CN116679213A - SOH estimation method for electric vehicle power battery based on integrated deep learning - Google Patents
SOH estimation method for electric vehicle power battery based on integrated deep learning Download PDFInfo
- Publication number
- CN116679213A CN116679213A CN202310631339.0A CN202310631339A CN116679213A CN 116679213 A CN116679213 A CN 116679213A CN 202310631339 A CN202310631339 A CN 202310631339A CN 116679213 A CN116679213 A CN 116679213A
- Authority
- CN
- China
- Prior art keywords
- data
- soh
- charging
- training
- learner
- 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 40
- 238000013135 deep learning Methods 0.000 title claims abstract description 22
- 238000012549 training Methods 0.000 claims abstract description 50
- 238000007781 pre-processing Methods 0.000 claims abstract description 5
- 239000012634 fragment Substances 0.000 claims description 22
- 230000006870 function Effects 0.000 claims description 20
- 238000012360 testing method Methods 0.000 claims description 17
- 238000013528 artificial neural network Methods 0.000 claims description 15
- 238000010606 normalization Methods 0.000 claims description 15
- 230000004913 activation Effects 0.000 claims description 14
- 238000005070 sampling Methods 0.000 claims description 9
- 238000006243 chemical reaction Methods 0.000 claims description 8
- 238000012545 processing Methods 0.000 claims description 8
- 230000007246 mechanism Effects 0.000 claims description 7
- 230000007787 long-term memory Effects 0.000 claims description 5
- 230000008569 process Effects 0.000 claims description 5
- 230000015654 memory Effects 0.000 claims 1
- 230000009286 beneficial effect Effects 0.000 abstract description 3
- 210000004027 cell Anatomy 0.000 description 13
- HBBGRARXTFLTSG-UHFFFAOYSA-N Lithium ion Chemical compound [Li+] HBBGRARXTFLTSG-UHFFFAOYSA-N 0.000 description 9
- 229910001416 lithium ion Inorganic materials 0.000 description 9
- 230000032683 aging Effects 0.000 description 4
- 230000036541 health Effects 0.000 description 4
- 238000010586 diagram Methods 0.000 description 3
- 230000000306 recurrent effect Effects 0.000 description 3
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 description 2
- 230000003213 activating effect Effects 0.000 description 2
- 230000007613 environmental effect Effects 0.000 description 2
- 238000002474 experimental method Methods 0.000 description 2
- 229910000625 lithium cobalt oxide Inorganic materials 0.000 description 2
- BFZPBUKRYWOWDV-UHFFFAOYSA-N lithium;oxido(oxo)cobalt Chemical compound [Li+].[O-][Co]=O BFZPBUKRYWOWDV-UHFFFAOYSA-N 0.000 description 2
- 230000006403 short-term memory Effects 0.000 description 2
- OKTJSMMVPCPJKN-UHFFFAOYSA-N Carbon Chemical compound [C] OKTJSMMVPCPJKN-UHFFFAOYSA-N 0.000 description 1
- 101001095088 Homo sapiens Melanoma antigen preferentially expressed in tumors Proteins 0.000 description 1
- 101000720958 Homo sapiens Protein artemis Proteins 0.000 description 1
- 102100037020 Melanoma antigen preferentially expressed in tumors Human genes 0.000 description 1
- 102100025918 Protein artemis Human genes 0.000 description 1
- 238000009825 accumulation Methods 0.000 description 1
- 230000006399 behavior Effects 0.000 description 1
- 229910052799 carbon Inorganic materials 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 230000007423 decrease Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 239000000446 fuel Substances 0.000 description 1
- 238000011835 investigation Methods 0.000 description 1
- 230000002427 irreversible effect Effects 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
- 210000005036 nerve Anatomy 0.000 description 1
- 210000002569 neuron Anatomy 0.000 description 1
- 230000011218 segmentation Effects 0.000 description 1
- 239000000126 substance Substances 0.000 description 1
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
- G01R31/00—Arrangements for testing electric properties; Arrangements for locating electric faults; Arrangements for electrical testing characterised by what is being tested not provided for elsewhere
- G01R31/36—Arrangements for testing, measuring or monitoring the electrical condition of accumulators or electric batteries, e.g. capacity or state of charge [SoC]
- G01R31/367—Software therefor, e.g. for battery testing using modelling or look-up tables
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
- G01R31/00—Arrangements for testing electric properties; Arrangements for locating electric faults; Arrangements for electrical testing characterised by what is being tested not provided for elsewhere
- G01R31/36—Arrangements for testing, measuring or monitoring the electrical condition of accumulators or electric batteries, e.g. capacity or state of charge [SoC]
- G01R31/392—Determining battery ageing or deterioration, e.g. state of health
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
- Y02T10/00—Road transport of goods or passengers
- Y02T10/60—Other road transportation technologies with climate change mitigation effect
- Y02T10/70—Energy storage systems for electromobility, e.g. batteries
Landscapes
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Secondary Cells (AREA)
Abstract
The invention discloses an electric vehicle power battery SOH estimation method based on integrated deep learning, and relates to the technical field of electric vehicle charging safety. Comprising the following steps: s1, a data acquisition step, a S2, a data division step, a S3, a data slicing step, a S4, a data preprocessing step, a S5, a base model training step, a S6, an estimation accuracy judgment step, a S7, a training cost judgment step, a S8, a meta-learner training step and a S9, and a battery SOH estimation step. The invention is scientific and reasonable, has strong applicability, is beneficial to ensuring the safety and reliability of the battery system, and has practical significance.
Description
Technical Field
The invention relates to the technical field of electric vehicle charging safety, in particular to an electric vehicle power battery SOH estimation method based on integrated deep learning.
Background
With the increasing shortage of global energy situations and urgent need for reducing carbon emissions, the development and popularization of new vehicles to replace traditional fuel automobiles has become a global consensus. The electric automobile has the advantages of high energy efficiency, low pollution, environmental friendliness and the like, and is becoming the first choice of future traffic modes. The lithium ion battery is widely applied to an electric automobile power battery system due to the advantages of small volume, high energy density, long service life, low cost and the like. However, the performance of the lithium ion battery may be degraded with the increase of the number of charge and discharge due to irreversible physical and chemical changes occurring inside the lithium ion battery during the charge and discharge. In order to ensure the safety and reliability of the operation of the lithium ion power battery after the performance decay, the performance of the lithium ion power battery needs to be evaluated.
Battery state of health (SOH) is used as an indicator of battery performance and age, which is generally defined as the ratio between the current maximum available capacity of the battery and the factory rated capacity.
SOH is affected by a number of factors, is related to the aging mechanism of the battery, has higher difficulty in estimating the battery, and is critical to the safe and efficient utilization of the lithium ion power battery. Most of the current online monitoring systems are difficult to evaluate the SOH of each battery cell online and correct the charging parameters of each battery cell, so that the battery cell with lower SOH is in an overcharged state. With the accumulation of overcharge behavior, the SOH of the battery cells is further reduced, so as to form a vicious circle, accelerate the aging of the battery pack, and even cause faults and fires due to frequent overcharging.
Therefore, the invention provides an electric vehicle power battery SOH estimation method based on integrated deep learning from two starting points of the actual working condition of an electric vehicle battery and the sampling precision and sampling frequency limitation of a sensor of a common battery management system, so as to solve the problems in the prior art, which are the problems to be solved by the skilled in the art.
Disclosure of Invention
In view of this, the invention provides an integrated deep learning-based SOH estimation method for an electric vehicle power battery, which is applicable to SOH estimation of battery cells with different charging modes.
In order to achieve the above purpose, the present invention adopts the following technical scheme:
the SOH estimation method for the electric vehicle power battery based on the integrated deep learning comprises the following steps:
s1, acquiring data: acquiring charging fragment data;
s2, data dividing: dividing the charging fragment data into a training data set and a test data set;
s3, data slicing: performing data slicing on the training data set and the test data set in the S2 according to the interval width and the sampling interval;
s4, data preprocessing: carrying out normalization processing on the data slices in the step S3 to obtain a preprocessed training data set and a preprocessed testing data set;
s5, training a base model: inputting the preprocessed training data set, outputting an estimated value of the SOH of the battery, adjusting weight values of each model through a back propagation algorithm according to the output estimated value and the actual value, and performing output matching;
s6, estimating accuracy judgment: judging whether the estimated value of the battery SOH output in the S5 meets the prediction precision, if so, carrying out the S7, otherwise, returning to the S5;
s7, training cost judgment: judging whether the estimated value of the battery SOH output in the S6 meets the training time and cost, if so, carrying out the S8, otherwise, returning to the S3;
s8, training the element learner: taking the basic model as a meta learner of the integrated model, taking the output value of each basic learner as the input value of the meta learner, and training the meta learner;
s9, estimating the SOH of the battery: inputting the N preprocessed test set data fragments into the corresponding N base learners to obtain SOH estimation results of the N base learners, and inputting the SOH estimation results of the N base learners into the meta-learner to obtain estimation results SOH of the integrated model e 。
In the above method, optionally, the charging segment data obtained in S1 is segment data of a determining length determining stage, specifically:
dividing the complete charging process into N charging phases, each charging phase having an equal width DeltaS e N satisfies formula (1):
ΔS e satisfy formula (2):
wherein DeltaS cmin The minimum value of the SOC increment after each electric automobile is charged;
and (3) discarding an incomplete part in the charging data, and converting a section of charging fragment data with any length and any stage into a plurality of sections of charging data with determined length and determined stage.
In the above method, optionally, the incomplete part of the charging data is discarded by using the charge state curve.
In the above method, optionally, the specific content of normalizing the charging segment data in S4 is: and (3) carrying out normalization treatment by adopting a min-max normalization method, wherein the normalization treatment is as shown in a formula (3):
wherein x is norm For normalized data, x is the original data, x max And x min Representing the maximum and minimum values of the raw data during the charging phase.
In the above method, optionally, in S5, a network model combining a long-term memory network and an attention mechanism is selected to train the base model.
The method can be used for improving the model robustness by adding a Recurrent dropout mechanism into the long-term and short-term memory network.
In the above method, optionally, in the step of training the S8-element learner, a BP neural network is selected as the element learner.
In the above method, optionally, the specific content of the step of estimating SOH of the battery in S9 is: the BP neural network comprises an input layer, a hidden layer and an output layer;
the input layer of the BP neural network receives the battery SOH estimated value given by each base learner;
carrying out weighted sum summation through the hidden layer, and carrying out nonlinear conversion on the weighted sum summation result through an activation function to obtain the output of the hidden layer;
the output of the hidden layer is subjected to the operation of weighting and summation, and is subjected to nonlinear conversion of an activation function again to obtain the output of the output layer, namely the estimated value SOH of the battery SOH e 。
Compared with the prior art, the invention provides the SOH estimation method for the electric vehicle power battery based on integrated deep learning, which has the following beneficial effects: firstly, comprehensively considering parameter availability and universality, and selecting the variation trend of battery voltage along with the state of charge (SOC) as a characteristic parameter of a model. Then, a power battery SOH estimation method based on integrated deep learning is proposed, which divides charging data into a plurality of segments, creates a plurality of base models based on the data of each segment to estimate SOH, and integrates SOH estimation values from each base model using an integrated framework on the basis of the estimated SOH to obtain a final SOH estimation value. The method provided by the invention is scientific and reasonable, has strong applicability, is beneficial to ensuring the safety and reliability of the battery system, and has practical significance in estimating the battery health of the electric automobile under the background of rapid popularization of the electric automobile.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings that are required to be used in the embodiments or the description of the prior art will be briefly described below, and it is obvious that the drawings in the following description are only embodiments of the present invention, and that other drawings can be obtained according to the provided drawings without inventive effort for a person skilled in the art.
FIG. 1 is a flowchart of an SOH estimation method of an electric vehicle power battery based on integrated deep learning;
fig. 2 is a u=f (SOC) curve corresponding to different SOHs according to an embodiment of the present invention;
FIG. 3 is a schematic diagram of a charging stage and a sampling point according to an embodiment of the present invention;
fig. 4 is a schematic diagram of dividing charging data according to an embodiment of the present invention;
fig. 5 is a BP nerve block diagram provided in an embodiment of the present invention.
Detailed Description
The following description of the embodiments of the present invention will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present invention, but 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.
Referring to fig. 1, the invention discloses an electric vehicle power battery SOH estimation method based on integrated deep learning, which comprises the following steps:
s1, acquiring data: acquiring charging fragment data;
s2, data dividing: dividing the charging fragment data into a training data set and a test data set;
s3, data slicing: performing data slicing on the training data set and the test data set in the S2 according to the interval width and the sampling interval;
s4, data preprocessing: carrying out normalization processing on the data slices in the step S3 to obtain a preprocessed training data set and a preprocessed testing data set;
s5, training a base model: inputting the preprocessed training data set, outputting an estimated value of the SOH of the battery, adjusting weight values of each model through a back propagation algorithm according to the output estimated value and the actual value, and performing output matching;
s6, estimating accuracy judgment: judging whether the estimated value of the battery SOH output in the S5 meets the prediction precision, if so, carrying out the S7, otherwise, returning to the S5;
s7, training cost judgment: judging whether the estimated value of the battery SOH output in the S6 meets the training time and cost, if so, carrying out the S8, otherwise, returning to the S3;
s8, training the element learner: taking the basic model as a meta learner of the integrated model, taking the output value of each basic learner as the input value of the meta learner, and training the meta learner;
s9, estimating the SOH of the battery: inputting the N preprocessed test set data fragments into the corresponding N base learners to obtain SOH estimation results of the N base learners, and inputting the SOH estimation results of the N base learners into the meta-learner to obtain estimation results SOH of the integrated model e 。
Further, the charging fragment data obtained in S1 is fragment data of a determining length determining stage, specifically:
dividing the complete charging process into N charging phases, each charging phase having an equal width DeltaS e N satisfies formula (1):
ΔS e satisfy formula (2):
wherein DeltaS cmin The minimum value of the SOC increment after each electric automobile is charged;
and (3) discarding an incomplete part in the charging data, and converting a section of charging fragment data with any length and any stage into a plurality of sections of charging data with determined length and determined stage.
Specifically, after the actual charging data is divided into a plurality of data parts according to the set charging stage, incomplete parts in the charging data are discarded. For example, the charge data shown in fig. 4 is divided into N parts by a dotted line, the parts thereof located in the charge stage 1 and the charge stage N are discarded, and the remaining N-2 parts. And processing the charging data according to the method, and converting a section of charging fragment data with any length and any stage into a plurality of sections of charging data with determined length and determined stage.
Further, the incomplete part of the charging data is discarded by adopting the charge state curve.
The state of charge (SOC) is defined as the ratio of the remaining capacity to the maximum available capacity, which is able to quantify the energy remaining inside the current battery. As shown in fig. 2, the u=f (SOC) curve becomes larger and larger in area as the battery SOH decreases.
Further, in S4, the specific content of the normalization processing for the charging fragment data is: and (3) carrying out normalization treatment by adopting a min-max normalization method, wherein the normalization treatment is as shown in a formula (3):
wherein x is norm For normalized data, x is the original data, x max And x min Representing the maximum and minimum values of the raw data during the charging phase.
Specifically, the normalized data is in interval [0,1].
Further, in S5, a network model combining a long-term memory network and an attention mechanism is selected for training the base model.
Furthermore, the method also comprises the step of adding a Recurrent dropout mechanism into the long-term and short-term memory network to improve the robustness of the model.
Further, in the step of training the S8-element learner, a BP neural network is selected as the element learner.
Further, the specific content of the step of estimating SOH of the battery in S9 is as follows: the BP neural network comprises an input layer, a hidden layer and an output layer;
the input layer of the BP neural network receives the battery SOH estimated value given by each base learner;
carrying out weighted sum summation through the hidden layer, and carrying out nonlinear conversion on the weighted sum summation result through an activation function to obtain the output of the hidden layer;
the output of the hidden layer is subjected to the operation of weighting and summation, and is subjected to nonlinear conversion of an activation function again to obtain the output of the output layer, namely the estimated value SOH of the battery SOH e 。
In a specific embodiment, the method for estimating the SOH of the power battery of the electric vehicle based on the integrated deep learning comprises the following steps:
(1) Data processing
And taking the voltage of the battery cell during normal charging and the corresponding SOC value as raw data, and dividing the raw data into a training set and a testing set. Determining the length DeltaS of the charging stage according to experience and actual working conditions e And sampling interval DeltaS n The training set data is divided into N segments, each segment containing N data points. And normalizing the training set data by using a min-max normalization method.
(2) Model training
Firstly, constructing an SOH estimation base model according to the battery SOH estimation method based on deep learning. Then, training N base models based on N data fragments divided in the data processing of the step (1) respectively by taking training set data; comparing the estimated value with the actual value of the model to obtain the prediction precision of each model; comprehensively considering the model estimation precision and the training cost, judging whether the charging stage length delta S needs to be reset e And sampling interval DeltaS n (as shown in fig. 3), when the estimation accuracy and training cost of the base model meet the requirements, the base model can be used as a meta learner of the integrated model. Finally, the method constructs the element learner, takes the output value of each element learner as the input value of the element learner, and trains the element learner.
(3) Battery SOH estimation
Firstly, the data of the test set is taken, and whether the base learner corresponding to a charging stage is activated is judged according to whether the data completely falls into the charging stage. Then, the data segment is input into the corresponding base learner. Finally, the SOH estimation results of each base learner are input into the element learner to obtain the estimation results SOH of the integrated model e 。
In another specific embodiment, the method comprises the steps of:
1. data preprocessing
In order to verify the SOH estimation method provided by the invention, the experiment is carried out based on an oxford battery aging data set. The oxford battery aging dataset was from 8 lithium-cobalt oxide batteries, each battery having a rated capacity of 740mAh. Researchers use the drive conditions of the ARTEMIS city to circularly charge and discharge the lithium-cobalt oxide ion battery at a constant environmental temperature of 40 ℃, and each 100 drive condition cycles are followed by 1C cycle and pseudo OCV cycle and the time, voltage, charge and temperature are recorded. Therefore, the voltage of the battery cell during normal charging and the corresponding SOC value can be calculated as original input data, and the SOH value of the battery after the cycle is taken as output data.
In the experiment, the data of 1C charging of the No. 7 lithium ion battery and the No. 8 lithium ion battery are used as test samples, and the rest 6 lithium ion batteries 1C charging data are used as training samples to train the model. According to the actual working condition obtained by investigation, delta S is calculated e And DeltaS n Set to 10% and 0.1%, respectively. From equation (1), it can be calculated that 10 charging phases are generated after segmentation, each phase containing 10 sampling points. The charging curves of the data sets are all SOC from 0 to 100%, and the charging data under the actual working condition are random fragments, so that the complete charging curve is cut into a plurality of fragments randomly for simulating the actual working condition. Furthermore, since some battery management systems only estimate the overall SOC of the battery pack, the SOC of each battery cell is not estimated.
In actual conditions, the battery cell SOC may not be obtained. When this is encountered, the present invention uses the overall SOC of the battery pack instead of the cell SOC. In practical working conditions, the SOC of a cell with relatively high health in the battery pack is slightly lower than the SOC of the battery pack, and the SOC value of a cell with relatively low health is slightly higher than the SOC of the battery pack. In order to simulate the error caused by using the whole SOC of the battery pack to replace the SOC of the battery cell, the invention introduces offset with the average value of 0 and the amplitude of 1-5% of the original data to the upper and lower bounds of each charging stage respectively, and finally expands the original sample data by 60 times.
2. Model parameter setting
Setting the number of base models as 10, setting the number of LSTM model input layer nodes as 100, and selecting a relu as an activation function. The first LSTM layer node number is 10,return sequences, the parameter is set to true, the Recurrent dropout value is set to 0.5, and the activation function is relu. The number of the nodes of the Attention layer is set to 10. The second LSTM layer node number 5,return sequences parameter is set to false, and the activating function is relu. The number of nodes of the full connection layer is 100, and the activation function is relu. The number of output layer nodes is 1, and the activation function selects tanh.
The objective function is set as a mean square error MSE, an Adam self-adaptive optimizer is adopted to minimize the objective function, the initial value of the learning rate of the optimizer is set to 0.001, the first-order momentum attenuation coefficient is 0.9, the second-order momentum attenuation coefficient is 0.999, the minimum batch is set to 200, and the iteration number is set to 10000. The setting of the parameters is preferably selected according to comparison of experimental results under different parameters. Because 10 different base models are required to be trained, in order to improve training efficiency and avoid over fitting, the invention adopts dynamic learning rate in the model training process, and the value of the learning rate is adaptively adjusted according to the change condition of mean square error in the training process. The training set data is input into the model, after training, the test set data is input into the model to obtain an SOH estimated value, and the average absolute percentage error (Mean Absolute Percentage Error, MAPE) of each base model is calculated according to the formula (4).
Wherein:SOH estimation given for model, SOH i For the true value of SOH, w is the number of data。
Each base model is then used as a base learner for the integrated model. The output of each basic learner is used as the input of the element learner, the node number of the input layer of the element learner is 10, and the activating function is relu. The number of hidden layer nodes is 18, and the activation function is relu. The number of output layer nodes is set to 1, and the activation function is relu. The objective function is set as an average absolute error MAE, an Adam self-adaptive optimizer is adopted to minimize the objective function, the learning rate of the optimizer is 0.001, the first-order momentum attenuation coefficient is 0.99, the second-order momentum attenuation coefficient is 0.999, the minimum batch is set as 1, and the iteration number is set as 1000. The setting of the parameters is preferably selected according to comparison of experimental results under different parameters. And inputting the test set data into the trained integrated model to obtain an SOH estimation result of the integrated model. The whole estimation flow is shown in fig. 1.
3. Model training
The input data is divided according to the charging stages, the charging stages where the data are located are determined according to the position information of the input data, and then the corresponding base learner is determined. The charging data as shown will activate the base learner 2 and the base learner 3. The activated base learner performs deep learning according to the input data characteristics, outputs a corresponding SOH estimated value, and the unactivated base learner outputs 0. Then, the SOH is further estimated and processed using the outputs of the base learners as inputs to the meta learner. Because the input data of the meta-learner has a large number of null values, the selected meta-learner has strong self-adaptability and self-learning capability.
BP neural networks are good at handling non-linearity problems, which allows it to better handle data with incomplete information, such as data containing a large number of nulls. Therefore, the present invention selects the BP neural network as the element learner, and as shown in FIG. 5, the basic structure of the BP neural network generally comprises an input layer, a hidden layer and an output layer.
During training of the network, the weights between each neuron are adjusted by a back-propagation algorithm so that the output of the network can be matched to the desired output. The BP neural network has good generalization capability and can accurately predict new data. The BP neural network input layer node number used in the invention is N, which represents N input values from a base learner. The number of hidden layer nodes can be compared and selected preferentially according to experimental results. The output layer has 1 node, which represents SOH value output by final integrated model.
4. Battery SOH estimation
The input layer of the BP neural network receives the battery SOH estimated values given by each base learner, performs weighting and summation operation through the hidden layer, and performs nonlinear conversion on the result through an activation function to obtain the output of the hidden layer. And then the output of the hidden layer is subjected to the operation of weighting and summation, and is subjected to nonlinear conversion of an activation function again, so that the output of the output layer is obtained.
The previous description of the disclosed embodiments is provided to enable any person skilled in the art to make or use the present invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the generic principles defined herein may be applied to other embodiments without departing from the spirit or scope of the invention. Thus, the present invention is not intended to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.
Claims (8)
1. The electric vehicle power battery SOH estimation method based on integrated deep learning is characterized by comprising the following steps of:
s1, acquiring data: acquiring charging fragment data;
s2, data dividing: dividing the charging fragment data into a training data set and a test data set;
s3, data slicing: performing data slicing on the training data set and the test data set in the S2 according to the interval width and the sampling interval;
s4, data preprocessing: carrying out normalization processing on the data slices in the step S3 to obtain a preprocessed training data set and a preprocessed testing data set;
s5, training a base model: inputting the preprocessed training data set, outputting an estimated value of the SOH of the battery, adjusting weight values of each model through a back propagation algorithm according to the output estimated value and the actual value, and performing output matching;
s6, estimating accuracy judgment: judging whether the estimated value of the battery SOH output in the S5 meets the prediction precision, if so, carrying out the S7, otherwise, returning to the S5;
s7, training cost judgment: judging whether the estimated value of the battery SOH output in the S6 meets the training time and cost, if so, carrying out the S8, otherwise, returning to the S3;
s8, training the element learner: taking the basic model as a meta learner of the integrated model, taking the output value of each basic learner as the input value of the meta learner, and training the meta learner;
s9, estimating the SOH of the battery: inputting the N preprocessed test set data fragments into the corresponding N base learners to obtain SOH estimation results of the N base learners, and inputting the SOH estimation results of the N base learners into the meta-learner to obtain estimation results SOH of the integrated model e 。
2. The method for estimating SOH of electric vehicle power battery based on integrated deep learning according to claim 1, wherein,
the charging fragment data acquired in S1 is fragment data of a determining length determining stage, specifically:
dividing the complete charging process into N charging phases, each charging phase having an equal width DeltaS e N satisfies formula (1):
ΔS e satisfy formula (2):
wherein DeltaS cmin SOC increment after charging for each electric automobileIs the minimum of (2);
and (3) discarding an incomplete part in the charging data, and converting a section of charging fragment data with any length and any stage into a plurality of sections of charging data with determined length and determined stage.
3. The method for estimating SOH of electric vehicle power battery based on integrated deep learning according to claim 2, wherein,
the state of charge curve is used to discard incomplete portions of the charge data.
4. The method for estimating SOH of electric vehicle power battery based on integrated deep learning according to claim 1, wherein,
the specific content of normalization processing of the charging fragment data in S4 is as follows: and (3) carrying out normalization treatment by adopting a min-max normalization method, wherein the normalization treatment is as shown in a formula (3):
wherein x is norm For normalized data, x is the original data, x max And x min Representing the maximum and minimum values of the raw data during the charging phase.
5. The method for estimating SOH of electric vehicle power battery based on integrated deep learning according to claim 1, wherein,
and S5, selecting a network model combining a long-term memory network and an attention mechanism to train the base model.
6. The method for estimating SOH of electric vehicle power battery based on integrated deep learning according to claim 1, wherein,
the method also comprises the step of adding a Recurrentdropout mechanism into the long-period memory network to improve model robustness.
7. The method for estimating SOH of electric vehicle power battery based on integrated deep learning according to claim 1, wherein,
in the training step of the S8-element learner, a BP neural network is selected as the element learner.
8. The method for estimating SOH of electric vehicle power battery based on integrated deep learning according to claim 1, wherein,
the specific content of the estimation step of the battery SOH in S9 is: the BP neural network comprises an input layer, a hidden layer and an output layer;
the input layer of the BP neural network receives the battery SOH estimated value given by each base learner;
carrying out weighted sum summation through the hidden layer, and carrying out nonlinear conversion on the weighted sum summation result through an activation function to obtain the output of the hidden layer;
the output of the hidden layer is subjected to the operation of weighting and summation, and is subjected to nonlinear conversion of an activation function again to obtain the output of the output layer, namely the estimated value SOH of the battery SOH e 。
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310631339.0A CN116679213A (en) | 2023-05-31 | 2023-05-31 | SOH estimation method for electric vehicle power battery based on integrated deep learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310631339.0A CN116679213A (en) | 2023-05-31 | 2023-05-31 | SOH estimation method for electric vehicle power battery based on integrated deep learning |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116679213A true CN116679213A (en) | 2023-09-01 |
Family
ID=87786560
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310631339.0A Pending CN116679213A (en) | 2023-05-31 | 2023-05-31 | SOH estimation method for electric vehicle power battery based on integrated deep learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116679213A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117310538A (en) * | 2023-11-27 | 2023-12-29 | 深圳市普裕时代新能源科技有限公司 | Energy storage battery electric quantity monitoring system capable of automatically detecting electric quantity residual conversion efficiency |
-
2023
- 2023-05-31 CN CN202310631339.0A patent/CN116679213A/en active Pending
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117310538A (en) * | 2023-11-27 | 2023-12-29 | 深圳市普裕时代新能源科技有限公司 | Energy storage battery electric quantity monitoring system capable of automatically detecting electric quantity residual conversion efficiency |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Jiang et al. | A review on the state of health estimation methods of lead-acid batteries | |
| CN110488202B (en) | Vehicle battery state of charge estimation method based on deep neural network | |
| Eddahech et al. | Behavior and state-of-health monitoring of Li-ion batteries using impedance spectroscopy and recurrent neural networks | |
| CN112034356B (en) | GP-UKF-based online SOH estimation method for electric vehicle power battery | |
| CN111007401A (en) | Electric vehicle battery fault diagnosis method and device based on artificial intelligence | |
| CN108872869B (en) | Lithium ion battery degradation classification method based on BP neural network | |
| CN115639481B (en) | Battery data preprocessing system and method based on big data prediction SOC | |
| CN111983474A (en) | Lithium ion battery life prediction method and system based on capacity decline model | |
| CN113687242A (en) | Lithium ion battery SOH estimation method for optimizing and improving GRU neural network based on GA algorithm | |
| CN116298936A (en) | Intelligent lithium ion battery health state prediction method in incomplete voltage range | |
| CN115308606B (en) | Lithium ion battery health state estimation method based on adjacent characteristics | |
| CN112881916A (en) | Method and system for predicting health state and remaining usable life of lithium battery | |
| CN116679213A (en) | SOH estimation method for electric vehicle power battery based on integrated deep learning | |
| CN111257753A (en) | Battery system fault diagnosis method | |
| CN112098874A (en) | Lithium ion battery electric quantity prediction method considering aging condition | |
| CN115308628A (en) | Battery health state monitoring method | |
| Cao et al. | A flexible battery capacity estimation method based on partial voltage curves and polynomial fitting | |
| Naguib et al. | Comparative Study between equivalent circuit and recurrent neural network battery voltage models | |
| CN113791351B (en) | Lithium battery life prediction method based on transfer learning and difference probability distribution | |
| CN114280490A (en) | Lithium ion battery state of charge estimation method and system | |
| Yu et al. | SOH estimation method for lithium-ion battery based on discharge characteristics | |
| CN117233635A (en) | Echelon utilization battery performance evaluation method based on two-way parallel network | |
| CN116774045A (en) | Lithium battery health state prediction method based on HHO-SVR | |
| CN116400224A (en) | Battery remaining service life prediction method based on working temperature correction | |
| CN115327389A (en) | Lithium battery SOC estimation method based on genetic algorithm improved double-Kalman filtering |
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 |