Battery health diagnosis analysis method based on big data

Abstract

The invention discloses a battery health diagnosis and analysis method based on big data. The method comprises the following steps of performing real-time monitoring and data acquisition through various sensors and data sources; performing multi-source data fusion based on feature selection of the multi-source data to obtain a key feature set; creating an adaptive deep learning model, and inputting features in the key feature set into the adaptive deep learning model to obtain a battery health state prediction result; generating a battery health score based on the battery health state prediction result and the real-time data of the battery; based on the real-time battery health score, an early warning threshold value is adjusted, a dynamic alarm is generated, and corresponding maintenance suggestions and early warning measures are provided according to different health scores; and continuously optimizing the adaptive deep learning model according to historical battery health data and a model prediction result. According to the method, the data weight is dynamically adjusted based on the data quality, the sensor precision and the influence degree of the sensor precision on health prediction, and more accurate battery health state prediction can be realized.

Description

Technical Field

[0001] The present invention relates to the technical field of battery health diagnosis, and in particular to a battery health diagnosis and analysis method based on big data. Background Art

[0002] With the widespread use of mobile devices, electric vehicles and other energy storage devices, batteries, as core components of energy storage and supply, are facing more and more health management challenges. The health status of the battery directly affects the operating efficiency and safety of the equipment. Traditional battery health management methods mostly rely on simple voltage, current and temperature monitoring, which cannot fully and accurately reflect the actual health status of the battery. With the continuous development of battery technology, the complexity of battery health management is increasing. How to accurately and real-time evaluate and predict battery health has become a technical problem that the industry needs to solve urgently.

[0003] At present, battery health management methods based on sensor monitoring and big data analysis have gradually become a research hotspot. Although there are some battery diagnosis technologies based on a single data source or simple statistical methods, since the battery health status is affected by many factors, a single data source and traditional methods are difficult to effectively handle complex battery behaviors, especially when the battery is in different use environments and different load conditions. The accuracy and real-time performance of battery health prediction face great challenges. Summary of the invention

[0004] The purpose of this section is to summarize some aspects of embodiments of the present invention and briefly introduce some preferred embodiments. Some simplifications or omissions may be made in this section and the specification abstract and the invention title of this application to avoid blurring the purpose of this section, the specification abstract and the invention title, and such simplifications or omissions cannot be used to limit the scope of the present invention.

[0005] In order to solve the above technical problems, the present invention provides the following technical solutions: a battery health diagnosis and analysis method based on big data, comprising the following steps:

[0006] Real-time monitoring and data collection are carried out through a variety of sensors and data sources, and the collected raw data are pre-processed;

[0007] Based on the feature selection of multi-source data, multi-source data fusion is performed to obtain the key feature set;

[0008] Create an adaptive deep learning model and input the key features in the key feature set into the adaptive deep learning model to obtain the battery health status prediction result;

[0009] Generate a battery health score based on the battery health status prediction results and the real-time data of the battery;

[0010] Adjust the warning threshold based on the real-time battery health score, generate dynamic alarms, and provide corresponding maintenance suggestions and warning measures according to different health scores;

[0011] Continuously optimize the adaptive deep learning model based on historical battery health data and model prediction results.

[0012] As a preferred embodiment of the battery health diagnosis and analysis method based on big data according to the present invention, wherein: the sensors include voltage sensors, current sensors, temperature sensors, charge and discharge frequency sensors, ambient humidity sensors, and external vibration sensors, the data sources include historical usage data, charge and discharge cycle times, load changes, and temperature fluctuations, and the historical data includes records of the charge and discharge cycles, load, and temperature changes of the battery.

[0013] As a preferred embodiment of the battery health diagnosis and analysis method based on big data according to the present invention, wherein: the multi-source data fusion is performed based on the feature selection of multi-source data to obtain a key feature set, including the following steps:

[0014] Classify the collected multi-source data to identify the data sources, including sensor data and historical data;

[0015] For each data source, determine its data type and data characteristics, and perform preprocessing and standardization on different types of data sources;

[0016] According to the characteristics of each data source and its importance for predicting the battery health status, perform feature selection to retain the features that have a significant impact on the battery health;

[0017] Assign initial weights to the features of each data source, and the determination basis of the weights includes the accuracy of the data source, the stability of the sensor, and the degree of influence on health prediction;

[0018] Perform data fusion through the weighted average method and the principal component analysis method, and dynamically adjust the weights during the data fusion process according to the sensor type, data accuracy, and the degree of influence on health prediction to obtain the fused key feature set.

[0019] As a preferred embodiment of the battery health diagnosis and analysis method based on big data according to the present invention, wherein: the dynamic adjustment of weights according to the sensor type, data accuracy, and the degree of influence on health prediction includes the following steps:

[0020] Determine the priority and contribution of each data source in battery health prediction by classifying and evaluating the characteristics of sensor data and historical data;

[0021] Preliminarily assign weights to each data source according to the accuracy, stability of the sensor, and the influence of the data on battery health prediction;

[0022] Monitor the quality changes of real-time monitoring data sources in real time, and dynamically adjust the weights of each data source according to sensor failures and data fluctuations;

[0023] Use machine learning algorithms to evaluate the importance of features, and dynamically adjust the weights according to their impact on health prediction, and regularly update the adjustment strategy;

[0024] Based on the dynamically adjusted weights, fuse multi-source data through the weighted average method or the principal component analysis method to improve the accuracy of health status prediction;

[0025] Through iterative learning and feedback mechanisms, continuously optimize the feature weight allocation.

[0026] As a preferred solution of the battery health diagnosis and analysis method based on big data described in the present invention, wherein: the regularly updated adjustment strategy includes the following steps:

[0027] Regularly check the quality changes of sensor data and historical data, evaluate the data fluctuation trend, and identify abnormal situations that may affect data accuracy; at the same time, analyze the time span and change rules of historical data to ensure that the data used by the model conforms to the long-term health prediction trend.

[0028] Use machine learning algorithms to regularly re-evaluate the contribution of each feature to health prediction, and dynamically adjust the weights of features according to changes in the battery operating state and environmental conditions, giving priority to features that have a greater impact on health prediction currently;

[0029] Based on the data quality monitoring and feature importance evaluation results, regularly correct the weight adjustment rules; for data sources with poor quality, reduce their weights in feature fusion, while for data sources with good quality, appropriately increase their weights;

[0030] Through a continuous feedback mechanism, collect the comparison information between the prediction results after each weight adjustment and the actual health status to evaluate the effect of the current strategy. According to the feedback results, fine-tune the weight adjustment strategy to optimize the contribution to health prediction;

[0031] Regularly retrain the adaptive deep learning model with new data and updated weights to enhance the adaptive ability of the model; at the end of each update cycle, evaluate the overall performance of the model, and further optimize the weight adjustment strategy according to the evaluation results;

[0032] Adopt an adaptive learning mechanism to enable the strategy to automatically adjust with changes in the battery operating environment. Combine new operating data and prediction results, and use artificial intelligence technology to continuously optimize the weight adjustment and strategy update process.

[0033] As a preferred solution of the battery health diagnosis and analysis method based on big data according to the present invention, wherein: the weight adjustment formula is as follows:

[0034]

[0035] w i (t) represents the weight of the i-th sensor data source at time t, and the value range is [0, 1];

[0036] α i represents the initial weight coefficient of the i-th sensor, which is set according to its basic importance in the prediction;

[0037] β i represents the data accuracy coefficient of the i-th sensor, reflecting the impact of its accuracy on health prediction. For sensors with higher accuracy, β i is smaller;

[0038] N is the total number of sensors;

[0039] t is the current time, used to represent the change of weight over time;

[0040] represents the dynamic weighting of the i-th sensor, which is an exponential decay based on the change of the data quality of this sensor over time;

[0041] represents the normalization of the weighting of all sensors, so that the sum of the weights of all sensors is 1. Through normalization, it is ensured that the contribution of each sensor to the prediction is distributed proportionally;

[0042] δ i (τ) represents the data quality change function of the i-th sensor at time τ, reflecting the quality change of the data source. When δ i (τ) is close to 1, it means that the sensor quality is stable or good. When δ i (τ) is close to 0, it means that the sensor may malfunction or have poor quality. It is calculated in real time according to the sensor failure or data quality fluctuation. The function expression of δ i (τ) is as follows:

[0043]

[0044] where, γ i is the performance degradation coefficient of the i-th sensor. The larger it is, the faster the sensor degrades; λ i reflects the change rate coefficient of the environment or the system itself, indicating the change speed of the data quality at time τ; t0 is the initial time point of the data quality, indicating the change of the sensor from the start of use to the current time.

[0045] As a preferred solution of the battery health diagnosis and analysis method based on big data according to the present invention, wherein: the creation of the adaptive deep learning model includes the following steps:

[0046] Divide the feature set after multi-source data fusion into training set, validation set and test set, which are respectively used for subsequent deep learning model training, hyperparameter tuning and performance evaluation; in addition, before data division, according to the weight adjustment formula w i (t), dynamically adjust the importance of features in the data set;

[0047] Select a suitable deep learning model architecture according to the task requirements. The model architecture includes long short-term memory network and convolutional neural network. When selecting the model architecture, combine the dynamic weighting scheme of features so that different sensor data have different influences during model training;

[0048] Design an adaptive mechanism. By introducing an adaptive optimization algorithm, the adaptive deep learning model can dynamically adjust its hyperparameters according to the changes of real-time data; after each update of the network weights of the model, the adaptive deep learning model will re-adjust the weights of each data according to the new model weight distribution, and further improve the prediction accuracy through the optimization algorithm;

[0049] Use the training set data to train the adaptive deep learning model, and after each round of training, use the validation set to verify the adaptive deep learning model, check the accuracy and generalization ability of the adaptive deep learning model on the validation set, and further adjust the hyperparameters of the adaptive deep learning model;

[0050] After training is completed, use the test set to conduct a final evaluation of the model. If the prediction result is not ideal, adjust the hyperparameters or network structure according to the feedback of the model on the validation set, or introduce more data sources according to the real-time changes of battery health data to enhance the prediction ability of the model;

[0051] During the evaluation process of the adaptive deep learning model, combine historical feedback information and real-time data, adjust the feature weights, and use the iterative learning method to fine-tune the weight adjustment strategy, so that the adaptive deep learning model can be continuously optimized in subsequent prediction tasks, and finally achieve more accurate battery health diagnosis.

[0052] As a preferred solution of the battery health diagnosis and analysis method based on big data according to the present invention, wherein: the generation of the battery health score includes the following steps:

[0053] Input the key features after preprocessing and feature selection into the adaptive deep learning model to calculate the predicted result of the battery health status, which is used as the preliminary basis for the battery health score;

[0054] Combine the real-time collected battery data with the health status prediction results;

[0055] According to the combination of the health status prediction results and the real-time data, use a preset scoring rule or model to calculate the final battery health score;

[0056] Perform normalization processing on the health score to ensure the consistency and comparability of the health score. The normalization is adjusted according to the battery type, usage environment, and different performance criteria, so that the battery health score can be comparable under different circumstances;

[0057] According to the generated battery health score, the battery health management system generates corresponding maintenance suggestions and warning measures to guide the user to take appropriate actions;

[0058] Dynamically adjust the health score according to the real-time monitoring data.

[0059] As a preferred solution of the battery health diagnosis and analysis method based on big data according to the present invention, wherein: the step of adjusting the warning threshold according to the real-time health score and generating a dynamic alarm includes the following steps:

[0060] The battery health management system continuously monitors the health status of the battery, and based on the health score calculated by the adaptive deep learning model, updates the health score of the battery in real time;

[0061] The battery health management system sets multiple warning thresholds according to the health score. The warning thresholds are defined according to the performance criteria of the battery and the range of the health score;

[0062] According to the real-time health score of the battery, the battery health management system dynamically adjusts the warning threshold according to a predetermined rule; for example, when the health score of the battery continuously decreases, the system will automatically lower the warning threshold to issue a warning before the battery health condition further deteriorates; conversely, when the health score recovers, the warning threshold will be appropriately increased;

[0063] When the health score of the battery approaches or is lower than the set warning threshold, the battery health management system automatically generates an alarm signal. The alarm signal is divided into different severity levels, such as "warning", "severe warning", or "emergency alarm", and issues a warning through text messages, APP notifications, or directly to the battery management system to remind the user or relevant system personnel to take necessary maintenance or replacement measures;

[0064] The battery health management system gradually optimizes the alarm strategy according to the historical alarm data, user feedback, and actual situations.

[0065] The present invention also provides a battery health diagnosis and analysis system based on big data, including:

[0066] Data acquisition module, which is used to obtain the real-time operation data of the battery from multiple sensors and data sources;

[0067] Data preprocessing module, which is used to preprocess the collected raw data to ensure the accuracy and consistency of the data;

[0068] Feature selection and fusion module, which is used to perform feature selection and fusion on the collected multi-source data, and extract key features closely related to the battery health status;

[0069] Adaptive deep learning module, which is used to establish a deep learning model to predict the battery health status;

[0070] Battery health score generation module, which is used to convert the prediction result of the battery health status into a health score, and dynamically adjust the health score according to the real-time data;

[0071] Dynamic warning and alarm module, which is used to dynamically adjust the warning threshold and generate real-time alarms;

[0072] Historical data feedback and optimization module, which is used to obtain the comparison between the historical battery health data and the model prediction result, and optimize the adaptive deep learning model and weight adjustment strategy of the system.

[0073] Advantages of the present invention:

[0074] 1. By fusing multi-source data, the present invention obtains a key feature set of the battery and inputs it into the adaptive deep learning model, which can accurately predict the battery health status and provide a more accurate health score. In addition, through real-time monitoring and dynamic adjustment of the warning threshold, the system can timely detect potential battery failures or health problems, ensuring the long-term operation and timely maintenance of the battery.

[0075] 2. By dynamically adjusting the weights of multi-source data based on data quality, sensor accuracy and their impact on health prediction, the present invention makes the battery health prediction result more accurate. The weight adjustment formula and feedback mechanism can automatically optimize the performance of the prediction model in different working environments and states, avoiding the dependence on fixed data sources in traditional methods, and thus better adapting to complex working environments and changing battery states.

[0076] 3. By generating dynamic alarms and real-time warning adjustments based on the battery health score, the present invention can automatically adjust the warning threshold according to the actual health status of the battery, and timely remind users or system maintenance personnel to take actions. At the same time, combining real-time data and historical feedback, the system can provide personalized maintenance suggestions and warning measures according to the battery health status, thereby optimizing the service life of the battery, reducing the risk of battery damage, and improving the reliability of the device and the user experience. Description of the Drawings

[0077] To more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings. Among them:

[0078] Figure 1 It is a flowchart of the battery health diagnosis and analysis method based on big data of the present invention. Specific embodiments

[0079] To make the above objects, features, and advantages of the present invention more obvious and understandable, the following will make a detailed description of the specific embodiments of the present invention in conjunction with the drawings of the specification.

[0080] In the following description, many specific details are set forth in order to fully understand the present invention. However, the present invention can also be implemented in other ways different from those described herein. Those skilled in the art can make similar promotions without departing from the connotation of the present invention. Therefore, the present invention is not limited by the specific embodiments disclosed below.

[0081] Secondly, the so-called "one embodiment" or "embodiment" herein refers to a specific feature, structure, or characteristic that can be included in at least one implementation manner of the present invention. The "in one embodiment" that appears in different places in this specification does not all refer to the same embodiment, nor is it a separate or alternative embodiment that mutually excludes other embodiments.

[0082] Furthermore, the present invention is described in detail in conjunction with schematic diagrams. When describing the embodiments of the present invention in detail, for the convenience of explanation, the cross-sectional views showing the device structure will be enlarged locally not in accordance with the general ratio, and the schematic diagrams are only examples, which should not limit the protection scope of the present invention herein. In addition, in actual production, three-dimensional spatial dimensions including length, width, and depth should be included.

[0083] Embodiment 1

[0084] Referring to Figure 1 , which is the first embodiment of the present invention, a battery health diagnosis and analysis method based on big data is provided, including the following steps:

[0085] S1: Real-time monitoring and data collection are carried out through a variety of sensors and data sources, and the collected raw data is preprocessed.

[0086] Specifically, the sensors include voltage sensors, current sensors, temperature sensors, charge and discharge frequency sensors, ambient humidity sensors, external vibration sensors, etc. The data sources include historical usage data, charge and discharge cycle times, load changes, and temperature fluctuations. The historical data includes the charge and discharge cycles, load, and temperature change records of the battery. The preprocessing includes operations such as denoising, standardization, and normalization to ensure the quality of the data to meet the input requirements of the deep learning model.

[0087] S2: Perform multi-source data fusion based on the feature selection of multi-source data (information from different sensors and historical data sources) to obtain a key feature set.

[0088] Specifically, performing multi-source data fusion based on the feature selection of multi-source data to obtain a key feature set includes the following steps:

[0089] Classify the collected multi-source data to identify the data sources, including sensor data (such as voltage, current, temperature, humidity, vibration, etc.) and historical data (such as charge and discharge cycle times, load changes, temperature fluctuations, etc.).

[0090] For each data source, determine its data type (numeric, time series, etc.) and data characteristics (for example, whether the current data is an instantaneous value or an average value, whether the temperature data has periodic changes, etc.), and perform preprocessing and standardization on different types of data sources.

[0091] According to the characteristics of each data source and its importance for predicting the battery health status, perform feature selection to retain the features that have a significant impact on the battery health.

[0092] Assign initial weights to the features of each data source. The determination of the weights is based on factors including the accuracy of the data source, the stability of the sensor, and the degree of influence on health prediction;

[0093] Perform data fusion through the weighted average method and the principal component analysis method, and during the data fusion process, dynamically adjust the weights according to the sensor type, data accuracy, and the degree of influence on health prediction to obtain the fused key feature set.

[0094] Furthermore, dynamically adjusting the weights according to the sensor type, data accuracy, and the degree of influence on health prediction includes the following steps:

[0095] By classifying and evaluating the characteristics of sensor data and historical data, determine the priority and contribution of each data source in battery health prediction.

[0096] According to the accuracy, stability of the sensor, and the influence of the data on battery health prediction, initially assign weights to each data source. The factors considered for the initial weights include:

[0097] Accuracy of the sensor: Sensors with higher accuracy are given higher weights;

[0098] Stability of the sensor: Sensors with higher stability are given higher weights;

[0099] Direct impact of data on battery health prediction: Data sources with greater impact are given higher weights (for example, current and voltage usually have a greater impact on battery health).

[0100] Monitor the quality changes of real-time monitoring data sources in real time, and dynamically adjust the weights of each data source according to sensor failures and data fluctuations. For example, when a sensor has a short-term failure, an increase in data fluctuations, or a decrease in quality, the weight of the corresponding data source will automatically decrease. For periodic changes in historical data (such as load, temperature changes, etc.), weight adjustments are made according to the time span and change pattern of the data. For example, during periods with large load changes or drastic temperature changes, higher weights can be assigned to the relevant data to capture rapid changes in the battery health state.

[0101] Use machine learning algorithms (such as decision trees, random forests, etc.) to evaluate the importance of features and dynamically adjust the weights according to their impact on health prediction. For example, for features with a greater impact (such as the impact of temperature on battery life), increase their weights; for features with a smaller impact (such as the impact of external vibration on the health state is small), reduce their weights accordingly. Regularly update the adjustment strategy, and adopt a dynamic learning mechanism to monitor the changes of different features in real time and adjust their contributions to health prediction. For example, if the prediction accuracy of the number of charge and discharge cycles of the battery improves, the weight of the charge and discharge data can be gradually increased.

[0102] Based on the dynamically adjusted weights, fuse multi-source data through the weighted average method or the principal component analysis method to improve the accuracy of health state prediction.

[0103] Through iterative learning and feedback mechanisms, continuously optimize the feature weight allocation. That is, according to the impact degree of each data source and its features on health prediction, perform real-time feedback and adjustment. Through continuous iterative training mechanisms, continuously optimize the weight allocation of data sources to ensure the accuracy of the model at different battery operation stages. If a certain feature or data source performs poorly in training, its weight is reduced through an adaptive algorithm until its stability is improved.

[0104] In the above steps, by dynamically adjusting the weights by combining the type of sensor, data accuracy, and the degree of impact of data on health prediction, the accuracy of multi-source data fusion can be improved, ensuring that the weights of each data source under different conditions can reflect their actual contributions to health diagnosis. In addition, real-time feedback on the battery health state is introduced to adjust the weights in real time, ensuring that the fusion of data sources and features is more accurate, and further improving the adaptability and prediction accuracy of the model.

[0105] Furthermore, the steps for regularly updating and adjusting the strategy include the following:

[0106] Regularly check the quality changes of sensor data and historical data, evaluate the data fluctuation trend, and identify abnormal conditions that may affect data accuracy (such as sensor aging, faults, or environmental interference); at the same time, analyze the time span and change rules of historical data to ensure that the data used by the model conforms to the long-term health prediction trend.

[0107] Regularly use machine learning algorithms (such as random forest, decision tree, etc.) to re-evaluate the contribution of each feature to health prediction, and identify which features have a greater impact on health prediction at the current stage. Dynamically adjust the weights of features according to the changes in battery operating status and environmental conditions, and give priority to the features that have a greater impact on health prediction currently.

[0108] Based on the results of data quality monitoring and feature importance assessment, regularly correct the weight adjustment rules; for data sources with poor quality, reduce their weights in feature fusion, while for data sources with good quality, appropriately increase their weights. When updating the weight adjustment strategy, consider the importance of different battery health states (such as charging, discharging, or standby states) for different data sources to ensure accurate prediction of battery health under different usage scenarios.

[0109] Through a continuous feedback mechanism, collect the comparison information between the prediction results and the actual health status after each weight adjustment to evaluate the effectiveness of the current strategy. According to the feedback results, fine-tune the weight adjustment strategy to optimize the contribution to health prediction and ensure that the system always maintains a high prediction accuracy.

[0110] Regularly retrain the adaptive deep learning model with new data and updated weights to enhance the adaptive ability of the model; at the end of each update cycle, evaluate the overall performance of the model and further optimize the weight adjustment strategy according to the evaluation results.

[0111] Adopt an adaptive learning mechanism so that the strategy can automatically adjust with the changes in the battery operating environment, combine new operating data and prediction results, and continuously optimize the weight adjustment and strategy update process using artificial intelligence technology.

[0112] The above steps can ensure the regular update of the weight adjustment strategy, enabling the battery health prediction system to be dynamically optimized according to actual changes and improving the prediction accuracy.

[0113] Furthermore, the weight adjustment formula is as follows:

[0114]

[0115] w i(t) represents the weight of the i-th sensor data source at time t, and the value range is [0,1];

[0116] α i represents the initial weight coefficient of the i-th sensor, which is set according to its basic importance in prediction;

[0117] β i represents the data accuracy coefficient of the i-th sensor, reflecting the impact of its accuracy on health prediction. For sensors with higher accuracy, β i is smaller;

[0118] N is the total number of sensors;

[0119] t is the current time, used to represent the change of weight over time;

[0120] represents the dynamic weighting of the i-th sensor, which is an exponential decay based on the change of the data quality of this sensor over time;

[0121] represents the normalization of the weighting of all sensors, making the sum of the weights of all sensors equal to 1. Through normalization, it is ensured that the contribution of each sensor to the prediction is allocated proportionally;

[0122] δ i (τ) represents the data quality change function of the i-th sensor at time τ, reflecting the quality change of the data source. When δ i (τ) is close to 1, it indicates that the sensor quality is stable or good. When δ i (τ) is close to 0, it indicates that the sensor may malfunction or have poor quality. It is calculated in real time according to the sensor failure or data quality fluctuation. The function expression of δ i (τ) is as follows:

[0123]

[0124] Among them, γ i is the performance decay coefficient of the i-th sensor. The larger it is, the faster the sensor decays; λ i reflects the change rate coefficient of the environment or the system itself, indicating the change speed of the data quality at time τ; t0 is the initial time point of the data quality, indicating the change of the sensor from the start of use to the current time.

[0125] In the above formula, the impact of the sensor quality changing over time is reflected by the exponential decay function, which can dynamically adjust the weights. The formula avoids simple weighted averaging and instead comprehensively considers multiple factors through complex functions (such as exponential decay, integration, etc.), thereby enhancing the flexibility and prediction accuracy of the model. This formula combines time series data and an adaptive adjustment mechanism to ensure the accuracy and real-time nature of the battery health diagnosis method.

[0126] S3: Create an adaptive deep learning model and input the key features in the key feature set into the adaptive deep learning model to obtain the battery health state prediction result.

[0127] Specifically, creating an adaptive deep learning model includes the following steps:

[0128] Partition the feature set after multi-source data fusion into a training set, a validation set, and a test set. Among them, the training set is used to train the deep learning model, the validation set is used to adjust the hyperparameters of the model, which helps to optimize the performance of the model and avoid overfitting, and the test set is used to evaluate the generalization ability and performance of the model. The independence of the test set ensures that the final evaluation result of the model will not be affected by data leakage during the training process. The training set, validation set, and test set are respectively used for subsequent deep learning model training, hyperparameter adjustment, and performance evaluation; in addition, before data partitioning, according to the weight adjustment formula w i (t), dynamically adjust the importance of features in the data set.

[0129] Select a suitable deep learning model architecture according to the task requirements. The model architecture includes long short-term memory networks (LSTM) and convolutional neural networks (CNN). When selecting the model architecture, combine the dynamic weighting scheme of features so that different sensor data have different influences during the model training process. For time series data, LSTM can be used to capture the time dependence of the data; for feature extraction, CNN can efficiently extract local patterns in the data; determine hyperparameters such as the number of network layers, the number of nodes in each layer, activation functions (such as ReLU, Sigmoid, etc.) to adapt to the complexity of the battery health state prediction task.

[0130] Design an adaptive mechanism. By introducing an adaptive optimization algorithm (such as Adam or RMSprop), the adaptive deep learning model can dynamically adjust its hyperparameters according to the changes in real-time data; after each update of the network weights of the model, the adaptive deep learning model will re-adjust the weights of each data according to the new model weight distribution and further improve the prediction accuracy through the optimization algorithm.

[0131] Among them, the hyperparameters include:

[0132] Learning rate: Determines the step size when updating the model weights each time, affecting the convergence speed and effect of training;

[0133] Number of network layers: Determines the depth of the neural network;

[0134] Number of neurons in each layer: The number of neurons in each layer, affecting the expressive power of the model.

[0135] Batch size: Refers to the number of samples input into the model at one time, affecting the stability and efficiency of the training process;

[0136] Regularization coefficient: Used to control the complexity of the model and prevent overfitting.

[0137] Use the training set data to train the adaptive deep learning model, calculate the loss of the model using the backpropagation algorithm, and update the weights of the adaptive deep learning model according to the gradient descent method (i.e., adjust the parameters of the connections in each layer of the neural network). An early stopping strategy can be set to avoid overfitting. After each round of training, use the validation set to verify the adaptive deep learning model and check the accuracy and generalization ability of the adaptive deep learning model on the validation set, and further adjust the hyperparameters of the adaptive deep learning model.

[0138] After training is completed, use the test set to conduct a final evaluation of the model, calculate indicators such as the accuracy, precision, and recall rate of battery health prediction to verify the effectiveness of the model. If the prediction results are not satisfactory, adjust the hyperparameters or network structure according to the feedback of the model on the validation set, or introduce more data sources according to the real-time changes of battery health data to enhance the prediction ability of the model.

[0139] During the evaluation process of the adaptive deep learning model, combine historical feedback information with real-time data, adjust the feature weights, and use the iterative learning method to fine-tune the weight adjustment strategy, so that the adaptive deep learning model can be continuously optimized in subsequent prediction tasks, and finally achieve more accurate battery health diagnosis.

[0140] In the above steps, by integrating dynamic weight adjustment into the training process of the adaptive deep learning model, the adaptive deep learning model can not only be adjusted in real time based on the current sensor data quality, but also continuously optimize the prediction results. This interaction mechanism ensures that the adaptive deep learning model can automatically adjust its model parameters according to the changes in battery health data and the fluctuations in sensor quality, improving the overall diagnostic accuracy and the adaptive ability of the system.

[0141] S4: Generate a battery health score based on the battery health status prediction result and the real-time data of the battery.

[0142] Specifically, generating a battery health score includes the following steps:

[0143] Input the key features after preprocessing and feature selection into an adaptive deep learning model (such as LSTM or CNN) to calculate the prediction result of the battery's state of health, which serves as the preliminary basis for the battery health score.

[0144] Combine the battery data (such as voltage, temperature, charge / discharge rate, etc.) collected in real time with the prediction result of the state of health. By fusing the prediction result with the real-time data, the current health condition of the battery can be reflected more comprehensively and accurately.

[0145] Based on the combination of the prediction result of the state of health and the real-time data, use a preset scoring rule or model to calculate the final battery health score. This score can be a numerical index, usually expressed as a score between 0 and 100. The higher the score, the better the battery health condition, and the lower the score, the worse the battery health condition.

[0146] Perform standardization processing on the health score to ensure the consistency and comparability of the health score. The standardization is adjusted according to the battery type, usage environment, and different performance criteria, so that the battery health score can be comparable under different circumstances.

[0147] According to the generated battery health score, the battery health management system generates corresponding maintenance suggestions and warning measures to guide the user to take appropriate actions.

[0148] Dynamically adjust the health score according to the real-time monitoring data. For example, when the performance of the battery significantly deteriorates, the health score will decrease accordingly, and vice versa.

[0149] The above steps combine the prediction result of the deep learning model with the fusion of real-time data to obtain the battery health score.

[0150] S5: Based on the real-time battery health score, adjust the warning threshold, generate a dynamic alarm, and provide corresponding maintenance suggestions and warning measures according to different health scores.

[0151] Specifically, adjusting the warning threshold according to the real-time health score and generating a dynamic alarm includes the following steps:

[0152] Continuously monitor the health condition of the battery, and based on the health score calculated by the adaptive deep learning model, update the health score of the battery in real time.

[0153] The battery health management system sets multiple warning thresholds according to the health score. The warning thresholds are defined according to the performance criteria of the battery and the range of the health score. For example, it can be divided into "good", "medium", "bad" and other states according to the high and low scores, and different warning levels are set for each state.

[0154] Based on the real-time battery health score, the battery health management system dynamically adjusts the warning threshold according to predetermined rules; for example, when the health score of the battery continuously decreases, the system will automatically lower the warning threshold to issue a warning before the battery health deteriorates further; conversely, when the health score recovers, the warning threshold will be appropriately increased.

[0155] When the health score of the battery approaches or is lower than the set warning threshold, the battery health management system automatically generates an alarm signal. The alarm signal is divided into different severity levels, such as "warning", "severe warning", or "emergency alert", and warns the user or relevant system personnel by sending text messages, APP notifications, or directly to the battery management system, reminding them to take necessary maintenance or replacement measures; the notification content should include the current health score of the battery, the predicted remaining service life, the specific warning level, the recommended action plan (such as checking immediately, performing repairs, or replacing the battery), and other relevant information.

[0156] The battery health management system gradually optimizes the alarm strategy based on historical alarm data, user feedback, and actual situations. For example, the accuracy of the warning can be improved by further training the model, or the alarm threshold settings can be adjusted to make the alarm more intelligent and accurate.

[0157] Adjusting the warning threshold according to the real-time health score and generating corresponding dynamic alarms in the above steps can ensure that the battery is timely warned in different health states, thus avoiding the battery from malfunctioning or deteriorating to an unusable state.

[0158] S6: Continuously optimize the adaptive deep learning model based on historical battery health data and model prediction results. That is, the battery health management system continuously optimizes the adaptive deep learning model according to historical battery health data and model prediction results, improves its prediction ability for the battery health status. When the health state of the battery changes significantly, the battery health management system updates the new data into the model through a feedback mechanism so that the model can be further adjusted and optimized.

[0159] In summary, the present invention fuses data from multiple sensors and data sources to obtain a key feature set of the battery, and uses an adaptive deep learning model to accurately predict the health state of the battery. By dynamically adjusting the weights of sensor data based on data quality, sensor accuracy, and their impact on health prediction, the present invention can achieve more accurate prediction of the battery health state. At the same time, the system monitors the battery health score in real time and dynamically adjusts the warning threshold, effectively avoiding the occurrence of battery failures and health problems. In addition, based on the real-time data and historical data of the battery health state, the present invention can also provide users with personalized maintenance suggestions and warning measures to ensure the stability and long-term effectiveness of the battery in different usage environments, and improve the reliability and service life of the device.

[0160] Embodiment 2, which is an embodiment of the present invention, provides a battery health diagnosis and analysis system based on big data, including: a data acquisition module for obtaining real-time operation data of the battery from multiple sensors and data sources; a data preprocessing module for preprocessing the collected raw data to ensure the accuracy and consistency of the data; a feature selection and fusion module for performing feature selection and fusion on the collected multi-source data to extract key features closely related to the battery health status; an adaptive deep learning module for establishing a deep learning model to predict the battery health status; a battery health score generation module for converting the prediction result of the battery health status into a health score and dynamically adjusting the health score according to real-time data; a dynamic warning and alarm module for dynamically adjusting the warning threshold and generating real-time alarms; a historical data feedback and optimization module for obtaining the comparison between historical battery health data and model prediction results to optimize the adaptive deep learning model and weight adjustment strategy of the system.

[0161] Embodiment 3, which is an embodiment of the present invention, provides a battery health diagnosis and analysis method based on big data. In order to verify the beneficial effects of the present invention, scientific demonstration is carried out through simulation experiments. The main objective of the experiment is to verify the prediction accuracy, real-time performance, and accuracy of battery health based on multi-source data fusion, an adaptive deep learning model, and dynamic warning threshold adjustment.

[0162] In this embodiment, the data acquisition comes from multiple sensors, including battery voltage, current, temperature, charge and discharge cycle times, etc. A big data platform is used for data storage and real-time analysis. In order to cover different usage conditions, the experiment includes battery data under different working environments (such as high temperature, low temperature, different loads).

[0163] In this embodiment, the method of the present invention is used, including multi-source data fusion, an adaptive deep learning model, and dynamic warning threshold adjustment. A traditional health assessment method using a single data source (such as battery voltage, current) without a dynamic warning mechanism is used as a control group. The duration of each group of experiments is two weeks. During the experiment, battery health data is collected regularly to monitor and predict the battery health status in real time.

[0164] Table 1 Battery Health Diagnosis and Analysis Result Table

[0165]

[0166] It can be seen from the summary table that the average health score of the experimental group is significantly higher than that of the control group, indicating that the adaptive deep learning model of the present invention can more accurately evaluate the health status of the battery. The prediction accuracy of the health status of the experimental group is 98%, which is much higher than 82% of the control group. This shows that the present invention can effectively utilize multi-source data and improve the accuracy of the prediction model. The dynamic warning response time of the experimental group is shorter, only 5.2 seconds, while that of the control group is 15.7 seconds. Real-time dynamic adjustment of the warning threshold and timely response are a major advantage of the present invention. The experimental group can detect potential faults 12 hours in advance, while the early warning time of the control group is only 4 hours. The dynamically adjusted warning threshold helps the system identify battery health problems earlier and reduce the occurrence of potential faults. The accuracy of the maintenance suggestions of the experimental group is 95%, showing the effectiveness of personalized suggestions and warning measures. In contrast, the accuracy of the maintenance suggestions of the control group is only 70%.

[0167] From the above experimental results, it can be seen that the big data-based battery health diagnosis and analysis method of the present invention can significantly improve the prediction accuracy and real-time performance of the battery health status. Compared with the traditional single data source method, the present invention not only improves the prediction accuracy through multi-source data fusion, adaptive deep learning model and dynamic warning threshold adjustment, but also can respond in real time and give early warnings, significantly improving the service life of the battery and the reliability of the equipment. In addition, the system can provide personalized maintenance suggestions according to the battery health status, further optimizing the battery management efficiency.

[0168] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical solutions of the present invention, and they should all be covered within the scope of the claims of the present invention.

Claims

1. A battery health diagnosis and analysis method based on big data, characterized in that, Including the following steps: Conduct real-time monitoring and data collection through a variety of sensors and data sources, and preprocess the collected raw data; Perform multi-source data fusion based on feature selection of multi-source data to obtain a key feature set; Create an adaptive deep learning model, and input the key features in the key feature set into the adaptive deep learning model to obtain the battery health status prediction result; Generate a battery health score based on the battery health status prediction result and the real-time data of the battery; Adjust the warning threshold based on the real-time battery health score to generate a dynamic alarm, and provide corresponding maintenance suggestions and warning measures according to different health scores; Continuously optimize the adaptive deep learning model according to the historical battery health data and the model prediction result.

2. The battery health diagnosis and analysis method based on big data according to claim 1, wherein: The sensors include a voltage sensor, a current sensor, a temperature sensor, a charge and discharge frequency sensor, an environmental humidity sensor, and an external vibration sensor. The data sources include historical usage data, the number of charge and discharge cycles, load changes, and temperature fluctuations. The historical data includes records of the battery's charge and discharge cycles, load, and temperature changes.

3. The battery health diagnosis and analysis method based on big data according to claim 1, wherein: The step of performing multi-source data fusion based on feature selection of multi-source data to obtain a key feature set includes the following steps: Classify the collected multi-source data to identify the data sources, including sensor data and historical data; For each data source, determine its data type and data characteristics, and preprocess and standardize different types of data sources; According to the characteristics of each data source and its importance for battery health status prediction, perform feature selection to retain features that have a significant impact on battery health; Assign initial weights to the features of each data source. The determination basis of the weights includes the accuracy of the data source, the stability of the sensor, and the degree of influence on health prediction; Perform data fusion through the weighted average method and the principal component analysis method, and dynamically adjust the weights during the data fusion process according to the sensor type, data accuracy, and the degree of influence on health prediction to obtain the fused key feature set.

4. The battery health diagnosis and analysis method based on big data according to claim 3, wherein: The step of dynamically adjusting the weights according to the sensor type, data accuracy, and the degree of influence on health prediction includes the following steps: Determine the priority and contribution of each data source in battery health prediction by classifying and evaluating the characteristics of sensor data and historical data; Preliminarily assign weights to each data source according to the accuracy, stability of the sensor, and the influence of the data on battery health prediction; Real-time monitor the quality change of the data source, and dynamically adjust the weights of each data source according to sensor failures and data fluctuations; Use machine learning algorithms to evaluate the importance of features, and dynamically adjust the weights according to their influence on health prediction, and regularly update the adjustment strategy; Based on the dynamically adjusted weights, fuse the multi-source data through the weighted average method or the principal component analysis method to improve the accuracy of health status prediction; Continuously optimize the feature weight assignment through iterative learning and feedback mechanisms.

5. The battery health diagnosis and analysis method based on big data according to claim 4, wherein: The step of regularly updating the adjustment strategy includes the following steps: Regularly check the quality changes of sensor data and historical data, evaluate the data fluctuation trend, and identify abnormal situations that may affect data accuracy; at the same time, analyze the time span and change rules of historical data to ensure that the data used by the model conforms to the long-term health prediction trend. Use machine learning algorithms to regularly re-evaluate the contribution of each feature to health prediction, and dynamically adjust the weights of features according to changes in battery operating status and environmental conditions, giving priority to features that have a greater impact on health prediction currently; Based on the results of data quality monitoring and feature importance assessment, regularly correct the weight adjustment rules; for data sources with poor quality, reduce their weights in feature fusion, while for data sources with good quality, appropriately increase their weights; Through a continuous feedback mechanism, collect the comparison information between the prediction results and the actual health status after each weight adjustment to evaluate the effectiveness of the current strategy. According to the feedback results, fine-tune the weight adjustment strategy to optimize the contribution to health prediction; Regularly retrain the adaptive deep learning model with new data and updated weights to enhance the adaptive ability of the model; at the end of each update cycle, evaluate the overall performance of the model, and further optimize the weight adjustment strategy according to the evaluation results; Adopt an adaptive learning mechanism so that the strategy can automatically adjust with changes in the battery operating environment. Combine new operating data and prediction results, and use artificial intelligence technology to continuously optimize the weight adjustment and strategy update process.

6. The battery health diagnosis and analysis method based on big data according to claim 4, characterized in that: The weight adjustment formula is as follows: w i (t) represents the weight of the i-th sensor data source at time t, and the value range is [0, 1]; α i Represents the initial weight coefficient of the i-th sensor, set according to its basic importance in prediction; β i represents the data accuracy coefficient of the i-th sensor, reflecting the impact of its accuracy on health prediction. For sensors with higher accuracy, β i is smaller; N is the total number of sensors; t is the current time, used to represent the change of weight over time; Represents the dynamic weighting of the i-th sensor, which is an exponentially decaying function based on the time-varying data quality of that sensor; Indicates that the weighting of all sensors is normalized so that the sum of the weights of all sensors is 1. Through normalization, it is ensured that the contribution of each sensor to the prediction is proportionally distributed; δ i $Q_i(\tau)$ represents the data quality change function of the $i$-th sensor at time $\tau$, reflecting the quality change of the data source. When $\delta$ i $Q_i(\tau)$ approaches 1, it indicates that the sensor quality is stable or good. When $\delta$ i $Q_i(\tau)$ approaches 0, it indicates that the sensor may malfunction or has poor quality. It is calculated in real time according to the sensor failure or data quality fluctuation. The functional expression of $\delta$ i $Q_i(\tau)$ is as follows: Among them, γ i is the performance degradation coefficient of the i-th sensor. The larger it is, the faster the sensor degrades; λ i reflects the change rate coefficient of the environment or the system itself, indicating the change speed of data quality over time τ; t0 is the initial time point of data quality, indicating the change of the sensor from the start of use to the current time.

7. The battery health diagnosis and analysis method based on big data according to claim 1, characterized in that: The creation of the adaptive deep learning model includes the following steps: Divide the feature set after multi-source data fusion into a training set, a validation set, and a test set, which are used for subsequent deep learning model training, hyperparameter tuning, and performance evaluation respectively; in addition, before data division, according to the weight adjustment formula w i (t), dynamically adjust the importance of features in the dataset; Select a suitable deep learning model architecture according to the task requirements. The model architecture includes long short-term memory networks and convolutional neural networks. When selecting the model architecture, combine the dynamic weighting scheme of features so that different sensor data have different influences during model training; Design an adaptive mechanism. By introducing an adaptive optimization algorithm, the adaptive deep learning model can dynamically adjust its hyperparameters according to changes in real-time data; after each update of the network weights of the model, the adaptive deep learning model will re-adjust the weights of each data according to the new model weight distribution, and further improve the prediction accuracy through the optimization algorithm; Use the training set data to train the adaptive deep learning model, and after each round of training, use the validation set to verify the adaptive deep learning model, check the accuracy and generalization ability of the adaptive deep learning model on the validation set, and further adjust the hyperparameters of the adaptive deep learning model; After training is completed, use the test set to conduct a final evaluation of the model. If the prediction results are not satisfactory, adjust the hyperparameters or network structure according to the feedback of the model on the validation set, or introduce more data sources according to the real-time changes in battery health data to enhance the prediction ability of the model; During the evaluation process of the adaptive deep learning model, historical feedback information is combined with real-time data to adjust the feature weights, and an iterative learning method is used to fine-tune the weight adjustment strategy, enabling the adaptive deep learning model to continuously optimize in subsequent prediction tasks and ultimately achieving more accurate battery health diagnosis.

8. The battery health diagnosis and analysis method based on big data according to claim 1, characterized in that: The generation of the battery health score includes the following steps: Input the key features after preprocessing and feature selection into the adaptive deep learning model to calculate the predicted battery health status, which serves as the preliminary basis for the battery health score. Combine the real-time collected battery data with the predicted health status. Based on the combination of the predicted health status and real-time data, calculate the final battery health score using a preset scoring rule or model. Perform normalization processing on the health score to ensure its consistency and comparability. The normalization is adjusted according to the battery type, usage environment, and different performance criteria, enabling the battery health score to be comparable under different circumstances. Based on the generated battery health score, the battery health management system generates corresponding maintenance suggestions and warning measures to guide users to take appropriate actions. Dynamically adjust the health score according to the real-time monitoring data.

9. The battery health diagnosis and analysis method based on big data according to claim 1, wherein: The adjustment of the warning threshold according to the real-time health score and the generation of dynamic alarms include the following steps: The battery health management system continuously monitors the health status of the battery and, based on the health score calculated by the adaptive deep learning model, updates the battery health score in real time. The battery health management system sets multiple warning thresholds according to the health score. The warning thresholds are defined according to the performance criteria of the battery and the range of the health score. According to the real-time battery health score, the battery health management system dynamically adjusts the warning threshold according to a predefined rule. For example, when the battery health score continues to decline, the system will automatically lower the warning threshold to issue a warning before the battery health deteriorates further; conversely, when the health score recovers, the warning threshold will be appropriately increased. When the battery health score approaches or is lower than the set warning threshold, the battery health management system automatically generates an alarm signal. The alarm signal is divided into different severity levels and warns users or relevant system personnel by sending text messages, APP notifications, or directly warning the battery management system, reminding them to take necessary maintenance or replacement measures. The battery health management system gradually optimizes the alarm strategy based on historical alarm data, user feedback, and actual situations.

10. A battery health diagnosis and analysis system based on big data, based on the big data-based battery health diagnosis and analysis method according to any one of claims 1-9, characterized in that, It includes: A data acquisition module for obtaining the real-time operating data of the battery from multiple sensors and data sources. A data preprocessing module for preprocessing the collected raw data to ensure the accuracy and consistency of the data. A feature selection and fusion module for performing feature selection and fusion on the collected multi-source data to extract key features closely related to the battery health status. An adaptive deep learning module for establishing a deep learning model to predict the battery health status. A battery health score generation module for converting the predicted result of the battery health status into a health score and dynamically adjusting the health score according to real-time data. Dynamic warning and alarm module, which is used to dynamically adjust the warning threshold and generate real-time alarms; Historical data feedback and optimization module, which is used to obtain the comparison between historical battery health data and model prediction results, and optimize the system's adaptive deep learning model and weight adjustment strategy.

Images