Battery SOH and RUL joint prediction method based on battery fusion data
Through the multi-layer perception machine model, the multi-objective prediction model is constructed, and the battery operation data and SOH/RUL data are integrated, which solves the problem of high training and maintenance costs of battery SOH and RUL prediction models, and achieves efficient and accurate battery health status and residual use. Joint prediction of lifespan.
Patent Information
- Application Number
- CN202411089355.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-09
- Publication Date
- 2024-11-29
AI Technical Summary
In the prior art, the SOH and RUL prediction models of the battery have high training and maintenance costs, and the data usage efficiency is not high, resulting in high model training and maintenance costs.
Multi-level perception machine model is used to build a multi-objective prediction model. By fusing battery operation data from different sources and corresponding SOH and RUL data, calculate the minimum common multiple of the data set for data repetition, build a shared feature representation and train the model to realize SOH and RUL joint prediction.
It reduces model training and maintenance costs, improves data usage efficiency, realizes high-performance and efficient joint prediction of battery SOH and RUL, reduces the risk of overfitting, and captures the interaction and dependence between SOH and RUL.
Smart Images

Figure CN119044774A_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the field of electrochemical technology, and in particular to a battery SOH and RUL joint prediction method based on battery fusion data. Background Art
[0002] Battery state of health (SOH) is an indicator that evaluates battery performance relative to the level of a new battery. It is used to monitor battery performance, estimate its life, evaluate safety, develop maintenance plans, and assist users in deciding whether to replace the battery. This indicator is crucial for the effective use and safe management of batteries.
[0003] The remaining useful life (RUL) of a battery is a parameter used to estimate how long the battery can continue to be used under the current conditions. It helps to implement preventive maintenance, reduce costs, ensure safety, arrange resources reasonably, improve user satisfaction, ensure system stability, improve product development, comply with regulatory requirements, and provide a reference for battery recycling. Usually, the RUL of a battery is expressed as the remaining number of charge and discharge cycles. Therefore, accurately predicting the SOH and RUL of a battery has extremely important practical application significance.
[0004] Battery operating data such as battery current, voltage, temperature, resistance and impedance are closely related to SOH and RUL. Using the above battery operating data to predict SOH and RUL is a feasible path.
[0005] In many cases, two different sets of data are collected separately to obtain the battery's current, voltage, temperature, impedance, and corresponding SOH and RUL data. Although these data are somewhat correlated, the quantity and dimension of the data are often different. Therefore, a separate prediction model is trained for each prediction target, which results in low data utilization efficiency and high model training and maintenance costs. Summary of the invention
[0006] In order to solve the above technical problems, the present invention proposes a method of integrating battery operation data under different situations and the corresponding SOH and RUL data, using a simple multi-layer perceptron model as a model example, constructing a multi-objective prediction model, and jointly predicting SOH and RUL based on the battery operation fusion data, so that the model training and maintenance costs are greatly reduced.
[0007] The technical solution adopted by the present invention to solve its technical problem is: A battery SOH and RUL joint prediction method based on battery fusion data includes the following implementation steps: Step S1. Test and collect battery data. The collected battery data comes from two sources: one is the battery current, voltage, temperature and SOH data; the other is the battery current, resistance, impedance and RUL data; Step S2: Fusing data from different sources, including fusing different battery operation data and corresponding SOH and RUL data.
[0008] The least common multiple of the numbers of different data sets is calculated, and then each data set is repeated so that the number of each data set increases to their least common multiple to obtain the fused data.
[0009] Furthermore, the data fusion step: first calculates the least common multiple of the numbers of different data sets; and then expands the number of each data set to the least common multiple through a repetitive method.
[0010] Step S3. Constructing a multi-objective prediction model, including: using multiple MLP models to extract battery operation data features; splicing the features of the battery operation data extracted by different MLP models to obtain a shared feature representation.
[0011] Furthermore, multiple MLP models are used to extract battery operation data features in the following two ways: Use multiple MLP models to extract features of different battery operation data respectively: use the fused different battery operation data as input data of different MLP models for feature extraction.
[0012] The features of the battery operation data are extracted independently through multiple MLP models: the fused different battery operation data sets are input into their respective MLP models for feature extraction.
[0013] Furthermore, the activation function of the MLP model of the present invention is the relu function, the number of intermediate layers is 1, and the number of neurons is 32.
[0014] Step S4: training a multi-objective prediction model. The two sets of data sets are divided into a training set and a test set in a ratio of 8:2. To ensure that there is no intersection between the training data and the test data, data fusion is performed on the two training sets and the test set.
[0015] Furthermore, the present invention implements model code and training on the Keras platform.
[0016] Step S5. Simultaneously predict the battery SOH and RUL: After the shared feature representation, there are two linear regression models, and the corresponding prediction values are the battery capacity, i.e., SOH, and the remaining number of cycles, i.e., RUL.
[0017] The beneficial effects of the present invention are: The model of the present invention takes into account the comprehensive advantages of data fusion, multi-target prediction, feature sharing, accuracy and efficiency. It is a high-performance and efficient battery SOH and RUL joint prediction model. Its technical advantages are mainly the following: 1. Multi-source data fusion: This method fuses battery operation data with certain correlations obtained under different conditions and the corresponding SOH and RUL data. 2. Joint prediction model: By constructing a model that simultaneously predicts SOH and RUL, the interaction and dependency between the two can be more accurately captured. 3. Shared feature representation: Extract shared feature representation for the fused battery operation data, and use the shared feature representation to improve the prediction performance of the two objectives and reduce the risk of overfitting. 4. MLP model feature extraction: The MLP model has a simple structure, is easy to learn, and has high training and prediction efficiency. BRIEF DESCRIPTION OF THE DRAWINGS
[0018] Figure 1 It is a structural diagram of the model for joint prediction of battery SOH and RUL based on battery fusion data of the present invention.
[0019] Figure 2 It is the overall flow chart of the joint prediction of battery SOH and RUL based on battery fusion data of the present invention.
[0020] Figure 3 This is a sample diagram of the battery SOH prediction results based on battery fusion data of the present invention.
[0021] Figure 4 This is a sample diagram of the battery RUL prediction results based on battery fusion data of the present invention. DETAILED DESCRIPTION
[0022] refer to Figure 1 The input data is the battery operation fusion data, which comes from different data sets. The figure takes two input data sets as an example, which can also be expanded to multiple data sets; followed by multiple MLP structures, with the activation function being relu, which are used to extract the features of different battery operation data; then, the features extracted by multiple MLPs are concatenated to obtain a shared feature representation; then, the shared feature representation is connected to two linear regression models, which are used to predict SOH and RUL, respectively.
[0023] refer to Figure 2 The specific process of implementing the present invention is as follows: Step S1: Test and collect battery data. The collected battery data has two sources: one is the battery current, voltage, temperature and SOH data; the other is the battery current, resistance, impedance and RUL data; the above two sets of data have a certain correlation with each other, but the data quantity and dimension are different.
[0024] Furthermore, when collecting data, the battery current, voltage and temperature data, as well as the battery capacity data SOH are recorded. In addition, the battery current, resistance and impedance data, as well as the battery RUL data are also recorded.
[0025] Step S2: Fusing data from different sources. Fusing different battery operation data and corresponding SOH and RUL data.
[0026] Since the battery operation data and the corresponding SOH and RUL are often tested under different test conditions, their quantity and dimensions are different. When building a multi-objective prediction model, the total amount of input data must be the same, and the number of training data in each batch must also be the same. Therefore, it is necessary to fuse the battery operation data and the corresponding SOH and RUL data from different sources.
[0027] The fusion method is: calculate the least common multiple of the quantities of different data sets, and then repeat each data set so that the quantity of each data set increases to their least common multiple to obtain the fused data.
[0028] In the multi-objective prediction model, data from different sources need to be fused to ensure that the model input data has the same quantity. When collecting battery operation data and the corresponding SOH and RUL, these data are usually generated under different conditions, resulting in differences in the quantity and dimension of the obtained data.
[0029] In order to build a multi-objective prediction model, it is necessary to ensure that the total amount of input data is consistent, and the amount of data in each training batch must also be consistent. Therefore, the battery operation data from different test environments and their related SOH and RUL data must be fused.
[0030] The process of data fusion involves the following steps: first, calculate the least common multiple of the number of different data sets; then expand the number of each data set to this least common multiple through repeated methods.
[0031] Step S3. Constructing a multi-objective prediction model, including: using multiple MLP models to extract battery operation data features; splicing the features of the battery operation data extracted by different MLP models to obtain a shared feature representation.
[0032] Among them, multiple MLP models are used to extract battery operation data features in the following two ways: The first method is to use multiple MLP models to extract features of different battery operation data respectively: the fused different battery operation data are used as input data of different MLP models for feature extraction.
[0033] The second method is to independently extract the features of battery operation data through multiple MLP models: the fused different battery operation data sets are input into their respective MLP models for feature extraction.
[0034] Furthermore, the activation function of the MLP model of the present invention is the relu function, the number of intermediate layers is 1, and the number of neurons is 32.
[0035] Generate shared feature representation: Concatenate the features of battery operation data extracted by different MLP models to obtain shared feature representation for input into subsequent prediction models.
[0036] Construct a unified feature representation: After multiple MLP models have completed feature extraction of battery operation data, the features obtained by these models are concatenated. This step integrates the scattered feature information specific to each MLP model into a comprehensive and shared feature representation. Such a feature representation not only contains the multi-dimensional spatial features of the original battery operation data, but also forms a richer and more comprehensive feature set through feature concatenation.
[0037] Step S4: training a multi-objective prediction model. The two sets of data sets are divided into a training set and a test set in a ratio of 8:2. To ensure that there is no intersection between the training data and the test data, data fusion is performed on the two training sets and the test set.
[0038] Furthermore, the present invention implements model code and training on the Keras platform.
[0039] Step S5. Simultaneously predict the battery SOH and RUL: After the shared feature representation, there are two linear regression models, and the corresponding prediction values are the battery capacity, i.e., SOH, and the remaining training times, i.e., RUL.
[0040] Since SOH and RUL are both continuous values, the loss function of each prediction target is the mean square error (MSE), and the overall loss function of the model is the sum of the two loss functions; the optimizer is rmsprop; the evaluation metric is the mean absolute error (MAE), which is the sum of the absolute values of the differences between the target value and the predicted value.
[0041] Furthermore, after the shared feature representation, there are two linear regression models, the predicted values are the battery capacity (SOH) and the remaining number of cycles (RUL), and the activation functions are both linear functions. The loss function of the model is the mean square error (MSE); the optimizer is rmsprop; the evaluation metric is the mean absolute error (MAE), which is the sum of the absolute values of the difference between the target value and the predicted value. Example
[0042] 1. Forecast data display: The test data of the present invention has two sources. One is the data of the discharge (type=discharge) phase of the B0018 battery in the NASA battery data set, and the battery current, voltage, temperature and SOH therein are used.
[0043] The second is the data of the impedance state (type=impedance) of the B0018 battery in the NASA battery data set, using the battery current, resistance, impedance and RUL.
[0044] Among them, there are 34,866 pieces of battery current, voltage, temperature and SOH data, and 2,544 pieces of battery current, resistance, impedance and RUL data.
[0045] The battery current, voltage, temperature and SOH data have a total of 5 features and 1 prediction target. The names and meanings of the features and prediction targets are shown in Table 1 below:
[0046] The predicted target Capacity represents the battery capacity, namely SOH.
[0047] The specific data sample is shown in Table 2 below:
[0048] The battery current, resistance, impedance and RUL test data have a total of 7 features and 1 prediction target. The names and meanings of the features and prediction targets are shown in Table 3 below:
[0049] The specific data sample is shown in Table 4 below:
[0050] Among them, the features such as Sense_current, Battery_current, Current_ratio, Battery_impedance and Rectified_impedance are divided into real data and imaginary data. The features ending with "_real" are real data, and the features ending with "_imag" are imaginary data.
[0051] 2. Data Fusion: The above two sets of data sets are divided into training set and test set in a ratio of 8:2, respectively, to obtain 27892 training data and 6974 test data of battery current, voltage, temperature and SOH, and 2035 training data and 509 test data of battery current, resistance, impedance and RUL.
[0052] In order to ensure that there is no intersection between the training data and the test data, data integration is performed on the two training sets and the test set respectively.
[0053] The least common multiple of the 27,892 training data of battery current, voltage, temperature and SOH and the 2,035 training data of battery current, resistance, impedance and RUL is 56,760,220, so the training sets are repeated respectively to increase the number of each training set to 56,760,220.
[0054] The least common multiple of 6974 test data of battery current, voltage, temperature and SOH and 509 test data of battery current, resistance, impedance and RUL is 3549766, so the test sets are repeated respectively, so that the number of each test set increases to 3549766.
[0055] 3. Training model: The present invention implements the model code on the Keras platform.
[0056] The multi-objective prediction model is trained on the training set and the effect is tested on the test set.
[0057] The sample batch size for model training is 1280, and the number of epochs of the model is 10.
[0058] 4.Prediction results display: 4.1SOH prediction results display: The mean of the test set of battery SOH is 1.5841, the standard deviation is 0.1571, the mean of the prediction set is 1.5683, and the standard deviation is 0.1367, which shows that the data distribution of the prediction results is basically consistent with the data distribution of the real data.
[0059] The root mean square error between the prediction set and the test set of battery SOH is only 0.0796, indicating that the prediction effect of the model is good.
[0060] The prediction results are shown in Table 5:
[0061] The prediction results are as follows: Figure 3 As shown, Figure 3 The solid line in the upper middle represents the true value, the dotted line represents the predicted value, and the solid square represents the error value. The numbers represent the true values. It can be seen from the figure that the prediction effect is good.
[0062] 4.2 RUL prediction results display: The mean of the test set is 139.7838, the standard deviation is 96.9046, the mean of the prediction set is 139.8524, and the standard deviation is 96.9486, which shows that the data distribution of the prediction results is basically consistent with the data distribution of the real data.
[0063] The root mean square error between the prediction set and the test set is 0.4681, and the model effect is good.
[0064] The prediction results are shown in Table 6 below:
[0065] The prediction results are as follows: Figure 4 As shown, Figure 4 The solid line in the upper middle represents the true value, the dotted line represents the predicted value, and the solid square represents the error value. The numbers represent the true values. Figure 4 It can be seen that the prediction effect is good.
Claims
1. A battery SOH and RUL joint prediction method based on battery fusion data, characterized in that: The implementation steps include: Step S1. Test and collect battery data: The collected battery data has two sources: one is the battery current, voltage, temperature and SOH data; the other is the battery current, resistance, impedance and RUL data; Step S2. Fusion of data from different sources: including fusion of different battery operation data and corresponding SOH and RUL data; Step S3. constructing a multi-objective prediction model, including: using multiple MLP models to extract battery operation data features; splicing the features of the battery operation data extracted by different MLP models to obtain a shared feature representation; Step S4. Training the multi-objective prediction model: the two sets of data sets are divided into a training set and a test set in a ratio of 8:2, and data fusion is performed on the two training sets and the test set respectively; Step S5. Simultaneously predict the battery SOH and RUL: After the shared feature representation, there are two linear regression models, and the corresponding prediction values are the battery capacity, i.e., SOH, and the remaining number of cycles, i.e., RUL.
2. According to claim 1, a battery SOH and RUL joint prediction method based on battery fusion data is characterized in that: In step S2, the least common multiple of the quantities of different data sets is calculated, and then each data set is repeated so that the quantity of each data set increases to their least common multiple, thereby obtaining the fused data.
3. The method for jointly predicting battery SOH and RUL based on battery fusion data according to claim 1 is characterized in that: In step S2, the data fusion step: first calculate the least common multiple of the numbers of different data sets; then expand the number of each data set to the least common multiple through a repetitive method.
4. The method for jointly predicting battery SOH and RUL based on battery fusion data according to claim 1 is characterized in that: In step S3, multiple MLP models are used to extract battery operation data features in two ways: using multiple MLP models to respectively extract features of different battery operation data, and using the fused different battery operation data as input data of different MLP models for feature extraction; independently extracting features of battery operation data through multiple MLP models, and inputting the fused different battery operation data sets into their respective MLP models for feature extraction.
5. The method for jointly predicting battery SOH and RUL based on battery fusion data according to claim 1 is characterized in that: In step S3, the activation function of the MLP model is the relu function, the number of intermediate layers is 1, and the number of neurons is 32.
Citation Information
Cited By
Method and system for indirectly predicting state of health of battery based on expansion stress
CN119936686A
Method and system for monitoring and predicting health state of storage battery based on multi-modal feature fusion
CN120294587A
Battery health status monitoring and prediction method and system based on multimodal feature fusion
CN120294587B
