A battery online monitoring method and system with adaptive charge and discharge compensation

Abstract

The present invention discloses a method and system for online monitoring of batteries with adaptive charge and discharge compensation, which relates to the field of battery management technology. The method includes: building a hybrid sensor architecture, collecting multi-dimensional battery operating data streams in real time, and performing defect identification in combination with verification charge and discharge data, determining a battery operating defect feature set, training a battery life prediction network to perform life prediction, obtaining safety hazard diagnostic information in combination with a battery safety hazard database, analyzing the life prediction results and the safety hazard diagnostic information, determining charge and discharge compensation parameters, and performing adaptive charge and discharge compensation control based on the parameters. The present invention solves the technical problems that existing battery monitoring methods are single, difficult to accurately identify defects and hazards, and lack dynamically optimized charge and discharge strategies, which affect battery life and operational safety. The present invention achieves the technical effect of realizing dynamic adaptive charge and discharge control and improving battery life and operational safety through multi-dimensional monitoring and intelligent analysis.

Description

Technical Field

[0001] The present invention relates to the technical field of battery management, and in particular to a method and system for online monitoring of batteries with adaptive charge and discharge compensation. Background Art

[0002] Batteries are widely used in power systems, communication base stations, electric vehicles, and other fields. Their performance and reliability directly impact the normal operation and safety of these devices. However, existing battery monitoring methods often rely on monitoring a single parameter (such as voltage or current), which fails to fully reflect the battery's status and accurately identify potential defects and safety hazards. Furthermore, existing technologies have shortcomings in battery life prediction and charge / discharge strategy optimization, lacking dynamic adjustment capabilities and failing to meet practical application requirements. Summary of the Invention

[0003] The present application provides a method and system for online monitoring of batteries with adaptive charge and discharge compensation, which is used to solve the technical problems that existing battery monitoring methods are single, difficult to accurately identify defects and hidden dangers, and lack dynamically optimized charge and discharge strategies, which affect battery life and operational safety.

[0004] The first aspect of the present application provides a method for online monitoring of batteries with adaptive charge and discharge compensation, the method comprising: building a hybrid sensor architecture, and acquiring a multidimensional battery operating data stream during the operation of a target battery through the hybrid sensor architecture; extracting and constructing a charge and discharge feature fingerprint matrix of the multidimensional battery operating data stream, and performing defect identification on the charge and discharge feature fingerprint matrix in combination with verification charge and discharge data to determine a battery operating defect feature set; training and constructing a battery life prediction network, and using the battery life prediction network to perform life prediction for the battery operating defect feature set within a preset time window to obtain a target battery life prediction result; establishing a battery safety hazard database, and comparing and matching the battery operating defect feature set with the battery safety hazard database to obtain battery safety hazard diagnostic information; performing charge and discharge strategy analysis on the target battery life prediction result and the battery safety hazard diagnostic information to determine target charge and discharge compensation parameters, and performing adaptive compensation control on the target battery based on the target charge and discharge compensation parameters.

[0005] The second aspect of the present application provides a battery online monitoring system with adaptive charge and discharge compensation, the system comprising: a battery working data stream acquisition module, the battery working data stream acquisition module is used to build a hybrid sensor architecture, and the hybrid sensor architecture is used to collect and obtain the multi-dimensional battery working data stream during the operation of the target battery; a working defect recognition module, the working defect recognition module is used to extract the charge and discharge feature fingerprint matrix for constructing the multi-dimensional battery working data stream, and perform defect recognition on the charge and discharge feature fingerprint matrix in combination with the verification charge and discharge data to determine the battery working defect feature set; a battery life prediction module, the battery life prediction module is used to train and construct the battery life A prediction network, which uses the battery life prediction network to predict the life of the battery working defect feature set in a preset time window to obtain a target battery life prediction result; a safety hazard diagnosis module, which is used to establish a battery safety hazard database, compare and match the battery working defect feature set with the battery safety hazard database to obtain battery safety hazard diagnosis information; an adaptive compensation module, which is used to perform charge and discharge strategy analysis on the target battery life prediction result and the battery safety hazard diagnosis information, determine target charge and discharge compensation parameters, and perform adaptive compensation control on the target battery based on the target charge and discharge compensation parameters.

[0006] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0007] The present application provides a method and system for online monitoring of batteries with adaptive charge and discharge compensation, which relate to the field of battery management technology. Multi-dimensional battery data is collected in real time through a hybrid sensor architecture, a charge and discharge feature fingerprint matrix is ​​constructed, defects are identified in combination with verification charge and discharge data, a battery life prediction network is trained to predict life, and the data is compared with a safety hazard database to obtain safety hazard diagnosis information. Based on the prediction results and hazard diagnosis, charge and discharge compensation parameters are determined and adaptive compensation control is performed. This method solves the technical problems of existing battery monitoring methods that are single, difficult to accurately identify defects and hazards, and lack dynamically optimized charge and discharge strategies, which affect battery life and operational safety. It achieves the technical effect of realizing dynamic adaptive charge and discharge control and improving battery life and operational safety through multi-dimensional monitoring and intelligent analysis. BRIEF DESCRIPTION OF THE DRAWINGS

[0008] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0009] Figure 1 A flowchart of a method for online monitoring of a battery with adaptive charge and discharge compensation provided in an embodiment of the present application;

[0010] Figure 2 A schematic structural diagram of a battery online monitoring system with adaptive charge and discharge compensation provided in an embodiment of the present application.

[0011] Explanation of the accompanying symbols: battery working data stream acquisition module 11, working defect identification module 12, battery life prediction module 13, safety hazard diagnosis module 14, adaptive compensation module 15. DETAILED DESCRIPTION

[0012] The present application provides a method and system for online monitoring of batteries with adaptive charge and discharge compensation, which is used to solve the technical problems that existing battery monitoring methods are single, difficult to accurately identify defects and hidden dangers, and lack dynamically optimized charge and discharge strategies, which affect battery life and operational safety.

[0013] The following will be combined with the accompanying drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only some of the embodiments of this application, not all of them. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0014] It should be noted that the terms "first", "second", etc. in the specification of the present application and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or server that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or modules that are not clearly listed or inherent to these processes, methods, products or devices.

[0015] Example 1, as Figure 1 As shown, the present application provides a battery online monitoring method with adaptive charge and discharge compensation, the method comprising:

[0016] P10: Build a hybrid sensor architecture to collect and obtain multi-dimensional battery operation data streams during the operation of the target battery through the hybrid sensor architecture.

[0017] Specifically, a hybrid sensor architecture must be built to provide comprehensive, real-time monitoring of the target battery's operating status. A hybrid sensor architecture organically combines multiple sensors of varying types and functions to form a collaborative sensor network for comprehensive battery status monitoring. The design of this architecture requires comprehensive consideration of factors such as sensor type, quantity, layout, and data acquisition synchronization to ensure accurate and comprehensive acquisition of the battery's multi-dimensional operating data stream.

[0018] These sensors typically include voltage sensors, current sensors, and temperature sensors, each responsible for monitoring different battery operating parameters. For example, voltage sensors monitor battery voltage fluctuations in real time. Voltage changes are a key indicator of battery health, reflecting whether the battery is experiencing charge-discharge imbalance or aging. Current sensors detect the battery's charge and discharge current, which is crucial for determining whether the battery is overcharged or over-discharged, as well as potential internal short circuits. Temperature sensors monitor battery temperature, as temperature changes are directly related to the battery's chemical reaction rate. Excessively high temperatures can accelerate battery capacity decay and even lead to safety issues.

[0019] In some special cases, where the battery application environment is complex, vibration sensors and ambient humidity sensors can also be introduced. Vibration sensors can monitor vibrations generated during device operation, which may affect the physical state and internal structure of the battery, while humidity sensors help monitor humidity changes in the battery's operating environment. These factors can affect the battery's lifespan and safety. Data collected by all sensors is transmitted to the data processing platform via wireless or wired communication, ensuring real-time and stable data transmission.

[0020] The hybrid sensor architecture described above collects multidimensional battery operating data streams in real time during the target battery's operation. A multidimensional data stream refers to a dynamic data sequence encompassing multiple dimensions (such as voltage, current, and temperature), which exhibits continuity and correlation over time. The data collection frequency should be appropriately set based on the battery's actual application scenario and monitoring requirements. For dynamic charging and discharging processes, the collection frequency should be high, for example, tens to hundreds of times per second, to capture rapidly changing signals. For static monitoring, the collection frequency can be appropriately lowered, for example, once per minute. By establishing a hybrid sensor architecture and collecting multidimensional battery operating data streams in real time, comprehensive monitoring of the battery's operating status is achieved, providing accurate data support for subsequent battery health diagnosis, lifespan prediction, and charge and discharge strategy adjustments, ensuring battery safety and reliability under various operating conditions.

[0021] P20: extracting and constructing a charge and discharge feature fingerprint matrix of the multi-dimensional battery operation data stream, performing defect identification on the charge and discharge feature fingerprint matrix in combination with verification charge and discharge data, and determining a battery operation defect feature set.

[0022] Furthermore, in order to extract and construct the charge and discharge feature fingerprint matrix of the multi-dimensional battery operation data stream, step P20 of the embodiment of the present application further includes:

[0023] P21: Analyze the noise characteristics of the multidimensional battery operating data stream to obtain the battery operating data noise characteristics, and determine the battery noise data threshold based on the battery operating data noise characteristics; P22: Perform noise data identification on the multidimensional battery operating data stream according to the battery noise data threshold to obtain a battery operating noise data set; P23: Initialize the Kalman filter to perform filtering preprocessing on the battery operating noise data set to obtain a denoised multidimensional battery operating data stream; P24: Perform data alignment and feature extraction on the denoised multidimensional battery operating data stream based on the sensor data sampling time to construct a charge and discharge feature fingerprint matrix.

[0024] It should be understood that in order to improve the quality of data and the accuracy of subsequent analysis, it is necessary to perform noise processing and feature extraction on the multi-dimensional battery operation data stream, construct a charge and discharge feature fingerprint matrix, and perform defect identification based on this.

[0025] First, noise characteristics analysis is performed on the collected multidimensional battery operating data stream. The goal of noise characteristics analysis is to identify noise components in the data and determine their statistical characteristics, such as the noise amplitude distribution and frequency characteristics. By analyzing the noise characteristics, a reasonable battery noise data threshold can be determined for subsequent noise data identification. Specifically, statistical analysis, frequency domain analysis, and time domain analysis can be used to identify noise characteristics. For example, by calculating statistical quantities such as the mean, variance, and standard deviation, the data's fluctuation range and distribution characteristics can be analyzed; by analyzing its spectral characteristics through Fourier transform, the distribution range of high-frequency noise can be identified; and by observing the data's time domain waveform, abnormal signals such as sudden changes and spikes can be identified. Based on these analysis results, the battery noise data threshold is determined. This threshold should be adjusted according to the actual application scenario and noise characteristics to ensure effective distinction between normal data and noise data.

[0026] Next, based on the determined battery noise data threshold, noise data identification is performed on the multidimensional battery operating data stream. For example, each data point is compared with the threshold, and data points exceeding the threshold range are identified as noise data. This results in a set containing all noise data, namely the battery operating noise dataset. This noise data identification process can be implemented using algorithms such as threshold comparison or sliding window methods. For example, for each data point x, if |x−μ|>threshold, then x is considered noise data, where μ is the mean of the data. To remove noise data and improve data quality and usability, a Kalman filter can be used to pre-process the battery operating noise dataset. The Kalman filter is a highly efficient recursive filter that estimates and filters noisy measurement data based on the system's dynamic model and observation model. During the filter initialization phase, the state vector, covariance matrix, system model, and observation model parameters must be set. After initialization, the Kalman filter is used to filter the battery operating noise data set. Through the two steps of prediction and update, the estimated values ​​of the state variables are gradually corrected to remove the influence of the noise data, thereby obtaining a denoised multidimensional battery operating data stream. This data stream removes the interference of noise and more accurately reflects the actual operating status of the battery.

[0027] After obtaining the denoised multidimensional battery operating data stream, data alignment and feature extraction are performed to construct a charge-discharge characteristic fingerprint matrix. Because the sampling times of different sensors may vary slightly, this data may not be perfectly aligned in time. Therefore, the denoised multidimensional battery operating data stream is first time-aligned based on the sensor data sampling times. Interpolation methods (such as linear interpolation or spline interpolation) can be used to time-align the data to ensure that data in different dimensions have corresponding values ​​at the same time point. Next, feature extraction is performed on the multidimensional battery operating data stream to extract key features that reflect the battery's charge and discharge characteristics from the raw data. Feature extraction methods can include time-domain features, frequency-domain features, statistical features, and custom features, such as the mean, variance, and extreme values ​​of voltage, the rate of change and fluctuation coefficient of current, the cumulative distribution function of temperature, and the correlation coefficient between voltage and current. The extracted feature values ​​are combined into a matrix, the charge-discharge characteristic fingerprint matrix, which comprehensively and accurately describes the battery's behavior during the charge and discharge process, similar to a battery "fingerprint" that is unique and identifiable.

[0028] Finally, defect identification is performed on the charge-discharge feature fingerprint matrix in conjunction with verification charge-discharge data. Verification charge-discharge data refers to data obtained from battery charge-discharge tests under standard conditions. This data reflects the battery's performance characteristics under ideal conditions. By comparing the eigenvalues ​​in the charge-discharge feature fingerprint matrix with the standard eigenvalues ​​in the verification charge-discharge data, potential battery defects can be identified. This is accomplished by calculating feature differences, threshold determination, and defect classification. For example, the difference between the eigenvalues ​​in the charge-discharge feature fingerprint matrix and the standard eigenvalues ​​in the verification charge-discharge data is calculated, using metrics such as Euclidean distance and cosine similarity. Based on a set threshold, it is determined whether the feature difference exceeds the normal range. Ultimately, the identified defect features are aggregated to form a battery operating defect feature set. This feature set details potential battery issues under the current operating state, providing an important basis for subsequent lifespan prediction and safety hazard diagnosis.

[0029] Furthermore, step P24 of the embodiment of the present application further includes:

[0030] P24-1: Arrange the denoised multidimensional battery operating data stream in sequence according to the sensor data sampling time to obtain a multidimensional battery operating timing data stream; P24-2: Perform linear interpolation and time alignment processing on the multidimensional battery operating timing data stream according to the data timing information to obtain a multidimensional battery operating aligned data stream; P24-3: Extract charging features and discharging features from the multidimensional battery operating aligned data stream in turn to obtain a charging feature data set and a discharging feature data set; P24-4: Arrange the feature vectors and construct matrix combinations of the charging feature data set and the discharging feature data set according to the number of charge and discharge cycles to generate the charging and discharging feature fingerprint matrix.

[0031] Optionally, the construction process of the charge-discharge feature fingerprint matrix may be further refined.

[0032] First, the denoised multidimensional battery operating data stream is sorted sequentially according to the sensor data sampling time. The main purpose of this process is to sort the data in time so that the data conforms to the actual acquisition order in the time dimension. Because the time points at which sensors collect data may be asynchronous, sorting by time can make the data more sequential in subsequent processing, thereby ensuring the temporal consistency of the data.

[0033] Next, the battery operating timing data stream is linearly interpolated and time-aligned using the data timing information. Because the sampling times of different sensors may not be completely consistent, direct analysis can result in data mismatches. Therefore, linear interpolation is used to fill in the data sampling gaps, achieving a smooth data transition and generating a multidimensional battery operating aligned data stream. Time alignment aligns the data from different sensors, ensuring data consistency in subsequent analysis and eliminating errors caused by sampling time differences.

[0034] Furthermore, charging and discharging feature extraction is performed sequentially on the multi-dimensional battery operating alignment data stream. Charging feature extraction involves extracting key characteristic data related to the battery's charging state, such as charging voltage, charging current, and charging rate. Discharging feature extraction involves extracting characteristic data related to the battery's discharge state, such as discharge voltage, discharge current, and discharge time. These charging and discharging feature datasets contain important information about the battery's charging and discharging processes and are key to constructing the charge and discharge feature fingerprint matrix.

[0035] Finally, the charge and discharge feature datasets are arranged by feature vector and matrix combination according to the number of charge and discharge cycles to generate a charge and discharge feature fingerprint matrix. The number of charge and discharge cycles is a critical parameter in battery life, reflecting the battery's usage and aging. By arranging and combining the feature data according to the number of charge and discharge cycles, the charge and discharge feature data at different cycle numbers can be integrated into a single matrix, forming a charge and discharge feature fingerprint matrix. This matrix displays the changes in battery characteristics under different charge and discharge cycles in a structured and systematic manner, providing important data support for subsequent battery status assessment, lifespan prediction, and defect identification, and facilitating a more accurate assessment of battery health status.

[0036] Furthermore, to determine the battery operating defect feature set, step P20 in this embodiment of the application further includes:

[0037] P25: Annotate the working status of the verification charge and discharge data to obtain battery charge and discharge status data, which includes normal operation, minor defects and serious defect status data; P26: Use a support vector machine to perform defect recognition training based on the battery charge and discharge status data to generate a battery working defect recognition network; P27: Based on the verification charge and discharge data, perform association alignment and fusion on the charge and discharge feature fingerprint matrix to obtain a charge and discharge fusion feature data set; P28: Perform defect recognition on the charge and discharge fusion feature data set based on the battery working defect recognition network to determine the battery working defect feature set.

[0038] In a possible embodiment of the present application, in order to further determine the battery operation defect feature set, the embodiment of the present application further introduces the processing and analysis of verification charge and discharge data.

[0039] First, the verification charge and discharge data is annotated with the operating status to obtain the battery's charge and discharge status data. The core of this process is to correlate the collected charge and discharge data with the actual operating status of the battery. Specifically, it includes categorizing and annotating the battery's operating status, which is generally divided into three categories: normal operation, minor defects, and severe defects. This annotation can clearly define the battery's status at different operating stages, providing clear identification and training data for subsequent defect identification. Under normal operating conditions, the battery's charge and discharge characteristics are close to the ideal state; under minor defect conditions, the battery may exhibit some anomalies, but the impact on overall performance is minimal; and severe defect conditions indicate significant problems with the battery, which may affect its safety or lifespan.

[0040] Next, a support vector machine (SVM) is used for defect recognition training, generating a battery defect recognition network based on battery charge and discharge status data. A support vector machine is a machine learning algorithm commonly used for classification problems. It can find the optimal separating hyperplane in a high-dimensional space to distinguish data in different states. In this step, the SVM is trained on charge and discharge status data to identify different battery operating states, thereby generating a battery defect recognition network. This network can determine whether a battery is in a defective state by analyzing the battery charge and discharge data, providing accurate defect recognition capabilities for subsequent steps.

[0041] Furthermore, based on the verification charge and discharge data, the charge and discharge feature fingerprint matrix is ​​aligned and fused. Since the charge and discharge feature fingerprint matrix is ​​extracted from actual operating data, while the verification charge and discharge data is obtained under standard conditions, there may be certain differences in data format and time series between the two. Therefore, it is necessary to associate and align the charge and discharge feature fingerprint matrix with the verification charge and discharge data to ensure their consistency in time series and feature dimensions. Through association alignment, it is ensured that the data in the fingerprint matrix can correspond to the actual charge and discharge process, and through data fusion technology, a charge and discharge fusion feature dataset is generated. This fused dataset integrates more battery operating characteristics, can more comprehensively reflect the battery's health status, and provide richer data support for defect identification.

[0042] Finally, based on the generated battery operational defect recognition network, defect recognition is performed on the charge-discharge fusion feature dataset, ultimately determining the battery operational defect feature set. At this stage, the defect recognition network analyzes the fused feature dataset, identifies potential battery defects, and classifies them by severity to generate a battery operational defect feature set. This feature set encompasses battery defect patterns under different operating conditions and can be used for subsequent battery health management, lifespan prediction, and safety hazard diagnosis.

[0043] Through the above steps, the final battery operating defect feature set obtained provides accurate basic data for battery health assessment and fault prediction, can effectively identify potential battery problems, and provide strong support for intelligent management and safety control of batteries.

[0044] P30: Train and construct a battery life prediction network, and use the battery life prediction network to perform life prediction for the battery operating defect feature set within a preset time window to obtain a target battery life prediction result.

[0045] Furthermore, step P30 in the embodiment of the present application further includes:

[0046] P31: Collect and obtain a battery working life database, which includes historical battery working data and corresponding battery life data; P32: Divide and sort the battery working life database according to a preset time window to obtain a battery working life time series data set; P33: Perform multi-logic layer training and merge connections on the battery working life time series data set to construct a battery life prediction network.

[0047] Optionally, a battery life prediction network is trained and constructed, and the battery life prediction network is used to predict the future life of the battery based on the battery operating defect feature set.

[0048] First, a battery lifespan database is collected, which includes historical battery operating data and corresponding battery lifespan data. Historical operating data refers to multiple monitoring data such as voltage, current, and temperature under different operating conditions, providing fundamental information for battery lifespan analysis. The corresponding battery lifespan data records the actual battery lifespan under specific conditions, including information such as continuous use time, number of uses, and occurrence of faults. By correlating this data with the actual operating conditions of the battery, a complete battery lifespan database is formed, providing rich training data for the subsequent lifespan prediction network.

[0049] Next, the battery life database is divided, sorted, and labeled according to preset time windows, thereby generating a battery life time series dataset. The preset time window refers to the time range selected when making life predictions, which can be several hours, days, months, or years, depending on the battery's usage cycle. By dividing, sorting, and labeling the database, historical data can be organized according to time series to form an ordered dataset. The data within each time window contains information about the battery's operating status during that time period, such as average voltage, cumulative charge and discharge volume, maximum temperature, and the corresponding remaining battery life. This method of constructing a time series dataset helps the model capture the patterns of battery life changes over time and provides structured data support for subsequent training.

[0050] Next, multi-logic-layer training and merge-connection are performed on the battery life time series dataset to construct a battery life prediction network. Multi-logic-layer training utilizes a multi-layer neural network structure from deep learning to extract and learn features from the data through multiple logical layers. Each logical layer extracts features at a different level, from low-level raw data features to high-level abstract features, gradually constructing a feature space that effectively characterizes battery life characteristics. During training, optimization algorithms (such as gradient descent) adjust the network's weights and bias parameters, enabling the network to learn the complex nonlinear relationship between input data (battery operating data) and output data (battery life). Merge-connection involves fusing features extracted from different logical layers. Through operations such as fully connected layers or convolutional layers, these multiple layers of features are combined into a comprehensive feature vector, which serves as the final prediction basis. The battery life prediction network constructed in this way can fully leverage the information in historical data to accurately predict battery life.

[0051] Finally, the trained battery life prediction network is used to predict the battery's remaining life within a preset time window using the battery defect feature set, yielding the target battery life prediction result. The battery defect feature set is fed into the prediction network, which, based on its learned patterns, outputs a predicted value for the battery's remaining life within the preset time window. This prediction provides a scientific basis for battery maintenance, replacement, and management, helping users plan battery usage and maintenance strategies in advance and reducing the risks and costs associated with battery failure.

[0052] Furthermore, step P33 of the embodiment of the present application also includes:

[0053] P33-1: Perform operation status data identification and service life data identification on the battery working life time series data set to obtain a battery operation status characteristic data set and a corresponding battery service life data set; P33-2: Use an LSTM network structure to evaluate and train the battery operation status characteristic data set to generate a battery operation status characteristic evaluation layer; P33-3: Use a deep neural network structure to predict and train the battery service life data set to obtain a battery life prediction logic layer; P33-4: Merge and connect the battery operation status characteristic evaluation layer and the battery life prediction logic layer to construct the battery life prediction network.

[0054] Specifically, in order to further improve the accuracy and reliability of the battery life prediction network, the construction process of the battery life prediction network can be further refined to ensure that the battery life prediction network finally constructed can effectively perform battery life prediction.

[0055] First, each time window data in the battery working life time series dataset is identified to obtain a battery operating status characteristic dataset and a corresponding battery service life dataset. Operating status data identification refers to the classification of the battery operating status within each time window. For example, the battery operating status is divided into different levels such as normal operation, mild aging, and severe aging. This classification can be achieved by analyzing the changing trends of battery parameters such as voltage, current, and temperature. Service life data identification refers to the quantification of the remaining battery service life corresponding to each time window, for example, expressed as the remaining number of cycles or remaining usage time of the battery. Through this identification method, complex battery operating data can be converted into a feature dataset with clear physical meaning, providing clear data input for subsequent model training.

[0056] Next, a long short-term memory (LSTM) network structure is used to evaluate and train the battery operating state characteristic dataset, generating a battery operating state characteristic evaluation layer. The LSTM network is a deep learning algorithm commonly used to process time series data, effectively capturing long-term dependencies within sequence data. Because battery operating state characteristics exhibit distinct time series characteristics—for example, battery performance changes evolve gradually over time—the LSTM network structure is well-suited for evaluating battery operating state. During training, the battery operating state characteristic dataset is input into the LSTM network. By learning the time series features in the data, the network accurately assesses the battery's operating state and generates a battery operating state characteristic evaluation layer. This evaluation layer outputs the battery's operating state level within each time window, providing important status information for subsequent lifespan prediction.

[0057] Next, a deep neural network structure is used to perform prediction training on the battery life dataset, resulting in a battery life prediction logic layer. Deep neural networks have powerful feature learning and nonlinear fitting capabilities, and can model complex input-output relationships. In this step, the battery life dataset is input into the deep neural network. Through multi-layer feature extraction and learning, the network is able to learn the complex relationship between battery life and operating status. Ultimately, the resulting battery life prediction logic layer can predict the remaining battery life based on the input battery operating status characteristics. This logic layer is the core component of the battery life prediction network, and its accuracy and reliability directly determine the performance of the entire prediction system.

[0058] Finally, the battery operating status characteristic evaluation layer and the battery life prediction logic layer are merged and connected to construct a battery life prediction network. Merged connection refers to fusing the outputs of the two network layers to form a complete prediction model. For example, the output of the battery operating status characteristic evaluation layer can be used as one of the input features of the battery life prediction logic layer, or the outputs of the two network layers can be merged by performing weighted summation through a fully connected layer. The battery life prediction network constructed in this way can make full use of the correlation between the battery operating status characteristics and service life data to achieve accurate prediction of battery life. The network can not only take into account the current operating status of the battery, but also combine its historical service life data to provide more accurate and reliable prediction results.

[0059] P40: Establish a battery safety hazard database, compare and match the battery operating defect feature set with the battery safety hazard database, and obtain battery safety hazard diagnosis information.

[0060] It should be understood that by establishing a battery safety hazard database and comparing and matching the battery operating defect feature set with the database, the diagnosis of battery safety hazards can be achieved.

[0061] First, establishing a battery safety hazard database is the starting point of this step. The battery safety hazard database should include a variety of safety issues and defect modes that batteries may face, such as overcharging, over-discharging, abnormal temperature, internal battery short circuits, leakage, swelling, etc. The sources of these safety hazards can be factors such as the material properties of the battery, the operating environment, and improper operation during the charging and discharging process. The database should include all possible fault and hazard characteristics of different battery types under different operating conditions, and provide standard data for subsequent comparisons. The construction of the database needs to be based on a large amount of historical data, experimental results, and industry standards, and it needs to be continuously updated to ensure its timeliness and accuracy.

[0062] Next, the battery operating defect feature set is compared and matched with the battery safety hazard database. This feature set includes the charge and discharge feature fingerprint matrix extracted in the previous steps, as well as the defect identification results. It records abnormal patterns that may occur during the battery's actual operation. These abnormal patterns may indicate potential safety hazards. By comparing this feature set with known defect patterns in the safety hazard database, the system can identify potential hazards under the battery's current operating state. This comparison process can use machine learning, pattern recognition, and other technologies to automatically match the battery's operating characteristics with standard hazard features in the database, quickly and accurately identifying potential problems.

[0063] Ultimately, through comparison and matching, diagnostic information on battery safety hazards is generated. This diagnostic information can provide information on current battery safety risks, such as overheating, overcharging or over-discharging, possible short circuits, or other safety failures. Furthermore, diagnostic information can be categorized by the severity of the hazard, providing a basis for subsequent early warning and maintenance. If a battery presents a significant safety hazard, the system will automatically issue a warning, prompting the operator to conduct a prompt inspection and address the issue to avoid an accident.

[0064] Through the above steps, potential safety hazards that may exist in the battery during use can be effectively identified, and safety hazard diagnostic information can be used to provide decision support for battery safety management, reduce the occurrence of battery failures and safety accidents, and improve battery safety and reliability.

[0065] P50: Performing a charge and discharge strategy analysis on the target battery life prediction result and the battery safety hazard diagnosis information, determining a target charge and discharge compensation parameter, and performing adaptive compensation control on the target battery based on the target charge and discharge compensation parameter.

[0066] Furthermore, step P50 in the embodiment of the present application further includes:

[0067] P51: Build a battery charge and discharge compensation strategy library, which includes historical battery life data, battery safety hazard diagnosis data and corresponding charge and discharge compensation strategy parameters; P52: Obtain the battery charge and discharge optimization target, and establish a charge and discharge compensation strategy fitness function based on the battery charge and discharge optimization target; P53: Obtain the charge and discharge compensation particle swarm space based on the battery charge and discharge compensation strategy library; P54: Use the charge and discharge compensation strategy fitness function to perform a global search and optimization in the charge and discharge compensation particle swarm space to determine the target charge and discharge compensation parameter with the largest particle fitness.

[0068] Optionally, in order to achieve adaptive charge and discharge compensation control of the target battery, the charge and discharge strategy analysis is performed on the target battery life prediction results and battery safety hazard diagnosis information, and then the target charge and discharge compensation parameters are determined, thereby optimizing the battery life and safety.

[0069] First, a battery charge-discharge compensation strategy library is built. This library should include three key elements: historical battery life data, battery safety hazard diagnostic data, and corresponding charge-discharge compensation strategy parameters. Historical battery life data provides information about battery life performance under different operating conditions, helping to understand the battery degradation process under specific conditions. Battery safety hazard diagnostic data records potential safety risks that may arise during actual battery operation, such as overcharging, over-discharging, and overheating. Charge-discharge compensation strategy parameters include compensation plans and adjustment strategies tailored to these lifespan and safety issues. By integrating this data, the battery charge-discharge compensation strategy library provides rich reference information for subsequent compensation strategy optimization.

[0070] Next, the battery charge and discharge optimization objectives are obtained, and a fitness function for the charge and discharge compensation strategy is established based on these objectives. These objectives typically include extending battery life, improving battery efficiency, reducing energy loss, and ensuring safety. Based on this, the fitness function is used to evaluate the effectiveness of different charge and discharge compensation strategies. Specifically, the fitness function quantifies the impact of different strategies on battery life and safety, providing a quantitative basis for strategy selection. By establishing this fitness function, the charge and discharge process can be optimized more precisely, ensuring that each charge and discharge cycle meets the battery's health and safety requirements as closely as possible.

[0071] Furthermore, the charge-discharge compensation particle swarm space is obtained based on the battery charge-discharge compensation strategy library. Particle swarm optimization (PSO) is an optimization algorithm based on swarm intelligence. In this step, the charge-discharge compensation strategy parameters are considered as particles in a particle swarm, with each particle representing a set of possible charge-discharge compensation parameters. The particle swarm space is the set of all possible combinations of charge-discharge compensation parameters. By initializing the particle swarm, each particle is assigned an initial position (i.e., an initial set of charge-discharge compensation parameters) and velocity, thus constructing the charge-discharge compensation particle swarm space.

[0072] Finally, a global search is performed within the charge-discharge compensation particle swarm space using the charge-discharge compensation strategy's fitness function to determine the target charge-discharge compensation parameters that maximize particle fitness. For example, using the particle swarm optimization algorithm, each particle adjusts its position and velocity based on its own fitness and information about other particles in the swarm, gradually approaching the optimal solution. In each iteration, the fitness of each particle is calculated, and both the individual optimal position and the global optimal position are updated. Through multiple iterations, the target charge-discharge compensation parameters that maximize particle fitness are ultimately determined. These parameters best meet battery charge-discharge optimization objectives, such as extending battery life and reducing safety risks.

[0073] Ultimately, adaptive compensation control is performed on the target battery based on the determined target charge and discharge compensation parameters. This control process is dynamic, continuously adjusting the charge and discharge strategy based on the battery's real-time operating status, lifespan prediction, and safety hazard diagnosis information to ensure the battery always operates in optimal conditions.

[0074] In summary, the embodiments of the present application have at least the following technical effects:

[0075] This application builds a hybrid sensor architecture to collect multi-dimensional battery operating data streams in real time, construct a charge and discharge feature fingerprint matrix, and combines it with verification charge and discharge data to identify defects and determine the battery operating defect feature set. Based on this feature set, a battery life prediction network is trained to predict lifespan, and compared with a battery safety hazard database to obtain safety hazard diagnostic information. Finally, by analyzing the lifespan prediction results and safety hazard diagnostic information, the charge and discharge compensation parameters are determined, and based on this, adaptive charge and discharge compensation control is performed to optimize battery performance and safety.

[0076] The technical effect of achieving dynamic adaptive charge and discharge control through multi-dimensional monitoring and intelligent analysis, and improving battery life and operational safety has been achieved.

[0077] Embodiment 2 is based on the same inventive concept as the method for online monitoring of a battery with adaptive charge and discharge compensation in the above embodiment. Figure 2 As shown, the present application provides a battery online monitoring system with adaptive charge and discharge compensation. The system and method embodiments in the present application are based on the same inventive concept. The system includes:

[0078] The battery operation data stream acquisition module 11 is used to build a hybrid sensor architecture, and collect and acquire the multi-dimensional battery operation data stream during the operation of the target battery through the hybrid sensor architecture.

[0079] The working defect identification module 12 is used to extract the charge and discharge feature fingerprint matrix that constructs the multi-dimensional battery working data stream, perform defect identification on the charge and discharge feature fingerprint matrix in combination with the verification charge and discharge data, and determine the battery working defect feature set.

[0080] The battery life prediction module 13 is used to train and construct a battery life prediction network, and use the battery life prediction network to predict the life of the battery working defect feature set in a preset time window to obtain a target battery life prediction result.

[0081] The safety hazard diagnosis module 14 is used to establish a battery safety hazard database, compare and match the battery operating defect feature set with the battery safety hazard database, and obtain battery safety hazard diagnosis information.

[0082] The adaptive compensation module 15 is used to perform a charge and discharge strategy analysis on the target battery life prediction result and the battery safety hazard diagnosis information, determine target charge and discharge compensation parameters, and perform adaptive compensation control on the target battery based on the target charge and discharge compensation parameters.

[0083] Furthermore, the work defect identification module 12 is further configured to perform the following steps:

[0084] Noise characteristics of the multidimensional battery operating data stream are analyzed to obtain battery operating data noise characteristics, and a battery noise data threshold is determined based on the battery operating data noise characteristics; noise data of the multidimensional battery operating data stream is identified according to the battery noise data threshold to obtain a battery operating noise data set; a Kalman filter is initialized to perform filtering preprocessing on the battery operating noise data set to obtain a de-noised multidimensional battery operating data stream; data alignment and feature extraction are performed on the de-noised multidimensional battery operating data stream based on sensor data sampling time to construct a charge and discharge feature fingerprint matrix.

[0085] Furthermore, the work defect identification module 12 is further configured to perform the following steps:

[0086] The denoised multidimensional battery operation data stream is arranged in sequence according to the sensor data sampling time to obtain a multidimensional battery operation timing data stream; the multidimensional battery operation timing data stream is linearly interpolated and time aligned according to the data timing information to obtain a multidimensional battery operation aligned data stream; the charging characteristics and the discharging characteristics of the multidimensional battery operation aligned data stream are extracted in sequence to obtain a charging characteristic data set and a discharging characteristic data set; the charging characteristic data set and the discharging characteristic data set are arranged by feature vectors and constructed by matrix combination according to the number of charge and discharge cycles to generate the charging and discharging characteristic fingerprint matrix.

[0087] Furthermore, the work defect identification module 12 is further configured to perform the following steps:

[0088] The verification charge and discharge data are annotated with working status to obtain battery charge and discharge status data, wherein the battery charge and discharge status data includes normal operation, minor defect and serious defect status data; a support vector machine is used to perform defect recognition training based on the battery charge and discharge status data to generate a battery working defect recognition network; based on the verification charge and discharge data, the charge and discharge feature fingerprint matrix is ​​associated, aligned and fused to obtain a charge and discharge fusion feature data set; defect recognition is performed on the charge and discharge fusion feature data set based on the battery working defect recognition network to determine the battery working defect feature set.

[0089] Furthermore, the battery life prediction module 13 is further configured to perform the following steps:

[0090] A battery service life database is acquired, the battery service life database including historical battery service data and corresponding battery service life data; the battery service life database is divided, sorted and identified according to a preset time window to obtain a battery service life time series data set; and the battery service life time series data set is trained and merged at multiple logical layers to construct a battery service life prediction network.

[0091] Furthermore, the battery life prediction module 13 is further configured to perform the following steps:

[0092] The battery operating life time series data set is subjected to operating status data identification and service life data identification to obtain a battery operating status characteristic data set and a corresponding battery service life data set; the battery operating status characteristic data set is evaluated and trained using an LSTM network structure to generate a battery operating status characteristic evaluation layer; the battery service life data set is predicted and trained using a deep neural network structure to obtain a battery service life prediction logic layer; the battery operating status characteristic evaluation layer and the battery service life prediction logic layer are merged and connected to construct the battery service life prediction network.

[0093] Furthermore, the adaptive compensation module 15 is further configured to perform the following steps:

[0094] A battery charge and discharge compensation strategy library is established, wherein the battery charge and discharge compensation strategy library includes historical battery life data, battery safety hazard diagnosis data, and corresponding charge and discharge compensation strategy parameters; a battery charge and discharge optimization target is obtained, and a charge and discharge compensation strategy fitness function is established based on the battery charge and discharge optimization target; a charge and discharge compensation particle swarm space is obtained based on the battery charge and discharge compensation strategy library; and a global search and optimization is performed within the charge and discharge compensation particle swarm space using the charge and discharge compensation strategy fitness function to determine the target charge and discharge compensation parameter with the maximum particle fitness.

[0095] It should be noted that the order in which the embodiments of the present application are presented is for illustrative purposes only and does not necessarily represent the superiority or inferiority of the embodiments. Furthermore, the foregoing descriptions of specific embodiments of this specification are provided. Furthermore, the processes depicted in the accompanying drawings do not necessarily require the specific order or sequential sequence shown to achieve the desired results. In certain embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0096] The above description is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present application should be included in the scope of protection of the present application.

[0097] This specification and drawings are merely illustrative of the present application and are intended to cover any and all modifications, variations, combinations, or equivalents within the scope of this application. Obviously, those skilled in the art may make various modifications and variations to this application without departing from the scope of this application. Thus, this application is intended to include such modifications and variations as fall within the scope of this application and its equivalents.

Claims

1. A battery online monitoring method with adaptive charge and discharge compensation, characterized in that: The method comprises: Building a hybrid sensor architecture to collect multi-dimensional battery operation data streams during the operation of the target battery; Extracting and constructing a charge and discharge feature fingerprint matrix of the multidimensional battery operation data stream, performing defect identification on the charge and discharge feature fingerprint matrix in combination with verification charge and discharge data, and determining a battery operation defect feature set; Training and constructing a battery life prediction network, using the battery life prediction network to perform life prediction for the battery operating defect feature set within a preset time window to obtain a target battery life prediction result; Establishing a battery safety hazard database, comparing and matching the battery operating defect feature set with the battery safety hazard database to obtain battery safety hazard diagnostic information; Performing a charge and discharge strategy analysis on the target battery life prediction result and the battery safety hazard diagnosis information, determining a target charge and discharge compensation parameter, and performing adaptive compensation control on the target battery based on the target charge and discharge compensation parameter; Determining the target charge-discharge compensation parameter includes: Building a battery charge and discharge compensation strategy library, which includes historical battery life data, battery safety hazard diagnosis data, and corresponding charge and discharge compensation strategy parameters; Obtaining a battery charge and discharge optimization target, and establishing a charge and discharge compensation strategy fitness function based on the battery charge and discharge optimization target; Obtaining a charge-discharge compensation particle swarm space according to the battery charge-discharge compensation strategy library; The charge-discharge compensation strategy fitness function is used to perform a global search and optimization in the charge-discharge compensation particle swarm space to determine the target charge-discharge compensation parameter with the maximum particle fitness.

2. The method for online monitoring of a battery with adaptive charge and discharge compensation according to claim 1, characterized in that: The extraction and construction of the charge and discharge feature fingerprint matrix of the multi-dimensional battery operation data stream includes: performing noise characteristic analysis on the multi-dimensional battery operating data stream to obtain a battery operating data noise characteristic, and determining a battery noise data threshold based on the battery operating data noise characteristic; Performing noise data identification on the multi-dimensional battery operation data stream according to the battery noise data threshold to obtain a battery operation noise data set; Initializing a Kalman filter and performing filtering preprocessing on the battery operation noise data set to obtain a denoised multidimensional battery operation data stream; Data alignment and feature extraction are performed on the denoised multi-dimensional battery operation data stream based on the sensor data sampling time to construct a charge and discharge feature fingerprint matrix.

3. The method for online monitoring of a battery with adaptive charge and discharge compensation according to claim 2, characterized in that: The construction of the charge-discharge characteristic fingerprint matrix includes: Arranging the denoised multidimensional battery operation data stream in sequence according to the sensor data sampling time to obtain a multidimensional battery operation time series data stream; Performing linear interpolation and time alignment processing on the multidimensional battery operation timing data stream according to the data timing information to obtain a multidimensional battery operation aligned data stream; Sequentially extracting charging features and discharging features from the multi-dimensional battery operation alignment data stream to obtain a charging feature data set and a discharging feature data set; The charge feature data set and the discharge feature data set are constructed by performing feature vector arrangement and matrix combination according to the number of charge and discharge cycles to generate the charge and discharge feature fingerprint matrix.

4. The method for online monitoring of a battery with adaptive charge and discharge compensation according to claim 1, wherein: The step of determining a battery operation defect feature set includes: Performing working status marking on the verification charge and discharge data to obtain battery charge and discharge status data, wherein the battery charge and discharge status data includes normal working, minor defect, and serious defect status data; Using a support vector machine to perform defect recognition training based on the battery charge and discharge status data to generate a battery working defect recognition network; Based on the verification charge and discharge data, the charge and discharge feature fingerprint matrix is ​​associated, aligned and fused to obtain a charge and discharge fusion feature data set; Defect recognition is performed on the charge-discharge fusion feature data set based on the battery operation defect recognition network to determine the battery operation defect feature set.

5. The method for online monitoring of a battery with adaptive charge and discharge compensation according to claim 1, characterized in that: The training and construction of the battery life prediction network includes: Collecting and acquiring a battery service life database, wherein the battery service life database includes historical battery service data and corresponding battery service life data; Dividing and sorting the battery service life database according to a preset time window to obtain a battery service life time series data set; Multi-logic layer training and merge connection are performed on the battery service life time series data set to build a battery life prediction network.

6. The method for online monitoring of a battery with adaptive charge and discharge compensation according to claim 5, characterized in that: The battery life prediction network is constructed, including: Performing operation state data identification and service life data identification on the battery operation life time series data set to obtain a battery operation state characteristic data set and a corresponding battery service life data set; Using an LSTM network structure to evaluate and train the battery operating state characteristic data set to generate a battery operating state characteristic evaluation layer; Performing prediction training on the battery life dataset using a deep neural network structure to obtain a battery life prediction logic layer; The battery operating state characteristic evaluation layer and the battery life prediction logic layer are merged and connected to construct the battery life prediction network.

7. A battery online monitoring system with adaptive charge and discharge compensation, characterized in that: The system comprises: A battery operating data stream acquisition module, which is used to build a hybrid sensor architecture and collect and acquire multi-dimensional battery operating data streams during the operation of the target battery through the hybrid sensor architecture; an operating defect identification module, the operating defect identification module being used to extract a charge and discharge feature fingerprint matrix for constructing the multi-dimensional battery operating data stream, perform defect identification on the charge and discharge feature fingerprint matrix in combination with verification charge and discharge data, and determine a battery operating defect feature set; A battery life prediction module, the battery life prediction module is used to train and construct a battery life prediction network, and use the battery life prediction network to perform life prediction for the battery operating defect feature set within a preset time window to obtain a target battery life prediction result; a safety hazard diagnosis module, the safety hazard diagnosis module being used to establish a battery safety hazard database, compare and match the battery operating defect feature set with the battery safety hazard database, and obtain battery safety hazard diagnosis information; an adaptive compensation module, configured to perform a charge-discharge strategy analysis on the target battery life prediction result and the battery safety hazard diagnosis information, determine a target charge-discharge compensation parameter, and perform adaptive compensation control on the target battery based on the target charge-discharge compensation parameter; When determining the target charge and discharge compensation parameters, the adaptive compensation module is further configured to: Building a battery charge and discharge compensation strategy library, which includes historical battery life data, battery safety hazard diagnosis data, and corresponding charge and discharge compensation strategy parameters; Obtaining a battery charge and discharge optimization target, and establishing a charge and discharge compensation strategy fitness function based on the battery charge and discharge optimization target; Obtaining a charge-discharge compensation particle swarm space according to the battery charge-discharge compensation strategy library; The charge-discharge compensation strategy fitness function is used to perform a global search and optimization in the charge-discharge compensation particle swarm space to determine the target charge-discharge compensation parameter with the maximum particle fitness.

Images