Power battery health state online monitoring and early warning method and system based on multi-data fusion

Abstract

The invention discloses a power battery health state on-line monitoring and early warning method and system based on multi-data fusion. The method comprises the following steps: synchronously acquiring multi-dimensional heterogeneous data in a battery operation process through a multi-sensor data acquisition module; carrying out standardization processing and preprocessing on the multi-dimensional heterogeneous data, extracting a comprehensive feature vector and carrying out multi-sensor data fusion; a multi-sensor data fusion algorithm is adopted to carry out weight distribution on the fused comprehensive feature vector, a dynamic weight coefficient is calculated according to the sensitivity and reliability of each parameter to the health state of the battery, and a fused battery state feature descriptor is obtained; and performing mode recognition and state classification on the fused battery state feature descriptors by using a pre-established evaluation model to realize quantitative evaluation of the health degree of the battery. According to the invention, real-time monitoring, evaluation and early warning of the health state of the battery in the whole life cycle are realized, and a reliable basis is provided for safe operation and maintenance decision of a battery system.

Description

Technical Field

[0001] This invention belongs to the field of online monitoring and early warning technology for the health status of power batteries, specifically relating to a method and system for online monitoring and early warning of the health status of power batteries based on multi-data fusion. Background Technology

[0002] As a core component of new energy vehicles, the health status of power batteries is directly related to the safety, reliability, and economy of the vehicle. Accurate monitoring and early warning of battery health status has become a key technological support for the development of the new energy vehicle industry.

[0003] Current battery health monitoring methods primarily rely on single parameters or simple combinations of parameters for evaluation. This approach often fails to comprehensively reflect the true health level of the battery, leading to misjudgments and omissions. Traditional monitoring systems lack sufficient adaptability and accuracy when facing complex operating conditions and changing environments, making it difficult to meet practical application needs.

[0004] The core challenge in battery health monitoring stems from the effective integration of multi-source data. Parameters such as voltage, current, temperature, and internal resistance generated during battery operation have different physical meanings, data formats, and variation patterns. These heterogeneous data exhibit complex coupling relationships and mutual influences. Due to the lack of an effective data fusion mechanism, information collected by various sensors cannot form a unified feature description system, making it difficult for monitoring systems to establish accurate battery state models. This data silo phenomenon further exacerbates the technical challenges of feature extraction and state assessment. Traditional single-dimensional analysis methods cannot capture the multi-dimensional characteristic changes in battery health, especially during battery aging, when the trends and correlations of different parameters dynamically evolve, making it difficult for existing static assessment models to adapt to these complex nonlinear changes.

[0005] How to build an online monitoring system that can effectively integrate multi-dimensional sensor data, achieve heterogeneous information fusion processing, and establish an accurate comprehensive assessment model reflecting the health status of batteries has become a key issue that urgently needs to be addressed in the field of power battery health management. Summary of the Invention

[0006] To address the problems existing in the prior art, this invention provides a method and system for online monitoring and early warning of the health status of power batteries based on multi-data fusion. This method realizes real-time monitoring, evaluation and early warning of the health status of batteries throughout their entire life cycle, providing a reliable basis for the safe operation and maintenance decisions of battery systems, effectively extending battery lifespan and improving the level of intelligence in battery management.

[0007] To achieve the above objectives, the present invention provides the following solution:

[0008] A method for online monitoring and early warning of the health status of power batteries based on multi-data fusion, the method comprising:

[0009] Multidimensional heterogeneous data during battery operation is acquired synchronously through a multi-sensor data acquisition module;

[0010] Standardize and preprocess multidimensional heterogeneous data, extract comprehensive feature vectors, and perform multi-sensor data fusion.

[0011] A multi-sensor data fusion algorithm is used to assign weights to the fused comprehensive feature vector. Dynamic weight coefficients are calculated based on the sensitivity and reliability of each parameter to the battery health status to obtain the fused battery state feature descriptor.

[0012] By using a pre-established evaluation model to perform pattern recognition and state classification on the fused battery state feature descriptors, a quantitative assessment of battery health can be achieved.

[0013] Based on the quantitative assessment results of battery health and historical status data, a trend analysis algorithm is used to calculate the changing trend and degradation rate of battery health status, and to determine the predicted value of the remaining battery life.

[0014] The predicted remaining battery life is compared with a preset safety threshold. If the predicted value is lower than the safety threshold, an early warning mechanism is triggered and maintenance suggestions are generated, thus obtaining complete battery health status monitoring and early warning information.

[0015] Preferably, the method for standardizing and preprocessing multidimensional heterogeneous data, extracting comprehensive feature vectors, and performing multi-sensor data fusion includes:

[0016] The multi-sensor data acquisition module synchronously acquires heterogeneous data such as voltage, current, temperature and internal resistance signals during battery operation. Based on the preset sampling frequency, the heterogeneous data is timestamped and format standardized to obtain a unified multi-dimensional sensor dataset.

[0017] Data preprocessing algorithms are used to perform noise filtering and outlier detection on multi-dimensional sensor datasets. If data anomalies are detected, interpolation methods are used to repair the data and obtain a cleaned and standardized data matrix.

[0018] Based on the physical characteristics and variation patterns of each parameter in the standardized data matrix, the derived feature parameters, including the voltage change rate, current fluctuation coefficient, temperature gradient, and internal resistance growth rate, are calculated using a feature extraction algorithm to determine a comprehensive feature vector containing both the original and derived parameters.

[0019] Preferably, the method for determining a comprehensive feature vector containing the original parameters and derived parameters by calculating derived feature parameters such as voltage change rate, current fluctuation coefficient, temperature gradient, and internal resistance growth rate through feature extraction algorithms based on the physical characteristics and variation patterns of each parameter in the standardized data matrix includes:

[0020] By matrixing the data, initial data is obtained from the original parameter set, and a standardized data set is obtained by preprocessing the data using a standardization method.

[0021] Based on the standardized basic dataset, and considering the physical characteristic values ​​and their variation patterns, feature extraction methods are applied to calculate the voltage change rate, current fluctuation coefficient, temperature gradient value, and internal resistance growth rate, thereby determining the derived characteristic values.

[0022] Key indicators are extracted from the derived feature values, and combined with the original parameter set to construct a comprehensive feature vector, thereby obtaining a complete feature set for subsequent analysis.

[0023] If the voltage change rate in the complete feature set exceeds the preset threshold range, the parameter is marked as abnormal, and the corresponding change regularity is recorded to obtain an abnormal feature subset.

[0024] Based on the abnormal feature subset, the correlation between the current fluctuation coefficient and the temperature gradient value is checked. If the correlation between the two parameters is lower than the preset standard, the data is calibrated a second time to determine the corrected feature subset.

[0025] By combining the modified feature subset with the internal resistance growth rate, the support vector machine algorithm is used to classify the comprehensive feature vector and determine the distribution of potential abnormal patterns.

[0026] The distribution of abnormal patterns after classification is obtained. For high-risk areas in the classification results, in-depth analysis is performed based on parameter correlation to obtain the final feature analysis results.

[0027] Preferably, the method of using a multi-sensor data fusion algorithm to assign weights to the fused comprehensive feature vector, and calculating dynamic weight coefficients based on the sensitivity and reliability of each parameter to the battery health state, to obtain the fused battery state feature descriptor includes:

[0028] The integrated feature vector is processed by a multi-sensor data fusion algorithm. Based on the parameter sensitivity and reliability, the dynamic coefficient is calculated to obtain the fused battery state feature description.

[0029] Based on the obtained battery state feature descriptions, and in response to the needs of health status and state assessment, a pre-established classification model is used to perform hierarchical processing on the feature descriptions to obtain hierarchical state distribution data.

[0030] By analyzing the stratified state distribution data and combining it with the needs of battery monitoring, the correlation between comprehensive characteristics and health status is analyzed to identify potential abnormal distribution areas.

[0031] If the potential abnormal distribution area exceeds the preset threshold range, the vector weights in that area are recalibrated to obtain the calibrated feature distribution data.

[0032] Based on the calibrated feature distribution data and combined with the characteristics of data fusion, the correlation index of the sensor source is extracted to determine whether there is data bias.

[0033] If the data deviation exceeds the preset threshold range, the data from the sensor source will be fused a second time to obtain a corrected state feature description.

[0034] By modifying the state feature description and combining it with the weight allocation logic, the dynamic coefficients are updated to determine the final battery health status assessment data.

[0035] Preferably, the method for quantitatively assessing battery health by using a pre-established evaluation model to perform pattern recognition and state classification on the fused battery state feature descriptors includes:

[0036] The battery capacity and the obtained fused health factor are preprocessed, and the processed data are divided into training set and test set. The fused health factor is used as the input of the model, and the corresponding battery capacity is used as the output of the model.

[0037] In the improved whale optimization algorithm, the initial population size, search dimension, maximum number of iterations, and upper and lower bounds of the parameter search range are set.

[0038] The fitness function is defined as the root mean square value between the predicted capacity value and the actual capacity value of the training set three-fold cross-validation.

[0039] The fitness value of each individual whale is obtained based on the fitness function. The current best search individual and its fitness value are saved. The optimal parameters are output by iteratively comparing the results until the maximum number of iterations is reached.

[0040] The obtained optimal parameters are input into the SVR model to complete the construction of the IWAO-SVR model. Then, the test set is input into the model to realize the health status assessment of lithium batteries.

[0041] Preferably, the improved whale optimization algorithm, based on the standard whale optimization algorithm, initializes the population through Logistic chaotic mapping to ensure that the initial population is evenly distributed in the search space; improves the convergence accuracy of the algorithm by adjusting the dynamic factor 'a'; improves the convergence speed of the algorithm by changing the adaptive weight 'ω'; and introduces the Levy flight strategy to enable it to escape local optima and search for the global optimum more efficiently. Its inputs are the maximum number of iterations T, the population size N, and the dimension D, and the output is the optimal value of the fitness function of the optimization object.

[0042] Preferably, the method for determining the predicted remaining battery life by calculating the changing trend and degradation rate of battery health status using a trend analysis algorithm based on the quantitative assessment results of battery health and historical state data includes:

[0043] Obtain battery health assessment results and historical status data, remove outliers through data cleaning, and obtain a standardized dataset;

[0044] A time series analysis algorithm is used to process the standardized dataset and calculate the changing trend of battery health status.

[0045] Based on the changing trend, the battery degradation rate is calculated using the exponential smoothing method to determine the degradation rate value;

[0046] If the degradation rate exceeds a preset threshold, the remaining useful life is predicted by fitting historical state data through a linear regression model.

[0047] Key features are extracted from the remaining useful life prediction values, and data fusion technology is used to combine the battery health assessment results to optimize the accuracy of the prediction values.

[0048] By periodically updating historical status data through a status monitoring mechanism, the changing trend and degradation rate are recalculated to obtain dynamic prediction values.

[0049] The monitoring frequency of battery health status is adjusted based on dynamic prediction values ​​to determine the final predicted remaining lifespan.

[0050] Preferably, the method of comparing the predicted remaining battery life with a preset safety threshold, and triggering an early warning mechanism and generating maintenance suggestions if the predicted value is lower than the safety threshold, to obtain complete battery health status monitoring and early warning information includes:

[0051]

[0052] Where RUL(Y) is the remaining lifespan predicted based on the battery state characteristics of Y, and T is a preset safety threshold.

[0053] The present invention also provides an online monitoring and early warning system for the health status of power batteries based on multi-data fusion. The system is used to implement the aforementioned method and includes: a multi-sensor data acquisition module, a feature extraction module, a data fusion module, a status assessment module, a trend analysis module, and an early warning mechanism module.

[0054] The multi-sensor data acquisition module is used to synchronously acquire multi-dimensional heterogeneous data during battery operation.

[0055] The feature extraction module is used to standardize and preprocess multidimensional heterogeneous data, extract comprehensive feature vectors, and perform multi-sensor data fusion.

[0056] The data fusion module is used to assign weights to the fused comprehensive feature vector using a multi-sensor data fusion algorithm, and to calculate dynamic weight coefficients based on the sensitivity and reliability of each parameter to the battery health status, thereby obtaining the fused battery status feature descriptor.

[0057] The state assessment module is used to perform pattern recognition and state classification on the fused battery state feature descriptors using a pre-established assessment model, so as to achieve a quantitative assessment of battery health.

[0058] The trend analysis module is used to calculate the changing trend and degradation rate of battery health status based on the battery health quantification assessment results and historical status data, and to determine the predicted value of the remaining battery life.

[0059] The early warning mechanism module is used to compare the predicted value of the remaining battery life with a preset safety threshold. If the predicted value is lower than the safety threshold, the early warning mechanism is triggered and maintenance suggestions are generated to obtain complete battery health status monitoring and early warning information.

[0060] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0061] This invention discloses a method and system for online monitoring and early warning of the health status of power batteries based on multi-data fusion. It synchronously acquires multi-dimensional parameters of the battery during operation through a multi-sensor data acquisition module, standardizes and preprocesses heterogeneous data, extracts comprehensive feature vectors, and performs multi-sensor data fusion. A pre-established evaluation model is used to perform pattern recognition and state classification on the fused feature descriptors, achieving a quantitative assessment of battery health. This invention also analyzes the trend of battery health status changes based on the evaluation results and historical data, predicts the remaining service life, and compares it with a safety threshold to trigger an early warning mechanism. This method realizes real-time monitoring, evaluation, and early warning of the battery's health status throughout its entire life cycle, providing a reliable basis for safe operation and maintenance decisions of the battery system, effectively extending battery life, and improving the intelligence level of battery management. Attached Figure Description

[0062] To more clearly illustrate the technical solution of the present invention, the drawings used in the embodiments are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0063] Figure 1 This is a schematic diagram of a method for online monitoring and early warning of the health status of a power battery based on multi-data fusion, according to an embodiment of the present invention.

[0064] Figure 2 This is a flowchart of the lithium battery health status estimation process based on IWOA-SVR according to an embodiment of the present invention. Detailed Implementation

[0065] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0066] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0067] Example 1

[0068] like Figure 1 As shown in the figure, this invention provides a method for online monitoring and early warning of the health status of power batteries based on multi-data fusion, the method comprising:

[0069] Multidimensional heterogeneous data during battery operation is acquired synchronously through a multi-sensor data acquisition module;

[0070] Standardize and preprocess multidimensional heterogeneous data, extract comprehensive feature vectors, and perform multi-sensor data fusion.

[0071] A multi-sensor data fusion algorithm is used to assign weights to the fused comprehensive feature vector. Dynamic weight coefficients are calculated based on the sensitivity and reliability of each parameter to the battery health status to obtain the fused battery state feature descriptor.

[0072] By using a pre-established evaluation model to perform pattern recognition and state classification on the fused battery state feature descriptors, a quantitative assessment of battery health can be achieved.

[0073] Based on the quantitative assessment results of battery health and historical status data, a trend analysis algorithm is used to calculate the changing trend and degradation rate of battery health status, and to determine the predicted value of the remaining battery life.

[0074] The predicted remaining battery life is compared with a preset safety threshold. If the predicted value is lower than the safety threshold, an early warning mechanism is triggered and maintenance suggestions are generated, thus obtaining complete battery health status monitoring and early warning information.

[0075] In this embodiment, the method for standardizing and preprocessing multidimensional heterogeneous data, extracting comprehensive feature vectors, and performing multi-sensor data fusion includes:

[0076] The multi-sensor data acquisition module synchronously acquires heterogeneous data such as voltage, current, temperature and internal resistance signals during battery operation. Based on the preset sampling frequency, the heterogeneous data is timestamped and format standardized to obtain a unified multi-dimensional sensor dataset.

[0077] Data preprocessing algorithms are used to perform noise filtering and outlier detection on multi-dimensional sensor datasets. If data anomalies are detected, interpolation methods are used to repair the data and obtain a cleaned and standardized data matrix.

[0078] Based on the physical characteristics and variation patterns of each parameter in the standardized data matrix, the derived feature parameters, including the voltage change rate, current fluctuation coefficient, temperature gradient, and internal resistance growth rate, are calculated using a feature extraction algorithm to determine a comprehensive feature vector containing both the original and derived parameters.

[0079] Among them, the method for determining the comprehensive feature vector containing the original parameters and derived parameters by calculating the derived feature parameters such as voltage change rate, current fluctuation coefficient, temperature gradient, and internal resistance growth rate through feature extraction algorithms based on the physical characteristics and variation laws of each parameter in the standardized data matrix includes:

[0080] By matrixing the data, initial data is obtained from the original parameter set, and a standardized data set is obtained by preprocessing the data using a standardization method.

[0081] Based on the standardized basic dataset, and considering the physical characteristic values ​​and their variation patterns, feature extraction methods are applied to calculate the voltage change rate, current fluctuation coefficient, temperature gradient value, and internal resistance growth rate, thereby determining the derived characteristic values.

[0082] Key indicators are extracted from the derived feature values, and combined with the original parameter set to construct a comprehensive feature vector, thereby obtaining a complete feature set for subsequent analysis.

[0083] If the voltage change rate in the complete feature set exceeds the preset threshold range, the parameter is marked as abnormal, and its corresponding change regularity is recorded to obtain an abnormal feature subset.

[0084] Based on the abnormal feature subset, the correlation between the current fluctuation coefficient and the temperature gradient value is checked. If the correlation between the two parameters is lower than the preset standard, the data is calibrated a second time to determine the corrected feature subset.

[0085] By combining the modified feature subset with the internal resistance growth rate, the support vector machine algorithm is used to classify the comprehensive feature vector and determine the distribution of potential abnormal patterns.

[0086] The distribution of abnormal patterns after classification is obtained. For high-risk areas in the classification results, in-depth analysis is performed based on parameter correlation to obtain the final feature analysis results.

[0087] Specifically, for processing multi-dimensional sensor datasets, data matrixing is a core step, aiming to transform raw data into a structured format. In a battery management system, the raw parameter set includes data such as voltage, current, temperature, and internal resistance. Data matrixing processes arrange these parameters according to time series, generating a matrix with multiple columns, each column corresponding to a parameter and each row corresponding to a time point.

[0088] Assuming data is collected at 1000 time points, the matrix is ​​1000 rows and 4 columns. The voltage column records values ​​from 3.2V to 4.2V, and the current column records variations from 0.1A to 5A. This matrix-based approach facilitates subsequent analysis and ensures a clear data structure. For standardized preprocessing, the data needs to be standardized in terms of units and format to eliminate differences in measurement ranges.

[0089] Voltage and temperature have different dimensions. Standardization converts data into dimensionless values ​​with a mean of 0 and a standard deviation of 1. Specifically, for voltage data, assuming the original values ​​are 3.6V, 3.8V, and 4.0V, the standardized values ​​are obtained by subtracting the mean of 3.8V and dividing by the standard deviation of 0.2. This method ensures that different parameters can be directly compared, improving the accuracy of feature extraction.

[0090] Feature extraction methods are used to calculate derived feature values, such as voltage change rate, current fluctuation coefficient, temperature gradient value, and internal resistance growth rate. For example, the voltage change rate can be calculated by dividing the voltage difference between adjacent time points by the time interval; assuming the voltage rises from 3.6V to 3.7V over a time interval of 1 second, the change rate is 0.1V / s. The current fluctuation coefficient can be calculated as the ratio of the standard deviation to the mean of the current data. The temperature gradient value reflects the rate of temperature change over time, and the internal resistance growth rate measures the battery aging trend. These derived features capture the dynamic characteristics of the data, providing a basis for anomaly detection. When constructing a comprehensive feature vector, it is necessary to integrate the original parameters and derived features.

[0091] Methods for extracting key indicators from derived feature values ​​and combining them with the original parameter set to construct a comprehensive feature vector include:

[0092] Suppose the original parameter set is X = (X1, X2, ..., X...). n Suppose that a set of derived feature values ​​Y = (Y1, Y2, ..., Yn) are calculated from the original parameter set using some method (such as statistical analysis, machine learning model, etc.). m Extract key indicators from the derived feature values ​​Y. This can be achieved through feature selection methods, such as model-based feature importance assessment (e.g., random forest, gradient boosting machine), or statistical feature selection (e.g., correlation coefficient, mutual information). Assume the extracted key indicators are Z = (Z1, Z2, ..., Z...). k ), where Z i ∈Y. The original parameter set X and key indicators Z are combined to form a comprehensive feature vector F. The formula for calculating the comprehensive feature vector can be expressed as: F = X∪Z, where U represents the vector merging operation. To ensure consistency in the dimensions of different features, the comprehensive feature vector F usually needs to be standardized. Standardization methods can include Min-Max standardization, Z-Score standardization, etc. Assuming Z-Score standardization is used, the formula for calculating the standardized comprehensive feature vector F' is: Where, μ i and σ i These are the mean and standard deviation of the i-th feature, respectively.

[0093] Methods for verifying the correlation between current fluctuation coefficient and temperature gradient value include:

[0094]

[0095] Where, x i y is the observed value of the i-th current fluctuation coefficient. i It is the observed value of the i-th temperature gradient. It is the average value of the current fluctuation coefficient. This is the average value of the temperature gradient. n is the number of observations. The correlation coefficient r ranges from -1 to 1. When r is close to 1, it indicates a strong positive correlation between the current fluctuation coefficient and the temperature gradient value; when r is close to -1, it indicates a strong negative correlation; when r is close to 0, it indicates no linear correlation between the two variables.

[0096] By using the corrected feature subset and combining it with the internal resistance growth rate, the support vector machine algorithm is employed to classify the comprehensive feature vector. Methods for determining the distribution of potential abnormal patterns include:

[0097] Kernel functions are key to transforming nonlinearly inseparable feature data in low-dimensional space into linearly separable data in high-dimensional space. Different kernel function forms result in different ways in which the support vector machine interacts with the samples.

[0098] Therefore, this invention improves the Gaussian kernel function by introducing an amplitude adjustment parameter, the expression of which is as follows:

[0099]

[0100] Where, for all samples x, x i ∈X, where X is the sample set and σ is the bandwidth.

[0101] The improved Gaussian kernel function of this invention enables sample data to decay faster near the support vectors. To further cluster the features near the support vectors in high-dimensional space, a radial width adjustment parameter c is introduced, allowing the function to decay faster while simultaneously concentrating the sample data near the support vectors. Its expression is as follows:

[0102]

[0103] A larger value of c corresponds to a faster decay rate of the curve, and the data becomes more clustered near the support vectors. Therefore, by controlling the parameter c, the decay rate of the kernel function and the degree of data clustering can be changed.

[0104] Methods for obtaining the distribution of abnormal patterns after classification, and for conducting in-depth analysis of high-risk areas in the classification results, combined with parameter correlation, to obtain the final feature analysis results include:

[0105] F = FeatureAnalysis(H, M);

[0106] Here, H is the set of high-risk regions, M is the parameter correlation matrix, and FeatureAnalysis represents the feature analysis function. The specific feature analysis method can be selected based on the actual situation, such as Principal Component Analysis (PCA) or Independent Component Analysis (ICA).

[0107] In this embodiment, a multi-sensor data fusion algorithm is used to assign weights to the fused integrated feature vector. Dynamic weight coefficients are calculated based on the sensitivity and reliability of each parameter to the battery health state to obtain the fused battery state feature descriptor. The method includes:

[0108] The integrated feature vector is processed by a multi-sensor data fusion algorithm. Based on the parameter sensitivity and reliability, the dynamic coefficient is calculated to obtain the fused battery state feature description.

[0109] Based on the obtained battery state feature descriptions, and in response to the needs of health status and state assessment, a pre-established classification model is used to perform hierarchical processing on the feature descriptions to obtain hierarchical state distribution data.

[0110] By analyzing the stratified state distribution data and combining it with the needs of battery monitoring, the correlation between comprehensive characteristics and health status is analyzed to identify potential abnormal distribution areas.

[0111] If the potential abnormal distribution area exceeds the preset threshold range, the vector weights in that area are recalibrated to obtain the calibrated feature distribution data.

[0112] Based on the calibrated feature distribution data and combined with the characteristics of data fusion, the correlation index of the sensor source is extracted to determine whether there is data bias.

[0113] If the data deviation exceeds the preset threshold range, the data from the sensor source will be fused a second time to obtain a corrected state feature description.

[0114] By modifying the state feature description and combining it with the weight allocation logic, the dynamic coefficients are updated to determine the final battery health status assessment data.

[0115] Among them, the method of processing the comprehensive feature vector through multi-sensor data fusion algorithm, calculating dynamic coefficients based on parameter sensitivity and reliability, and obtaining the fused battery state feature description includes:

[0116]

[0117] In the formula, X i S represents the feature vector of the i-th sensor. i and R iThese represent the parameter sensitivity and reliability of each sensor, respectively.

[0118] Based on the obtained battery state feature descriptions, and to meet the needs of health status and state assessment, a pre-established classification model is used to perform hierarchical processing on the feature descriptions. The methods for obtaining the hierarchical state distribution data include:

[0119] P = M(Y);

[0120] Where M is the classification model and Y is the feature description of the battery state.

[0121] In this embodiment, the method for quantitatively assessing battery health by using a pre-established evaluation model to perform pattern recognition and state classification on the fused battery state feature descriptors includes:

[0122] 1) Data preprocessing: The battery capacity and the obtained fused battery state feature descriptor are preprocessed, and the processed data is divided into training set and test set. The fused battery state feature descriptor is used as the input of the model, and the corresponding battery capacity is used as the output of the model.

[0123] 2) Parameter initialization: In the improved whale optimization algorithm, the initial population size, search dimension, maximum number of iterations, and upper and lower bounds of the parameter search range are set.

[0124] 3) Set the fitness function: The fitness function is defined as the root mean square value between the predicted capacity value and the actual capacity value of the training set three-fold cross-validation.

[0125] 4) Search for optimal parameters: Calculate the fitness value of each whale individual based on the fitness function, save the current optimal search individual and fitness value, and continuously iterate and compare until the maximum number of iterations is reached, and output the optimal parameters.

[0126] 5) Lithium-ion battery SOH assessment: First, the obtained optimal parameters are input into the SVR model to complete the IWAO-SVR model construction. Then, the test set is input into the model to achieve lithium battery health status assessment. Figure 2 As shown.

[0127] Specifically, the improved whale optimization algorithm, based on the standard whale optimization algorithm, initializes the population through a Logistic chaotic mapping to ensure that the initial population is evenly distributed in the search space and guarantees population diversity; it improves the convergence accuracy of the algorithm by adjusting the dynamic factor 'a'; it improves the convergence speed of the algorithm by changing the adaptive weight 'ω'; and it introduces the Levy flight strategy to help the algorithm escape local optima and search for the global optimum more efficiently. Its inputs are the maximum number of iterations T, the population size N, and the dimension D, and its output is the optimal value of the fitness function of the optimization object.

[0128] In this embodiment, the method for determining the predicted remaining battery life based on the battery health quantification assessment results and historical state data, using a trend analysis algorithm to calculate the changing trend and degradation rate of the battery health status, includes:

[0129] Obtain battery health assessment results and historical status data, remove outliers through data cleaning, and obtain a standardized dataset;

[0130] A time series analysis algorithm is used to process the standardized dataset and calculate the changing trend of battery health status.

[0131] Based on the changing trend, the battery degradation rate is calculated using the exponential smoothing method to determine the degradation rate value;

[0132] If the degradation rate exceeds a preset threshold, the remaining useful life is predicted by fitting historical state data through a linear regression model.

[0133] Key features are extracted from the remaining useful life prediction values, and data fusion technology is used to combine the battery health assessment results to optimize the accuracy of the prediction values.

[0134] By periodically updating historical status data through a status monitoring mechanism, the changing trend and degradation rate are recalculated to obtain dynamic prediction values.

[0135] The monitoring frequency of battery health status is adjusted based on dynamic prediction values ​​to determine the final predicted remaining lifespan.

[0136] In this embodiment, the method of comparing the predicted remaining battery life with a preset safety threshold, and triggering an early warning mechanism and generating maintenance suggestions if the predicted value is lower than the safety threshold, to obtain complete battery health status monitoring and early warning information includes:

[0137]

[0138] Where RUL(Y) is the remaining lifespan predicted based on the battery state characteristics of Y, and T is a preset safety threshold.

[0139] Example 2

[0140] The present invention also provides an online monitoring and early warning system for the health status of power batteries based on multi-data fusion. The system is used to implement the method described in Embodiment 1. The system includes: a multi-sensor data acquisition module, a feature extraction module, a data fusion module, a status assessment module, a trend analysis module, and an early warning mechanism module.

[0141] A multi-sensor data acquisition module is used to synchronously acquire multi-dimensional heterogeneous data during battery operation.

[0142] The feature extraction module is used to standardize and preprocess multidimensional heterogeneous data, extract comprehensive feature vectors, and perform multi-sensor data fusion.

[0143] The data fusion module is used to assign weights to the fused comprehensive feature vector using a multi-sensor data fusion algorithm. It calculates dynamic weight coefficients based on the sensitivity and reliability of each parameter to the battery health status, and obtains the fused battery state feature descriptor.

[0144] The status assessment module is used to perform pattern recognition and status classification on the fused battery status feature descriptors using a pre-established assessment model, so as to achieve a quantitative assessment of battery health.

[0145] The trend analysis module is used to calculate the changing trend and degradation rate of battery health status based on the quantitative assessment results of battery health and historical status data, and to determine the predicted value of the remaining battery life.

[0146] The early warning mechanism module compares the predicted remaining battery life with a preset safety threshold. If the predicted value is lower than the safety threshold, the early warning mechanism is triggered and maintenance suggestions are generated, thus obtaining complete battery health status monitoring and early warning information.

[0147] The embodiments described above are merely preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Various modifications and improvements made to the technical solutions of the present invention by those skilled in the art without departing from the spirit of the present invention should fall within the protection scope defined by the claims of the present invention.

Claims

1. A method for online monitoring and early warning of the health status of power batteries based on multi-data fusion, characterized in that, The method includes: Multidimensional heterogeneous data during battery operation is acquired synchronously through a multi-sensor data acquisition module; Standardize and preprocess multidimensional heterogeneous data, extract comprehensive feature vectors, and perform multi-sensor data fusion. A multi-sensor data fusion algorithm is used to assign weights to the fused comprehensive feature vector. Dynamic weight coefficients are calculated based on the sensitivity and reliability of each parameter to the battery health status to obtain the fused battery state feature descriptor. By using a pre-established evaluation model to perform pattern recognition and state classification on the fused battery state feature descriptors, a quantitative assessment of battery health can be achieved. Based on the quantitative assessment results of battery health and historical status data, a trend analysis algorithm is used to calculate the changing trend and degradation rate of battery health status, and to determine the predicted value of the remaining battery life. The predicted remaining battery life is compared with a preset safety threshold. If the predicted value is lower than the safety threshold, an early warning mechanism is triggered and maintenance suggestions are generated, thus obtaining complete battery health status monitoring and early warning information.

2. The method according to claim 1, characterized in that, Methods for standardizing and preprocessing multidimensional heterogeneous data, extracting comprehensive feature vectors, and fusing multi-sensor data include: The multi-sensor data acquisition module synchronously acquires heterogeneous data such as voltage, current, temperature and internal resistance signals during battery operation. Based on the preset sampling frequency, the heterogeneous data is timestamped and format standardized to obtain a unified multi-dimensional sensor dataset. Data preprocessing algorithms are used to perform noise filtering and outlier detection on multi-dimensional sensor datasets. If data anomalies are detected, interpolation methods are used to repair the data and obtain a cleaned and standardized data matrix. Based on the physical characteristics and variation patterns of each parameter in the standardized data matrix, the derived feature parameters, including the voltage change rate, current fluctuation coefficient, temperature gradient, and internal resistance growth rate, are calculated using a feature extraction algorithm to determine a comprehensive feature vector containing both the original and derived parameters.

3. The method according to claim 2, characterized in that, Based on the physical characteristics and variation patterns of each parameter in the standardized data matrix, methods for determining the comprehensive feature vector containing both original and derived parameters include: calculating derived feature parameters such as voltage change rate, current fluctuation coefficient, temperature gradient, and internal resistance growth rate using feature extraction algorithms; and using these methods. By matrixing the data, initial data is obtained from the original parameter set, and a standardized data set is obtained by preprocessing the data using a standardization method. Based on the standardized basic dataset, and considering the physical characteristic values ​​and their variation patterns, feature extraction methods are applied to calculate the voltage change rate, current fluctuation coefficient, temperature gradient value, and internal resistance growth rate, thereby determining the derived characteristic values. Key indicators are extracted from the derived feature values, and combined with the original parameter set to construct a comprehensive feature vector, thereby obtaining a complete feature set for subsequent analysis. If the voltage change rate in the complete feature set exceeds the preset threshold range, the parameter is marked as abnormal, and the corresponding change regularity is recorded to obtain an abnormal feature subset. Based on the abnormal feature subset, the correlation between the current fluctuation coefficient and the temperature gradient value is checked. If the correlation between the two parameters is lower than the preset standard, the data is calibrated a second time to determine the corrected feature subset. By combining the modified feature subset with the internal resistance growth rate, the support vector machine algorithm is used to classify the comprehensive feature vector and determine the distribution of potential abnormal patterns. The distribution of abnormal patterns after classification is obtained. For high-risk areas in the classification results, in-depth analysis is performed based on parameter correlation to obtain the final feature analysis results.

4. The method according to claim 1, characterized in that, The method for obtaining the fused battery state feature descriptor by weighting the fused comprehensive feature vector using a multi-sensor data fusion algorithm and calculating dynamic weight coefficients based on the sensitivity and reliability of each parameter to the battery health state includes: The integrated feature vector is processed by a multi-sensor data fusion algorithm. Based on the parameter sensitivity and reliability, the dynamic coefficient is calculated to obtain the fused battery state feature description. Based on the obtained battery state feature descriptions, and in response to the needs of health status and state assessment, a pre-established classification model is used to perform hierarchical processing on the feature descriptions to obtain hierarchical state distribution data. By analyzing the stratified state distribution data and combining it with the needs of battery monitoring, the correlation between comprehensive characteristics and health status is analyzed to identify potential abnormal distribution areas. If the potential abnormal distribution area exceeds the preset threshold range, the vector weights in that area are recalibrated to obtain the calibrated feature distribution data. Based on the calibrated feature distribution data and combined with the characteristics of data fusion, the correlation index of the sensor source is extracted to determine whether there is data bias. If the data deviation exceeds the preset threshold range, the data from the sensor source will be fused a second time to obtain a corrected state feature description. By modifying the state feature description and combining it with the weight allocation logic, the dynamic coefficients are updated to determine the final battery health status assessment data.

5. The method according to claim 1, characterized in that, Methods for quantitatively assessing battery health by using a pre-established evaluation model to perform pattern recognition and state classification on the fused battery state feature descriptors include: The battery capacity and the obtained fused health factor are preprocessed, and the processed data are divided into training set and test set. The fused health factor is used as the input of the model, and the corresponding battery capacity is used as the output of the model. In the improved whale optimization algorithm, the initial population size, search dimension, maximum number of iterations, and upper and lower bounds of the parameter search range are set. The fitness function is defined as the root mean square value between the predicted capacity value and the actual capacity value of the training set three-fold cross-validation. The fitness value of each individual whale is obtained based on the fitness function. The current best search individual and its fitness value are saved. The optimal parameters are output by iteratively comparing the results until the maximum number of iterations is reached. The obtained optimal parameters are input into the SVR model to complete the construction of the IWAO-SVR model. Then, the test set is input into the model to realize the health status assessment of lithium batteries.

6. The method according to claim 5, characterized in that, The improved whale optimization algorithm, based on the standard whale optimization algorithm, initializes the population through a Logistic chaotic mapping to ensure a uniform distribution of the initial population in the search space; improves the convergence accuracy by adjusting the dynamic factor 'a'; improves the convergence speed by changing the adaptive weight 'ω'; and introduces the Levy flight strategy to help the algorithm escape local optima and search for the global optimum more efficiently. Its inputs are the maximum number of iterations T, the population size N, and the dimension D, and its output is the optimal value of the fitness function of the optimization object.

7. The method according to claim 1, characterized in that, Based on the quantitative assessment results of battery health and historical state data, the method for determining the predicted value of battery remaining lifespan by using trend analysis algorithms to calculate the changing trend and degradation rate of battery health status includes: Obtain battery health assessment results and historical status data, remove outliers through data cleaning, and obtain a standardized dataset; A time series analysis algorithm is used to process the standardized dataset and calculate the changing trend of battery health status. Based on the changing trend, the battery degradation rate is calculated using the exponential smoothing method to determine the degradation rate value; If the degradation rate exceeds a preset threshold, the remaining useful life is predicted by fitting historical state data through a linear regression model. Key features are extracted from the remaining useful life prediction values, and data fusion technology is used to combine the battery health assessment results to optimize the accuracy of the prediction values. By periodically updating historical status data through a status monitoring mechanism, the changing trend and degradation rate are recalculated to obtain dynamic prediction values. The monitoring frequency of battery health status is adjusted based on dynamic prediction values ​​to determine the final predicted remaining lifespan.

8. The method according to claim 1, characterized in that, The method of comparing the predicted remaining battery life with a preset safety threshold, and triggering an early warning mechanism and generating maintenance suggestions if the predicted value is lower than the safety threshold, to obtain complete battery health status monitoring and early warning information includes: Where RUL(Y) is the remaining lifespan predicted based on the battery state characteristics of Y, and T is a preset safety threshold.

9. A power battery health status online monitoring and early warning system based on multi-data fusion, the system being used to implement the method described in any one of claims 1-8, characterized in that, The system includes: a multi-sensor data acquisition module, a feature extraction module, a data fusion module, a status assessment module, a trend analysis module, and an early warning mechanism module; The multi-sensor data acquisition module is used to synchronously acquire multi-dimensional heterogeneous data during battery operation. The feature extraction module is used to standardize and preprocess multidimensional heterogeneous data, extract comprehensive feature vectors, and perform multi-sensor data fusion. The data fusion module is used to assign weights to the fused comprehensive feature vector using a multi-sensor data fusion algorithm, and to calculate dynamic weight coefficients based on the sensitivity and reliability of each parameter to the battery health status, thereby obtaining the fused battery status feature descriptor. The state assessment module is used to perform pattern recognition and state classification on the fused battery state feature descriptors using a pre-established assessment model, so as to achieve a quantitative assessment of battery health. The trend analysis module is used to calculate the changing trend and degradation rate of battery health status based on the battery health quantification assessment results and historical status data, and to determine the predicted value of the remaining battery life. The early warning mechanism module is used to compare the predicted value of the remaining battery life with a preset safety threshold. If the predicted value is lower than the safety threshold, the early warning mechanism is triggered and maintenance suggestions are generated to obtain complete battery health status monitoring and early warning information.

Images