Energy storage battery cycle life diving point prediction method and system

Abstract

The invention discloses an energy storage battery cycle life diving point prediction method and system. The method comprises the steps of obtaining first historical data of a target lithium battery, and performing first preprocessing on the first historical data; establishing a first prediction model according to the first historical data after the first preprocessing, wherein the input of the first prediction model is the first feature parameter; the first characteristic parameter is any characteristic parameter selected from a first characteristic parameter set; and acquiring first real-time data of a target lithium battery, and predicting the cycle life diving point of the energy storage battery in combination with the first prediction model. The method not only improves the prediction accuracy, but also has good real-time performance and practicability, and has important practical application value for prolonging the service life of the lithium battery, reducing the maintenance cost and improving the operation safety of equipment.

Description

Technical Field

[0001] The present invention relates to the technical field of energy storage battery cycle life dive point prediction, and in particular to a method and system for predicting a dive point in the cycle life of an energy storage battery. Background Art

[0002] Lithium batteries play a key role in modern energy storage and electric devices. Their cycle life directly affects the cost, safety and reliability of related equipment. The diving point in the cycle life of lithium batteries, that is, the stage where the capacity suddenly drops sharply, is a complex phenomenon that is difficult to accurately predict. Existing prediction methods often have problems such as low accuracy and insufficient data feature mining, which makes it difficult to meet the growing demand for high-performance lithium battery applications. However, existing technologies face many challenges in making such predictions:

[0003] Calculation of characteristic parameters: For key parameters that can reflect the cycle life characteristics of lithium batteries, such as capacity decay rate (CDR), voltage change rate (VDR), and internal resistance change rate (IDR), there is a lack of systematic and accurate calculation methods, or there are difficulties in extracting and calculating these parameters from a large amount of cycle test data, and they cannot be well used for subsequent model analysis.

[0004] Model construction and training: In the past, the models used to predict the drop point of the cycle life of energy storage batteries may not fully consider the time series characteristics of lithium battery cycle life data, making it difficult to capture the long-term dependencies in the data, thus affecting the accuracy of the prediction. At the same time, in the process of model training, there is often a lack of scientific and effective determination methods for the selection of hyperparameters (such as the number of hidden layers, learning rate, number of iterations, batch size, etc.). Most of them rely on experience or simple attempts, which can easily lead to underfitting or overfitting of the model, resulting in poor generalization ability of the model, and unable to stably and accurately predict the number of drop point cycles of different lithium batteries. Summary of the invention

[0005] 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.

[0006] In view of the above existing problems, the present invention is proposed.

[0007] Therefore, the present invention provides a method and system for predicting the diving point of the cycle life of an energy storage battery, which can solve the problems mentioned in the background technology.

[0008] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0009] In a first aspect, the present invention provides a method for predicting a drop point in the cycle life of an energy storage battery, comprising:

[0010] Acquire first historical data of a target lithium battery, and perform first preprocessing on the first historical data;

[0011] Establishing a first prediction model according to the first historical data after the first preprocessing, wherein the input of the first prediction model is the first characteristic parameter;

[0012] The first characteristic parameter is any characteristic parameter selected from the first characteristic parameter set;

[0013] The first real-time data of the target lithium battery is obtained, and the diving point of the cycle life of the energy storage battery is predicted by combining the first prediction model.

[0014] As a preferred solution of the method for predicting the drop point of the cycle life of an energy storage battery according to the present invention, the first prediction model includes:

[0015] The first prediction model is any model whose input is the first characteristic parameter and whose output is the lithium battery cycle life diving point or the parameters related to the lithium battery cycle life diving point that can be directly or indirectly obtained.

[0016] As a preferred solution of the method for predicting the drop point of the cycle life of an energy storage battery according to the present invention, wherein: the first prediction model is established according to the first historical data after the first preprocessing, comprising:

[0017] Selecting a first characteristic parameter, and performing a second preprocessing on the first characteristic parameter;

[0018] The first characteristic parameter after the second preprocessing and the lithium battery cycle life diving point corresponding to the first characteristic parameter or related parameters that can directly or indirectly obtain the lithium battery cycle life diving point are used as a training set to train the first prediction model.

[0019] As a preferred solution of the method for predicting the drop point of the cycle life of an energy storage battery according to the present invention, the step of obtaining the first historical data of the target lithium battery includes:

[0020] The first historical data includes at least voltage, current, temperature data and capacity data;

[0021] Preset several cycles for data collection and set the collection duration to a fixed time;

[0022] The cycle includes charging at a constant current to a set voltage upper limit, then switching to constant voltage charging until the current decreases to a set threshold, and discharging at a constant current to a set voltage lower limit.

[0023] As a preferred solution of the method for predicting the drop point of the cycle life of an energy storage battery according to the present invention, the first preprocessing includes:

[0024] removing spike noise data from the first historical data;

[0025] Sort the removed data by size and take the middle value as the valid data;

[0026] The processed data is transmitted according to the set format and communication interface.

[0027] As a preferred solution of the method for predicting the diving point of the cycle life of an energy storage battery described in the present invention, wherein: the first characteristic parameter at least includes the capacity decay rate, the voltage change rate and the internal resistance change rate.

[0028] As a preferred solution of the method for predicting the diving point of the cycle life of an energy storage battery described in the present invention, the second preprocessing at least includes a standardization operation on the first characteristic parameter.

[0029] In a second aspect, the present invention provides a system for predicting the drop point of cycle life of an energy storage battery, comprising:

[0030] A data processing module, used for acquiring first historical data of a target lithium battery and performing first preprocessing on the first historical data;

[0031] A model building module, used to build a first prediction model according to the first historical data after the first preprocessing, wherein the input of the first prediction model is the first characteristic parameter;

[0032] The first characteristic parameter is any characteristic parameter selected from the first characteristic parameter set;

[0033] The prediction module is used to obtain the first real-time data of the target lithium battery and predict the diving point of the cycle life of the energy storage battery in combination with the first prediction model.

[0034] In a third aspect, the present invention provides a computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of the above-mentioned method when executing the computer program.

[0035] In a fourth aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program implements the steps of the method described above when executed by a processor.

[0036] Compared with the prior art, the present invention has the following beneficial effects: the present invention proposes a method and system for predicting the diving point of the cycle life of an energy storage battery, obtains the first historical data of the target lithium battery, and performs a first preprocessing on the first historical data; establishes a first prediction model based on the first historical data after the first preprocessing, and the input of the first prediction model is the first characteristic parameter; the first characteristic parameter is any characteristic parameter selected from the first characteristic parameter set; obtains the first real-time data of the target lithium battery, and predicts the diving point of the cycle life of the energy storage battery in combination with the first prediction model. It not only improves the accuracy of the prediction, but also has good real-time and practicality, and has important practical application value for extending the service life of lithium batteries, reducing maintenance costs, and improving the safety of equipment operation. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for describing the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative labor. Among them:

[0038] Figure 1 A method flow chart of a method and system for predicting a drop point in the cycle life of an energy storage battery provided by an embodiment of the present invention;

[0039] Figure 2 An internal structural diagram of a computer device of a method and system for predicting a drop point in the cycle life of an energy storage battery provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0040] In order to make the above-mentioned purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in the art without creative work should fall within the scope of protection of the present invention.

[0041] Example 1

[0042] Reference Figure 1-Figure 2 , which is the first embodiment of the present invention, provides a method and system for predicting the drop point of cycle life of an energy storage battery, including:

[0043] There are some problems in the existing related technologies. For example, the prediction of battery cycle life usually relies on empirical formulas or simple statistical models, which often cannot accurately capture the complex dynamic changes in the battery aging process. In addition, these traditional methods are inefficient in processing large-scale data and are difficult to adapt to the needs of high-precision life prediction for rapidly developing battery technology.

[0044] The present application provides a method that can effectively solve the above-mentioned problems. Next, multiple embodiments will be combined to explain in detail how to implement the energy storage battery cycle life diving point prediction method;

[0045] Figure 1 A method flow chart of a method and system for predicting the drop point of cycle life of an energy storage battery is shown, including:

[0046] S101, acquiring first historical data of a target lithium battery, and performing first preprocessing on the first historical data;

[0047] In an optional embodiment, the target lithium battery is placed in a controlled test environment that simulates the temperature, humidity and load changes under actual use conditions. By precisely controlling these parameters, the accuracy and repeatability of the test data can be ensured. During the test, key parameters such as battery voltage, current and temperature are monitored and recorded in real time to capture detailed information about battery performance changes over time.

[0048] In an optional embodiment, it may also include periodic charge and discharge tests on the battery to evaluate its cycle performance and capacity decay. Through these comprehensive tests, the battery cycle life diving point, that is, the critical point where the battery performance drops sharply, can be predicted more accurately.

[0049] In an embodiment of the present application, obtaining the first historical data of the target lithium battery includes:

[0050] The first historical data includes at least voltage, current, temperature data and capacity data;

[0051] Preset several cycles for data collection and set the collection duration to a fixed time;

[0052] The cycle includes charging at a constant current to a set voltage upper limit, then switching to constant voltage charging until the current decreases to a set threshold, and discharging at a constant current to a set voltage lower limit.

[0053] In an optional embodiment, several cycles are preset for data collection, and the collection duration is set to a fixed time, which can be designed according to the actual needs of relevant technical personnel. For example, data can be collected every 5 minutes, or adjusted to every 10 minutes or every 20 minutes according to the battery usage and test purpose.

[0054] In an optional embodiment, the acquisition duration can be determined based on the battery charge and discharge rate and the level of detail of the test. For example, for a battery that charges and discharges quickly, a shorter acquisition duration may be required to capture the details of performance changes. For batteries that need to operate stably for a long time, a longer acquisition duration may be more appropriate. By flexibly adjusting these parameters, it is possible to ensure that the test results are both detailed and efficient, meeting the needs of different application scenarios.

[0055] In an embodiment of the present application, the first preprocessing includes:

[0056] removing spike noise data from the first historical data;

[0057] Sort the removed data by size and take the middle value as the valid data;

[0058] The processed data is transmitted according to the set format and communication interface.

[0059] Exemplarily, the present application selects high-precision data acquisition equipment to ensure that it has the function of simultaneously collecting voltage, current, temperature and capacity data, with a voltage measurement accuracy of ±0.01% FS, a current measurement accuracy of ±0.05% FS, and a temperature measurement accuracy of ±0.5°C.

[0060] In an optional embodiment, the data acquisition device is correctly connected to the lithium battery test device. The connection line should have good electrical performance to reduce signal interference and transmission loss. Voltage acquisition connects the voltage probe to the positive and negative terminals of the lithium battery; current acquisition connects the current sensor in series in the battery charging and discharging circuit; the temperature sensor should fit tightly to the temperature sensitive area on the battery surface; the capacity data is calculated by integrating the current, and the data acquisition device should have the corresponding integration algorithm function module and set the parameters correctly.

[0061] In an optional embodiment, the data acquisition device is initialized and set. The sampling rate is set according to the accuracy and frequency characteristics of the required data; the data storage format can select the commonly used CSV format to facilitate subsequent data processing and analysis; the communication interface is set to an Ethernet interface to ensure that the data can be stably transmitted to the computer or storage device.

[0062] In the embodiment of the present application, the lithium battery cycle test is started, and according to the constant current-constant voltage (CC-CV) charge and discharge mode, the battery is first charged with a constant current to the set voltage upper limit, and then switched to constant voltage charging until the current decreases to the set threshold; the discharge process is discharged with a constant current to the set voltage lower limit. During the cycle test, data is collected according to the set rule of collecting data once every 5 cycles, and each collection time is 10 minutes. During the collection period, the data acquisition equipment continuously records the voltage, current, and temperature data, and calculates the capacity data in real time, by integrating the current: Where t1 and t2 are the start and end time of the collection period.

[0063] In an optional embodiment, the data acquisition device performs preliminary processing on the collected raw data, uses median filtering to remove spike noise data during the acquisition process, sorts a set of collected data by size, takes the middle value as valid data, and then transmits the processed data to a computer or storage device according to a set format and communication interface.

[0064] It should be noted that obtaining the first historical data of the target lithium battery and performing the first preprocessing on the first historical data can effectively eliminate outliers and noise and improve the accuracy and reliability of the data. Through preprocessing, the quality of input data for subsequent analysis and prediction models can be ensured, thereby improving the accuracy of predictions. In addition, the preprocessing of the first historical data also includes data normalization or standardization processing to eliminate the influence of different dimensions and magnitudes on the analysis results. After such processing, the data can be made more suitable for the training of machine learning algorithms, laying a solid foundation for accurately predicting the cycle life diving point of lithium batteries.

[0065] S102, establishing a first prediction model according to the first historical data after the first preprocessing, wherein the input of the first prediction model is the first characteristic parameter;

[0066] In an embodiment of the present application, the first prediction model includes:

[0067] The first prediction model is any model whose input is the first characteristic parameter and whose output is the lithium battery cycle life diving point or the parameters related to the lithium battery cycle life diving point that can be directly or indirectly obtained.

[0068] In an optional embodiment, the first prediction model can be trained by machine learning algorithms, such as support vector machines (SVM), random forests, neural networks, etc. Through these algorithms, the model can learn complex patterns and relationships in historical data, thereby accurately predicting the cycle life of lithium batteries. During the training process, the model will continuously adjust its internal parameters to minimize the prediction error. After the training is completed, the first prediction model will be able to predict the cycle life diving point of the lithium battery, that is, the critical point where the battery performance drops sharply, based on the input characteristic parameters. The prediction results can provide a scientific basis for the maintenance and replacement of the battery, thereby extending the service life of the battery and reducing maintenance costs.

[0069] In an optional embodiment, the first prediction model can further improve the accuracy of the prediction through an integrated learning method. Integrated learning can effectively reduce the risk of overfitting and improve the generalization ability of the model by constructing and combining multiple learners to complete the prediction task. For example, integrated algorithms such as gradient boosting decision tree (GBDT) or extreme gradient boosting (XGBoost) can be used. These algorithms build a strong learner by combining multiple weak learners, thereby achieving higher accuracy and stability when predicting the cycle life of lithium batteries. In addition, cross-validation and other techniques can be used to optimize model parameters to ensure that the model maintains good prediction performance on different data sets. Through these methods, the first prediction model can not only accurately predict the cycle life diving point of lithium batteries, but also provide reliable data support for the battery management system to help realize intelligent management of batteries.

[0070] In an optional embodiment, the first prediction model can also process time series data by establishing a long short-term memory network (LSTM) to capture the complex pattern of battery performance changes over time. The LSTM network is particularly suitable for prediction problems of sequence data because it can remember long-term dependencies and can effectively process and predict long-term dependency problems in time series. In this way, the model can learn key information in the battery charging and discharging history, and then predict future trends in battery performance.

[0071] In an optional embodiment, other deep learning techniques, such as convolutional neural network (CNN), can be combined to extract the characteristics of the battery charge and discharge curve to further improve the accuracy of the prediction. Through the integration of these deep learning methods, the first prediction model can more accurately predict the cycle life diving point of the lithium battery, providing more scientific decision support for battery maintenance and management.

[0072] In the embodiment of the present application, the first characteristic parameter is any characteristic parameter selected from the first characteristic parameter set;

[0073] In an optional embodiment, the first characteristic parameter set is a set of characteristic parameters of the cycle life of a lithium battery, for example, including but not limited to the number of charge and discharge times, charge and discharge rate, temperature, voltage and current of the battery. These parameters can reflect the health status and aging degree of the battery, and are key factors in predicting the battery cycle life drop point. By analyzing the changing trends of these parameters over time, the first prediction model can more accurately identify early signals of battery performance degradation, thereby providing timely warnings for battery maintenance and replacement.

[0074] In an optional embodiment, the model can also flexibly adjust the characteristic parameter set according to different battery types and application scenarios to adapt to different prediction requirements.

[0075] In an embodiment of the present application, establishing a first prediction model according to the first historical data after the first preprocessing includes:

[0076] Selecting a first characteristic parameter, and performing a second preprocessing on the first characteristic parameter;

[0077] The first characteristic parameter after the second preprocessing and the lithium battery cycle life diving point corresponding to the first characteristic parameter or related parameters that can directly or indirectly obtain the lithium battery cycle life diving point are used as a training set to train the first prediction model.

[0078] In the embodiment of the present application, the second preprocessing at least includes performing a standardization operation on the first characteristic parameter.

[0079] In an embodiment of the present application, the first characteristic parameter includes at least a capacity attenuation rate, a voltage change rate, and an internal resistance change rate.

[0080] In the embodiment of the present application, the capacity attenuation rate, voltage change rate and internal resistance change rate in the first characteristic parameter set are selected as the first characteristic parameter, and the specific operation is as follows:

[0081] For the capacity decay rate (CDR) calculation, after data collection and sorting are completed, the initial capacity C0 of the lithium battery is read from the database, and the capacity value recorded at the beginning of the test is read from the database in the order of the number of cycles. n According to the capacity attenuation rate calculation formula Calculate the capacity decay rate at each cycle. First calculate the molecular part C0-C i , then divide the result by C0, and finally multiply by 100% to get the capacity decay rate CDR i The calculated CDR value for each cycle number is stored in a new data column or array for subsequent use.

[0082] For the voltage change rate (VDR) calculation, read the voltage data of two adjacent cycles from the database in the order of the number of cycles. Read the voltage value V of the n-1th cycle n-1 , and then read the voltage value V of the nth cycle n According to the voltage change rate calculation formula Calculate the voltage change rate at each cycle. When calculating, first calculate the numerator part V n -V n-1 , and then divide the result by V n-1 , and finally multiply by 100% to get the voltage change rate VDR n The value of .

[0083] For the calculation of the internal resistance change rate (IDR), the voltage value V at the same cycle time is read from the database n and current value I n According to Ohm's law Calculate the internal resistance at each cycle number According to the order of the number of cycles, read the internal resistance value R of two adjacent cycles in turn. n-1 and R n According to the internal resistance change rate calculation formula Calculate the change rate of internal resistance at each cycle number.

[0084] In an optional embodiment, a Z-score standardization method is used as the second preprocessing method to standardize the calculated characteristic parameters (CDR, VDR, IDR), and the formula is: where x i is the original characteristic parameter value, μ is the mean value of the characteristic parameter, and σ is the standard deviation of the characteristic parameter. The standardized characteristic parameters can eliminate the dimension effect and facilitate subsequent model processing.

[0085] In the embodiment of the present application, the first prediction model is trained by establishing a long short-term memory network (LSTM), and the specific steps may be as follows:

[0086] First, given that lithium battery cycle life data has the characteristics of a time series, LSTM can well capture the long-term dependencies in the sequence, so it is selected as the core prediction model structure.

[0087] Secondly, determine the number of input layer nodes. Since the input features include the standardized capacity decay rate (CDR), voltage change rate (VDR), and internal resistance change rate (IDR), the number of input layer nodes is set to 3 to receive data in these three dimensions.

[0088] Furthermore, the number of hidden layers is set based on experience and a preliminary assessment of the complexity of the data. Start with a smaller number of nodes and gradually adjust. The number of nodes in each hidden layer affects the expressiveness and complexity of the model. Considering that the output is the number of cycles of the diving point, the number of nodes in the output layer is set to 1, representing the single numerical result of the predicted number of cycles. According to the structure of the input layer, hidden layer, and output layer determined above, connect each layer in turn.

[0089] Furthermore, the calculated standardized CDR, VDR, IDR data and all the corresponding diving point cycle number labels are collected and organized, and regarded as an overall data set. The total number of samples in the data set is set to (N).

[0090] In an optional embodiment, the data set is divided according to a set ratio (eg, 80% as a training set and 20% as a validation set).

[0091] Furthermore, the number of training set samples: Calculate the number of training set samples (n train ), the calculation formula is Indicates rounding down operation), that is, the integer number of samples corresponding to 80% of the total number of samples is taken as the number of samples in the training set.

[0092] Furthermore, the number of samples in the validation set: the number of samples in the validation set (n val =Nn train ), subtract the number of samples in the training set from the total number of samples to get the number of samples in the validation set.

[0093] In an optional embodiment, in order to ensure the randomness of data division and avoid the deviation caused by the order of the data itself, the entire data set is first randomly shuffled. This allows each sample to have an equal chance of appearing in a different position after shuffling. According to the number of samples calculated above, the first (n train ) samples as the input features of the training set (denoted as (X train )) and the corresponding output label (denoted as (y train )); select the remaining (n val ) samples as the input features of the validation set (denoted as (X val )) and the corresponding output label (denoted as (y val )).

[0094] Furthermore, the value range of each hyperparameter is determined.

[0095] Furthermore, based on experience and preliminary judgment of the complexity of the problem, the number of hidden layers is set to range from 1 to 4 layers.

[0096] Furthermore, considering that the learning rate is usually set in a small range to ensure the stability and effectiveness of model training, the learning rate range is determined to be 0.0001-0.001. This range allows us to try the impact of different update step sizes on the model training effect during the random search process, and avoid poor training conditions caused by too large or too small learning rates.

[0097] In an optional embodiment, the number of iterations is initially fixed to 50 times.

[0098] In an optional embodiment, the batch size is set to a fixed value of 32, which represents the number of samples selected from the training set each time a gradient calculation and weight update operation is performed.

[0099] In an optional embodiment, the total number of random searches is determined, and a total of 15 random searches are performed to construct a corresponding number of models for training and evaluation.

[0100] Furthermore, random hyperparameter combinations are selected and models are constructed.

[0101] In an optional embodiment, within the set range of 1-4 hidden layers, a specific number of layers is randomly generated as the number of hidden layers of this model. The number of hidden layers extracted is set to (h), and the value of (h) is an integer of 1, 2, 3 or 4.

[0102] In an optional embodiment, within the determined learning rate value range of 0.0001-0.001, a method that can generate uniformly distributed random numbers within this interval is used to obtain the learning rate of this model. Assuming that the generated random number uniformly distributed between 0 and 1 is (r), the calculation formula for mapping it to the learning rate value range is as follows: [lr = 0.0001 + (0.001-0.0001) × r]. The obtained (lr) value is the learning rate randomly drawn this time, which is within the range of 0.0001-0.001, and it will determine the step size of each weight update during the model training process.

[0103] In an optional embodiment, the constructed model is trained using pre-divided training set data. The training process is repeated for a fixed number of 50 iterations, and a fixed number of 32 samples (batch size) are selected from the training set each time for gradient calculation and weight update operations. The specific step size of each weight update is determined based on the extracted learning rate (lr).

[0104] In an optional embodiment, after the model completes 50 iterations of training, the model is evaluated using independent validation set data to calculate common evaluation indicators, root mean square error RMSE and mean absolute error MAE. After completing all 15 random search training and evaluations, the RMSE and MAE values ​​of each model on the validation set are compared to minimize the value, and the relatively optimal hyperparameter combination is selected. Through the random search hyperparameter process, combined with the analysis of model training effects and evaluation indicators, a relatively suitable hyperparameter combination can be found to a certain extent, thereby improving the accuracy and generalization ability of the model in predicting the diving point of the cycle life of energy storage batteries.

[0105] Furthermore, the sample set is used to enter a training loop (for each iteration).

[0106] It should be noted that forward propagation. For each batch of data in the training set, the input features (X train ) in the sample, the shape is (batch s ize×3)(3 is the number of features), and is sent to the input layer of the model. The data is calculated and transformed through each hidden layer in turn according to the set model structure, and the output result (batch shape) is obtained. s ize×1), corresponding to the predicted number of diving point cycles). In each layer, the calculation process involves the corresponding activation function. The fully connected layer uses the ReLU activation function for calculation, which performs nonlinear transformation on the input data to enhance the expression ability of the model.

[0107] In an optional embodiment, the input layer receives an input feature vector (x = (x1, x2, ..., x n )), the input layer and the hidden layer are connected by a weight matrix (W [1] )(dimension is (m×n), where (m) is the number of hidden layer neurons and (n) is the number of input features) connected, and the hidden layer neurons also have a bias term (b [1] )(dimension is (m×1)).

[0108] Furthermore, the output of the hidden layer (z [1] ) is calculated as:

[0109] [z [1] =W [1] x+b [1] ]

[0110] Furthermore, after the activation function of the hidden layer (g(·)), the commonly used ReLU activation function, the activation value of the hidden layer (a [1] ): [a [1] =g(z [1] )]

[0111] Furthermore, the hidden layer and the output layer are connected by a weight matrix (W [2] )(dimension is (1×m), the output layer has only one neuron, and the output is the number of cycles of the diving point) connection, and the output layer also has a bias term (b [2] )(dimension is (1×1)).

[0112] Furthermore, the input of the output layer (z [2] ) is calculated as:

[0113] [z [2] =W [2] a [1] +b [2] ]

[0114] Furthermore, the final prediction output is obtained through the activation function of the output layer

[0115]

[0116] In an optional embodiment, the loss is calculated (LossCalculation). The selected loss function, root mean square error (RMSE), and mean absolute error (MAE) are used to measure the degree of difference between the model prediction result and the true output label.

[0117] In an optional embodiment, the root mean square error (RMSE) calculation formula is:

[0118]

[0119] Where (m) is the number of samples in the validation set, (y i ) is the number of real diving point cycles of the (i)th sample in the validation set, is the number of dive point cycles predicted by the model for the (i)th sample.

[0120] In an optional embodiment, the mean absolute error (MAE) is calculated as:

[0121] Going further, backpropagation and weight updating.

[0122] It should be noted that a first prediction model is established based on the first historical data after the first preprocessing, and the input of the first prediction model is the first characteristic parameter, which can more accurately capture the early signal of battery performance degradation. By analyzing the internal resistance change rate and other characteristic parameters, the first prediction model can predict the performance diving point that may occur in the battery in future cycles. In addition, the model can also provide personalized maintenance suggestions based on the battery's usage history and current status, thereby extending the battery's service life. In practical applications, the establishment and training of the first prediction model requires a large amount of historical data support to ensure the accuracy and reliability of the prediction results.

[0123] S103, obtaining first real-time data of the target lithium battery, and predicting the drop point of the cycle life of the energy storage battery in combination with the first prediction model.

[0124] In an optional embodiment, after the prediction is completed, the system will automatically compare the prediction result with the current number of cycles of the battery, and determine whether the battery is about to reach a performance drop point based on a preset threshold. If the prediction result indicates that the battery is about to reach or has entered a critical area of ​​performance degradation, the system will trigger an early warning mechanism and issue a recommendation to the user to maintain or replace the battery.

[0125] In an optional embodiment, an optimal charging and discharging strategy may be provided according to the battery usage and prediction results to slow down the degradation of battery performance and extend its service life.

[0126] It should be noted that in this way, users can take timely measures to avoid losses caused by sudden battery failure.

[0127] In an optional embodiment, the specific step of threshold determination may include setting thresholds for one or more performance parameters, such as battery capacity, internal resistance, charge and discharge efficiency, etc. The system will compare the real-time monitoring data of these parameters with the preset thresholds. When any parameter reaches or falls below the preset threshold, the system will consider that the battery performance has entered the critical attenuation region.

[0128] In an optional embodiment, the threshold value may be dynamically adjusted according to the usage history and current status of the battery to adapt to individual differences of different batteries and changes in usage environment.

[0129] It should be noted that through this flexible threshold judgment mechanism, the performance drop point of the battery can be predicted more accurately, thereby providing users with more precise maintenance or replacement recommendations.

[0130] In summary, the present invention proposes a method for predicting the diving point of the cycle life of an energy storage battery, obtaining the first historical data of a target lithium battery, and performing a first preprocessing on the first historical data; establishing a first prediction model based on the first historical data after the first preprocessing, the input of the first prediction model is the first characteristic parameter; the first characteristic parameter is any characteristic parameter selected from the first characteristic parameter set; obtaining the first real-time data of the target lithium battery, and predicting the diving point of the cycle life of the energy storage battery in combination with the first prediction model. Not only the accuracy of the prediction is improved, but also it has good real-time and practicality, and has important practical application value for extending the service life of lithium batteries, reducing maintenance costs, and improving the safety of equipment operation.

[0131] Example 2

[0132] In an optional embodiment, the loss function is calculated for the output layer weights (W [2] ) and bias (b [2] ). According to the chain rule, we first calculate the gradient of the loss function (L) on the output layer input (z [2] ) Using the mean square error loss function,

[0133]

[0134] Furthermore, based on this result, the loss function is calculated for the output layer weight (W [2] )

[0135]

[0136] Furthermore, similarly, the loss function is calculated for the output layer bias (b [2] )

[0137]

[0138] Furthermore, the loss function is calculated for the hidden layer weights (W [1] ) and bias (b [1] ), using the chain rule, we first calculate the gradient of the loss function on the hidden layer output (z [1] )

[0139]

[0140] in ((g ′ (·)) is the derivative of the hidden layer activation function (g(·))), so:

[0141]

[0142] Furthermore, the loss function is calculated for the hidden layer weights (W [1] )

[0143]

[0144] Furthermore, the loss function is calculated for the hidden layer bias (b [1] )

[0145]

[0146] Furthermore, after calculating the gradient of each layer's weight, the weight parameters are updated according to the set learning rate (η). The weight update formula is as follows:

[0147] For the output layer weights (W [2] )'s update:

[0148]

[0149] For the output layer bias (b [2] )'s update:

[0150]

[0151] For the hidden layer weights (W [1] )'s update:

[0152]

[0153] For the hidden layer bias (b [1] )'s update:

[0154]

[0155] Furthermore, the training cycle is repeated: the above forward propagation, loss calculation, backpropagation and weight update process are repeated continuously until the set number of iterations is reached.

[0156] Furthermore, the real-time collected data of the lithium battery to be predicted is processed according to the above steps to obtain standardized feature parameters, which are then input into the trained and optimized LSTM model. The model output is the predicted result of the number of cycles that may occur when the diving point occurs in the cycle life of the lithium battery.

[0157] Example 3

[0158] This embodiment also provides a system for predicting the drop point of the cycle life of an energy storage battery, including:

[0159] A data processing module, used for acquiring first historical data of a target lithium battery and performing first preprocessing on the first historical data;

[0160] A model building module, used to build a first prediction model according to the first historical data after the first preprocessing, wherein the input of the first prediction model is the first characteristic parameter;

[0161] The first characteristic parameter is any characteristic parameter selected from the first characteristic parameter set;

[0162] The prediction module is used to obtain the first real-time data of the target lithium battery and predict the diving point of the cycle life of the energy storage battery in combination with the first prediction model.

[0163] The above-mentioned unit modules may be embedded in or independent of a processor in a computer device in the form of hardware, or may be stored in a memory in a computer device in the form of software, so that the processor can call and execute operations corresponding to the above-mentioned modules.

[0164] This embodiment also provides a computer device, which may be a terminal, and its internal structure diagram may be as shown in FIG. Figure 2 As shown. The computer device includes a processor, a memory, a communication interface, a display screen and an input device connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be achieved through WIFI, an operator network, NFC (near field communication) or other technologies. When the computer program is executed by the processor, a method for predicting a diving point of the cycle life of an energy storage battery is implemented. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device can be a touch layer covered on the display screen, or a key, trackball or touchpad set on the computer device housing, or an external keyboard, touchpad or mouse, etc.

[0165] This embodiment further provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the following steps are implemented:

[0166] Acquire first historical data of a target lithium battery, and perform first preprocessing on the first historical data;

[0167] Establishing a first prediction model according to the first historical data after the first preprocessing, wherein the input of the first prediction model is the first characteristic parameter;

[0168] The first characteristic parameter is any characteristic parameter selected from the first characteristic parameter set;

[0169] The first real-time data of the target lithium battery is obtained, and the diving point of the cycle life of the energy storage battery is predicted by combining the first prediction model.

[0170] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

[0171] Those skilled in the art will appreciate that the embodiments of the present application can be provided as methods, systems, or computer program products. Therefore, the present application can adopt the form of complete hardware embodiments, complete software embodiments, or embodiments in combination with software and hardware. Moreover, the present application can adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that contain computer-usable program code. The scheme in the embodiments of the present application can be implemented in various computer languages, for example, object-oriented programming language Java and literal scripting language JavaScript, etc.

[0172] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0173] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0174] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0175] Although the preferred embodiments of the present application have been described, those skilled in the art may make other changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications falling within the scope of the present application.

[0176] Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the spirit and scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the claims of the present application and their equivalents, the present application is also intended to include these modifications and variations.

Claims

1. A method for predicting the drop point of cycle life of energy storage batteries, characterized in that: include: Acquire first historical data of a target lithium battery, and perform first preprocessing on the first historical data; Establishing a first prediction model according to the first historical data after the first preprocessing, wherein the input of the first prediction model is the first characteristic parameter; The first characteristic parameter is any characteristic parameter selected from the first characteristic parameter set; The first real-time data of the target lithium battery is obtained, and the diving point of the cycle life of the energy storage battery is predicted by combining the first prediction model.

2. The method for predicting the drop point of the cycle life of an energy storage battery according to claim 1, characterized in that: The first prediction model includes: The first prediction model is any model whose input is the first characteristic parameter and whose output is the lithium battery cycle life diving point or the parameters related to the lithium battery cycle life diving point that can be directly or indirectly obtained.

3. The method for predicting the drop point of the cycle life of an energy storage battery according to claim 2, characterized in that: The establishing of the first prediction model according to the first preprocessed first historical data comprises: Selecting a first characteristic parameter, and performing a second preprocessing on the first characteristic parameter; The first characteristic parameter after the second preprocessing and the lithium battery cycle life diving point corresponding to the first characteristic parameter or related parameters that can directly or indirectly obtain the lithium battery cycle life diving point are used as a training set to train the first prediction model.

4. The method for predicting the drop point of the cycle life of an energy storage battery according to claim 3, characterized in that: The obtaining of the first historical data of the target lithium battery comprises: The first historical data includes at least voltage, current, temperature data and capacity data; Preset several cycles for data collection and set the collection duration to a fixed time; The cycle includes charging at a constant current to a set voltage upper limit, then switching to constant voltage charging until the current decreases to a set threshold, and discharging at a constant current to a set voltage lower limit.

5. The method for predicting the drop point of the cycle life of an energy storage battery according to claim 4, characterized in that: The first preprocessing comprises: removing spike noise data from the first historical data; Sort the removed data by size and take the middle value as the valid data; The processed data is transmitted according to the set format and communication interface.

6. The method for predicting the drop point of the cycle life of an energy storage battery according to claim 5, characterized in that: The first characteristic parameter includes at least a capacity attenuation rate, a voltage change rate, and an internal resistance change rate.

7. The method for predicting the drop point of the cycle life of an energy storage battery according to claim 6, characterized in that: The second preprocessing at least includes performing a standardization operation on the first characteristic parameter.

8. A system for predicting the drop point of cycle life of energy storage batteries, characterized in that: include: A data processing module, used for acquiring first historical data of a target lithium battery and performing first preprocessing on the first historical data; A model building module, used to build a first prediction model according to the first historical data after the first preprocessing, wherein the input of the first prediction model is the first characteristic parameter; The first characteristic parameter is any characteristic parameter selected from the first characteristic parameter set; The prediction module is used to obtain the first real-time data of the target lithium battery and predict the diving point of the cycle life of the energy storage battery in combination with the first prediction model.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images