An intelligent evaluation method and system for the health state of a mobile energy storage power supply

Abstract

The present invention discloses an intelligent evaluation method and system for the health state of a mobile energy storage power supply. The multi-parameter monitoring data of the power supply is input into a battery physical model for preliminary evaluation to obtain a first evaluation value. Feature extraction and feature selection are performed on the acquired monitoring data and environmental data, and the obtained principal component matrix is input into a health state evaluation model for prediction to obtain a second evaluation value. The first evaluation value is compared with the second evaluation value, and the parameters of the battery physical model are adjusted by using the health state evaluation model to obtain an optimized battery physical model. A third evaluation value is obtained based on the optimized battery physical model. The final battery health state evaluation result is obtained based on the third evaluation value and the second evaluation value. By collecting various monitoring data and environmental data in real time, the health state of the battery can be comprehensively understood, thereby improving the accuracy of the evaluation. In addition, the established physical model and the health state evaluation model are combined to improve the accuracy and robustness of the evaluation result.

Description

Technical Field

[0001] The present invention relates to the technical field of battery management, and in particular, to an intelligent evaluation method and system for the health state of a mobile energy storage power supply. Background Art

[0002] The statements in this section merely provide background technical information related to the present disclosure and do not necessarily constitute prior art.

[0003] With the rapid development of fields such as portable electronic devices, renewable energy storage, and electric transportation, mobile energy storage power supplies play an increasingly important role in daily life and industrial applications. An efficient, safe, and reliable mobile energy storage power supply is not only the key to meeting user needs but also an important foundation for promoting sustainable development.

[0004] However, the health state of a mobile energy storage power supply directly affects its performance and safety. During the use of the battery, it may be affected by various factors such as charge and discharge cycles, environmental temperature, and usage habits, resulting in problems such as battery capacity attenuation and internal resistance increase. If these problems are not identified in a timely manner, they may lead to battery failures, equipment damage, and even safety hazards, bringing economic losses and safety risks to users. Therefore, it is particularly important to conduct accurate health state assessment.

[0005] In the prior art, there are already various health state assessment methods, such as simple monitoring methods based on parameters such as voltage, current, and temperature. However, these methods often have defects such as insufficient monitoring accuracy, limited data analysis capabilities, and failure to consider environmental factors, resulting in inaccurate and incomplete assessment results. For example, the assessment method based on a statistical or empirical data assessment model. However, these methods lack the interpretability of the assessment results. Summary of the Invention

[0006] To overcome the deficiencies of the above prior art, the present invention provides an intelligent evaluation method and system for the health state of a mobile energy storage power supply. By collecting various monitoring data and environmental data in real time, the health state of the battery can be comprehensively understood, thereby improving the accuracy of the assessment; and by combining the established physical model and the health state assessment model, the accuracy and robustness of the assessment results are further improved.

[0007] To achieve the above object, one or more embodiments of the present invention provide the following technical solutions:

[0008] In the first aspect, the present invention provides an intelligent evaluation method for the health state of a mobile energy storage power supply, including:

[0009] Obtain multi-parameter monitoring data of the mobile energy storage power supply and input it into the battery physical model for preliminary evaluation to obtain the first evaluation value of the battery health state;

[0010] Obtain multi-parameter monitoring data and environmental data of the mobile energy storage power supply, perform feature extraction and feature selection on the multi-parameter monitoring data and environmental data, and use the principal component analysis method to fuse the selected features to obtain a principal component matrix; input the principal component matrix into a pre-trained health status evaluation model for further prediction to obtain a second evaluation value of the battery health status;

[0011] Compare the first evaluation value with the second evaluation value, and use the health status evaluation model to adjust the parameters of the battery physical model to obtain an optimized battery physical model;

[0012] Perform re-evaluation based on the optimized battery physical model to obtain a third evaluation value of the battery health status; perform weighted averaging on the third evaluation value and the second evaluation value to obtain the final battery health status evaluation result.

[0013] In a further technical solution, use multi-scale isolation forests to identify and remove abnormal data points in the monitoring data, specifically:

[0014] Perform data cleaning and noise removal on the monitoring data, and decompose the monitoring data into different time scales;

[0015] Train isolation forest models on different time scales respectively to obtain abnormal scores for each time scale;

[0016] Fuse the abnormal scores of different time scales to generate a comprehensive abnormal score;

[0017] Dynamically adjust the threshold, mark and output abnormal data points exceeding the threshold, and remove the marked abnormal data points.

[0018] In a further technical solution, the battery physical model includes an internal resistance model, a capacity attenuation model, and an open-circuit voltage model.

[0019] In a further technical solution, obtain multi-parameter monitoring data of the mobile energy storage power supply and input it into the battery physical model for preliminary evaluation to obtain a first evaluation value of the battery health status. Specifically: obtain the first evaluation value of the battery health status by weighted fusion of the calculation results of the internal resistance model, the capacity attenuation model, and the open-circuit voltage model, which is expressed by the following formula:

[0020]

[0021] Among them, is the first evaluation value; , , are weight coefficients; is the current internal resistance; is the initial internal resistance; is the current capacity; is the initial capacity; is the current temperature; is the maximum safe operating temperature.

[0022] A further technical solution is that the feature selection adopts a dynamic feature selection mechanism, specifically:

[0023] The filtering method is used for preliminary feature screening to select features with high correlation;

[0024] Recursive feature elimination is used to further refine the selected features, and features with low importance are gradually removed;

[0025] Dynamic feature selection is carried out in the changes of real-time multi-parameter monitoring data and environmental data, and the feature set is adjusted according to the changes in data distribution.

[0026] A further technical solution is that the health state evaluation model adopts an improved long short-term memory network (LSTM) model, including an input layer, a multi-layer perceptron layer, multiple LSTM layers, and an output layer connected in sequence, and residual connections are added between each LSTM layer.

[0027] A further technical solution is to use the health state evaluation model to correct the parameters in the battery physical model through a loss function and an optimization algorithm, and iteratively update the model.

[0028] In a second aspect, the present invention provides a mobile energy storage power supply health state intelligent evaluation system, including:

[0029] A physical model evaluation module, which is configured to: obtain multi-parameter monitoring data of the mobile energy storage power supply and input it into the battery physical model for preliminary evaluation to obtain a first evaluation value of the battery health state;

[0030] A network model evaluation module, which is configured to: obtain multi-parameter monitoring data and environmental data of the mobile energy storage power supply, perform feature extraction and feature selection on the multi-parameter monitoring data and environmental data, and use the principal component analysis method to fuse the selected features to obtain a principal component matrix; input the principal component matrix into a pre-trained health state evaluation model for further prediction to obtain a second evaluation value of the battery health state;

[0031] A physical model optimization module, which is configured to: compare the first evaluation value with the second evaluation value, and use the health state evaluation model to adjust the parameters of the battery physical model to obtain an optimized battery physical model;

[0032] A health status assessment module, which is configured to: re-evaluate based on an optimized battery physical model to obtain a third evaluation value of the battery health status; perform a weighted average of the third evaluation value and the second evaluation value to obtain a final battery health status assessment result.

[0033] In a third aspect, the present invention provides a computer-readable storage medium, on which a computer program is stored, and when the program is executed by a processor, it implements the steps in an intelligent assessment of the health status of a mobile energy storage power supply as described in the first aspect.

[0034] In a fourth aspect, the present invention provides a computer device, including a memory, a processor, and a computer program stored on the memory and executable on the processor, and when the processor executes the program, it implements the steps in an intelligent assessment of the health status of a mobile energy storage power supply as described in the first aspect.

[0035] The above one or more technical solutions have the following beneficial effects:

[0036] By combining a physical model and a health status assessment model, the present invention improves the accuracy and robustness of the assessment results. The physical model can provide basic mechanism information of the battery, while the data-driven model can capture complex patterns and non-linear relationships in real-time data, and adjust the parameters in the physical model through the health status assessment model, so as to show higher accuracy and adaptability in the real environment. Moreover, the physical model provides a transparent and interpretable basis, and when combined with the health status assessment model, it can provide more interpretability for the assessment results. For example, through the physical model, it is possible to clearly analyze how factors such as temperature, charge and discharge cycles, and voltage changes affect the performance of the battery, thus providing more accurate theoretical support for the health status assessment.

[0037] By collecting various monitoring data (such as voltage, current, temperature, internal resistance, and charge and discharge cycle times) and environmental data in real-time, the present invention can comprehensively understand the health status of the battery, thereby improving the accuracy of the assessment; using an improved isolation forest to identify outliers in the monitoring data can capture the abnormal characteristics of the data at different time scales, which is more suitable for the complex, multi-dimensional characteristics and real-time requirements of mobile energy storage power supplies. Especially in real-time data, multi-scale detection helps to improve the accuracy of anomaly detection and reduce false alarms and missed detections; using various feature extraction methods (such as statistical features, time decay features, frequency domain features, etc.) can effectively capture the changes in the battery state and enhance the prediction ability of the model; and using an improved feature selection method, namely a dynamic feature selection mechanism and a real-time update mechanism, can dynamically adjust the feature set according to the changes in new data, ensuring that the health status assessment model remains effective in a changing environment.

[0038] By inputting the principal component matrix into the health status evaluation model, the output health status score can provide clear battery health status information for users, helping them make maintenance or replacement decisions. The system can provide maintenance suggestions in a timely manner according to the health status of the battery, improve the safety and reliability of users' use of the battery, reduce the failure rate and maintenance cost by optimizing the use of mobile energy storage power supplies, and enhance safety. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] The accompanying drawings forming a part of this invention are used to provide a further understanding of the invention. The schematic embodiments and descriptions thereof of the invention are used to explain the invention and do not constitute an improper limitation of the invention.

[0040] Figure 1 It is a flowchart of the evaluation method of the embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0041] It should be noted that the following detailed descriptions are all exemplary and are intended to provide further explanations of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by those of ordinary skill in the technical field to which the present invention belongs.

[0042] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit the exemplary embodiments according to the present invention. As used herein, unless the context clearly indicates otherwise, the singular forms are also intended to include the plural forms. In addition, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof.

[0043] In the case of no conflict, the embodiments in the present invention and the features in the embodiments can be combined with each other.

[0044] During the use of mobile energy storage power supplies, their monitoring data will change dynamically with time, load, and environmental conditions. These changes have an important impact on battery health. For example, mobile energy storage power supplies can charge multiple loads, and excessive or frequent load changes can cause the battery to be over-discharged and damage battery health. In recent years, the rapid development of artificial intelligence and big data analysis technologies has provided new solutions for health status evaluation. The data-driven intelligent evaluation method can comprehensively consider various influencing factors and achieve more accurate health status evaluation. Therefore, developing an efficient and intelligent health status evaluation method and system for mobile energy storage power supplies has become an urgent need to improve equipment performance, ensure user safety, and promote market development.

[0045] Embodiment 1

[0046] Such as Figure 1As shown in the figure, this embodiment discloses an intelligent evaluation method for the health state of a mobile energy storage power supply, including:

[0047] S1: Obtain the multi-parameter monitoring data of the mobile energy storage power supply and input it into the battery physical model for preliminary evaluation to obtain the first evaluation value of the battery health state.

[0048] S101: Collect the multi-parameter monitoring data of the mobile energy storage power supply and perform preprocessing.

[0049] Install a variety of sensors on the battery unit of the mobile energy storage power supply to collect parameter monitoring data such as voltage, current, temperature, internal resistance, charge and discharge cycle times, and remaining power in real time. Combine different types of sensors (such as temperature sensors, voltage sensors, internal resistance sensors, charge and discharge cycle counters, etc.) to form a comprehensive monitoring system to obtain more comprehensive battery state data.

[0050] Dynamically adjust the data collection frequency according to the operating state of the battery. For example, increase the sampling frequency during the charging or discharging peak period, and decrease the frequency during the steady-state operation to optimize the data collection efficiency.

[0051] Since the obtained monitoring data may contain noise, missing values, or outliers, data preprocessing is required. Perform data preprocessing on the collected data, including data cleaning, noise removal, and outlier rejection, to ensure the accuracy of the data. Use the multi-scale isolation forest to effectively identify abnormal data points when rejecting outliers, providing strong support for the health assessment of the mobile energy storage power supply.

[0052] Improve the isolation forest to obtain the multi-scale isolation forest, which can capture the abnormal characteristics of the data at different time scales, and is more suitable for the complex, multi-dimensional characteristics and real-time requirements of the mobile energy storage power supply. Especially in real-time data, multi-scale detection helps to improve the accuracy of abnormal detection, reduce false alarms and missed reports, which will better support the accurate assessment of the battery health state. Use the multi-scale isolation forest to detect abnormal data points, specifically including:

[0053] (1) Perform data cleaning and noise removal on the collected data (i.e., monitoring data) to ensure data quality; decompose the time series data (i.e., monitoring data) into different time scales to capture short-term and long-term change trends; select appropriate parameters for the isolation forest model, such as voltage, current, temperature, etc., and calculate features such as moving average and standard deviation within short-term and long-term time windows.

[0054] (2) Train the isolation forest model on different time scales respectively to obtain the abnormal score for each time scale.

[0055] 1) Build isolation forest models for different time scales, such as short-term and long-term, respectively;

[0056] 2) Each isolation forest model will calculate an anomaly score for the data where is the anomaly score of the isolation forest model trained on the th time scale for the data point , is the isolation forest model trained on the th time scale, and is the time scale is the time scale of the isolation forest model for the data point .

[0057] (3) Use the weighted average method to fuse the anomaly scores of different time scales to generate a comprehensive anomaly score, improving the robustness of anomaly data point detection.

[0058] (4) Dynamically adjust the threshold to adapt to the real-time fluctuations of multi-parameter monitoring data, further reducing the false alarm and miss rate. The threshold can be dynamically set based on the historical statistical characteristics of the anomaly scores. For example, the mean and standard deviation of recent anomaly scores can be used to define the threshold.

[0059] (5) Mark and output the anomaly data points exceeding the threshold for use by the subsequent health status assessment module.

[0060] By generating anomaly scores at different time scales and fusing the detection results with dynamically adjusted thresholds, the multi-scale isolation forest can better adapt to the complexity in real-time monitoring of battery health status.

[0061] At the same time, it is necessary to ensure the temporal synchronization of data from different sensors. Different frequency data can be aligned to a unified time point through interpolation methods (such as linear interpolation) for subsequent analysis.

[0062] S102: Establish a battery physical model, including an internal resistance model, a capacity attenuation model, and an open circuit voltage model.

[0063] Internal resistance model: The internal resistance of the battery increases with factors such as the usage time, the number of charge-discharge cycles, and the temperature. The internal resistance model can be expressed by the following formula:

[0064]

[0065] where is the initial internal resistance of the battery, is the change in internal resistance over time, which can be modeled based on factors such as the number of charge-discharge cycles and the temperature.

[0066] Capacity attenuation model: The capacity attenuation of the battery is mainly affected by the number of charge-discharge cycles and the usage environment (especially the temperature). The capacity attenuation model can be expressed by the following formula:

[0067]

[0068] wherein, is the capacity of the battery at time t; is the initial capacity of the battery; is the capacity attenuation coefficient, which depends on the type of the battery, etc.; is the number of charge and discharge cycles of the battery.

[0069] Open-circuit voltage model: The open-circuit voltage is another important indicator for evaluating the battery health, and is usually proportional to the remaining capacity of the battery. According to the chemical reaction of the battery and the characteristics of the electrolyte, the open-circuit voltage can be expressed by the following relationship:

[0070]

[0071] wherein, is the open-circuit voltage of the battery, is the capacity of the battery, is the temperature of the battery, represents a mathematical function that depends on the battery capacity and temperature, and is used to reflect the comprehensive influence of the battery capacity and temperature on the open-circuit voltage, and is used to predict the open-circuit voltage of the battery.

[0072] S103: Calculate through the above physical model, and obtain the first evaluation value of the battery health state after weighted fusion of the calculation results of the internal resistance model, the capacity attenuation model and the open-circuit voltage model. It is expressed by the following formula:

[0073]

[0074] wherein, is the health state score calculated by the physical model, that is, the first evaluation value; , , are the weight coefficients, which are set according to the actual situation; is the current internal resistance; is the initial internal resistance; is the current capacity; is the initial capacity; is the current temperature; is the maximum safe operating temperature.

[0075] The battery physical model comprehensively considers the influence of factors such as internal resistance, capacity, and temperature, evaluates the health state of the battery, and through multi-factor modeling, can more accurately reflect the true state of the battery, avoid the deviation caused by a single parameter, especially in a complex usage environment, and provides a framework and basis for the preliminary evaluation.

[0076] S2: Obtain the multi-parameter monitoring data and environmental data of the mobile energy storage power supply, perform feature extraction and feature selection on the multi-parameter monitoring data and environmental data, and use the principal component analysis method to fuse the selected features to obtain a principal component matrix; input the principal component matrix into a pre-trained health status evaluation model for further prediction to obtain a second evaluation value of the battery health status.

[0077] S201: Obtain the preprocessed multi-parameter monitoring data in S1, and collect external factors such as temperature, humidity, and air pressure in the environment where the mobile energy storage power supply is located. When collecting environmental data, each sensor can be placed near the battery shell or outside the device to ensure that it can reflect the changes in the surrounding environment.

[0078] S202: Perform feature extraction and enhancement on the preprocessed multi-parameter monitoring data and environmental data to obtain enhanced features.

[0079] In this embodiment, the feature extraction specifically includes:

[0080] Extract the statistical features of voltage, current, temperature, and internal resistance in the multi-parameter monitoring data. The statistical features include mean, variance, maximum value, minimum value, peak value, skewness, etc.; extract the time decay features of the multi-parameter monitoring data, that is, each time series data, and use weighted average decay feature extraction, which is specifically expressed as:

[0081]

[0082] Among them, is the observed value at the most recent n moments, is the corresponding weight, usually using an exponential decay weight to give higher weight to the most recent data, is the decay factor (selected as 0.9 in this embodiment).

[0083] Extract the frequency domain features of the multi-parameter monitoring data, that is, each time series data. Use the fast Fourier transform FFT to convert the time series data into the frequency domain, extract the main frequency components, and extract the main frequency and the corresponding amplitude from the fast Fourier result.

[0084] Specifically expressed as:

[0085]

[0086] Among them, represents the amplitude and phase of the signal at frequency , represents the nth sample value of the time domain signal, N represents the number of data points, represents the frequency, j is the imaginary unit, representing complex number operations; It is a complex exponential function, representing the rotation of the phase in the frequency domain.

[0087] Extract the frequency and stability characteristics of the charge-discharge cycles. The cycle frequency is specifically expressed as:

[0088]

[0089] Where is the number of charge-discharge cycles per unit time, is the length of the time window.

[0090] The cycle stability is specifically expressed as:

[0091]

[0092] Where is the maximum value of the number of charge-discharge cycles, is the minimum value of the number of charge-discharge cycles, is the average value of the number of charge-discharge cycles.

[0093] Extract the environmental factor characteristics of the environmental data, including the temperature fluctuation range and the humidity influence characteristics. The temperature fluctuation range is specifically expressed as:

[0094] .

[0095] The humidity influence characteristics can be obtained through a regression model and are specifically expressed as:

[0096]

[0097] Where is the degree of influence of humidity on the battery health state; is the intercept of the regression model, representing the basic influence of humidity when the humidity H and temperature T are 0; is the regression coefficient of humidity, representing the change in the influence of humidity on the health state when the humidity H increases by one unit; H is the value of the current environmental humidity; is the regression coefficient of temperature, representing the change in the influence of temperature on the health state when the temperature T increases by one unit; T is the value of the current environmental temperature.

[0098] The present invention designs a flexible and dynamic feature selection mechanism, combining the filtering method, recursive feature elimination RFE, and dynamic feature selection, to ensure the effectiveness and accuracy of the model in a changing data environment. The specific steps are as follows:

[0099] (1) Use the filtering method for preliminary feature screening to select highly relevant features. Remove irrelevant or low-variance features through statistical methods (such as correlation analysis, variance threshold, chi-square test, etc.) to reduce the number of features.

[0100] (2) Further refine the selected features by using recursive feature elimination, gradually removing features with low importance.

[0101] 1. Select a linear regression model as the basic model and train the regression model on the filtered feature set; 2. Evaluate the importance of features, and the coefficients or importance scores of the model can be used; 3. Remove the feature with the lowest importance, and then retrain the model; Repeat steps 2 and 3 until the required number of features is reached. For the linear regression model, the feature importance is represented by the regression coefficient.

[0102] (3) Perform dynamic feature selection in the real-time or periodic changes of the mobile energy storage power supply data, and adjust the feature set according to the changes in the data distribution to maintain the adaptability and accuracy of the health status assessment model.

[0103] Regularly monitor the distribution of the monitoring data and environmental data of the mobile energy storage power supply (such as changes in the mean and variance). If the changes exceed a certain threshold, trigger dynamic feature selection; Reapply the filtering method and recursive feature elimination when the data changes significantly to generate an updated feature set; Automatically adjust the feature selection frequency and selection method according to the data changes. For example, the feature selection frequency can be increased when the real-time data changes greatly.

[0104] The present invention combines the filtering method, recursive feature elimination, and dynamic feature selection, and can update the feature set in real time or periodically according to data changes, so as to maintain the robustness of the model in a dynamic environment and avoid the decline of model performance caused by data distribution changes; This feature selection mechanism can reduce the number of features on the premise of ensuring the model effect, thereby reducing the computational cost; The above improvements help to improve the applicability and accuracy of the health status assessment model in the health status assessment of the mobile energy storage system and adapt to complex data dynamic changes.

[0105] Use the principal component analysis method PCA to fuse and enhance the selected features, which can effectively fuse the selected features into principal components to improve the prediction effect of the health status assessment model. The specific steps are as follows:

[0106] 1) Before performing principal component analysis, it is first necessary to standardize the data to ensure that each feature is on the same scale.

[0107] 2) Calculate the covariance matrix of the standardized feature matrix to evaluate the linear relationship between features. It is expressed as:

[0108]

[0109] Among them, is the covariance matrix, n is the number of samples, is the standardized feature matrix.

[0110] 3) Perform eigenvalue decomposition on the covariance matrix to find the eigenvalues and eigenvectors. It is expressed as:

[0111]

[0112] where, represents the eigenvector, is the eigenvalue, representing the variance of the corresponding eigenvector.

[0113] 4) According to the magnitudes of the eigenvalues, select the top k eigenvectors with the largest eigenvalues. Usually, the number of principal components to be selected is determined by the cumulative variance ratio (Cumulative Explained Variance).

[0114] 5) Use the selected eigenvectors to transform the standardized feature data into the new principal component space to obtain the principal component matrix. It is expressed as:

[0115]

[0116] where, Z is the principal component matrix, is the matrix containing the top k eigenvectors.

[0117] S203: Input the principal component matrix into a pre-trained health state evaluation model for prediction, and output the second evaluation value of the battery health state.

[0118] In this embodiment, the health state evaluation model adopts an improved long short-term memory network LSTM model structure. The health state of the battery is affected by long-term use and historical data, and LSTM can effectively capture the long-term battery health change trend.

[0119] The improved long short-term memory network LSTM model includes an input layer, a fully connected layer (MLP), multiple LSTM layers, and an output layer connected in sequence. In the battery health state evaluation, it can effectively capture the long-term dependencies in the time series data. The input layer receives the principal component matrix; the fully connected layer (using a multi-layer perceptron MLP) is used to perform non-linear mapping on the principal components to extract potential features; multiple LSTM layers are connected in series to capture the time dependence of the battery health data. The multi-layer LSTM can extract complex features in the battery health state more deeply, improving the fitting ability and generalization ability of the model, and is particularly suitable for data with high dimensions and complex non-linear relationships; the output layer outputs the battery health state evaluation value, that is, the second evaluation value.

[0120] Residual connections are added between each LSTM layer in multiple LSTM layers to avoid the difficulty of gradient flow caused by overly deep networks; through residual connections, the health state assessment model can be trained more easily because the gradient can be transmitted more effectively, which can mitigate the overfitting phenomenon that appears in deep networks and enhance the generalization ability of the model. At the same time, Dropout layers are added after the LSTM layer and the fully connected layer to prevent the model from overly relying on specific nodes during training; L2 regularization is added between the LSTM layer and the fully connected layer, which can penalize large weights during training, thereby preventing the model from overfitting to the training data.

[0121] Through the input of the principal component matrix, the dimension of the input data is effectively reduced, while the most important features are retained, reducing noise and redundancy. First, a multi-layer perceptron (MLP) is used for feature mapping, and the data after dimensionality reduction is further mapped to a new feature space through the MLP so that the model can be more easily understood and learned; a multi-layer LSTM network is used to capture the time dependence, local features, and changes over a long time span in the battery health data, and finally the evaluation value of the battery health state is output, so as to be able to more accurately evaluate the health state of the battery, provide more interpretable prediction results, and enhance the robustness and reliability of the system. Therefore, in addition to the features extracted from multi-parameter monitoring data, the health state assessment model also combines the features extracted from environmental data (external features) as input features to capture the changes of the battery under different environments. It can further improve the model's understanding of the battery health state, help the model more accurately predict the battery health state in complex environments, and is applicable to the health state assessment affected by multiple factors.

[0122] S3: Compare the first evaluation value with the second evaluation value, and use the health state assessment model to adjust the parameters of the battery physical model to obtain an optimized battery physical model.

[0123] Use the health state assessment model to correct the parameters in the battery physical model through the loss function and optimization algorithm, and iteratively update the model to improve the overall evaluation accuracy.

[0124] S301: Define the loss function: In the battery health state assessment, the loss function is used to measure the difference between the output of the battery physical model and the prediction result of the health state assessment model.

[0125] Assume is the output of the battery physical model for the i-th sample, is the predicted output of the health state assessment model for the i-th sample, then the loss function is expressed as:

[0126]

[0127] where N is the number of samples, They are the parameters of the battery physical model that need to be optimized.

[0128] S302: Use the gradient descent method to minimize the loss function to correct the parameters in the battery physical model. Calculate the gradient of the loss function with respect to each parameter. According to the calculated gradient, use the gradient descent method to update the parameters in the battery physical model. Multiple iterations are required until the loss function converges.

[0129] S4: Conduct a re-evaluation based on the optimized battery physical model to obtain the third evaluation value of the battery health state; perform a weighted average of the third evaluation value and the second evaluation value to obtain the final battery health state evaluation result.

[0130] Combine the re-evaluation result of the optimized battery physical model and the prediction result of the health state evaluation model to obtain the final health state evaluation result through weighted average. It can be expressed as:

[0131]

[0132] Where, is the final health state evaluation result, is the re-evaluation result of the optimized battery physical model, is the prediction result of the health state evaluation model, is the weighting coefficient, which controls the influence weights of the physical model and the health state evaluation model.

[0133] The present invention improves the accuracy and robustness of the evaluation result by combining the physical model and the health state evaluation model. The physical model can reflect the basic physical characteristics of the battery and provides a preliminary framework, while the health state evaluation model can learn complex non-linear relationships from historical data and can adjust the parameters in the physical model according to the health state evaluation model, thus showing higher accuracy and adaptability in the real environment, and the physical model provides a transparent and interpretable basis.

[0134] Embodiment 2

[0135] This embodiment discloses an intelligent evaluation system for the health state of a mobile energy storage power supply, including:

[0136] A physical model evaluation module, which is configured to: obtain multi-parameter monitoring data of the mobile energy storage power supply and input it into the battery physical model for preliminary evaluation to obtain the first evaluation value of the battery health state;

[0137] The network model evaluation module is configured to: obtain multi-parameter monitoring data and environmental data of the mobile energy storage power supply, perform feature extraction and feature selection on the multi-parameter monitoring data and environmental data, and fuse the selected features using the principal component analysis method to obtain a principal component matrix; input the principal component matrix into a pre-trained health state evaluation model for further prediction to obtain a second evaluation value of the battery health state;

[0138] The physical model optimization module is configured to: compare the first evaluation value with the second evaluation value, and use the health state evaluation model to adjust the parameters of the battery physical model to obtain an optimized battery physical model;

[0139] The health state evaluation module is configured to: perform re-evaluation based on the optimized battery physical model to obtain a third evaluation value of the battery health state; perform weighted averaging on the third evaluation value and the second evaluation value to obtain a final battery health state evaluation result.

[0140] Embodiment III

[0141] The purpose of this embodiment is to provide a computing device, including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the program, the steps of the method in Embodiment I are implemented.

[0142] Embodiment IV

[0143] The purpose of this embodiment is to provide a computer-readable storage medium, on which a computer program is stored. When the program is executed by a processor, the steps of the method in Embodiment I are executed.

[0144] The steps involved in the devices in the above Embodiments III and IV correspond to those in Method Embodiment I. For the specific implementation manners, reference may be made to the relevant description part of Embodiment I. The term "computer-readable storage medium" should be understood to include a single medium or multiple media including one or more instruction sets; it should also be understood to include any medium that can store, encode, or carry an instruction set for execution by a processor and enable the processor to execute any method in the present invention.

[0145] Those skilled in the art should understand that the above-mentioned modules or steps of the present invention can be implemented by a general-purpose computer device. Optionally, they can be implemented by program codes executable by a computing device, so that they can be stored in a storage device and executed by the computing device, or they can be separately fabricated into individual integrated circuit modules, or multiple modules or steps among them can be fabricated into a single integrated circuit module for implementation. The present invention is not limited to any specific combination of hardware and software.

[0146] The above are only the preferred embodiments of the present invention and are not intended to limit the present invention. For those skilled in the art, the present invention may have various changes and modifications. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

[0147] Although the specific implementation manners of the present invention have been described above in conjunction with the accompanying drawings, it is not a limitation on the protection scope of the present invention. Those skilled in the art should understand that, based on the technical solutions of the present invention, various modifications or deformations that can be made by those skilled in the art without creative efforts are still within the protection scope of the present invention.

Claims

1. A method for intelligently evaluating the health status of a mobile energy storage power supply, characterized in that: include: Obtain multi-parameter monitoring data of the mobile energy storage power source and input it into the battery physical model for preliminary evaluation to obtain the first evaluation value of the battery health status; Acquire multi-parameter monitoring data and environmental data of the mobile energy storage power supply, perform feature extraction and feature selection on the multi-parameter monitoring data and environmental data, and fuse the selected features using principal component analysis to obtain a principal component matrix; input the principal component matrix into a pre-trained health status assessment model for further prediction to obtain a second assessment value of the battery health status; Comparing the first evaluation value with the second evaluation value, and adjusting the parameters of the battery physical model using the health status evaluation model to obtain an optimized battery physical model; Re-evaluate based on the optimized battery physical model to obtain a third battery health status evaluation value; perform weighted average of the third evaluation value and the second evaluation value to obtain a final battery health status evaluation result; The optimized battery physical model is as follows: Define the loss function, assuming is the output of the battery physical model for the i-th sample, is the predicted output of the health status assessment model of the i-th sample, then the loss function is expressed as: Where N is the number of samples, are the battery physical model parameters that need to be optimized; The gradient descent method is used to minimize the loss function to correct the parameters in the battery physical model. The gradient of the loss function for each parameter is calculated, and the gradient descent method is used to update the parameters in the battery physical model according to the calculated gradient. Multiple iterations are required until the loss function converges. Multi-scale isolation forest is used to identify abnormal data points of the monitoring data and remove them, specifically: Performing data cleaning and noise removal on the monitoring data, and decomposing the monitoring data into different time scales; Train the isolation forest model at different time scales to obtain the anomaly score at each time scale; The anomaly scores at different time scales are fused using a weighted average method to generate a comprehensive anomaly score; Dynamically adjust the threshold, mark and output abnormal data points that exceed the threshold, and remove the marked abnormal data points; The battery physical model includes an internal resistance model, a capacity decay model and an open circuit voltage model; the multi-parameter monitoring data of the mobile energy storage power supply is obtained and input into the battery physical model for preliminary evaluation to obtain a first evaluation value of the battery health state: the calculation results of the internal resistance model, the capacity decay model and the open circuit voltage model are weighted and integrated to obtain the first evaluation value of the battery health state, which is expressed by the following formula: in, is the first evaluation value; , , is the weight coefficient; is the current internal resistance; is the initial internal resistance; is the current capacity; is the initial capacity; is the current temperature; is the maximum safe operating temperature.

2. A method for intelligently evaluating the health status of a mobile energy storage power supply according to claim 1, characterized in that: The feature selection adopts a dynamic feature selection mechanism, specifically: The filtering method is used for preliminary feature screening to select features with high relevance; Recursive feature elimination is used to further refine the selected features and gradually remove features with low importance; Dynamic feature selection is performed in the changes of real-time multi-parameter monitoring data and environmental data, and the feature set is adjusted according to the changes in data distribution.

3. A method for intelligently evaluating the health status of a mobile energy storage power supply according to claim 1, characterized in that: The health status assessment model adopts an improved long short-term memory network LSTM model, including an input layer, a multi-layer perceptron layer, multiple LSTM layers and an output layer connected in sequence, and a residual connection is added between each LSTM layer.

4. A mobile energy storage power supply health status intelligent assessment system, characterized in that: include: A physical model evaluation module is configured to: obtain multi-parameter monitoring data of a mobile energy storage power source, and input the data into a battery physical model for preliminary evaluation to obtain a first evaluation value of the battery health status; The network model evaluation module is configured to: obtain multi-parameter monitoring data and environmental data of the mobile energy storage power supply, perform feature extraction and feature selection on the multi-parameter monitoring data and environmental data, and fuse the selected features using the principal component analysis method to obtain a principal component matrix; input the principal component matrix into a pre-trained health status evaluation model for further prediction to obtain a second evaluation value of the battery health status; A physical model optimization module, configured to: compare the first evaluation value with the second evaluation value, and adjust the parameters of the battery physical model using the health status evaluation model to obtain an optimized battery physical model; A health status assessment module is configured to: perform a reassessment based on the optimized battery physical model to obtain a third battery health status assessment value; and perform a weighted average of the third assessment value and the second assessment value to obtain a final battery health status assessment result; The optimized battery physical model is as follows: Define the loss function, assuming is the output of the battery physical model for the i-th sample, is the predicted output of the health status assessment model of the i-th sample, then the loss function is expressed as: Where N is the number of samples, are the battery physical model parameters that need to be optimized; The gradient descent method is used to minimize the loss function to correct the parameters in the battery physical model. The gradient of the loss function for each parameter is calculated, and the gradient descent method is used to update the parameters in the battery physical model according to the calculated gradient. Multiple iterations are required until the loss function converges. Multi-scale isolation forest is used to identify abnormal data points of the monitoring data and remove them, specifically: Performing data cleaning and noise removal on the monitoring data, and decomposing the monitoring data into different time scales; Train the isolation forest model at different time scales to obtain the anomaly score at each time scale; The anomaly scores at different time scales are fused using a weighted average method to generate a comprehensive anomaly score; Dynamically adjust the threshold, mark and output abnormal data points that exceed the threshold, and remove the marked abnormal data points; The battery physical model includes an internal resistance model, a capacity decay model and an open circuit voltage model; the multi-parameter monitoring data of the mobile energy storage power supply is obtained and input into the battery physical model for preliminary evaluation to obtain a first evaluation value of the battery health state: the calculation results of the internal resistance model, the capacity decay model and the open circuit voltage model are weighted and integrated to obtain the first evaluation value of the battery health state, which is expressed by the following formula: in, is the first evaluation value; , , is the weight coefficient; is the current internal resistance; is the initial internal resistance; is the current capacity; is the initial capacity; is the current temperature; is the maximum safe operating temperature.

5. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps in the method for intelligently evaluating the health status of a mobile energy storage power supply as described in any one of claims 1 to 3 are implemented.

6. A computer device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the steps in the method for intelligently evaluating the health status of a mobile energy storage power supply as described in any one of claims 1-3 are implemented.

Images