Inspection dynamic updating method combined with battery degradation prediction

Abstract

The application provides a battery deterioration prediction combined with a dynamic updating method of inspection, relates to the technical field of battery management, and comprises the following steps: calling historical operation data of a battery pack, performing battery deterioration risk prediction, and establishing a basic deterioration prediction score; performing prediction analysis on each battery in the battery pack, and extracting a characteristic space deviation index; performing data analysis, and establishing a data confidence index; performing deterioration prediction backtracking on each battery in the battery pack, and establishing a historical consistency index; establishing a battery confidence degree, and constructing an inspection priority; performing deterioration conduction analysis, and constructing an additional priority; after compensation of the inspection priority, distributing an inspection task according to a compensation result and a battery distribution. The application solves the technical problem that, in the prior art, an inspection task in battery inspection usually depends on a static health score or a regular inspection cycle, cannot be flexibly adjusted according to the actual situation of the battery, leads to low inspection efficiency, and increases the risk of battery failure.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of battery management, in particular to a method for dynamically updating inspection combining battery degradation prediction. BACKGROUND

[0002] Batteries will gradually deteriorate during long-term use, leading to performance degradation and even failure. In order to improve the safety and reliability of battery packs, timely and effective battery inspection and maintenance are crucial. However, there are some technical problems in battery degradation prediction and inspection management in the prior art. In the battery inspection of the prior art, the inspection task usually relies on static health scores or regular inspection cycles. Once the inspection task is set, it cannot be adjusted and optimized according to real-time data, which leads to a lack of flexibility and real-time response capability in the inspection plan, and the inspection plan cannot respond to changes in battery state and environmental conditions. When the state of the battery pack changes, the inspection task cannot be updated in time, resulting in low efficiency of the inspection and increasing the risk of battery failure. SUMMARY

[0003] The present application provides a method for dynamically updating inspection combining battery degradation prediction, aiming to solve the technical problem that in the battery inspection of the prior art, the inspection task usually relies on static health scores or regular inspection cycles, and cannot be flexibly adjusted according to the actual situation of the battery, resulting in low inspection efficiency and increasing the risk of battery failure.

[0004] The method for dynamically updating inspection combining battery degradation prediction disclosed in the present application comprises the following steps: calling historical operation data of a battery pack, using the historical operation data to perform battery degradation risk prediction based on a general battery degradation prediction model, and establishing a basic degradation prediction score; performing prediction analysis on each battery in the battery pack, and extracting a characteristic space deviation index; performing data analysis on the historical operation data, and establishing a data confidence index; performing degradation prediction backtracking on each battery in the battery pack, and establishing a historical consistency index; establishing a battery confidence degree according to the characteristic space deviation index, the data confidence index and the historical consistency index, constructing an inspection priority according to the battery confidence degree and the basic degradation prediction score; using the basic degradation prediction score to perform degradation conduction analysis in combination with the spatial distribution and coupling structure of the battery pack, and constructing an additional priority; and according to the compensation result and the battery distribution, constructing an inspection task after compensating the inspection priority using the additional priority.

[0005] One or more technical solutions provided in the present application have at least the following beneficial effects:

[0006] By calling the historical operation data of the battery pack and based on the general battery degradation prediction model, the degradation risk of the battery can be predicted, which can provide accurate prediction for the future health status of the battery. The establishment of the basic degradation prediction score provides a scientific basis for the formulation of the inspection task, so that the battery inspection work can be based on the health status of the battery to be prioritized, avoiding inefficient inspection or missing potential faulty batteries. By performing independent degradation prediction analysis on each battery in the battery pack and extracting the feature space deviation index, the degradation degree of each battery relative to the standard distribution can be quantitatively evaluated. This process helps to accurately identify the degradation characteristics and abnormal changes of the battery, improves the identification accuracy of individual batteries and the pertinence of the inspection, and ensures that key batteries can be identified and inspected first. By analyzing the historical operation data, a data confidence index can be established to evaluate the data quality and reliability. The establishment of the data confidence index makes the subsequent degradation prediction and inspection decision based on historical data more reliable, preventing false predictions and decisions caused by data quality problems, and effectively avoiding the waste of inspection resources caused by data noise. By comparing the degradation prediction results of the battery with the actual observation values, a historical consistency index can be established, which can evaluate the long-term stability and consistency of the battery degradation prediction model, ensure the prediction accuracy of the model, and effectively identify the bias or instability of the prediction model, thereby enhancing the prediction ability and adaptability of the model. By combining the feature space deviation index, the data confidence index and the historical consistency index, a battery confidence is established, and an inspection priority is constructed based on the basic degradation prediction score. Through the evaluation of the battery confidence, the health status and degradation risk of the battery can be accurately identified, and the inspection resources can be preferentially allocated to the batteries that need attention. This method improves the scientificity and accuracy of the inspection, effectively avoids inspection omissions and excessive inspection of low-priority batteries. By combining the spatial distribution and coupling structure of the battery pack with the basic degradation prediction score, a degradation conduction analysis is performed to construct an additional priority. This process can capture the coupling effect between batteries, further improve the accuracy of the inspection task setting through spatial position and mutual influence between batteries; and through the use of the additional priority to compensate the inspection priority and according to the battery distribution to configure the inspection task, the dynamic optimization and real-time adjustment of the inspection task are ensured. Through compensation adjustment, the allocation of the inspection task can be dynamically changed according to the actual operating state of the battery and the influence of the surrounding batteries, so that the battery inspection always matches the real status of the battery, avoiding the waste or uneven distribution of inspection resources.

[0007] The above description is only a summary of the technical solutions of the present application. In order to more clearly understand the technical means of the present application, the present application can be implemented in accordance with the content of the specification, and in order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the following specific embodiments of the present application are described. BRIEF DESCRIPTION OF DRAWINGS

[0008] Figure 1 A flowchart of a method for dynamically updating inspection in combination with battery degradation prediction is provided for the embodiments of the present application.

[0009] Figure 2 A flowchart of extracting a characteristic space deviation index in a method for dynamically updating inspection in combination with battery degradation prediction is provided for the embodiments of the present application. DETAILED DESCRIPTION

[0010] The embodiments of the present application provide a method for dynamically updating inspection in combination with battery degradation prediction, which solves the technical problem of low inspection efficiency and increased risk of battery failure in the prior art, in which the inspection task usually relies on static health scores or regular inspection cycles, and cannot be flexibly adjusted according to the actual situation of the battery.

[0011] After introducing the basic principles of the present application, various non-limiting embodiments of the present application will be specifically introduced in combination with the drawings of the specification.

[0012] As shown in Figure 1 The embodiments of the present application provide a method for dynamically updating inspection in combination with battery degradation prediction, which includes:

[0013] The historical running data of the battery pack is called, and the battery degradation risk prediction based on the general battery degradation prediction model is performed using the historical running data to establish a basic degradation prediction score.

[0014] The historical running data accumulated by the battery pack in the actual running process, such as temperature, charge and discharge times, use environment, power output, etc., is called through the battery management system or the cloud data platform. These historical running data are stored in the form of time series and cover multiple cycles of the battery pack, reflecting the running state of the battery pack under different working conditions.

[0015] The general battery degradation prediction model is used to analyze the historical running data, evaluate the degradation degree and risk of the battery, and the general battery degradation prediction model includes a deep learning model, a support vector machine, a random forest, or a model based on physical and chemical processes. Through these models, the key factors affecting the performance of the battery can be extracted from the historical running data, and the risk of future degradation can be predicted, and a basic degradation prediction score is established. The basic degradation prediction score is generated based on the evaluation of the battery's charge capacity, internal resistance, cycle life, etc., and represents the degradation degree of the battery pack.

[0016] Each battery in the battery pack is predicted and analyzed, and a characteristic space deviation index is extracted.

[0017] The individual analysis of each battery in the battery pack is different from the overall battery pack degradation prediction. Different batteries may have different degradation processes, so they need to be handled separately. Specifically, a training data set for a general battery degradation prediction model is obtained, which includes multi-dimensional features such as temperature, charging cycle, discharge depth, etc. The data set is used to train the model to predict the degradation trend of the battery in a specific environment. In order to more efficiently analyze and predict battery degradation, the high-dimensional data is reduced in dimension. After dimension reduction, the data is mapped to a low-dimensional space, which is called model perception space. Each point in the space represents the state characteristics of a battery. The historical operation data of each battery in the battery pack is also reduced in dimension so as to be mapped into the model perception space. The data points after dimension reduction serve as the embedding points of the batteries. In the model perception space, the deviation of each battery embedding point from the mean or center point of the model training data is calculated. This deviation reflects the difference between the current state of the battery and the normal state of the historical data. A feature space deviation index is established by the numerical value of the deviation. This index reflects the abnormality of each battery in the feature space. The larger the deviation, the farther the performance and health state of the battery deviates from the normal state.

[0018] Data analysis is performed on the historical operation data to establish a data confidence index.

[0019] Data analysis is performed on the historical operation data to ensure the accuracy and reliability of the data. Specifically, the historical operation data is preprocessed, including time alignment, outlier removal, and normalization. The preprocessed data is subjected to multi-dimensional feature consistency analysis to identify the consistency of the operating characteristics of different batteries in the battery pack. A first confidence index is established. Missing value analysis is performed on the historical operation data to calculate the missing rate of the data. At the same time, the jitter phenomenon in the data, i.e. large fluctuations in the data, is analyzed. According to these calculation results, a second confidence index is established. The first and second confidence indexes are integrated to comprehensively establish a data confidence index, which reflects the reliability of the data. The higher the data confidence index, the more stable, complete and reliable the data is. Otherwise, there may be a large uncertainty, which affects subsequent prediction and decision-making.

[0020] Degradation prediction is performed on each battery in the battery pack to establish a historical consistency index.

[0021] For each battery in the battery pack, a degradation prediction sequence is established, which includes prediction results at different time points reflecting the degradation trend of the battery at different historical stages, which are predicted based on historical operation data and general battery degradation model. By comparing the deviation between the degradation prediction sequence of the battery and the actual observation value obtained through real-time data acquisition and monitoring of the battery, which represents the actual operating condition of the battery, the deviation authentication is carried out, and the accuracy of the degradation prediction model is evaluated. If the predicted value is significantly different from the actual value, it means that the model has deviation and needs to be adjusted.

[0022] Taking the current time node as the starting point of search, a plurality of backtracking windows are constructed, each of which represents a historical stage for analyzing the performance and prediction consistency of the battery in these stages. In these backtracking windows, the prediction ability of the model in different time periods is determined by comparing the predicted value and the actual value. For each backtracking window, the deviation authentication results in multiple windows are combined through weighted joint authentication to establish a historical consistency index of the battery, which can be a weighted average or other comprehensive evaluation method to quantify the consistency and reliability of the battery in historical prediction. If the historical consistency is poor, it means that the degradation prediction of the battery has a large uncertainty and needs to be re-evaluated or strengthened.

[0023] According to the feature space deviation index, the data confidence index and the historical consistency index, a battery confidence is established, and a patrol priority is constructed according to the battery confidence and the basic degradation prediction score.

[0024] The feature space deviation index, the data confidence index and the historical consistency index obtained before are combined to comprehensively establish the battery confidence, and the high and low of the battery confidence reflects the accuracy and reliability of the battery prediction result. If the feature space deviation is large, the data confidence is low or the historical consistency is poor, the confidence is low, which means that the battery needs more attention. According to the battery confidence and the previously established basic degradation prediction score, a patrol priority is constructed, and the basic degradation prediction score reflects the health status of the battery, and the battery confidence reflects the accuracy of the prediction. By combining these two factors, it can be determined which battery needs to be prioritized. Among them, the lower the battery confidence and the worse the basic degradation prediction score, the higher the patrol priority of the battery. This method helps to efficiently allocate patrol resources and ensure that problem batteries are detected in time.

[0025] Using the basic degradation prediction score, combined with the spatial distribution and coupling structure in the battery pack, a degradation conduction analysis is carried out to construct an additional priority.

[0026] Obtain the spatial coordinates, electrical connection relationship graph, thermal channel and cooling path of each battery in the battery pack, construct a battery pack structure dataset, which can be represented by a network topology structure to describe the interaction between batteries and its impact on performance. In the battery pack, the degradation of a battery can affect other batteries through electrical connection or heat conduction, etc. For example, over-discharge or overheating of a battery can cause the degradation of surrounding batteries to accelerate. Analyze the spatial distribution and coupling structure of the battery pack to evaluate the mutual influence between different batteries. This process includes calculating a coupling influence adjacency matrix, which describes the degradation propagation path and strength between batteries.

[0027] Using the coupling influence adjacency matrix in combination with the basic degradation prediction score, degradation conduction analysis is performed. Through this analysis, it can be determined whether the degradation of certain batteries will accelerate or affect the state of other batteries. For example, a battery with more severe degradation may affect other batteries through electrical connection or heat conduction. The degradation conduction analysis process uses a graph propagation algorithm to simulate the propagation of degradation status within the battery pack. Based on the results of the degradation conduction analysis, additional priority is constructed. If the degradation of a certain battery may affect the health status of other batteries, a higher additional priority is assigned to that battery.

[0028] After compensating the inspection priority using the additional priority, the inspection task is configured according to the compensation results and the battery distribution.

[0029] The additional priority is combined with the inspection priority to perform compensation. Through this compensation process, it is ensured that the inspection task gives priority to batteries with more severe degradation and considers the coupling influence between batteries. The compensation method can be direct weighted fusion, adding the additional priority and the inspection priority according to certain weights, or through an adaptive threshold triggering mechanism to adjust the inspection priority when certain conditions are met. After completing the priority compensation, the inspection task is configured according to the new priority order and the battery distribution, ensuring that the batteries that need to be checked are detected in a timely manner, avoiding larger-scale problems caused by battery failure.

[0030] Further, as shown in Figure 2 The predicted analysis of each battery in the battery pack is performed to extract the feature space deviation index, including:

[0031] Obtain the training dataset of the general battery degradation prediction model; after extracting the high-dimensional feature vector from the training dataset, perform dimension reduction processing to construct a model perception space; extract the feature vector from the historical operation data, and after synchronously performing dimension reduction processing on the feature vector, embed it into the model perception space; in the model perception space, calculate the feature deviation of the target battery embedding point and the reference distribution center, normalize the feature deviation, and establish the feature space deviation index.

[0032] The training data set is derived from a large number of battery packs in different operating environments. The historical operating data includes the temperature, charge and discharge cycle, power, capacity attenuation, internal resistance change and other characteristics of the battery, which can reflect the degradation trend of the battery.

[0033] For each data in the training data set, high-dimensional feature extraction is performed to convert it into a high-dimensional feature vector. The high-dimensional feature vector represents the performance of the battery in each state, including multiple dimensions of features such as battery voltage, temperature, capacity, etc. High-dimensional data may cause computational burden and have dimension disaster problem, therefore, dimension reduction techniques are used to reduce feature dimension, simplify data structure, and dimension reduction methods such as principal component analysis, t-SNE, etc. can effectively reduce the dimension of feature space and retain the most important features in the data. The data after dimension reduction is mapped to a low-dimensional space, called model perception space, which can effectively represent the degradation trend and state of the battery, so that the model can more easily understand the state of the battery.

[0034] For each battery in the battery pack, the feature vector related to degradation is extracted from the historical operating data, including the number of battery charges, discharge depth, temperature change, internal resistance, capacity loss, etc. The same as the training data set, the feature vector of the historical operating data is also subjected to dimension reduction processing. Since the feature dimensions of the training data set and the historical operating data are the same, the same dimension reduction method can be used for synchronous dimension reduction. The purpose of synchronous dimension reduction is to compress the feature vector of the historical operating data into the same low-dimensional space to ensure that the historical operating data can be compared and analyzed with the feature vector of the training data set in the same space. After dimension reduction, the feature vector of the historical operating data is embedded into the model perception space constructed earlier, so that subsequent analysis can be carried out in a unified space.

[0035] The target battery embedding point refers to the position of each battery in the model perception space, representing the degradation state and operating characteristics of the battery; the reference distribution center refers to the mean or median of the feature vectors of all batteries in the training data, representing the average state of normal or healthy batteries. The reference distribution center is the standard reference point of the battery pack, which is used to judge the deviation of each battery. The distance between the target battery embedding point and the reference distribution center is calculated, such as the Euclidean distance or Manhattan distance, to obtain the feature deviation of the battery. The larger the feature deviation, the farther the state of the battery from the normal or healthy state.

[0036] In order to compare the deviation degrees of different batteries, the calculated characteristic deviation degrees are normalized, for example, using the min-max normalization method to compress the characteristic deviation degrees to a fixed range, for example, between 0 and 1, to ensure that the characteristic deviation degrees of different batteries can be compared on the same scale. The normalized characteristic deviation degrees constitute the characteristic space deviation index, which reflects the abnormality degree of the battery. The higher the characteristic space deviation index, the worse the health condition of the battery, which needs to be checked first.

[0037] Further, the data analysis on the historical operation data to establish a data confidence index comprises:

[0038] The historical operation data is preprocessed, and the data preprocessing comprises time alignment, outlier elimination, and normalization. The preprocessed historical operation data is subjected to consistency analysis of multi-dimensional features to establish a first confidence index. The preprocessed historical operation data is subjected to data missing rate and jitter rate index calculation to establish a second confidence index. The data confidence index is established according to the first confidence index and the second confidence index.

[0039] Since the historical operation data comes from different batteries or different sensors, it may have different time stamps. In order to ensure that the data can be effectively compared and analyzed, time alignment is first performed, that is, the historical operation data of different batteries is aligned on the same time axis, so that the data points of each battery correspond to the same time point. This can be achieved by interpolation methods such as linear interpolation or spline interpolation, to ensure that the data points are consistent in time.

[0040] There may be outliers in the historical operation data of the battery pack, which are caused by sensor failure, unstable battery performance, data acquisition error, etc. Outliers need to be eliminated. The strategy for outlier elimination includes setting a reasonable threshold range, determining values outside the range as outliers and deleting them, and using statistical methods such as box plot method and Z-score to identify and eliminate extreme values in the data.

[0041] The purpose of normalization is to eliminate the influence of dimensional differences between different features and ensure that all features are compared on the same scale. The normalization method includes the min-max normalization method, which compresses data values to a fixed range, for example, between 0 and 1. Through the three steps of data preprocessing, the historical operation data can be aligned in time, consistent and free of abnormalities, facilitating subsequent analysis.

[0042] The historical operation data in the battery pack usually contains multiple features, such as temperature, voltage, charging times, discharge depth, etc., and there is a certain relationship between these features. Under normal circumstances, the data changes of different features should be consistent with each other. Consistency analysis can be achieved by calculating the correlation between different features, for example, using Pearson correlation coefficient to measure the linear relationship between two variables. If the correlation between features is high, it means that the performance of these features in the battery pack is consistent, and the data quality is good. If the correlation is low, it means that the data has high inconsistency. According to the consistency analysis results between multiple features, the consistency of each feature is integrated to establish a first confidence index. The value of this index reflects the overall consistency of the data. The larger the value, the more stable and reliable the data is. The smaller the value, the more likely there is a large fluctuation or inconsistency in the data.

[0043] The data missing rate refers to the proportion of missing values in the historical operation data. Missing values are caused by sensor failure, communication problems or battery damage, etc. The calculation method of data missing rate is the number of missing data divided by the total number of data points. High data missing rate means that the data quality is poor, which affects subsequent analysis and prediction.

[0044] The data jitter rate refers to the fluctuation amplitude of the data, especially in a short period of time. The performance fluctuation of the battery during operation will cause the jitter of the data. The calculation of the jitter rate index is based on the change amplitude of the data, which can be measured by calculating the variance or standard deviation of the data. A higher jitter rate index means that the state of the battery is unstable, which affects the accuracy of the degradation prediction. Based on the calculated data missing rate and jitter rate index, a second confidence index is established. This index can be obtained by weighted calculation of the data missing rate and jitter rate index, reflecting the integrity and stability of the data.

[0045] Finally, the first confidence index and the second confidence index are combined to comprehensively obtain the data confidence index, which is used to quantify the overall credibility of the data. The higher the value of the data confidence index, the better the data quality. Conversely, it means that the data quality is poor and needs further processing or enhancement of monitoring.

[0046] Further, the degradation prediction of each battery in the battery pack is backtracked, and a historical consistency index is established, including:

[0047] A degradation prediction sequence of each battery in the battery pack is established, and offset authentication is performed according to the degradation prediction sequence and the actual observation value to establish an offset authentication sequence. The current time node is taken as the starting point of search to construct multiple backtracking windows. The weighted joint authentication of the offset authentication sequence is performed under multiple backtracking windows to establish the historical consistency index.

[0048] For each battery within the battery pack, according to its historical operation data and the degradation prediction model, the degradation prediction value of the battery at different time points is calculated, which reflects the future state of the battery, such as capacity attenuation, internal resistance increase, etc. The degradation prediction values are arranged in chronological order to obtain the degradation prediction sequence, wherein the degradation prediction value at each time point represents the degradation degree of the battery at that time. The actual observation value is the battery state data obtained by the battery management system or other monitoring equipment, such as actual capacity, actual internal resistance, etc., which reflects the current health status of the battery.

[0049] By comparing the degradation prediction sequence with the actual observation value, the deviation between them is calculated, which is realized by calculating the difference between the prediction value and the actual value. The offset authentication sequence is a sequence composed of the deviation between the prediction value and the actual observation value of each battery, which is used to evaluate the accuracy of the prediction model. If the deviation is large, it means that the error of the degradation prediction model is large.

[0050] The current time node represents the current time, and the current time node is used as the starting point of the search, which means that the historical performance of the battery is reviewed from the current time. The backtracking window is used to review the historical data of a certain time period from the current time node, and each backtracking window represents a time period, such as 1 day, 1 week, 1 month, etc. in the past.

[0051] In each backtracking window, the weighted joint authentication of the offset authentication sequence is performed, which means that the offset degree in each backtracking window is weighted and integrated. The purpose of weighting is to adjust the weight of different backtracking windows in the final authentication according to their time period and importance, for example, the recent backtracking window has a higher weight because it can better reflect the current degradation state of the battery, while the distant backtracking window has a lower weight because it has less impact on the current state.

[0052] For the data in each backtracking window, the weighted average offset degree is calculated, and the offset degrees in all backtracking windows are synthesized according to the weight to establish the historical consistency index of the battery, which reflects the consistency degree of the battery's degradation prediction and actual performance. The higher the historical consistency index, the more stable the degradation process of the battery, and the more it follows the rules of the degradation prediction model. On the contrary, if the historical consistency index is low, it means that there is a large deviation in the degradation prediction of the battery, and the accuracy of the degradation prediction model needs to be re-evaluated.

[0053] Further, the use of the basic degradation prediction score, combined with the spatial distribution and coupling structure within the battery pack for degradation conduction analysis, constructs additional priority, including:

[0054] Obtaining the spatial coordinates of the batteries in the battery pack, the electrical connection relationship graph, the heat channel and the cooling path, constructing a battery pack structure dataset; performing coupling influence analysis on the battery pack structure dataset to establish a coupling influence adjacency matrix; using the coupling influence adjacency matrix and the basic degradation prediction score to perform influence graph propagation iteration to establish an additional priority.

[0055] The position of each battery in the battery pack affects its degradation conduction effect. The spatial coordinates are represented by three-dimensional coordinates (x, y, z) to reflect the specific position of each battery in the physical space. Adjacent or neighboring batteries will influence each other in terms of heat transfer, current conduction, etc. Obtaining spatial coordinates helps to understand these interactions.

[0056] The electrical connection relationship between batteries describes how batteries are connected together through electrical circuits to form a battery pack. These electrical connection relationships affect the current flow, power distribution, and possible electrical fault conduction of the batteries. The electrical connection relationship graph is represented by an adjacency matrix or graph data structure in graph theory, where each battery represents a node in the graph, and the electrical connection between batteries is represented as an edge in the graph.

[0057] In the battery pack, due to heat generation during charging and discharging, the battery will accelerate degradation due to overheating. Obtain the heat channel and cooling path. The heat channel refers to the heat transfer channel between batteries or between the battery and the external system. The cooling path refers to the path through which the battery is cooled by the cooling system during operation. By simulating these heat channels and cooling paths, it can be identified which batteries are easily affected by overheating of adjacent batteries.

[0058] Integrate the above information into a comprehensive battery pack structure dataset. This dataset can be represented in graph or network structure, where nodes represent batteries and edges represent electrical connections, heat conduction or cooling paths between batteries, providing a structural basis for subsequent coupling influence analysis.

[0059] Coupling influence refers to the fact that the degradation or failure of one battery may affect adjacent or connected batteries through electrical connections, heat conduction, etc. For example, a battery may be affected by heat conduction or current transmission due to high temperature or electrical failure. In the coupling influence analysis, the electrical connection relationship, heat channel and cooling path of the battery are considered comprehensively. For each pair of batteries, the mutual influence between them is analyzed to determine the coupling strength between them, including electrical coupling, thermal coupling and cooling path coupling. The coupling influence adjacency matrix is a graph data structure used to represent the mutual coupling influence relationship between batteries in the battery pack. In this coupling influence adjacency matrix, each element represents the coupling influence strength between two batteries.

[0060] The influence graph propagation is a process of information propagation based on graph structure. In the battery pack, through the influence graph propagation iteration, the degradation information of the battery will be propagated according to the coupling influence adjacency matrix, so as to establish the additional priority of the battery, that is, the basic degradation prediction score of each battery is propagated according to the state of its adjacent battery, for example, if the degradation score of a battery is high, it will affect the adjacent battery, so that the priority of the adjacent battery also rises. Through the influence graph propagation algorithm, such as graph convolution network, graph propagation algorithm, etc., the degradation information between the batteries is propagated to the whole battery pack by using the coupling influence adjacency matrix, and the propagated influence information is weighted according to the coupling degree of the battery, so as to ensure that the stronger the coupling relationship is, the greater the influence will be. The propagation process is iterated until the degradation state of all batteries is updated and converged. In this process, the basic degradation prediction score is used as the initial value of the current degradation state of the battery, and the basic degradation prediction score is adjusted and updated according to the mutual influence between the batteries as the influence graph propagation proceeds. After the influence graph propagation process is completed, the additional priority of each battery reflects its degradation propagation influence in the battery pack.

[0061] Further, after the additional priority is used to compensate the inspection priority, the inspection task is configured according to the compensation result and the battery distribution, comprising:

[0062] The additional priority and the inspection priority are synchronized to a joint priority analysis channel to establish a real priority, the joint priority analysis channel comprising a weighted fusion sub-channel, an adaptive threshold triggering sub-channel and an inspection path insertion sub-channel; the inspection battery pack is positioned according to the real priority, an inspection path and a detection task are established, and the inspection path and the detection task are output as the inspection task.

[0063] The additional priority and the inspection priority are synchronized to a joint priority analysis channel for processing to comprehensively consider the state of the battery and the influence of its adjacent batteries, wherein the additional priority and the inspection priority are weighted and fused in a weighted fusion sub-channel, specifically, different weights are given to each priority according to the influence degree of each priority, for example, if the coupling effect of the additional priority is more important, the additional priority is given a higher weight, and after weighted fusion, a comprehensive priority is obtained; in an adaptive threshold triggering sub-channel, the priority adjustment is triggered according to the set adaptive threshold, the adaptive threshold is dynamically set and can be automatically adjusted according to the overall condition of the battery pack, the actual operating environment or real-time monitoring data, if any priority of the battery exceeds a certain threshold, the corresponding inspection operation is automatically triggered; in a path insertion sub-channel, the inspection path is arranged according to the spatial position of the battery and the priority, the optimization of the path considers the position of the battery, the priority of the inspection task and the structure of the battery pack, and the path insertion sub-channel ensures that the inspection task can be most efficiently executed in the battery pack. Through the analysis and calculation of the above three sub-channels, the real priority is finally generated, which represents the final priority order of the inspection task, and the real priority comprehensively considers the health state of the battery, the mutual coupling influence, the spatial distribution and other factors, thereby ensuring that the most important battery is given priority for inspection.

[0064] Using the real priority, each battery in the battery pack is sorted, and the batteries with higher priority are concentrated for inspection to ensure that the batteries that may have problems are detected and maintained in a timely manner. The task of positioning the battery pack for inspection is to determine the batteries that need to be inspected according to the priority, spatial distribution and actual operating requirements of the batteries, and the batteries with high priority, poor health status or serious degradation should be given priority to enter the inspection queue.

[0065] The inspection path refers to the route that the inspection personnel or automatic inspection equipment follow according to the spatial distribution and priority of the battery pack. A reasonable inspection path can improve the inspection efficiency and reduce unnecessary time waste. The detection task includes the inspection content, method and time arrangement of each inspection battery, for example, for each battery, the key indicators such as voltage, current, temperature and internal resistance need to be detected, or more detailed degradation state evaluation is needed, and the detection task of each battery is based on its health status, historical data and prediction model to ensure the accuracy and comprehensiveness of the inspection.

[0066] Finally, combined with the real priority sorting, the inspection path planning and the specific detection task, a complete inspection task is output, which guides the inspection personnel or automatic inspection equipment to perform inspection according to the set path and conduct comprehensive detection on the battery pack to ensure accurate and efficient inspection work.

[0067] Further, the positioning of the battery pack for inspection according to the real priority includes:

[0068] The batteries that fail to enter the positioning and inspection battery group are prioritized and spatially clustered to establish a prioritized and spatially clustered result. The prioritized and spatially clustered result is subjected to high-priority cluster identification to configure high-priority cluster identification. The battery group with the high-priority cluster identification is added to the inspection battery group.

[0069] The batteries that fail to enter the positioning and inspection battery group can be due to a long distance, slightly lower priority, or not included in the first inspection range due to spatial or temporal limitations. Prioritizing and spatially clustering these batteries means that, according to their priority and spatial distribution, these batteries are clustered. Through clustering algorithms such as K-means, DBSCAN, or hierarchical clustering, these batteries are classified according to their priority and spatial coordinates. Batteries with high priority are classified into a class, and batteries with close locations are also clustered into the same class. Through clustering, a prioritized and spatially clustered result is established, where each clustering result contains a group of batteries with similar priority and spatial location. Through clustering, it can effectively identify which batteries have a strong correlation in space and priority, thereby providing a basis for subsequent inspection arrangements.

[0070] In the prioritized and spatially clustered result, some batteries have high priority and are concentrated in certain areas. These areas are identified as high-priority cluster areas by analyzing the average priority of each cluster. If the batteries in a cluster mostly have high priority, the cluster is identified as a high-priority cluster area. The identified high-priority cluster areas are configured with high-priority cluster area identification, which serves as a basis for subsequent inspection tasks.

[0071] The battery group configured with high-priority cluster area identification is extracted from the clustering result and added to the list of inspection battery groups. Inspection tasks are arranged according to priority to ensure the accuracy and efficiency of the inspection.

[0072] Further, the positioning and inspection battery group according to the true priority further comprises:

[0073] The base degradation prediction score is authenticated and backtracked to establish a change rate trend indicator. If the change rate trend indicator meets a preset breakthrough threshold, the corresponding battery is added to the inspection battery group.

[0074] The basic degradation prediction score reflects the health status of the battery. In order to verify the effectiveness of the basic degradation prediction score, authentication backtracking of battery degradation prediction is needed, which means comparing the actual observation data in the past period of time with the historical prediction score to check the accuracy of the prediction. For example, if the prediction score of the battery predicts that the capacity attenuation of the battery is a certain value in a certain period of time, the difference between the actual capacity attenuation value and the predicted value is the result of authentication backtracking. During authentication backtracking, the change rate of the battery degradation score in each time period is first calculated, for example, the difference between the prediction score of a certain time period and the score of the previous time period divided by the score of the previous time period, the change rate is calculated and the trend is identified, and the change rate trend index is obtained. This index reflects the change trend in the battery degradation process. If the degradation rate of the battery suddenly accelerates, the change rate shows a sharp rise, indicating that the battery may have a risk of failure.

[0075] The preset breakthrough threshold is used to determine whether the change of the battery degradation is abnormal. The preset breakthrough threshold can be a statistical value obtained from historical data analysis or a fixed value set according to experience. For example, if the change rate of the battery is greater than a certain set threshold (such as 20%), it indicates that the degradation trend of the battery has a larger fluctuation or acceleration, which may cause the performance of the battery to be unstable and needs to be checked first.

[0076] If the change rate trend index of the battery exceeds the preset breakthrough threshold, it indicates that the degradation process of the battery has abnormal changes, and the corresponding battery is added to the inspection battery group and its priority is increased to ensure that these batteries are processed first in the next inspection task. The purpose of this is to discover potential battery failures as soon as possible and avoid more serious damage or failure of the battery.

[0077] Further, after configuring the inspection task according to the compensation result and the battery distribution, the method further comprises:

[0078] Based on the inspection task, an inspection management is performed to establish an inspection data set; the consistency of the compensation result is authenticated using the inspection data set to generate an authentication feedback, and the inspection self-optimization management is performed according to the authentication feedback.

[0079] According to the start of the inspection task, the actual inspection operation of the battery group is performed, including checking the actual health status of the battery, such as measuring the voltage, current, temperature, internal resistance, capacity attenuation and other key indicators of the battery, as well as physical inspection, electrical connection inspection, cooling system inspection, etc., to ensure that the battery is in a safe and effective working state. During the inspection process, all detection data is recorded to form an inspection data set.

[0080] Consistency authentication refers to verifying whether the compensation adjustment is effective by comparing the matching degree of the compensated result and the actual inspection data. For example, if the health status of a certain battery is poor and the inspection priority is adjusted according to the compensation result, the subsequent actual inspection result should be consistent with this adjustment. The purpose of consistency authentication is to ensure the matching degree between the compensation measures and the actual inspection data in the inspection process, as well as the effectiveness of the compensation measures. Authentication feedback is an analysis output based on the consistency authentication result. If the consistency authentication shows that the compensation result is consistent with the actual inspection result, it is determined that the compensation adjustment is successful, otherwise the compensation strategy needs to be re-evaluated. Inspection self-optimization management is a process of continuously optimizing the inspection task based on authentication feedback. According to the authentication feedback result, the inspection strategy is adjusted, including adjusting the priority of the inspection task and modifying the inspection path. By adjusting the inspection strategy after each inspection, the inspection efficiency and the accuracy of battery management are continuously improved.

[0081] Further, the configuration of the inspection task according to the compensation result and the battery distribution includes:

[0082] A multi-objective evaluation function is established, which includes a total path evaluation target and a priority priority target. Based on the multi-objective evaluation function, the path optimization is performed according to the compensation result and the battery distribution, and the inspection task is configured based on the path optimization result.

[0083] A multi-objective evaluation function is established, wherein the total path evaluation target aims to optimize the length and efficiency of the inspection path, and the total path evaluation target is used to measure the total path length or time consumption of the inspection task, to ensure that the inspection process is as efficient as possible and to reduce unnecessary inspection time and path extension. For example, if the inspection task distribution is relatively scattered, the role of the total path evaluation target is to ensure that the inspection personnel can optimize the inspection path as much as possible and reduce repeated backtracking. The priority priority target aims to ensure that the priority of the inspection task is correctly focused, and the battery with high priority should be inspected first. This target weighs the priority of different batteries and allocates inspection resources according to the priority. The battery with higher priority should be given priority consideration in path planning, for example, in the planning of the inspection path, the battery with higher priority should be placed at the front end of the path or on the path to ensure that it can be checked as early as possible.

[0084] The path optimization is performed by an optimization algorithm, such as a genetic algorithm, a particle swarm optimization, a simulated annealing, or the like, according to a multi-objective evaluation function. The optimization algorithm calculates an optimal inspection path according to the distribution of the batteries, the priorities, and the length of the inspection path. For example, the algorithm arranges the batteries with high priorities in the shortest path or the shortest time according to the spatial distribution of the batteries in the battery pack, and reduces unnecessary path repetition and backtracking according to the total path evaluation target, so as to ensure the optimal use of resources in the inspection process. According to the path optimization result, the inspection task is configured to ensure that the inspection task is efficient and meets the priority requirement, so that each inspection task can be completed in a limited time, and the high-priority batteries are focused on.

[0085] In summary, the method for dynamically updating the inspection according to the battery degradation prediction has the following technical effects:

[0086] By calling the historical operation data of the battery pack and predicting the battery degradation risk based on the general battery degradation prediction model, the future health status of the battery can be accurately predicted, the establishment of the basic degradation prediction score provides a scientific basis for the formulation of the inspection task, and the inspection of the battery can be based on the health status of the battery. The low-efficiency inspection or the omission of the potential fault battery is avoided; by independently predicting and analyzing the degradation of each battery in the battery pack and extracting the feature space deviation index, the degradation degree of each battery relative to the standard distribution can be quantitatively evaluated, which helps to accurately identify the degradation characteristics and abnormal changes of the battery, improves the identification accuracy of the single battery and the pertinence of the inspection, and ensures that the key battery can be identified and inspected in priority; by analyzing the historical operation data, the data confidence index is established, which can evaluate the data quality and reliability, and the establishment of the data confidence degree makes the subsequent degradation prediction and inspection decision based on the historical data more reliable, prevents the wrong prediction and decision caused by the data quality problem, and effectively avoids the waste of inspection resources caused by data noise; by comparing the degradation prediction results of the battery with the actual observation value, the historical consistency index is established, which can evaluate the long-term stability and consistency of the battery degradation prediction model, ensure the prediction accuracy of the model, and the historical consistency analysis can effectively identify the deviation or instability of the prediction model, thereby enhancing the prediction ability and adaptability of the model; by comprehensively considering the feature space deviation index, the data confidence index and the historical consistency index, the battery confidence degree is established, and the inspection priority is constructed combined with the basic degradation prediction score. Through the evaluation of the battery confidence degree, the health status and degradation risk of the battery can be accurately identified, and the inspection resources are preferentially allocated to the battery that needs attention, which improves the scientificity and accuracy of the inspection, effectively avoids the omission of the inspection and the over-inspection of the low-priority battery; by combining the spatial distribution and coupling structure of the battery pack with the basic degradation prediction score, the degradation conduction analysis is performed, and the additional priority is constructed. This process can capture the coupling effect between the batteries, and through the spatial position and the mutual influence between the batteries, the accuracy of the inspection task setting is further improved; by using the additional priority to compensate the inspection priority, and according to the battery distribution, the inspection task is configured, the dynamic optimization and real-time adjustment of the inspection task are ensured, and through the compensation adjustment, the allocation of the inspection task can be dynamically changed according to the actual operation status of the battery and the influence of the surrounding batteries, so that the battery inspection is always matched with the real status of the battery, and the waste of the inspection resources or the uneven allocation is avoided.

[0087] The foregoing description of the disclosed embodiments enables a person skilled in the art to make or use the application. Modifications of these embodiments will occur to persons of skill in the art, and that the generic principles defined herein can be applied to other embodiments without departing from the spirit or scope of the application. Therefore, the present application is not intended to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method of updating a patrol dynamically in connection with a battery deterioration prediction, characterized by, The method comprises: Call the historical operation data of the battery pack, use the historical operation data to perform battery degradation risk prediction based on a general battery degradation prediction model, and establish a basic degradation prediction score; Perform prediction analysis on each battery in the battery pack, and extract a characteristic space deviation degree index; Perform data analysis on the historical operation data, and establish a data confidence index; Perform degradation prediction backtracking on each battery in the battery pack, and establish a historical consistency index; establish a degradation prediction sequence of each battery in the battery pack, perform offset authentication according to the degradation prediction sequence and the actual observation value, and establish an offset authentication sequence; Take the current time node as a search starting point, and construct a plurality of backtracking windows; Perform weighted joint authentication of the offset authentication sequence under the plurality of backtracking windows, and establish a historical consistency index; Establish a battery confidence degree according to the characteristic space deviation degree index, the data confidence index and the historical consistency index, and construct an inspection priority according to the battery confidence degree and the basic degradation prediction score; Use the basic degradation prediction score to perform degradation conduction analysis in combination with the spatial distribution and coupling structure in the battery pack, and construct an additional priority; After compensating the inspection priority by using the additional priority, configure an inspection task according to the compensation result and the battery distribution.

2. The method of claim 1, wherein the method is characterized by, The prediction analysis on each battery in the battery pack and the extraction of the characteristic space deviation degree index comprise: Obtain a training data set of a general battery degradation prediction model; After extracting a high-dimensional feature vector from the training data set, perform dimension reduction processing to construct a model perception space; Extract a feature vector in the historical operation data, and embed the feature vector into the model perception space after synchronous dimension reduction processing of the feature vector; In the model perception space, calculate the characteristic deviation degree of the target battery embedding point and the reference distribution center, normalize the characteristic deviation degree, and establish a characteristic space deviation degree index.

3. The method of claim 1, wherein the method is characterized by, The data analysis on the historical operation data and the establishment of the data confidence index comprise: Perform data preprocessing on the historical operation data, wherein the data preprocessing comprises time alignment, outlier elimination and normalization processing; Perform consistency analysis of multi-dimensional features on the preprocessed historical operation data to establish a first confidence index; Perform data missing rate and jitter rate index calculation on the preprocessed historical operation data to establish a second confidence index; Establish a data confidence index according to the first confidence index and the second confidence index.

4. The method of claim 1, wherein the method is characterized by, The degradation conduction analysis in combination with the spatial distribution and coupling structure in the battery pack by using the basic degradation prediction score to construct an additional priority comprises: Obtain the spatial coordinates, electrical connection relationship diagram, heat channel and cooling path of the batteries in the battery pack to construct a battery pack structure data set; Perform coupling influence analysis on the battery pack structure data set to establish a coupling influence adjacency matrix; Perform influence graph propagation iteration by using the coupling influence adjacency matrix and the basic degradation prediction score to establish an additional priority.

5. The method of claim 1, wherein the method is characterized by, The configuration of an inspection task according to the compensation result and the battery distribution after compensating the inspection priority by using the additional priority comprises: Synchronize the additional priority and the patrol priority to a joint priority analysis channel to establish a real priority, the joint priority analysis channel including a weighted fusion sub-channel, an adaptive threshold trigger sub-channel, and a patrol path insertion sub-channel; Position a patrol battery pack according to the real priority, establish a patrol path and a detection task, and output the patrol path and the detection task as a patrol task.

6. The method of claim 5, wherein the method is characterized by, The positioning of the patrol battery pack according to the real priority includes: Perform priority space clustering on a battery that fails to enter the positioned patrol battery pack, and establish a priority space clustering result; Perform high-priority aggregation area identification on the priority space clustering result, and configure a high-priority aggregation area identifier; Add a battery pack with the high-priority aggregation area identifier as the patrol battery pack.

7. The method of claim 5, wherein the method is characterized by, The positioning of the patrol battery pack according to the real priority further includes: Perform authentication backtracking on the basic degradation prediction score, and establish a change rate trend index; If the change rate trend index meets a preset breakthrough threshold, add a corresponding battery as the patrol battery pack.

8. The method of claim 1, wherein the method is characterized by, After the configuration of the patrol task according to the compensation result and the battery distribution, the method includes: Perform patrol management based on the patrol task, and establish a patrol data set; Perform consistency authentication of the compensation result by using the patrol data set, generate authentication feedback, and perform patrol self-optimization management according to the authentication feedback.

9. The method of claim 1, wherein the method is characterized by, The configuration of the patrol task according to the compensation result and the battery distribution includes: Establish a multi-objective evaluation function, the multi-objective evaluation function including a total path evaluation objective and a priority priority objective; Perform path optimization based on the multi-objective evaluation function according to the compensation result and the battery distribution, and configure the patrol task based on a path optimization result.

Images