An Adaptive Health Sensing Method for the Degradation of an Energy Storage Battery Stack throughout Its Entire Life

Abstract

The present invention discloses a method for adaptively perceiving the health of a degraded energy storage battery stack throughout its entire life cycle. First, through feature fusion and the construction of a cyclic window, a cyclic window for normalized analysis is constructed, combined with a dual-scale non-stationary degradation analysis framework. Then, through a designed degradation component decoupling algorithm for decoupling and characterization, the input signal is decomposed into inherent features that remain unchanged during the degradation process and drift features that change significantly during the degradation process. By using the extracted degradation decoupling characterization information, real anomalies and normal degradation phenomena are distinguished, and thus an adaptive monitoring scheme for the entire life cycle of a lithium battery stack with degradation perception ability is realized. The present invention for the first time establishes a degradation decoupling characterization and an adaptive health perception scheme under the entire life cycle. By identifying aging behaviors, degradation and real anomalies are carefully decoupled, thereby effectively improving the fault detection rate and reducing the false alarm rate, providing practical support for the fine health management and safe operation of the energy storage battery stack.

Description

Technical Field

[0001] The present invention relates to the technical field of health management of energy storage power stations, especially to the technical field of continuous degradation problems of lithium-ion battery stacks during operation in energy storage power stations, and particularly relates to a method for all-life adaptive health perception of energy storage battery stack degradation. Background Art

[0002] Benefiting from higher energy density and less pollution, lithium-ion batteries have become the most widely used choice in energy storage systems. And the safety problems of lithium-ion batteries have also emerged continuously, and their safety problems have attracted much attention. In recent decades, the rapid development of Internet of Things and artificial intelligence technologies has given rise to many data-based abnormal monitoring methods for lithium-ion batteries. Among these mainstream methods, a certain normal range is determined through historical data. Once the operation data deviates from the preset range, the abnormal behavior of the lithium battery can be identified in time.

[0003] However, the lithium battery module will inevitably experience performance degradation during long-term operation, such as the decrease of capacity and power, etc., which will lead to the deviation of its operation characteristics. Therefore, the operation data will have a slow time-varying problem, and the monitoring model developed for new batteries may gradually fail, resulting in a large number of false alarms. Although adaptive monitoring methods have been significantly developed in other industrial fields, these methods face challenges in terms of "whether to update". On the one hand, it is difficult for these methods to distinguish whether the online samples after degradation represent normal data drift or abnormality. When it is normal data drift, it needs to be updated, and when it is abnormal, the update should be terminated and an alarm should be triggered. On the other hand, these methods rely on the surface observation of online data to judge "whether to update", which leads to the ambiguity of phenomenon recognition and the unreliability of judgment. In addition, these methods are not suitable for modeling lithium battery modules. Lithium battery modules usually experience frequent charge and discharge cycles during operation, showing significant non-stationary characteristics. Existing methods have shown efficiency in separating non-stationary components from the steady state, however, they have failed to further identify non-stationary trends driven by different factors (such as degradation or charge and discharge behavior). Therefore, how to correctly characterize the degradation process in the whole life cycle of lithium batteries and establish a reliable adaptive monitoring model is still an urgent problem to be solved. Summary of the Invention

[0004] The purpose of the present invention is to provide a method for all-life adaptive health perception of energy storage battery stack degradation in view of the deficiencies of existing adaptive abnormal monitoring methods in capturing non-stationary data trends caused by degradation and distinguishing abnormalities.

[0005] The purpose of the present invention is achieved by the following technical solutions: A method for all-life adaptive health perception of energy storage battery stack degradation, comprising the following steps:

[0006] (1)Obtain the original offline data of the entire life cycle of the historical operation of the energy storage battery stack, and perform cluster feature fusion and cyclic window construction processing on the original offline data in the normal state to obtain the fusion feature sample data of multiple cyclic windows;

[0007] (2)Use the fusion feature sample data of the cyclic window for analysis through a dual-scale non-stationary degradation analysis framework. First, at the inner cycle scale, identify the charging behavior or discharging behavior within each charge-discharge cycle as an inner cycle non-stationary trend through a Gaussian mixture model. Then, at the cross-cycle scale, use the degradation component decoupling algorithm to decouple the fusion feature sample data to obtain the inherent projection matrix and the drift projection matrix;

[0008] (3)Obtain the historical service data of the energy storage battery stack to be monitored, and perform cluster feature fusion processing on it to obtain the corresponding fusion feature data; use the inherent projection matrix and the drift projection matrix to obtain the inherent features and drift features corresponding to the fusion feature data, and calculate the monitoring statistics of the inherent features and drift features respectively through a Gaussian mixture model. Use kernel density estimation to obtain the monitoring statistic control limits corresponding to the inherent features and drift features;

[0009] (4)When performing online adaptive anomaly monitoring on the energy storage battery stack to be monitored, take the service data collected in real time with the battery module as the basic analysis unit, and make a judgment according to the monitoring statistic control limits corresponding to the inherent features and drift features to determine the state of the energy storage battery stack to be monitored; and design an adaptive model update strategy to judge whether an update is required based on the aging degree of the online service data.

[0010] Further, the obtaining of the original offline data of the entire life cycle of the historical operation of the energy storage battery stack specifically includes:

[0011] The energy storage battery stack includes multiple battery modules. Taking the battery module as the smallest analysis unit, each battery module is composed of multiple battery cells connected in series and parallel. Each battery cell is a lithium-ion battery, which is the smallest unit of energy storage;

[0012] Collect the service data of all battery modules during the entire life cycle of the historical operation. The service data of the battery module includes the service state variable data of each battery cell in the battery module. The service state variables include the state of charge index and the measurement point variables. The measurement point variables include temperature, voltage, and current.

[0013] Further, the performing of cluster feature fusion and cyclic window construction processing on the original offline data in the normal state to obtain the fusion feature sample data of multiple cyclic windows specifically includes:

[0014] First, perform cluster feature fusion processing on the original offline data in the normal state. Specifically: Denote the service state variables of the battery module at any sampling time as , where represents the total number of battery cells in the battery module, and represents the total number of service state variables of the battery cells; After sequentially concatenating the service state variables of the battery cells, obtain the individual feature of the battery cell, denoted as , where represents the individual feature vector of the i-th battery cell in the battery module, and represents the j-th service state variable of the i-th battery cell in the battery module; After sequentially concatenating the individual features of all battery cells in the battery module, obtain the individual feature of the battery module; Calculate statistical indicators based on the individual feature of the battery module, and extract the cluster feature of the battery module, where represents the column vector of the -th statistical indicator, and represents the total number of statistical indicator types; Concatenate the individual feature of the battery module and its corresponding cluster feature to form a fusion feature, denoted as , where represents the total number of features of the fusion feature;

[0015] Then, construct a circular window for the obtained fusion feature, create a circular window containing multiple charge-discharge cycles, and include all the fusion feature samples collected from adjacent charge-discharge cycles in a circular window to obtain the fusion feature sample data of multiple circular windows. Among them, the fusion feature sample data of the -th circular window is denoted as , represents the total number of fusion feature samples in the -th circular window, represents the index of the first charge-discharge cycle within this circular window, and represents the number of fusion feature samples in the -th charge-discharge cycle.

[0016] Further, at the inner loop scale, identify the charging behavior or discharging behavior within each charge-discharge cycle as an inner loop non-stationary trend through a Gaussian mixture model. Specifically, it includes:

[0017] Identify the charging behavior or discharging behavior within each charge-discharge cycle as an inner loop non-stationary trend, and use the Gaussian mixture model to describe the data distribution of each circular window. For the Fusion feature sample data of a cyclic window , use the Gaussian mixture model to model it, and give the probability density function of the well-modeled Gaussian mixture model, expressed as:

[0018]

[0019] Among them, represents the probability density function of the Gaussian mixture model, represents the th single fusion feature sample data in a cyclic window, , and respectively represent the mixing weight, mean vector and covariance matrix of the mth Gaussian cost in the th charge-discharge cycle, and M represents the total number of Gaussian components, represents the Gaussian probability density function; among them, the mixing weight satisfies the constraint condition .

[0020] Furthermore, at the cross-cycle scale, use the degenerate component decoupling algorithm to decouple the fusion feature sample data to obtain the intrinsic projection matrix and the drift projection matrix, specifically including:

[0021] First, construct a cross-cycle intrinsic Gaussian mixture model based on the probability density function of the Gaussian mixture model of each cyclic window at the inner-cycle scale, so that the intrinsic features obtained from the original fusion feature sample data show similar data distributions in different cyclic windows; the probability density function of this cross-cycle intrinsic Gaussian mixture model is:

[0022]

[0023] Among them, represents the probability density function of the cross-cycle intrinsic Gaussian mixture model, , and respectively represent the mixing weight, mean vector and covariance matrix of the mth Gaussian cost of the cross-cycle intrinsic Gaussian mixture model;

[0024] Then, given the original model parameters of the cross-cycle intrinsic Gaussian mixture model of each cyclic window as , where represents the original model parameters of the cross-cycle intrinsic Gaussian mixture model of the th cyclic window, , and respectively represent the The original mixing weights, original mean vectors, and original covariance matrices of the m-th Gaussian component of the cross-cycle intrinsic Gaussian mixture model for each cycle window; the model parameters of the cross-cycle intrinsic Gaussian mixture model for each cycle window are estimated by the expectation-maximization algorithm. After Q iterations of estimation, the final model parameters of the cross-cycle intrinsic Gaussian mixture model for each cycle window are , which is simplified to , where , and respectively represent the final mixing weights, final mean vectors, and final covariance matrices of the m-th Gaussian component of the cross-cycle intrinsic Gaussian mixture model for the -th cycle window;

[0025] Secondly, solve the linear projection coefficient matrix according to the final model parameters of the cross-cycle intrinsic Gaussian mixture model for all cycle windows. Its optimization problem and constraint conditions are equivalent to:

[0026]

[0027] where represents the trace of the matrix, represents the identity matrix, W is the total number of cycle windows, and respectively represent the average of the mean vectors and the average of the covariance matrices of the m-th Gaussian component of the cross-cycle intrinsic Gaussian mixture model, are all intermediate parameters without practical significance;

[0028] Finally, use the singular value decomposition method to solve its optimization problem under the constraint conditions of the linear projection coefficient matrix, and obtain the closed-form solution of the optimization problem in the above formula. The closed-form solution is eigenvectors. Select the eigenvectors corresponding to the smallest R eigenvalues from the eigenvectors, and use them as the intrinsic projection matrix, denoted as ; The remaining eigenvectors among the eigenvectors form the drift projection matrix, denoted as ; where represents the r-th coefficient vector in, R is the dimension of the intrinsic features, is the dimension of the drift features.

[0029] Furthermore, the model parameters of the cross-cycle intrinsic Gaussian mixture model for each cycle window are estimated by the expectation-maximization algorithm, specifically including:

[0030] First, set the expectation step, specifically: at the -thWithin a cycle window, calculate the th training iteration of the th fused feature sample The posterior probability from the m-th Gaussian component, and its calculation formula is:

[0031]

[0032] where , and respectively represent the mixing weight, mean vector, and covariance matrix of the m-th Gaussian component of the cross-cycle intrinsic Gaussian mixture model of the th cycle window estimated in the th iteration;

[0033] Then set the maximum step, specifically: According to the value in the th iteration, estimate the Gaussian mixture model parameters in the th iteration, and its calculation formula is:

[0034]

[0035] where , and respectively represent the mixing weight, mean vector, and covariance matrix of the m-th Gaussian component of the cross-cycle intrinsic Gaussian mixture model of the th cycle window estimated in the th iteration.

[0036] Furthermore, step (3) specifically includes the following sub-steps:

[0037] (3.1) Obtain the historical service data of the energy storage battery stack to be monitored; where the energy storage battery stack includes multiple battery modules, and taking the battery module as the minimum analysis unit, the service data of the battery module includes the service status variable data of each single battery in the battery module, and the service status variables include the state of charge index and the measurement point variables;

[0038] (3.2) Perform cluster feature fusion processing on the historical service data of each battery module to be monitored to obtain the corresponding fused feature data , where represents the total number of fused feature data of the battery module to be monitored;

[0039] (3.3) Use the intrinsic projection matrix and the drift projection matrix Obtain the inherent characteristics and drift characteristics of the battery module to be monitored, and their calculation formulas are as follows:

[0040]

[0041]

[0042] Wherein, and respectively represent the inherent characteristics and drift characteristics of the battery module to be monitored;

[0043] (3.4) Calculate the monitoring statistics of the inherent characteristics and the drift characteristics respectively through the Gaussian mixture model. The monitoring statistic is the Bayesian inference distance, and its calculation formula is:

[0044]

[0045]

[0046] Wherein, is the Bayesian inference distance index, representing the monitoring statistic of the inherent characteristic or the drift characteristic ; represents the Gaussian mixture model parameter of the m-th Gaussian component, represents the mixing weight in the Gaussian mixture model parameter of the m-th Gaussian component; represents the inherent characteristic when represents the drift characteristic when; represents to the -th Mahalanobis distance of the Gaussian component, represents the posterior probability corresponding to the inherent characteristic or the drift characteristic;

[0047] (3.5) Use the kernel density estimation method to statistically analyze the monitoring statistics of the inherent characteristics and drift characteristics calculated in step (3.4) to obtain the monitoring statistic control limits corresponding to the inherent characteristics and drift characteristics.

[0048] Furthermore, step (4) specifically includes:

[0049] When performing online adaptive anomaly monitoring on the energy storage battery stack to be monitored, the service data collected in real time is taken as the basic analysis unit of the battery module. The service data of the battery module includes the service state variable data of each battery cell, and the service state variables include the state of charge index and the measurement point variable; when a new service data sample arrives, first perform cluster feature fusion processing to obtain a new fused feature vector , and use the inherent projection matrix and the drift projection matrix Calculate its inherent characteristics and drift characteristics, and their calculation formulas are respectively:

[0050]

[0051]

[0052] Wherein, and respectively represent the inherent characteristics and drift characteristics of the new fused feature vector ;

[0053] Then calculate the monitoring statistics of the inherent characteristics and the drift characteristics , and their calculation formula is:

[0054]

[0055] Wherein, represents the monitoring statistic of the inherent characteristics of the new in-service data sample, represents the monitoring statistic of the drift characteristics of the new in-service data sample; represents the total number of Gaussian components in the inherent Gaussian mixture model, represents the total number of Gaussian components in the drift Gaussian mixture model; represents the model parameter of the m-th Gaussian component in the inherent Gaussian mixture model, represents the model parameter of the m-th Gaussian component in the drift Gaussian mixture model; represents to the Mahalanobis distance of the m-th Gaussian component in the inherent Gaussian mixture model, represents to the Mahalanobis distance of the m-th Gaussian component in the drift Gaussian mixture model; represents the posterior probability of the inherent Gaussian mixture model corresponding to the inherent feature distribution, represents the posterior probability of the drift Gaussian mixture model corresponding to the drift feature;

[0056] Then judge whether the monitoring statistic of the inherent characteristics of the calculated new in-service data sample is greater than the monitoring statistic control limit corresponding to the inherent characteristics, and whether the monitoring statistic of the drift characteristics of the new in-service data sample is greater than the monitoring statistic control limit corresponding to the drift characteristics. If is less than or equal to the monitoring statistic control limit corresponding to the inherent characteristics, and is less than or equal to the monitoring statistic control limit corresponding to the drift characteristics, then it is considered that the state of the energy storage battery stack to be monitored is normal; otherwise, it is considered that the state of the energy storage battery stack to be monitored is abnormal.

[0057] Furthermore, the designed adaptive model update strategy determines whether an update is needed based on the aging degree of the online service data, specifically including:

[0058] For the inherent characteristics, a Gaussian mixture model for monitoring is designed, which is called the inherent Gaussian mixture model; for the drift characteristics, a recursive Gaussian mixture model is established, which is called the drift Gaussian mixture model.

[0059] An adaptive model update strategy is designed for the inherent Gaussian mixture model and the drift Gaussian mixture model. Specifically, by analyzing the degradation components, it is determined whether the inherent Gaussian mixture model or the drift Gaussian mixture model needs to be updated and an alarm is triggered.

[0060] Among them, the adaptive model update strategy specifically includes:

[0061] Case 1: When the monitoring statistic of the inherent characteristics of the calculated service data sample is less than or equal to the monitoring statistic control limit corresponding to the inherent characteristics, and the monitoring statistic of the drift characteristics is less than or equal to the monitoring statistic control limit corresponding to the drift characteristics, the state of the energy storage battery stack to be monitored at this time is regarded as the normal state, and new service data samples are collected to update the model parameters of the drift Gaussian mixture model. When the size of the collected data sample block reaches the preset value the model parameters of the drift Gaussian mixture model and the monitoring statistic control limit corresponding to the drift characteristics are updated recursively.

[0062] Case 2: When the monitoring statistic of the inherent characteristics of the calculated service data sample is greater than the monitoring statistic control limit corresponding to the inherent characteristics, and the monitoring statistic of the drift characteristics is less than or equal to the monitoring statistic control limit corresponding to the drift characteristics, the state of the energy storage battery stack to be monitored at this time is regarded as the abnormal state, and the update of the drift Gaussian mixture model is stopped and an alarm is triggered.

[0063] Case 3: When the monitoring statistic of the inherent characteristics of the calculated service data sample is less than or equal to the monitoring statistic control limit corresponding to the inherent characteristics, and the monitoring statistic of the drift characteristics is greater than the monitoring statistic control limit corresponding to the drift characteristics, a recursive strategy is used with the collected service data samples, and then the monitoring statistic of the drift characteristics is recalculated and the monitoring statistic control limit of the drift characteristics is judged again; afterwards, if the monitoring statistic of the drift characteristics of the calculated service data sample is still greater than the monitoring statistic control limit corresponding to the drift characteristics, an alarm is triggered; if the monitoring statistic of the drift characteristics of the calculated service data sample is less than or equal to the monitoring statistic control limit corresponding to the drift characteristics, it is considered that the energy storage battery stack to be monitored is in the normal state.

[0064] Case 4: When the monitoring statistic of the inherent feature of the calculated in-service data sample is greater than the monitoring statistic control limit corresponding to the inherent feature, and the monitoring statistic of the drift feature is greater than the monitoring statistic control limit corresponding to the drift feature, the update of the Gaussian mixture model is stopped and an alarm is triggered.

[0065] The beneficial effects of the present invention are as follows: The present invention first establishes a decoupled degradation representation and an adaptive health perception scheme under the full life cycle. By identifying aging behaviors, it carefully decouples degradation and real anomalies, provides a basis for model update, and thus effectively improves the fault detection rate and reduces the false alarm rate, providing practical support for the fine health management and safe operation of the energy storage battery stack; The present invention can jointly identify and process the non-stationary trends caused by charging or discharging behaviors and aging, and at the same time, clearly observe the degradation process; On this basis, the present invention establishes an adaptive monitoring scheme with reliable discrimination between decline and anomalies, realizes a clear representation of the degradation process of lithium battery units, and provides an integrated update strategy for the monitoring model based on aging perception. After the lithium battery ages, it maintains a high-sensitivity recognition ability for real anomalies. Brief Description of the Drawings

[0066] Figure 1 is a flowchart of the method for adaptive health perception of the degradation of the energy storage battery stack of the present invention;

[0067] Figure 2 is an architecture flowchart of the adaptive model update strategy of the present invention;

[0068] Figure 3 is an example diagram of the monitoring result of the present invention in Example 1; among them, Figure 3 in (a) is the monitoring result diagram of the inherent feature, Figure 3 in (b) is the monitoring result diagram of the drift feature;

[0069] Figure 4 is an example diagram of the monitoring result of the present invention in Example 2; among them, Figure 4 in (a) is the monitoring result diagram of the inherent feature, Figure 4 in (b) is the monitoring result diagram of the drift feature;

[0070] Figure 5 is the monitoring effect diagram of the Gaussian mixture model method of the present invention in Example 1;

[0071] Figure 6 is the monitoring effect diagram of the slow feature analysis method of the present invention in Example 1;

[0072] Figure 7 is the monitoring effect diagram of the Gaussian mixture model method of the present invention in Example 2;

[0073] Figure 8 Monitoring effect diagram of the slow feature analysis method of the present invention in Example 2. Detailed implementation manners

[0074] Here, the exemplary embodiments will be described in detail, and the examples are shown in the drawings. When the following description refers to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The implementation manners described in the following exemplary embodiments do not represent all implementation manners consistent with the present invention. On the contrary, they are merely examples of devices and methods consistent with some aspects of the present invention as detailed in the appended claims.

[0075] The terms used in the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention. The singular forms "a", "said" and "the" used in the present invention and the appended claims are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term "and / or" used herein refers to and includes any or all possible combinations of one or more of the associated listed items.

[0076] It should be understood that although the terms first, second, third, etc. may be used in the present invention to describe various information, such information should not be limited to these terms. These terms are only used to distinguish the same type of information from each other. For example, without departing from the scope of the present invention, the first information may also be referred to as the second information, and similarly, the second information may also be referred to as the first information. Depending on the context, the word "if" as used herein may be interpreted as "when" or "while" or "in response to determining".

[0077] The present invention will be described in detail below with reference to the drawings. Without conflict, the features in the following embodiments and implementation manners can be combined with each other.

[0078] Refer to Figure 1 , the all-life adaptive health perception method for degradation of the energy storage battery stack of the present invention specifically includes the following steps:

[0079] (1) Obtain the original off-line data of the entire life cycle of the historical operation of the energy storage battery stack, and perform cluster feature fusion and cyclic window construction processing on the original off-line data in the normal state to obtain the fusion feature sample data of multiple cyclic windows for subsequent analysis.

[0080] It should be understood that when performing cluster feature fusion and cyclic window construction processing on the original offline data, only the original offline data in the normal state is processed as described above, and discrete data in abnormal / fault conditions is not included. This is because using data in the normal state for offline modeling and setting control limits based on the normal state data facilitates subsequent online application phases, where the health of the battery can be judged online by determining whether the sample exceeds the limit.

[0081] Furthermore, obtain the original offline data of the entire life cycle of the historical operation of the energy storage battery stack, specifically including: the energy storage battery stack includes multiple battery modules. Taking the battery module as the smallest analysis unit, each battery module is composed of multiple battery cells connected in series and parallel, and each battery cell is a lithium-ion battery, which is the smallest unit of energy storage; collect the service data of all battery modules during the entire life cycle of the historical operation. The service data of the battery module includes the service state variable data of each battery cell in the battery module, and the service state variables include the state of charge index and measurement point variables, and the measurement point variables include temperature, voltage, current, etc.

[0082] Exemplarily, in this embodiment, obtain offline modeling samples: Collect 1.5 years of original offline data from the battery module as offline training data. This battery module is composed of 16 battery cells connected in series, and each battery cell is a lithium-ion battery, which is the smallest unit of energy storage; set the sampling interval to 1 minute and regularly collect the service data of this battery module. The service data of this battery module includes the service state variable data of 16 battery cells, and the service state variables include the state of charge index and measurement point variables (such as temperature, voltage, current, etc.).

[0083] Furthermore, perform cluster feature fusion and cyclic window construction processing on the original offline data in the normal state to obtain the fused feature sample data of multiple cyclic windows, specifically including: First, perform cluster feature fusion processing on the original offline data in the normal state. Specifically: Denote the service state variables of the battery module at any sampling time as , where represents the total number of battery cells in the battery module, represents the total number of service state variables of the battery cell; After sequentially splicing the service state variables of the battery cell, obtain the individual feature of this battery cell, denoted as , where represents the individual feature vector of the i-th battery cell in the battery module, represents the j-th service state variable of the i-th battery cell in the battery module; After sequentially splicing the individual features of all battery cells in the battery module, obtain the individual feature ; In addition, in order to describe the overall state of the entire battery module, statistical metrics such as mean, variance, etc. are calculated based on the individual characteristics of the battery module, and the cluster characteristics of the battery module are extracted according to the statistical metrics , where represents the column vector of the th statistical metric, represents the total number of statistical metric types; the individual characteristics of the battery module and its corresponding cluster characteristics are concatenated to form a fused feature, denoted as , where represents the total number of features of the fused feature. In this embodiment, is 16; is 3; four statistical metrics, namely mean, variance, maximum value, and minimum value, are used to extract the cluster characteristics of the battery module, that is ; is 60. Then, a cyclic window is constructed for the obtained fused feature, creating a cyclic window containing multiple charge-discharge cycles as the basic unit for subsequent analysis. All fused feature samples collected from adjacent charge-discharge cycles are included in one cyclic window, obtaining fused feature sample data for multiple cyclic windows, where the fused feature sample data of the th cyclic window is denoted as , represents the total number of fused feature samples in the th cyclic window, represents the index of the first charge-discharge cycle within the cyclic window, represents the number of fused feature samples in the th charge-discharge cycle. In this embodiment, 4 cyclic windows are uniformly sampled from the offline data of the battery module, and each window contains 12 charge-discharge cycles, that is is 12.

[0084] (2) Using the fused feature sample data of the cyclic window, analysis is performed through a dual-scale non-stationary degradation analysis framework. First, at the inner-loop scale, the charging behavior or discharging behavior within each charge-discharge cycle is identified as an inner-loop non-stationary trend through a Gaussian mixture model (GMM). Then, at the cross-cycle scale, the fused feature sample data is decoupled using a degradation component decoupling algorithm to obtain an inherent projection matrix and a drift projection matrix. Among them, the dual-scale non-stationary degradation analysis framework includes an inner-loop scale and a cross-cycle scale.

[0085] Further, at the inner loop scale, the charging behavior or discharging behavior within each charge-discharge cycle is identified as an inner loop non-stationary trend through the Gaussian mixture model, specifically including: identifying the charging behavior or discharging behavior within each charge-discharge cycle as an inner loop non-stationary trend, using the Gaussian mixture model to describe the data distribution of each cycle window. For the fused feature sample data of the th cycle window , the Gaussian mixture model is used to model it, and the probability density function of the well-modeled Gaussian mixture model is given, expressed as:

[0086]

[0087] where represents the probability density function of the Gaussian mixture model, represents the single fused feature sample data in the th cycle window, , and respectively represent the mixing weight, mean vector and covariance matrix of the mth Gaussian cost in the th charge-discharge cycle. M represents the total number of Gaussian components, represents the Gaussian probability density function; where the mixing weight satisfies the constraint condition .

[0088] Further, at the cross-cycle scale, the degenerate component decoupling algorithm is used to decouple the fused feature sample data to obtain the inherent projection matrix and the drift projection matrix, specifically including: first, a set of original feature fusion sample data is found, and a cross-cycle inherent Gaussian mixture model is constructed based on the probability density function of the Gaussian mixture model of each cycle window at the corresponding inner loop scale, so that the inherent features of the original fused feature sample data present similar data distributions in different cycle windows; this cross-cycle inherent Gaussian mixture model can reveal the inherent features of the fused feature sample data in different cycle windows, and the probability density function of the cross-cycle inherent Gaussian mixture model is:

[0089]

[0090] where represents the probability density function of the cross-cycle inherent Gaussian mixture model, , and respectively represent the mixing weight, mean vector and covariance matrix of the mth Gaussian cost of the cross-cycle inherent Gaussian mixture model. Then, the original model parameters of the cross-cycle inherent Gaussian mixture model for each cycle window are given as , where Denote the original model parameters of the cross-cycle intrinsic Gaussian mixture model for the th cycle window, , and respectively denote the original mixing weight, original mean vector, and original covariance matrix of the m-th Gaussian component of the cross-cycle intrinsic Gaussian mixture model for the th cycle window; the model parameters of the cross-cycle intrinsic Gaussian mixture model for each cycle window are estimated by the expectation-maximization algorithm. After Q iterations of estimation, the final model parameters of the cross-cycle intrinsic Gaussian mixture model for each cycle window are , which are simplified to , where , and respectively denote the final mixing weight, final mean vector, and final covariance matrix of the m-th Gaussian component of the cross-cycle intrinsic Gaussian mixture model for the th cycle window. After Q parameter estimations, the cross-cycle intrinsic Gaussian mixture model can be used to reveal the original data distribution. Secondly, assume that the linear projection coefficient matrix from the original fused feature sample data to the intrinsic features and drift features is . The corresponding intrinsic features and drift features are obtained using the corresponding linear projection coefficient matrix, that is, the linear projection coefficient matrix is solved based on the final model parameters of the cross-cycle intrinsic Gaussian mixture model for all cycle windows. Its optimization problem and constraint conditions are equivalent to:

[0091]

[0092] where denotes the trace of the matrix, denotes the identity matrix, W is the total number of cycle windows, and respectively denote the average of the mean vectors and the average of the covariance matrices of the m-th Gaussian component of the cross-cycle intrinsic Gaussian mixture model, are all intermediate parameters without practical significance. Finally, the singular value decomposition method is used to solve its optimization problem under the constraint conditions of the linear projection coefficient matrix, and a closed-form solution to the optimization problem in the above formula is obtained. This closed-form solution is eigenvectors. Among the eigenvectors, the eigenvectors corresponding to the smallest R eigenvalues are selected and used as the intrinsic projection matrix, denoted as ; the remaining eigenvectors among the eigenvectors form the drift projection matrix, denoted as ; where denotes the r-th coefficient vector in , and R is the dimension of the intrinsic features. is the dimension of the drift feature.

[0093] It should be noted that the intrinsic projection matrix is ​​obtained and the drift projection matrix Then, use the intrinsic projection matrix Get fusion feature sample data Inherent characteristics , using the drift projection matrix Get fusion feature sample data Drift characteristics Among them, the intrinsic characteristics and drift characteristics obtained show significant deviations in the degradation process of lithium-ion batteries; the intrinsic characteristics are used to characterize the long-term constant characteristics of lithium-ion batteries, and the drift characteristics are used to characterize the degradation behavior of lithium-ion batteries. In addition, through the intrinsic projection matrix and the drift projection matrix The inherent characteristics and drift characteristics of the fused feature sample data can be obtained, and its data distribution can also be described by a Gaussian mixture model, where the model parameters of the cross-cycle inherent Gaussian mixture model are: , , ; The model parameters of the cross-cycle drift Gaussian mixture model are: , , .

[0094] Furthermore, the model parameters of the cross-cycle intrinsic Gaussian mixture model of each cycle window are estimated by the expectation maximization algorithm, which specifically includes: first setting the expected step, specifically: In the cycle window, calculate the In the training iteration Fusion feature samples The posterior probability from the mth Gaussian component , and its calculation formula is:

[0095]

[0096] in, , and Respectively represent The estimated value of the first The mixing weights, mean vectors and covariance matrices of the mth Gaussian component of the cross-cycle intrinsic Gaussian mixture model of the cycle window. Then set the maximum step, specifically: according to the In iteration The value of The Gaussian mixture model parameters for the iteration , and its calculation formula is:

[0097]

[0098] Among them, 、 and respectively represent the mixing weight, mean vector and covariance matrix of the m-th Gaussian component of the cross-cycle intrinsic Gaussian mixture model of the th iteration estimate of the th cycle window.

[0099] (3) Obtain the historical service data of the energy storage battery stack to be monitored, and perform cluster feature fusion processing on it to obtain the corresponding fusion feature data; use the intrinsic projection matrix and the drift projection matrix to obtain the intrinsic features and drift features corresponding to the fusion feature data, and calculate the monitoring statistics of the intrinsic features and drift features respectively through the Gaussian mixture model, and obtain the monitoring statistic control limits corresponding to the intrinsic features and drift features by using kernel density estimation.

[0100] (3.1) Obtain the historical service data of the energy storage battery stack to be monitored. Similarly, the energy storage battery stack includes multiple battery modules. Taking the battery module as the smallest analysis unit, the service data of the battery module includes the service state variable data of each single battery in the battery module. The service state variables include the state of charge index and the measured point variables, which are the same as the offline data.

[0101] (3.2) Perform cluster feature fusion processing on the historical service data of each battery module to be monitored to obtain the corresponding fusion feature data , where represents the total number of fusion feature data of the battery module to be monitored.

[0102] (3.3) Use the intrinsic projection matrix and the drift projection matrix to obtain the intrinsic features and drift features of the battery module to be monitored. Their calculation formulas are respectively:

[0103]

[0104]

[0105] Among them, and respectively represent the intrinsic features and drift features of the battery module to be monitored.

[0106] (3.4) Calculate the monitoring statistics of the intrinsic feature and the drift feature respectively through the Gaussian mixture model. The monitoring statistic is the Bayesian inference distance, and its calculation formula is:

[0107]

[0108]

[0109] Among them, is the Bayesian inference distance index, representing the monitoring statistic of the inherent feature or the drift feature ; represents the Gaussian mixture model parameters of the m-th Gaussian component, and represents the mixing weight in the Gaussian mixture model parameters of the m-th Gaussian component; represents the inherent feature when and represents the drift feature when represents to the -th Mahalanobis distance of the Gaussian component, and represents the posterior probability corresponding to the inherent feature or the drift feature.

[0110] (3.5) Use the kernel density estimation method to statistically analyze the monitoring statistics of the inherent feature and the drift feature calculated in step (3.4), and obtain the monitoring statistic control limits corresponding to the inherent feature and the drift feature. In this embodiment, the confidence levels of the monitoring statistic control limits corresponding to the inherent feature and the drift feature are both 99.9%.

[0111] When performing online adaptive anomaly monitoring on the energy storage battery stack to be monitored, the service data collected in real time is taken as the basic analysis unit with the battery module, and the state of the energy storage battery stack to be monitored is judged according to the monitoring statistic control limits corresponding to the inherent feature and the drift feature; and an adaptive model update strategy is designed to judge whether an update is required based on the aging degree of the online service data, as Figure 1 shown.

[0112] Specifically, when performing online adaptive anomaly monitoring on the energy storage battery stack to be monitored, the service data collected in real time is taken as the basic analysis unit with the battery module. The service data of the battery module includes the service state variable data of each battery cell, and the service state variables include the state of charge index and the measurement point variable, which are consistent with the offline data. When a new service data sample arrives, first perform cluster feature fusion processing to obtain a new fused feature vector , and use the inherent projection matrix and the drift projection matrix to calculate its inherent feature and drift feature, and their calculation formulas are respectively:

[0113]

[0114]

[0115] Among them, and respectively represent the inherent feature and the drift feature of the new fusion feature vector . Then calculate the monitoring statistics of the inherent feature and the drift feature . The calculation formula is as follows:

[0116]

[0117] Among them, represents the monitoring statistic of the inherent feature of the new in-service data sample, represents the monitoring statistic of the drift feature of the new in-service data sample; represents the total number of Gaussian components in the inherent Gaussian mixture model, represents the total number of Gaussian components in the drift Gaussian mixture model; represents the model parameter of the m-th Gaussian component in the inherent Gaussian mixture model, represents the model parameter of the m-th Gaussian component in the drift Gaussian mixture model; represents to the Mahalanobis distance to the m-th Gaussian component in the inherent Gaussian mixture model, represents to the Mahalanobis distance to the m-th Gaussian component in the drift Gaussian mixture model; represents the posterior probability of the inherent Gaussian mixture model corresponding to the inherent feature distribution, represents the posterior probability of the drift Gaussian mixture model corresponding to the drift feature; the inherent Gaussian mixture model is a Gaussian mixture model for monitoring designed for the inherent feature, and the drift Gaussian mixture model is a recursive Gaussian mixture model established for the drift feature. In this embodiment, the size of m is determined by the Bayesian information criterion. Then judge whether the monitoring statistic of the inherent feature of the calculated new in-service data sample is greater than the monitoring statistic control limit corresponding to the inherent feature, and whether the monitoring statistic of the drift feature of the new in-service data sample is greater than the monitoring statistic control limit corresponding to the drift feature. If is less than or equal to the monitoring statistic control limit corresponding to the inherent feature, and is less than or equal to the monitoring statistic control limit corresponding to the drift feature, it is considered that the state of the energy storage battery stack to be monitored is normal; otherwise, in any other case, it is considered that the state of the energy storage battery stack to be monitored is abnormal.

[0118] It should be understood that the modeling of the inherent Gaussian mixture model and the drift Gaussian mixture model is similar to that of the cross-cycle inherent Gaussian mixture model and the cross-cycle drift Gaussian mixture model, and the included model parameters are also similar, both including mixture weights, mean vectors, and covariance matrices, except that the numerical values of their respective model parameters are different.

[0119] Furthermore, an adaptive model update strategy is designed to determine whether an update is required based on the aging degree of the online service data. As Figure 2 shown, it specifically includes: for the inherent characteristics, a Gaussian mixture model for monitoring is designed and referred to as the inherent Gaussian mixture model; for the drift characteristics, a recursive Gaussian mixture model is established and referred to as the drift Gaussian mixture model. An adaptive model update strategy is designed for the inherent Gaussian mixture model and the drift Gaussian mixture model, specifically by analyzing the degraded components to determine whether an update and alarm are required for the inherent Gaussian mixture model or the drift Gaussian mixture model. Among them, the adaptive model update strategy specifically includes the following four main situations and their corresponding model update strategies:

[0120] Situation 1: When the monitoring statistic of the inherent characteristics of the calculated service data sample is less than or equal to the monitoring statistic control limit corresponding to the inherent characteristics, and the monitoring statistic of the drift characteristics is less than or equal to the monitoring statistic control limit corresponding to the drift characteristics, the state of the energy storage battery stack to be monitored at this time is regarded as the normal state, and new service data samples are collected to update the model parameters of the drift Gaussian mixture model. When the size of the collected data sample block reaches the preset value , the model parameters of the drift Gaussian mixture model and the monitoring statistic control limit corresponding to the drift characteristics are updated in a recursive manner.

[0121] Furthermore, the model parameters of the drift Gaussian mixture model and the monitoring statistic control limit corresponding to the drift characteristics are updated in a recursive manner. Specifically, in the th iteration, the mixture weights are refreshed according to each drift characteristic within the data block :

[0122]

[0123] Among them, is the forgetting ratio, and the superscript represents the parameter in the bth iteration. The mixture weights are normalized to satisfy the constraint . The mean and covariance of the mth Gaussian component are updated as:

[0124]

[0125] Among them, represents the mean of the mth Gaussian component in the bth iteration, represents the covariance of the m-th Gaussian component in the b-th iteration. In this embodiment, is 200, is 0.01.

[0126] Case 2: When the monitoring statistic of the inherent feature of the calculated in-service data sample is greater than the monitoring statistic control limit corresponding to the inherent feature, and the monitoring statistic of the drift feature is less than or equal to the monitoring statistic control limit corresponding to the drift feature, this situation indicates that an abnormal situation may have occurred that breaks the long-term stable operation rule. At this time, the state of the energy storage battery stack to be monitored is regarded as an abnormal state, and the update of the drift Gaussian mixture model is stopped and an alarm is triggered.

[0127] Case 3: When the monitoring statistic of the inherent feature of the calculated in-service data sample is less than or equal to the monitoring statistic control limit corresponding to the inherent feature, and the monitoring statistic of the drift feature is greater than the monitoring statistic control limit corresponding to the drift feature, it means that the long-term stable operation rule remains normal, but there is a significant deviation in the drift part. In this case, there are two possibilities: one is that the operation deviation caused by degradation significantly exceeds the control limit, and the other is that an abnormality has occurred, resulting in abnormal data features of the drift feature. Therefore, it is necessary to further identify the operation state. Specifically, a recursive strategy is used with the collected in-service data samples, and then the monitoring statistic of the drift feature is recalculated and the monitoring statistic control limit of the drift feature is judged again; afterwards, if the monitoring statistic of the drift feature of the calculated in-service data sample is still greater than the monitoring statistic control limit corresponding to the drift feature, an alarm is triggered; if the monitoring statistic of the drift feature of the calculated in-service data sample is less than or equal to the monitoring statistic control limit corresponding to the drift feature, it is considered that the energy storage battery stack to be monitored is in a normal state.

[0128] Case 4: When the monitoring statistic of the inherent feature of the calculated in-service data sample is greater than the monitoring statistic control limit corresponding to the inherent feature, and the monitoring statistic of the drift feature is greater than the monitoring statistic control limit corresponding to the drift feature, it means that abnormal operations may have occurred in both the inherent part and the drift part of the energy storage battery stack to be monitored, and the update of the Gaussian mixture model is stopped and an alarm is triggered.

[0129] Exemplarily, the data of two different battery modules (i.e., battery module 2 and battery module 3) are respectively used for abnormal monitoring analysis below to illustrate the effect of the method of the present invention.

[0130] A total of 8331 unaged normal operation samples were collected from battery module 2 and battery module 3 respectively for the initialization of the monitoring model.

[0131] In terms of preventing false alarms for normal aging samples, Example 1 includes 1124 samples from battery module 2, collected after one year of operation, for testing the adaptive monitoring scheme. The results are asFigure 3 as shown in (a) of Figure 3 and (b) of Figure 4 as shown in (a) of Figure 4 and (b) of

[0132] To more clearly demonstrate the superiority of the method described in the present invention in the anomaly monitoring task, the Gaussian Mixture Model (GMM) method, the Slow Feature Analysis (SFA) method and the method described in the present invention are selected for comparison here. The monitoring statistic BID of the GMM method on two instances is statistically analyzed to verify the fault monitoring effect of the GMM method, and the results are as shown in Figure 5 and Figure 7 ; the significance difference index of the SFA method on two instances is statistically analyzed to verify the fault monitoring effect of the SFA method, and the results are as shown in Figure 6 and Figure 8 It can be found from this that in the GMM method in Instance 1, due to the data offset caused by degradation, there will be certain false alarms between about the 700th and 800th samples; in Instance 2, neither of the two traditional methods can detect the fault and give an alarm in time.

[0133] For a more intuitive comparison, Table 1 lists the false alarm rate and monitoring accuracy rate of different methods in Instance 1, and Table 2 lists the missed alarm rate and monitoring accuracy rate of different methods in Instance 2.

[0134] Table 1: False Alarm Rate and Monitoring Accuracy Rate of Different Methods in Instance 1

[0135]

[0136] Table 2: Missed Alarm Rate and Monitoring Accuracy Rate of Different Methods in Instance 2

[0137]

[0138] As can be seen from Table 1, in Example 1, the method of the present invention can effectively reduce the abnormal false alarms caused by aging compared with other methods. As can be seen from Table 2, in Example 2, the method of the present invention discovers abnormalities more accurately than other methods, and the number of missed alarm samples is significantly reduced. These two examples prove the effectiveness of the method of the present invention.

[0139] The above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements for some of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A full-life adaptive health perception method for energy storage battery stack degradation, characterized in that: The following steps are involved: (1) Obtaining the original offline data of the entire life cycle of the historical operation of the energy storage battery stack, and performing cluster feature fusion and cyclic window construction processing on the original offline data in the normal state to obtain fused feature sample data of multiple cyclic windows; (2) Utilizing the fused feature sample data of the cycle window, the dual-scale non-stationary degradation analysis framework is used for analysis. First, at the inner cycle scale, the charging behavior or discharging behavior in each charge and discharge cycle is identified as an inner cycle non-stationary trend through the Gaussian mixture model. Then, at the cross-cycle scale, the degradation component decoupling algorithm is used to decouple the fused feature sample data to obtain the intrinsic projection matrix and the drift projection matrix. The method uses a degenerate component decoupling algorithm to decouple the fused feature sample data under the cross-cycle scale to obtain an intrinsic projection matrix and a drift projection matrix, specifically including: First, a cross-loop intrinsic Gaussian mixture model is constructed based on the probability density function of the Gaussian mixture model of each loop window under the inner loop scale, so that the intrinsic feature d obtained by the original fusion feature sample data in Similar data distribution is presented in different cycle windows; the probability density function of the cross-cycle intrinsic Gaussian mixture model is: Among them, p(d in ) represents the probability density function of the cross-cyclic intrinsic Gaussian mixture model, and denote the mixture weight, mean vector and covariance matrix of the mth Gaussian cost of the cross-cyclic intrinsic Gaussian mixture model respectively; Then the original model parameters of the cross-cyclic intrinsic Gaussian mixture model for each cyclic window are given by in represents the original model parameters of the cross-cyclic intrinsic Gaussian mixture model for the w-th cyclic window, and They represent the original mixing weight, original mean vector and original covariance matrix of the mth Gaussian component of the cross-cycle intrinsic Gaussian mixture model of the wth cycle window respectively; the model parameters of the cross-cycle intrinsic Gaussian mixture model of each cycle window are estimated by the expectation maximization algorithm. After Q iterative estimates, the final model parameters of the cross-cycle intrinsic Gaussian mixture model of each cycle window are obtained as Simplify it to in and denote the final mixing weight, final mean vector and final covariance matrix of the m-th Gaussian component of the cross-cycle intrinsic Gaussian mixture model of the w-th cycle window respectively; Secondly, the linear projection coefficient matrix A is solved according to the final model parameters of the cross-cyclic intrinsic Gaussian mixture model of all cyclic windows. The optimization problem and constraints are equivalent to: A=argminTr[ASA T ] Where Tr[·] represents the trace of the matrix, I represents the identity matrix, W is the total number of cyclic windows, and denote the mean vector average and covariance matrix average of the mth Gaussian component of the cross-cyclic intrinsic Gaussian mixture model, S, They are all intermediate parameters with no practical significance; Finally, the singular value decomposition method is used to solve the optimization problem under the constraints of the linear projection coefficient matrix, and the closed-form solution of the optimization problem in the above formula is obtained. The closed-form solution is J e feature vectors, in J e Select the eigenvectors corresponding to the smallest R eigenvalues ​​from the eigenvectors and use them as the intrinsic projection matrix, expressed as J e The remaining eigenvectors in the eigenvectors form the drift projection matrix, which is expressed as where a r Indicates A in The rth coefficient vector in, R is the dimension of the inherent feature, J e -R is the dimension of the drift feature; (3) Obtain the historical service data of the energy storage battery stack to be monitored, and perform cluster feature fusion processing on it to obtain the corresponding fused feature data; use the inherent projection matrix and the drift projection matrix to obtain the inherent features and drift features corresponding to the fused feature data, and calculate the monitoring statistics of the inherent features and drift features respectively through the Gaussian mixture model, and use kernel density estimation to obtain the monitoring statistic control limits corresponding to the inherent features and drift features; The step (3) specifically includes the following sub-steps: (3.1) Obtaining historical service data of the energy storage battery stack to be monitored; wherein the energy storage battery stack includes a plurality of battery modules, with the battery module being the smallest analysis unit, and the service data of the battery module includes service status variable data of each single battery in the battery module, and the service status variable includes a state of charge index and a measurement point variable; (3.2) Perform cluster feature fusion processing on the historical service data of each battery module to be monitored to obtain the corresponding fusion feature data Where N f Indicates the total number of fused feature data of the battery module to be monitored; (3.3) Using the intrinsic projection matrix A in and the drift projection matrix A dr Obtain the inherent characteristics and drift characteristics of the battery module to be monitored, and the calculation formulas are: D in =A in X D dr =A dr X Among them, D in and D dr They respectively represent the inherent characteristics and drift characteristics of the battery module to be monitored; (3.4) Calculate the intrinsic features D respectively through the Gaussian mixture model in and drift characteristics D dr The monitoring statistic is the Bayesian inference distance, and its calculation formula is: Among them, BID is the Bayesian inference distance indicator, which represents the inherent feature D in Or drift characteristic D dr Monitoring statistics of m represents the Gaussian mixture model parameter of the mth Gaussian component, α m represents the mixing weight in the Gaussian mixture model parameters of the mth Gaussian component; d∈D in When , it represents the inherent characteristics, d∈D dr When D l (d,Θ m ) represents the Mahalanobis distance from d to the mth Gaussian component, p(Θ m |d) represents the posterior probability corresponding to the inherent feature or drift feature; (3.5) Using the kernel density estimation method to perform statistics on the monitoring statistics of the inherent characteristics and drift characteristics calculated in step (3.4), obtain the monitoring statistics control limits corresponding to the inherent characteristics and drift characteristics; (4) When performing online adaptive abnormal monitoring on the energy storage battery stack to be monitored, the service data collected in real time is analyzed with the battery module as the basic unit, and the status of the energy storage battery stack to be monitored is judged according to the control limits of the monitoring statistics corresponding to the inherent characteristics and drift characteristics; and an adaptive model update strategy is designed to determine whether an update is needed based on the aging degree of the online service data.

2. The method for adaptively sensing the degradation of energy storage battery stack over its entire life cycle according to claim 1 is characterized in that: The acquisition of the original offline data of the entire life cycle of the historical operation of the energy storage battery stack specifically includes: The energy storage battery stack includes a plurality of battery modules, with the battery module being the smallest analysis unit. Each battery module is composed of a plurality of battery cells connected in series and parallel. Each battery cell is a lithium-ion battery, which is the smallest unit of energy storage. The service data of all battery modules during the entire life cycle of historical operation are collected. The service data of the battery module includes the service status variable data of each battery cell in the battery module. The service status variables include charge state indicators and measurement point variables. The measurement point variables include temperature, voltage, and current.

3. The method for adaptively sensing the degradation of energy storage battery stack over its entire life cycle according to claim 1 is characterized in that: The cluster feature fusion and cyclic window construction processing are performed on the original offline data in the normal state to obtain fused feature sample data of multiple cyclic windows, specifically including: First, the original offline data in the normal state is subjected to cluster feature fusion processing, specifically: the service state variable of the battery module at any sampling time is recorded as Where I represents the total number of battery cells in the battery module, and J represents the total number of service status variables of the battery cells. The service status variables of the battery cells are sequentially spliced ​​to obtain the individual characteristics of the battery cells, which are expressed as where v i represents the individual feature vector of the ith battery cell in the battery module, v ij Represents the jth service state variable of the ith battery cell in the battery module; the individual characteristics of all battery cells in the battery module are sequentially spliced ​​to obtain the individual characteristics of the battery module Calculate statistical indicators based on the individual characteristics of the battery module, and extract the cluster characteristics of the battery module based on the statistical indicators in represents the column vector of the lth statistical indicator, and L represents the total number of statistical indicator types; the individual feature v′ of the battery module and its corresponding cluster feature s are concatenated to form a fusion feature, which is expressed as Among them J e =(I+L)J represents the total number of features of the fused features; Then, the obtained fusion features are subjected to cycle window construction to create a cycle window containing multiple charge and discharge cycles. w All fused feature samples collected in the wth charge-discharge cycle are included in a cycle window, and the fused feature sample data of multiple cycle windows are obtained, where the fused feature sample data of the wth cycle window is recorded as represents the total number of fused feature samples in the wth cycle window, c0 represents the index of the first charge-discharge cycle in the cycle window, N c Represents the number of fused feature samples in the cth charge and discharge cycle.

4. The method for adaptively sensing the degradation of energy storage battery stack over its entire life cycle according to claim 1 is characterized in that: In the inner cycle scale, the charging behavior or discharging behavior in each charge and discharge cycle is identified as an inner cycle non-stationary trend through a Gaussian mixture model, specifically including: The charging or discharging behavior in each charge and discharge cycle is identified as an internal cycle non-stationary trend, and the Gaussian mixture model is used to describe the data distribution of each cycle window. For the fusion feature sample data X of the w-th cycle window w , use the Gaussian mixture model to model it, and give the probability density function of the modeled Gaussian mixture model, which is expressed as: Among them, p(x) represents the probability density function of the Gaussian mixture model, x∈X w represents a single fusion feature sample data in the w-th cycle window, α m,c , μ m,c and Σ m,c denote the mixture weight, mean vector and covariance matrix of the mth Gaussian cost in the cth charge-discharge cycle, respectively. M denotes the total number of Gaussian components. represents the Gaussian probability density function; where the mixing weight α m,c Satisfy constraints 5. The method for adaptively sensing the degradation of energy storage battery stack over its entire life cycle according to claim 1, characterized in that: The model parameters of the cross-cycle intrinsic Gaussian mixture model of each cycle window are estimated by an expectation maximization algorithm, specifically including: First, set the expected step, specifically: in the w-th cycle window, calculate the n-th fusion feature sample x in the q-th training iteration n The posterior probability from the mth Gaussian component The calculation formula is: in, and They represent the mixing weight, mean vector and covariance matrix of the m-th Gaussian component of the cross-cycle intrinsic Gaussian mixture model of the w-th cycle window estimated at the q-th iteration respectively; Then set the maximum step, specifically: according to the qth iteration The value of , estimates the Gaussian mixture model parameters of the q+1th iteration The calculation formula is: in, and They respectively represent the mixing weight, mean vector and covariance matrix of the m-th Gaussian component of the cross-cyclic intrinsic Gaussian mixture model of the w-th cyclic window estimated at the q+1-th iteration.

6. The method for adaptively sensing the degradation of energy storage battery stack over its entire life cycle according to claim 1, characterized in that: The step (4) specifically comprises: When performing online adaptive abnormal monitoring on the energy storage battery stack to be monitored, the service data collected in real time is analyzed with the battery module as the basic unit. The service data of the battery module includes the service status variable data of each battery cell. The service status variables include the state of charge index and the measurement point variable. When a new service data sample arrives, the cluster feature fusion processing is first performed to obtain a new fusion feature vector x new , and using the intrinsic projection matrix A in and the drift projection matrix A dr Calculate its inherent characteristics and drift characteristics, and the calculation formulas are: d new =A in x new d dr,new =A dr x new Among them, d new and d dr,new Respectively represent the new fusion feature vector x new The inherent characteristics and drift characteristics of Then calculate the inherent characteristic d new and drift characteristics d dr,new The monitoring statistic is calculated as follows: Among them, BID in,new The monitoring statistic representing the inherent characteristics of the new service data sample, BID dr,new M is the monitoring statistic representing the drift characteristics of the new service data sample; in represents the total number of Gaussian components in the intrinsic Gaussian mixture model, M dr Represents the total number of Gaussian components in the drift Gaussian mixture model; represents the model parameters of the mth Gaussian component in the intrinsic Gaussian mixture model, Represents the model parameters of the mth Gaussian component in the drift Gaussian mixture model; Indicates d new The Mahalanobis distance to the mth Gaussian component in the intrinsic Gaussian mixture model, Indicates d new The Mahalanobis distance to the mth Gaussian component in the drift Gaussian mixture model; represents the posterior probability of the intrinsic Gaussian mixture model corresponding to the intrinsic feature distribution, Represents the posterior probability of the drift Gaussian mixture model corresponding to the drift feature; Then determine the monitoring statistic BID of the inherent characteristics of the calculated new service data sample in,new Is it greater than the control limit of the monitoring statistic corresponding to the inherent characteristics, and the monitoring statistic BID of the drift characteristics of the new service data sample? dr,new Is it greater than the monitoring statistic control limit corresponding to the drift characteristic? If BID in,new is less than or equal to the control limit of the monitoring statistic corresponding to the inherent characteristic, and BID dr,new If the value is less than or equal to the control limit of the monitoring statistic corresponding to the drift characteristic, the state of the energy storage battery stack to be monitored is considered normal; otherwise, the state of the energy storage battery stack to be monitored is considered abnormal.

7. The method for adaptively sensing the degradation of a storage battery stack over its entire life cycle according to claim 6, characterized in that: The design of the adaptive model update strategy determines whether an update is required based on the aging degree of the online service data, and specifically includes: For the intrinsic features, a Gaussian mixture model for monitoring is designed, which is called the intrinsic Gaussian mixture model; for the drift features, a recursive Gaussian mixture model is established, which is called the drift Gaussian mixture model; Design an adaptive model update strategy for the intrinsic Gaussian mixture model and the drifting Gaussian mixture model, specifically by analyzing the degradation components to determine whether the intrinsic Gaussian mixture model or the drifting Gaussian mixture model needs to be updated and alarmed; Among them, the adaptive model update strategy specifically includes: Case 1: When the calculated monitoring statistic of the inherent characteristics of the service data sample is less than or equal to the monitoring statistic control limit corresponding to the inherent characteristics, and the monitoring statistic of the drift characteristic is less than or equal to the monitoring statistic control limit corresponding to the drift characteristic, the state of the energy storage battery stack to be monitored at this time is regarded as a normal state, and new service data samples are collected to update the model parameters of the drift Gaussian mixture model. When the size of the collected data sample block reaches the preset value B, the model parameters of the drift Gaussian mixture model and the monitoring statistic control limit corresponding to the drift characteristic are updated recursively; Case 2: When the calculated monitoring statistic of the inherent characteristics of the service data sample is greater than the monitoring statistic control limit corresponding to the inherent characteristics, and the monitoring statistic of the drift characteristics is less than or equal to the monitoring statistic control limit corresponding to the drift characteristics, the state of the energy storage battery stack to be monitored at this time is regarded as an abnormal state, the update of the drift Gaussian mixture model is stopped and an alarm is triggered; Case 3: When the monitoring statistic of the inherent characteristic of the calculated service data sample is less than or equal to the monitoring statistic control limit corresponding to the inherent characteristic, and the monitoring statistic of the drift characteristic is greater than the monitoring statistic control limit corresponding to the drift characteristic, the collected service data samples are used for a recursive strategy, and then the monitoring statistic of the drift characteristic is recalculated and the monitoring statistic control limit of the drift characteristic is determined again; thereafter, if the monitoring statistic of the drift characteristic of the calculated service data sample is still greater than the monitoring statistic control limit corresponding to the drift characteristic, an alarm is triggered; if the monitoring statistic of the drift characteristic of the calculated service data sample is less than or equal to the monitoring statistic control limit corresponding to the drift characteristic, it is considered that the energy storage battery stack to be monitored is in a normal state; Case 4: When the calculated monitoring statistic of the inherent characteristic of the service data sample is greater than the monitoring statistic control limit corresponding to the inherent characteristic, and the monitoring statistic of the drift characteristic is greater than the monitoring statistic control limit corresponding to the drift characteristic, the update of the Gaussian mixture model is stopped and an alarm is triggered.

Images