(8) Initializing a network model and various parameters. The training weights are randomly generated, parameters (such as learning rate is 0.01, batch size is 5, round epoch is 50, number of network stages S is 3, number of nodes in stage ks is 6) needed in network training are set, a group of CNN network structures are randomly initialized, a part of possible stage structures in a single network are shown in figure 3, and one CNN network structure is formed by combining a plurality of stages. Considering a set of network structures herein as a population, an individual in the population represents a CNN network structure, initializing population size 40 and generation number 20.
(9) Grouping the processed training set data, wherein every 5 training sets are grouped into one group, circularly inputting each group of training sets into the CNN network structure, and training each CNN network, wherein the specific steps are as follows.
a. Inputting a set of data into each CNN network;
b. forward propagation calculation is carried out to obtain a loss function loss;
c. counter-propagating the calculated gradient;
d. iteratively updating the weight parameters of the network with the portion of the gradient;
the above four steps are continuously circulated until all rounds of training are completed.
(10) The population is evolved through a genetic algorithm to realize automatic learning of the network structure, and a group of good CNN network structures are obtained.
The initialized network structure is encoded as shown in fig. 4.
Here, the number of stages s=3 is set, and the nodes in each stage are (K1, K2, K3) = (6, 6), respectively. And (3) placing a maximum pooling layer with the step length of 2 after the first stage and the second stage, extracting main features, setting a global average pooling layer after the last stage, carrying out average sampling, coding each stage, and combining each coded stage into a fixed binary character string. Each coded network architecture is called an individual, and all network architectures constitute a population, set to a population size N, and remain unchanged for each generation. Each binary string length L is 45, meaning that there are 2 45 The number of possible individuals (network structures) is thus searched for a total of 1200 network architectures.
And evaluating the fitness value of the individual.
Inputting training set data into each network structure (individual) in the population for training, and performing iterative training on the network structure (individual) on the training set by using a standard random gradient descent (SGD) back propagation algorithm and a cosine annealing learning rate scheme until a proper weight is obtained, so that a loss function is reduced, and mse between a predicted value and a true value obtained in the training process is used as an fitness value of the network structure (individual).
And generating a new network structure through selection, intersection and variation, evaluating the fitness value of the network structures in the population, adopting the average value of the historical fitness of the old network structure as the fitness, and calculating mse as the fitness value of the new network structure. And each iteration keeps the individuals with low fitness value, deletes the individuals with high fitness value, keeps the population size as N all the time until the maximum generation number is reached, ends the iteration, and outputs a group of good CNN network structures.
(11) And inputting the processed data of the test set into a group of CNN network structures obtained at present, and selecting a network structure with the minimum mse (mean square error) between a true value and a predicted value as a final prediction model. When the voltage data of the B6 battery is inputted, the SOH of the battery can be rapidly and accurately calculated from the capacity data predicted by the model. Therefore, in practical application, only the charging voltage data of the constant current of the battery is needed to be recorded, and the trained model can be directly used for estimating the current SOH of the battery.
Claims (5)
1. The lithium battery health state estimation method based on the genetic convolutional neural network is characterized by comprising the following steps of:
step 1, aiming at different types of lithium batteries, measuring and calculating rated capacity of the lithium batteries when leaving the factory;
step 2, charging and discharging are carried out on lithium batteries of different types under constant current conditions, voltage data under charging work are recorded in real time until the end of service life of the batteries is recorded, a constant current charging voltage curve of the lithium batteries is formed according to the recorded data, and battery aging characteristics are obtained through the voltage curve;
step 3, after each time of charging the battery in the step 2, determining the current capacity of the battery as a true value of the CNN model, and continuously reducing the error between the true value and the predicted value to train the optimal CNN model;
step 4, sampling 8 points of each voltage curve recorded in the step 2 by using a dichotomy as characteristic points, and representing one voltage curve by using the 8 characteristic points and using the 8 characteristic points as input data of a CNN model;
step 5, processing the characteristic points in the step 4;
step 6, dividing the data set;
step 7, estimating SOH of the current lithium battery through the ratio of the current capacity of the battery to the rated capacity of the battery when leaving the factory;
step 8, initializing a network structure and various parameters;
step 9, grouping the processed training set data, wherein every 5 training sets are in a group, circularly inputting each group of training sets into a CNN network structure, and training each CNN network;
step 10, realizing automatic learning of a network structure by means of population evolution through a genetic algorithm, and obtaining a group of good CNN network structures;
step 11, inputting the processed test set data into a group of CNN network structures, and selecting a network structure with the minimum mean square error between a true value and a predicted value as a final prediction model; when the voltage data of another battery is input again, the residual capacity of the battery can be predicted through the trained model, and the SOH of the battery can be calculated rapidly and accurately.
2. The method for estimating the health state of a lithium battery based on a genetic convolutional neural network according to claim 1, wherein the specific steps of the step 5 process are as follows: according to 135 voltage curves, 8 characteristic points are extracted from each voltage curve to form a 135-row 8-column matrix, and 8 data of each row are converted into a 2-row 4-column matrix and are input into a CNN model as images.
3. The method for estimating the health of a lithium battery based on a genetic convolutional neural network according to claim 1, wherein in the step 6, the data set is divided into 80% training set and 20% test set.
4. The method for estimating the health of a lithium battery according to claim 1, wherein in the step 8, a set of CNN network structures is randomly initialized, the set of network structures is regarded as a population, each individual in the population represents a CNN network structure, and the population size and the generation number are initialized.
5. The method for estimating the health state of a lithium battery based on a genetic convolutional neural network according to claim 1, wherein the step 10 is specifically:
step 10.1, coding network structures, dividing each network structure, namely an individual, into different stages by taking a pooling layer as a boundary, combining each stage into a fixed binary character string by coding, and expressing each network structure by adopting genotypes;
step 10.2, evaluating the fitness value of the individual, inputting the voltage data into the individual for training, and taking MSE (mean square error) between the true value and the predicted value as the fitness value of the individual;
step 10.3: and (3) generating a new network structure through selection, intersection and variation, evaluating the fitness value of the new network structure, maintaining individuals with low fitness values in each iteration, deleting individuals with high fitness values, always maintaining the population size as N until the maximum generation number is reached, ending the iteration, and outputting a group of good CNN network structures.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011434822.2A CN112684346B (en) | 2020-12-10 | 2020-12-10 | Lithium battery health state estimation method based on genetic convolutional neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011434822.2A CN112684346B (en) | 2020-12-10 | 2020-12-10 | Lithium battery health state estimation method based on genetic convolutional neural network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112684346A CN112684346A (en) | 2021-04-20 |
| CN112684346B true CN112684346B (en) | 2023-06-20 |
Family
ID=75446612
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011434822.2A Active CN112684346B (en) | 2020-12-10 | 2020-12-10 | Lithium battery health state estimation method based on genetic convolutional neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112684346B (en) |
Families Citing this family (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113376540B (en) * | 2021-05-18 | 2023-07-07 | 西安理工大学 | LSTM battery health state estimation method based on evolutionary attention mechanism |
| CN114114052A (en) * | 2021-11-25 | 2022-03-01 | 福州大学 | Method for rapidly estimating SOH and SOC of battery based on shock response characteristic |
| CN114184861A (en) * | 2021-11-28 | 2022-03-15 | 辽宁石油化工大学 | Fault diagnosis method for oil-immersed transformer |
| CN115015760B (en) * | 2022-05-10 | 2024-06-14 | 香港中文大学（深圳） | Lithium battery health state assessment method based on neural network and migration integrated learning |
| CN115616415B (en) * | 2022-12-06 | 2023-04-07 | 北京志翔科技股份有限公司 | Method, device and equipment for evaluating state of battery pack and storage medium |
| CN117074964B (en) * | 2023-10-13 | 2024-01-16 | 快电动力(北京)新能源科技有限公司 | Method, device, system and component for monitoring battery health status |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110687452A (en) * | 2019-09-05 | 2020-01-14 | 南京理工大学 | Lithium battery capacity online prediction method based on K-means clustering and Elman neural network |
| CN110824364A (en) * | 2019-10-24 | 2020-02-21 | 重庆邮电大学 | Lithium battery SOH estimation and RUL prediction method based on AST-LSTM neural network |
| CN111638465A (en) * | 2020-05-29 | 2020-09-08 | 浙大宁波理工学院 | Lithium battery health state estimation method based on convolutional neural network and transfer learning |
| WO2020191800A1 (en) * | 2019-03-27 | 2020-10-01 | 东北大学 | Method for predicting remaining service life of lithium-ion battery employing wde-optimized lstm network |
-
2020
- 2020-12-10 CN CN202011434822.2A patent/CN112684346B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2020191800A1 (en) * | 2019-03-27 | 2020-10-01 | 东北大学 | Method for predicting remaining service life of lithium-ion battery employing wde-optimized lstm network |
| CN110687452A (en) * | 2019-09-05 | 2020-01-14 | 南京理工大学 | Lithium battery capacity online prediction method based on K-means clustering and Elman neural network |
| CN110824364A (en) * | 2019-10-24 | 2020-02-21 | 重庆邮电大学 | Lithium battery SOH estimation and RUL prediction method based on AST-LSTM neural network |
| CN111638465A (en) * | 2020-05-29 | 2020-09-08 | 浙大宁波理工学院 | Lithium battery health state estimation method based on convolutional neural network and transfer learning |
Non-Patent Citations (1)
| Title |
|---|
| 基于GA优化BP神经网络的微电网蓄电池健康状态评估;邓伟锋;李振璧;;电测与仪表(第21期);全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112684346A (en) | 2021-04-20 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112684346B (en) | Lithium battery health state estimation method based on genetic convolutional neural network | |
| CN110059377B (en) | Fuel cell life prediction method based on deep convolutional neural network | |
| CN113267733B (en) | Automatic configuration method for lithium battery health state estimation based on Gaussian process regression | |
| CN110568359B (en) | Lithium battery residual life prediction method | |
| CN110082682B (en) | Lithium battery state of charge estimation method | |
| CN113176518B (en) | Lithium battery health state estimation method | |
| CN112198434B (en) | Identification method, system, equipment, device and medium for battery capacity attenuation model parameters | |
| CN111898689A (en) | Image classification method based on neural network architecture search | |
| CN113376540B (en) | LSTM battery health state estimation method based on evolutionary attention mechanism | |
| CN111812515A (en) | XGboost model-based lithium ion battery state of charge estimation | |
| CN114757340A (en) | Lithium battery health state prediction method and system based on neural network integration | |
| CN114726045A (en) | Lithium battery SOH estimation method based on IPEA-LSTM model | |
| CN115276177A (en) | Energy storage battery charging and discharging power control method and device and battery control system | |
| CN115856633A (en) | Lithium ion battery capacity estimation method based on graph neural network | |
| CN113537474A (en) | Multi-precision quantitative neuron neural network training and precision selection method | |
| CN112287597B (en) | Lead-acid storage battery SOH estimation method based on VPGA-GPR algorithm | |
| CN117171588A (en) | Method for detecting gradient utilization faults of power battery | |
| CN117007974A (en) | Solid-state battery SOC estimation method based on model fusion | |
| CN115993537A (en) | Lithium battery capacity prediction method based on correlation analysis and WOA-LSTM | |
| CN115201684A (en) | Method for evaluating service life of lithium ion battery in different degradation modes | |
| CN115327389A (en) | Lithium battery SOC estimation method based on genetic algorithm improved double-Kalman filtering | |
| CN115358484A (en) | Battery pack capacity prediction method and related equipment | |
| CN115409317A (en) | Transformer area line loss detection method and device based on feature selection and machine learning | |
| CN115481796A (en) | Method for predicting remaining service life of battery based on Bayesian hybrid neural network | |
| CN115032544A (en) | Method for predicting residual life of power lithium battery under condition of incomplete data |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |