Lithium battery monitoring method and system based on transfer learning algorithm, medium and equipment

Abstract

The invention discloses a lithium battery monitoring method and system based on a transfer learning algorithm, a medium and equipment, and relates to the technical field of battery monitoring. The method comprises the following steps: constructing a standard prediction model based on a first environment parameter of a first area and battery data, wherein the standard prediction model comprises a standard model parameter; second environment parameters corresponding to the second areas are obtained, and at least one target area is determined based on the second environment parameters and the first environment parameters; constructing a basic prediction model corresponding to the target area based on a transfer learning algorithm and the standard model parameters, obtaining historical environment parameters and historical battery data corresponding to the target area, adjusting the basic prediction model based on the historical environment parameters and the historical battery data, and obtaining a target prediction model corresponding to the target area, and monitoring the lithium battery in the target area based on the target prediction model. By implementing the technical scheme provided by the invention, the comprehensiveness of the recommendation result can be improved, so that the user satisfaction is improved.

Description

Technical Field

[0001] The present application relates to the field of battery monitoring technology, and in particular to a lithium battery monitoring method, system, medium and device based on a transfer learning algorithm. Background Art

[0002] In modern industrial and consumer electronics, lithium batteries are widely used due to their high energy density and long cycle life. With the rapid development of high-tech products such as electric vehicles and wearable devices, the demand for performance monitoring and management of lithium batteries is increasing, and ensuring battery safety and efficiency is crucial.

[0003] Lithium battery performance monitoring usually uses collected battery data to evaluate the battery's health status and predict its future performance. Currently, many battery management systems (BMS) use fixed algorithm models to process this data and judge the battery status based on preset thresholds. Since environmental factors have a great impact on battery performance, when the lithium battery is in a different environment, the accuracy and adaptability of the existing model may be greatly reduced, resulting in a low prediction accuracy of the actual battery performance. Summary of the invention

[0004] The present application provides a lithium battery monitoring method, system, medium and device based on a transfer learning algorithm, which can enable the model to adapt well to different working environments and improve the accuracy of prediction.

[0005] In a first aspect, the present application provides a lithium battery monitoring method based on a transfer learning algorithm, the method comprising: Acquire a first environmental parameter of a first area and battery data under the first environmental parameter, and construct a standard prediction model based on the first environmental parameter and the battery data, wherein the standard prediction model includes standard model parameters; Acquire second environmental parameters corresponding to each of the second areas, and determine at least one target area based on each of the second environmental parameters and the first environmental parameters; Based on the transfer learning algorithm and the standard model parameters, a basic prediction model corresponding to the target area is constructed; Acquire historical environmental parameters and historical battery data corresponding to the target area, adjust the basic prediction model based on the historical environmental parameters and the historical battery data, obtain a target prediction model corresponding to the target area, and monitor the lithium batteries in the target area based on the target prediction model.

[0006] By adopting the above technical solution, a standard prediction model is constructed as a baseline model using the battery data and environmental parameters of the first area, which can more accurately predict the battery health status. By comparing the difference in environmental parameters between the second area and the first area, the target area where a new model needs to be constructed is determined, ensuring that a new model is only established when the environment undergoes significant changes, avoiding the inefficiency of repeated modeling. For the determined target area, a transfer learning algorithm is introduced, and the standard model parameters in the standard prediction model are used to quickly construct a basic prediction model for the target area. Compared with training the model from scratch, transfer learning greatly saves model training time and data collection. The basic prediction model is further adjusted using the historical environmental parameters and historical battery data actually collected in the target area, and a target prediction model specifically suitable for the target area is obtained, ensuring that the target prediction model can adapt well to the actual working environment of the target area and improving the accuracy of the prediction.

[0007] Optionally, constructing a standard prediction model based on the first environmental parameter and the battery data includes: performing feature extraction and data preprocessing on the first environmental parameter and the battery data, respectively, to obtain sample data, wherein the sample data includes battery feature labels and corresponding feature values ​​of the same lithium battery in different time periods; dividing the sample data into a training set and a test set; training an initial prediction model based on the training set; and verifying the initial prediction model according to the test set, stopping the training until the accuracy of the initial prediction model reaches an accuracy threshold, to obtain a standard prediction model.

[0008] By adopting the above technical solution, the original data is extracted and preprocessed to obtain the battery feature labels and corresponding feature values ​​of the same lithium battery in different time periods, forming a sample data set, and then training and verification are performed based on the sample data set, which can improve the accuracy of the prediction model.

[0009] Optionally, determining at least one target area based on each of the second environmental parameters and the first environmental parameters includes: respectively calculating the parameter difference between each dimensional parameter in the first environmental parameter and each dimensional parameter in the corresponding second environmental parameter; if there is any abnormal environmental parameter in the second environmental parameter whose parameter difference is greater than a preset difference range, then the area corresponding to the abnormal environmental parameter is used as the target area.

[0010] By adopting the above technical solution, by quantifying the differences in statistical environmental parameters, the second area with a large environmental difference from the first area can be effectively identified, and these areas with large environmental differences from the first area can be determined as target areas. This can make subsequent model adjustments and migrations more concentrated on these areas with large environmental differences, avoiding indifferent model migration of all areas.

[0011] Optionally, constructing a basic prediction model corresponding to the target area based on the transfer learning algorithm and the standard model parameters includes: using the standard model parameters as the basic model parameters corresponding to the target area; and migrating parameters of a fixed network layer in the basic prediction model according to the basic model parameters based on the transfer learning algorithm, so that the fixed network layer is adjusted according to the basic model parameters to obtain a basic prediction model.

[0012] By adopting the above technical solution, the parameters optimized by the first area environment training are used as the parameters of the basic model. Compared with directly migrating the preset fixed model parameters, it can better adapt to the environment of the target area. At the same time, by only adjusting the fixed network layer of the model, the initial structure of the model is kept unchanged, reducing the risk of over-adjusting the model.

[0013] Optionally, the basic prediction model is adjusted based on the historical environmental parameters and the historical battery data to obtain a target prediction model corresponding to the target area, including: taking the historical environmental parameters and the historical battery data corresponding to the target area as model adjustment samples, and training the feature processing layer in the basic prediction model based on the model adjustment samples to obtain a trained feature processing layer; combining the trained feature processing layer and the fixed network layer after parameter adjustment to obtain the target prediction model.

[0014] By adopting the above technical solution and using the historical data of the target area itself to adjust the model, the feature processing layer can adapt to the data distribution of the target area, improve the adaptability of the model to the specific environment, and at the same time retain the parameter migration results of the fixed network layer, maintain the prediction effect of the model, and improve the data adaptability to the target environment.

[0015] Optionally, after monitoring the lithium batteries in the target area based on the target prediction model, the method further includes: screening out sample batteries in the target area, obtaining target battery data corresponding to the sample batteries, the target battery data including first battery data in a first time period and second battery data in a second time period, the first time period being earlier than the second time period; obtaining a prediction result of the target prediction model for the first battery data; and adjusting the target prediction model based on the prediction result and the second battery data.

[0016] By adopting the above technical solution, the time series data of the sample battery can be used to determine the model's prediction effect on the state change of the specific battery. By comparing the prediction results with the actual situation, the parameter adjustment direction of the model can be more specifically guided, so that the model can better adapt to the actual state change law of the sample battery and improve the accuracy of the prediction.

[0017] Optionally, the method also includes: performing an accuracy test on the standard prediction model in real time; when the accuracy of the standard prediction model is lower than an accuracy threshold, re-determining a new target battery in the first area and obtaining new battery data corresponding to the target battery; and correcting the standard prediction model based on the new battery data to obtain a new standard prediction model.

[0018] By adopting the above technical solution, a closed-loop monitoring mechanism for the performance of the standard model can dynamically evaluate the degree of adaptation of the model in the first area, and update the model by re-acquiring sample data, so that the standard model can continue to adapt to changes in the environment of the first area, thereby improving the long-term applicability of the model.

[0019] In a second aspect of the present application, a lithium battery monitoring system based on a transfer learning algorithm is provided, the system comprising: A standard model building module, used to obtain a first environmental parameter of a first area and battery data under the first environmental parameter, and build a standard prediction model based on the first environmental parameter and the battery data, wherein the standard prediction model includes standard model parameters; a target area determination module, configured to obtain second environmental parameters corresponding to each of the second areas, and determine at least one target area based on each of the second environmental parameters and the first environmental parameters; A parameter migration module, used to construct a basic prediction model corresponding to the target area based on a migration learning algorithm and the standard model parameters; A target model construction module is used to obtain historical environmental parameters and historical battery data corresponding to the target area, adjust the basic prediction model based on the historical environmental parameters and historical battery data, obtain a target prediction model corresponding to the target area, and monitor the lithium batteries in the target area based on the target prediction model.

[0020] In a third aspect of the present application, a computer storage medium is provided, wherein the computer storage medium stores a plurality of instructions, wherein the instructions are suitable for being loaded by a processor and executing the above method steps.

[0021] In a fourth aspect of the present application, an electronic device is provided, comprising: a processor and a memory; wherein the memory stores a computer program, and the computer program is suitable for being loaded by the processor and executing the above-mentioned method steps.

[0022] In summary, one or more technical solutions provided in the embodiments of the present application have at least the following technical effects or advantages: A standard prediction model was constructed as a baseline model using the battery data and environmental parameters of the first area, which can predict the battery health status more accurately. By comparing the differences in environmental parameters between the second area and the first area, the target area where a new model needs to be constructed is determined, ensuring that a new model is only established when the environment changes significantly, avoiding the inefficiency of repeated modeling. For the determined target area, a transfer learning algorithm was introduced, and the basic prediction model of the target area was quickly constructed using the standard model parameters in the standard prediction model. Compared with training the model from scratch, transfer learning greatly saves model training time and data collection. The basic prediction model was further adjusted using the historical environmental parameters and historical battery data actually collected in the target area, and a target prediction model specifically suitable for the target area was obtained, ensuring that the target prediction model can adapt well to the actual working environment of the target area and improving the accuracy of the prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 It is a flow chart of a lithium battery monitoring method based on a transfer learning algorithm provided in an embodiment of the present application; Figure 2 It is a module schematic diagram of a lithium battery monitoring system based on a transfer learning algorithm provided in an embodiment of the present application; Figure 3 It is a structural schematic diagram of an electronic device provided in an embodiment of the present application.

[0024] Description of reference numerals: 300, electronic device; 301, processor; 302, communication bus; 303, user interface; 304, network interface; 305, memory. DETAILED DESCRIPTION

[0025] In order to enable technicians in this field to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below in conjunction with the drawings in the embodiments of this specification. Obviously, the described embodiments are only part of the embodiments of this application, not all of the embodiments.

[0026] In the description of the embodiments of the present application, words such as "for example" or "for example" are used to indicate examples, illustrations or explanations. Any embodiment or design described as "for example" or "for example" in the embodiments of the present application should not be interpreted as being more preferred or more advantageous than other embodiments or designs. Specifically, the use of words such as "for example" or "for example" is intended to present related concepts in a specific way.

[0027] In the description of the embodiments of the present application, the meaning of the term "multiple" refers to two or more. For example, multiple systems refer to two or more systems, and multiple screen terminals refer to two or more screen terminals. In addition, the terms "first" and "second" are used for descriptive purposes only and cannot be understood as indicating or implying relative importance or implicitly indicating the indicated technical features. Thus, the features defined as "first" and "second" may explicitly or implicitly include one or more of the features. The terms "include", "comprise", "have" and their variations all mean "including but not limited to", unless otherwise specifically emphasized.

[0028] The following will provide a clear and complete description of the technical solutions in the embodiments of the present application in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all of the embodiments.

[0029] The embodiments of the present application can be applied but not limited to various scenarios with high requirements for battery environment monitoring and changeable environmental conditions. For example, different environmental parameters such as temperature, humidity, altitude, etc. in different regions will affect the performance and life of electric vehicle power batteries, or energy storage battery packs in power grids, factories, buildings and other places. Different application scenarios have inconsistent environmental requirements for battery packs, such as temperature, humidity, vibration, etc.

[0030] As an example, the application scenario of the embodiment of the present application can be: a company that manages lithium batteries provides lithium battery health status assessment and remaining life prediction services to downstream customers, and the company's customers are spread all over the country. Due to differences in environmental conditions in different regions, such as temperature, humidity, altitude, battery usage patterns, etc., even the same model of lithium batteries has different aging conditions and remaining life when used in different regions. Through the solution provided by this application, it is possible to adapt to different environments and improve model migration capabilities to ensure the accuracy of battery performance evaluation and improve the safety and efficiency of battery use.

[0031] Please refer to Figure 1 , a flow chart of a lithium battery monitoring method based on a transfer learning algorithm is proposed. The method can be implemented by a computer program, can be implemented by a single-chip microcomputer, or can be run on a lithium battery monitoring system based on a transfer learning algorithm. The computer program can be integrated in a control terminal or run as an independent tool application. Specifically, the method includes steps 10 to 40, and the above steps are as follows: Step 10: Acquire a first environmental parameter of a first area and battery data under the first environmental parameter, and construct a standard prediction model based on the first environmental parameter and the battery data, wherein the standard prediction model includes standard model parameters.

[0032] The first area in the embodiment of the present application refers to a reference area for obtaining battery data and environmental parameters. In this area, the battery is in a known working environment, and data of the battery under different working conditions can be obtained. These data are used to train and build a standard prediction model. Therefore, the first area can be regarded as a reference area or a benchmark area. The number of lithium batteries used in this area must reach a certain scale to ensure that there is more sample training data.

[0033] The first environmental parameters refer to the parameter values ​​of various environmental factors that affect battery performance and aging in the first region, such as temperature, humidity, altitude, vibration, current and other working environment parameters. These environmental parameters will directly affect the performance and data of the battery in this region and are the basis for building a standard prediction model.

[0034] Specifically, a representative first region needs to be selected, and the environmental conditions of the first region, such as temperature, humidity, altitude and other first environmental parameters, need to be kept relatively stable. In the first region, for a specific model of lithium batteries, long-term monitoring and collection of their battery data under the first environmental parameters are performed, including time series data of parameters such as voltage, current, and temperature.

[0035] The first environmental parameter and the collected battery data are subjected to feature extraction and data preprocessing to construct a sample data set containing battery feature labels and corresponding feature values. The sample data obtained through feature engineering can better reflect the intrinsic relationship between the battery health status and various influencing factors. The sample data set is divided into a training set and a test set. Based on the training set data, an initial prediction model is trained using machine learning algorithms such as neural networks. The initial model is verified and trained, and the model parameters are continuously adjusted on the test set until its prediction accuracy reaches a preset threshold, and finally a standard prediction model with sufficient generalization ability is obtained. The model parameters in the standard prediction model are actually the mapping relationship between battery performance attenuation and various influencing factors under the environmental conditions of the first area. Because it is trained in a relatively stable environment, the standard model parameters can more accurately describe and predict the health status of the battery in this area.

[0036] Based on the above embodiment, as an optional embodiment, the step of building a standard prediction model based on battery data may further include the following steps: Step 101: extracting features and preprocessing the first environmental parameter and the battery data to obtain sample data. The sample data includes battery feature labels and corresponding feature values ​​of the same lithium battery in different time periods.

[0037] Specifically, the first environmental parameter generally includes environmental data in multiple dimensions such as temperature, humidity, and atmospheric pressure. The original battery data is usually a sequence of some original physical parameters, such as curve data of voltage, current, temperature, etc. that change over time. However, these raw data cannot be directly applied to machine learning model training because there are complex nonlinear relationships between them and they may contain noise and redundant information, which makes it difficult to directly reflect the health status of the battery. Therefore, it is necessary to transform these raw data into feature data that can better characterize the law of battery performance degradation through feature engineering.

[0038] First, the raw data needs to be preprocessed, such as filtering and denoising, interpolation processing, etc., and corresponding cleaning and filling processing is required to eliminate the interference of these abnormal data on the model. In addition, through the method of feature engineering, some high-order composite features can be constructed in order to obtain more valuable information features. In this way, the original first environmental parameters and battery data are converted into a series of feature vectors. These feature vectors are arranged in chronological order and form sample data pairs with the corresponding battery status indicator (such as power, internal resistance, etc.) label values. Since the same lithium battery will have multiple data samples in different time periods, the final sample data contains the battery feature labels of the same lithium battery in different time periods, as well as the corresponding environmental and other feature values.

[0039] That is to say, for each battery sample, it is necessary to record its characteristic data at different stages (such as new, after being used for a period of time, severely aged, etc.).

[0040] Step 102: Divide the sample data into a training set and a test set.

[0041] Specifically, machine learning algorithms usually adopt a supervised training method, which requires labeled training data to learn the model and optimize parameters to obtain better prediction performance. However, if all data is used for training, it is impossible to evaluate the generalization ability of the model on unseen new data, which is easy to cause overfitting. Therefore, it is necessary to clearly divide two independent data subsets, the training set and the test set, from the sample data. The training set is used in the training process of the model, and the model parameters will be continuously optimized and adjusted on this data set to minimize the prediction error, while the test set is used as new data to evaluate the generalization ability of the model to avoid overfitting. There are many specific ways to divide the data. The sample data can be randomly shuffled first, and then divided into a training set and a test set according to a certain ratio (such as 7:3 or 8:2). In addition to random division, strategies such as stratified sampling can also be considered to ensure that the training set and the test set contain representative data samples of different working conditions and aging degrees.

[0042] Step 103: Train an initial prediction model based on the training set.

[0043] Specifically, the raw battery data has been converted into a structured sample data set through data preprocessing and feature extraction, and appropriate machine learning algorithms such as artificial neural networks, support vector machines, and decision trees have been selected. Different algorithms have different abilities in processing nonlinear and high-dimensional data, and need to be selected in combination with the characteristics of the actual problem. Taking the neural network as an example, the network structure needs to be determined first, including the number of input layer nodes (corresponding to the number of sample features), the number of hidden layers and nodes, and the number of output layer nodes (corresponding to the number of predicted targets). Then, based on the training set data, optimization algorithms such as back propagation are used to continuously adjust the weight parameters of the inter-layer connections in the network so that the error between the model's predicted output on the training set and the true label value is minimized. During the training process, some hyperparameters such as learning rate and regularization strength need to be tuned to improve the convergence speed and generalization performance of the model. In addition, residual connections and attention mechanisms can be introduced to further improve the model performance. Through the above training process, an initial prediction model with good prediction accuracy on the training set can be obtained.

[0044] Step 104: Perform verification training on the initial prediction model according to the test set until the accuracy of the initial prediction model reaches the accuracy threshold, then stop the training to obtain a standard prediction model.

[0045] Specifically, in order to solve the possible overfitting problem, although the initial model performs well on the training set, due to the limited training data and the complexity of the model itself, it is likely to have an overfitting phenomenon where the model has high accuracy on the training data but poor performance on the new data. Overfitting the training data will lead to the lack of generalization ability of the model and will not be able to effectively deal with new data encountered in the actual working environment. Therefore, it is necessary to verify the initial model on an unseen test set to evaluate its actual prediction performance. First, the performance indicators of the initial model are calculated based on the sample data of the test set, such as root mean square error, R square, accuracy, etc. If these indicators do not meet expectations, it means that overfitting has occurred and the model needs to be adjusted. There are many ways to adjust, such as introducing regularization terms to reduce model complexity, increasing the amount of training data, updating optimization algorithm parameters, etc. Taking neural networks as an example, the learning rate and early stopping can be adjusted to alleviate overfitting. Based on the adjusted model, training is performed again on the training set, and repeatedly verified on the test set until the evaluation indicator on the test set reaches the corresponding threshold, that is, the training is stopped when the accuracy of the initial prediction model reaches the accuracy threshold. Through this alternating training and verification process, the model will continue to improve and eventually determine the optimal parameter configuration that meets expectations. The standard prediction model includes standard model parameters, such as feature weight parameters, network connection weights, bias parameters, kernel function parameters, regularization parameters, etc.

[0046] Step 20: Obtain second environmental parameters corresponding to each second area, and determine at least one target area based on each second environmental parameter and the first environmental parameter.

[0047] The second area in the embodiment of the present application refers to other areas except the first area, and their environmental parameters are different from those of the first area.

[0048] The parameters of the standard prediction model are learned based on the specific environmental conditions of the first region. When the environmental conditions change significantly, the prediction accuracy of the model may drop significantly. Therefore, it is necessary to first evaluate the degree of difference between the environmental parameters of the new region and the first region to determine whether the model needs to be adjusted to adapt to the new environment.

[0049] Specifically, it is necessary to collect environmental parameter data of each second area, such as temperature, humidity, altitude, electromagnetic environment, etc. These parameters will significantly affect the working performance and attenuation behavior of the battery. Then compare the environmental parameter data of each second area with the corresponding data of the first area, and calculate the difference measurement value between them. The difference measurement method can be a simple absolute value or square difference, or a complex weighted comprehensive distance function can be designed to assign different weights to different environmental parameters. According to the pre-set difference threshold, it is judged whether the environmental difference between each second area and the first area exceeds the threshold. Taking the area where the difference exceeds the threshold as the target area, it indicates that the applicability of the standard prediction model in this area will be affected to a certain extent, and the model needs to be regionalized to improve the prediction accuracy. For areas where the difference is within the threshold, the standard model can be directly applied without additional adjustment. While ensuring the prediction accuracy, the workload of repeated modeling is reduced as much as possible, which improves the efficiency of regional expansion.

[0050] Based on the above embodiment, as an optional embodiment, the step of obtaining the second environmental parameters corresponding to each second area and determining at least one target area based on each second environmental parameter and the first environmental parameter may further include the following steps: Step 201: Calculate the parameter difference between each dimensional parameter in the first environment parameter and each dimensional parameter in the corresponding second environment parameter.

[0051] In order to quantify the degree of environmental difference between the first area and each second area, since the performance degradation of the battery is closely related to a variety of environmental parameters, when these influencing factors change, directly applying the standard prediction model trained in the first area may result in a decrease in accuracy. Therefore, it is necessary to evaluate the difference between the new and old environmental parameters to determine whether the model needs to be adjusted in a targeted manner.

[0052] Specifically, it is necessary to determine the common environmental parameter dimensions contained in both the first environmental parameter and the second environmental parameter, such as temperature, humidity, altitude, electromagnetic environment, etc. Then calculate the specific numerical difference between each environmental parameter dimension in the first area and the second area respectively. For continuous environmental parameters, such as temperature, humidity, etc., the absolute difference or other distance measurement between the two values ​​can be directly calculated. For discrete environmental parameters, such as working environment levels, the distance weights between different levels can be pre-set, and then the weighted difference can be calculated. In addition to considering the differences in a single parameter dimension, it is also necessary to design a comprehensive scoring mechanism to weightedly sum the differences in all parameter dimensions to obtain a comprehensive environmental difference score between the first area and each second area. The weights can be assigned based on the prior knowledge of the degree of influence of different environmental parameters on battery attenuation.

[0053] Step 202: If there is any abnormal environmental parameter in the second environmental parameter whose parameter difference is greater than a preset difference range, the area corresponding to the abnormal environmental parameter is used as the target area.

[0054] When certain key environmental parameters change significantly, it will directly affect the working performance and attenuation law of the battery. If the standard model trained in the first area is still directly used for prediction, the accuracy will not meet the requirements. Therefore, it is necessary to identify these areas with abnormal environmental parameters in advance and list them as target areas where the model needs to be adjusted first.

[0055] Specifically, it is necessary to pre-set a reasonable difference threshold range for each environmental parameter dimension, indicating the range within which the environmental parameter can vary relative to the first area without significantly affecting battery performance. This threshold range can be set based on the degree of impact of the parameter on battery performance, historical experience data, etc. For example, the threshold range of temperature parameters may be relatively small, because temperature changes have a significant impact on batteries, while the threshold range of altitude may be appropriately relaxed. Traverse each second area to determine whether there is an abnormality in the difference between it and the first area in each environmental parameter dimension that exceeds the corresponding threshold range. As long as an over-threshold anomaly occurs in any parameter dimension, the area is directly listed as a target area where the model needs to be adjusted first.

[0056] Step 30: Based on the transfer learning algorithm and standard model parameters, a basic prediction model corresponding to the target area is constructed.

[0057] The parameters of the standard prediction model are obtained based on training of the first regional environment. Direct application in the target area where the environment has undergone significant changes may result in a decrease in accuracy. However, through the transfer learning algorithm, the standard model parameters are used as prior knowledge to quickly build a basic model of the target area, which can greatly improve the modeling efficiency and lay a good foundation for subsequent adjustments.

[0058] Specifically, it is necessary to first determine the type of transfer learning algorithm, such as weight migration. Taking weight migration as an example, the parameters of the standard model can be directly used as initialization parameters to construct the initial model structure of the basic model of the target area. For neural network models, the standard model parameters include network connection weights, bias parameters, etc., which can be directly migrated to the corresponding hierarchical structure of the newly constructed basic model. For kernel function models, etc., the kernel function type and related parameters of the standard model are migrated. While migrating the standard parameters, it is also necessary to make necessary adjustments to the model structure according to the special environmental conditions of the target area, such as adding or subtracting certain input features, adjusting the number of network layers, etc., to better adapt to the new environment. Through this method of parameter migration and structural fine-tuning, a basic prediction model with initialization prediction capabilities can be quickly constructed based on the standard model parameters. This basic model already contains the prior knowledge of the first area and can better describe the basic mode of battery performance degradation.

[0059] Based on the above embodiment, as an optional embodiment, the step of constructing a basic prediction model corresponding to the target area based on the transfer learning algorithm and the standard model parameters may further include the following steps: Step 301: Use the standard model parameters as the basic model parameters corresponding to the target area.

[0060] Step 302: Based on the transfer learning algorithm, parameters of the fixed network layer in the basic prediction model are migrated according to the basic model parameters, so that the fixed network layer is adjusted according to the basic model parameters to obtain the basic prediction model.

[0061] Specifically, assuming that a neural network is used as a machine learning model, when constructing the basic neural network of the target area, it is necessary to keep the structure of a part of the fixed network layers unchanged in advance, such as the underlying convolutional layer, pooling layer, etc. These fixed network layers are mainly used to automatically extract common underlying features from the original input data.

[0062] The parameters of the corresponding fixed layers in the standard prediction model, such as convolution kernel weights and biases, are directly migrated and assigned to the parameters of the same layer in the newly constructed base model. Through this parameter migration, the fixed layers of the base model "inherit" the patterns learned by the standard model when processing the underlying features. For non-fixed layers, such as fully connected layers, due to environmental differences in different regions, the parameters of these layers may need to be retrained to adapt to new data, so the parameters are not directly migrated from the standard model during initialization. After the parameter migration is completed, the fixed layer is adjusted according to the configuration of the standard model parameters, thereby providing the base model with better underlying feature representation capabilities.

[0063] Step 40: Obtain historical environmental parameters and historical battery data corresponding to the target area, adjust the basic prediction model based on the historical environmental parameters and historical battery data, obtain a target prediction model corresponding to the target area, and monitor the lithium batteries in the target area based on the target prediction model.

[0064] The target area in the embodiment of the present application may be an area with less lithium battery usage, and the historical battery data may be a small amount of battery data. Due to the lack of sufficient historical battery data for model training, the migration algorithm in the embodiment of the present application is used to avoid building a prediction model for this area from scratch. Even if the target area has only a small amount of historical battery data, the standard prediction model can be first built using the sufficient data in the first area, and then migrated to the basic prediction model for the target area through the transfer learning algorithm. It only needs to fine-tune the model with the limited historical data of the target area to obtain an accurate target prediction model for the area.

[0065] Specifically, environmental data and battery operation data within a certain period of time are collected from the target area as historical environmental parameters and historical battery data. These data reflect the actual working environment of the area and the performance of the lithium battery in this environment, and can be used as the basis for model adjustment. These historical data are input into the basic prediction model as adjustment samples. Since the basic model is inherited from the standard model, the parameters of its fixed network layer have preliminarily fitted the mapping relationship between the general environment and battery performance. Therefore, in the model adjustment stage, the feature processing layer of the model is mainly trained. By using the historical data of the target area as a training set, the parameters of the feature processing layer are learned and optimized, so that the model can adapt to the actual environmental distribution and battery performance characteristics of the area. After the training is completed, the optimized feature processing layer is combined with the fixed network layer to obtain the target prediction model corresponding to the target area. The target prediction model can better fit the real data distribution of the target area, so that in the subsequent battery monitoring process, the state and future trends of the lithium batteries in the area can be more accurately predicted and evaluated.

[0066] After successfully building a target prediction model for the target area, the lithium batteries in the target area can be monitored and evaluated in real time based on the target prediction model. As a widely used electrochemical energy storage device, the performance degradation and failure of lithium batteries will directly affect the reliable operation of downstream equipment. Timely discovery and early warning of potential battery failure risks are crucial to ensuring system safety and extending battery life. In the actual application environment of the target area, it is necessary to continuously obtain current environmental parameter data, such as temperature and humidity. And collect the latest operating data from the lithium batteries in service, such as power, voltage, internal resistance, etc. After processing these latest data into feature vectors according to the previous feature engineering method, they can be sent as input to the target prediction model. The output of the target prediction model is the predicted future state of the battery, such as remaining power, health score, etc. The prediction model corresponding to each target area has undergone knowledge transfer of the standard model and fine-tuning training of the target area data, so it can more accurately reflect the actual degradation trend of lithium batteries in the area. Based on the model output, the current condition and future trend of the battery can be monitored and evaluated. If a battery is found to be in poor health or has a high risk of failure, an early warning can be issued to arrange maintenance or replace it in advance to avoid serious consequences caused by the failure.

[0067] Based on the above embodiment, as an optional embodiment, this step may further include the following steps: Step 401: Use historical environmental parameters and historical battery data corresponding to the target area as model adjustment samples, and train the feature processing layer in the basic prediction model based on the model adjustment samples to obtain a trained feature processing layer.

[0068] Step 402: Combine the trained feature processing layer and the fixed network layer after parameter adjustment to obtain a target prediction model.

[0069] Specifically, in the deep neural network model, the underlying feature extraction layer is responsible for automatically learning feature representations from the original input, while the high-level network maps and predicts based on these features. Due to the different data distributions of the target area and the first area, the optimal feature representations to be learned may also be different. Therefore, by adjusting the parameters of the feature processing layer, the model can learn feature representations that are more suitable for the environmental distribution of the target area, thereby improving the accuracy of the prediction. First, the historical environmental parameters and battery data of the target area need to be preprocessed according to the previous feature engineering method to obtain a feature vector form that matches the model input. Then, these data are input into the basic prediction model as adjustment samples, and the feature processing layer of the model is trained specifically. During the training process, the parameters of the feature processing layer will be continuously iteratively optimized, trying to learn the optimal feature representation from the input target area samples, so that these learned features can better fit the mapping relationship between the environment and battery performance in the area. After the training is completed, the feature processing layer optimized for the target area is obtained. Finally, this adjusted feature processing layer is combined with the original fixed network layer obtained by parameter migration in the basic model to form the final target prediction model.

[0070] Based on the above embodiment, as an optional embodiment, after the lithium batteries in the target area are monitored based on the target prediction model, the following process may also be included: While real-time monitoring of lithium batteries in the target area is being performed, the target prediction model can also be continuously optimized using sample battery data. Specifically, some sample batteries are selected in the target area as a basis for modeling. The selection principle of these sample batteries can be based on factors such as the length of use time and differences in working environment, with the aim of covering the diversity of battery operating states in the area. It is necessary to obtain the actual operating data of these sample batteries in different time periods as input data for model optimization. It is assumed that the first battery data of the first time period and the second battery data of the second time period are obtained, where the second time period is later than the first time period. The first battery data of the first time period is input into the current target prediction model to obtain the prediction result of the battery state of the target prediction model for the time period. Since the time period is historical data in the past, we can compare the prediction result of the model with the real second battery data of the latter time period. If there is a large deviation between the prediction result and the actual situation, it means that during this period, the mapping relationship between the environment and battery performance in the target area may have changed to a certain extent. In order to track and capture this trend of change, it is necessary to further adjust the target prediction model based on the first battery data and the second battery data.

[0071] The specific adjustment process is to use the first battery data and the second battery data as new training samples and retrain the feature processing layer in the target prediction model. Under the supervision of these subsequent real data, the model will update the feature representation and parameters in a timely manner to continue to fit the latest data distribution, thereby obtaining the corresponding adjusted new model. The online optimization and self-correction of the prediction model are realized. Using the subsequent monitoring data of the sample battery, the target prediction model can continuously track the dynamic changes in the mapping relationship between battery performance and the environment, and update the model parameters through continuous training, thereby ensuring the long-term accuracy and effectiveness of the prediction results.

[0072] Based on the above embodiment, as an optional embodiment, the lithium battery monitoring method based on the transfer learning algorithm may also include the following process: Since the standard model is trained based on the historical data of the first area, the newly collected lithium battery data in the first area can be used to evaluate its prediction accuracy. If it is found that the accuracy of the standard prediction model is lower than the preset accuracy threshold, it means that due to the long-term changes in the environment or usage conditions, the model is no longer valid and needs to be corrected accordingly. New target batteries are re-determined in the first area, and the latest operating data of these new batteries in the current environment are obtained. Then, these new battery data are used as training samples to retrain the original standard prediction model from scratch to obtain a revised new standard prediction model. Since the target prediction model is fine-tuned based on the standard model, the mapping logic of its high-level network is still derived from the original standard model. Therefore, if the standard model itself has a large deviation, it cannot be completely corrected by simple fine-tuning, and a new standard model must be rebuilt.

[0073] See also Figure 2 , is a module schematic diagram of a lithium battery monitoring system based on a transfer learning algorithm provided in an embodiment of the present application. The lithium battery monitoring system based on the transfer learning algorithm may include: a standard model building module, a target area determination module, a parameter migration module and a target model building module, wherein: A standard model building module, used to obtain a first environmental parameter of a first area and battery data under the first environmental parameter, and build a standard prediction model based on the first environmental parameter and the battery data, wherein the standard prediction model includes standard model parameters; a target area determination module, configured to obtain second environmental parameters corresponding to each of the second areas, and determine at least one target area based on each of the second environmental parameters and the first environmental parameters; A parameter migration module, used to construct a basic prediction model corresponding to the target area based on a migration learning algorithm and the standard model parameters; A target model construction module is used to obtain historical environmental parameters and historical battery data corresponding to the target area, adjust the basic prediction model based on the historical environmental parameters and the historical battery data, obtain a target prediction model corresponding to the target area, and monitor the lithium batteries in the target area based on the target prediction model.

[0074] Optionally, the standard model building module is also used to perform feature extraction and data preprocessing on the first environmental parameter and battery data respectively to obtain sample data, wherein the sample data includes battery feature labels and corresponding feature values ​​of the same lithium battery in different time periods; divide the sample data into a training set and a test set; train an initial prediction model based on the training set; and verify the initial prediction model according to the test set, and stop training until the accuracy of the initial prediction model reaches an accuracy threshold to obtain a standard prediction model.

[0075] Optionally, the target area determination module is also used to respectively calculate the parameter difference between each dimensional parameter in the first environmental parameter and each dimensional parameter in the corresponding second environmental parameter; if there is any abnormal environmental parameter in the second environmental parameter whose parameter difference is greater than a preset difference range, the area corresponding to the abnormal environmental parameter is used as the target area.

[0076] Optionally, the parameter migration module is also used to use the standard model parameters as the basic model parameters corresponding to the target area; based on the transfer learning algorithm, the parameters of the fixed network layer in the basic prediction model are migrated according to the basic model parameters, so that the fixed network layer is adjusted according to the basic model parameters to obtain the basic prediction model.

[0077] Optionally, the target model construction module is also used to use the historical environmental parameters corresponding to the target area and the historical battery data as model adjustment samples, and train the feature processing layer in the basic prediction model based on the model adjustment samples to obtain a trained feature processing layer; combine the trained feature processing layer and the fixed network layer after parameter adjustment to obtain a target prediction model.

[0078] Optionally, a lithium battery monitoring based on a transfer learning algorithm also includes a target model correction module, which is used to screen out sample batteries in the target area and obtain target battery data corresponding to the sample batteries, the target battery data including first battery data in a first time period and second battery data in a second time period, the first time period being earlier than the second time period; obtain a prediction result of the target prediction model for the first battery data; and adjust the target prediction model based on the prediction result and the second battery data.

[0079] Optionally, a lithium battery monitoring based on a transfer learning algorithm also includes a standard model correction module, which is used to perform accuracy verification on the standard prediction model in real time. When the accuracy of the standard prediction model is lower than an accuracy threshold, a new target battery is re-determined in the first area, and new battery data corresponding to the target battery is obtained; the standard prediction model is corrected based on the new battery data to obtain a new standard prediction model.

[0080] It should be noted that: when the system provided in the above embodiment realizes its functions, only the division of the above functional modules is used as an example. In actual applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device is divided into different functional modules to complete all or part of the functions described above. In addition, the system and method embodiments provided in the above embodiment belong to the same concept, and the specific implementation process is detailed in the method embodiment, which will not be repeated here.

[0081] An embodiment of the present application also provides a computer storage medium, which can store multiple instructions. The instructions are suitable for being loaded by a processor and executing a lithium battery monitoring method based on a transfer learning algorithm of the above embodiment. The specific execution process can be found in the specific description of the above embodiment, which will not be repeated here.

[0082] Please refer to Figure 3 The application also discloses an electronic device. Figure 3 The electronic device 300 may include: at least one processor 301 , at least one network interface 304 , a user interface 303 , a memory 305 , and at least one communication bus 302 .

[0083] The communication bus 302 is used to realize the connection and communication between these components.

[0084] The user interface 303 may include a display screen (Display) and a camera (Camera). Optionally, the user interface 303 may also include a standard wired interface and a wireless interface.

[0085] The network interface 304 may optionally include a standard wired interface or a wireless interface (such as a WI-FI interface).

[0086] Among them, the processor 301 may include one or more processing cores. The processor 301 uses various interfaces and lines to connect various parts in the entire server, and executes various functions of the server and processes data by running or executing instructions, programs, code sets or instruction sets stored in the memory 305, and calling data stored in the memory 305. Optionally, the processor 301 can be implemented in at least one hardware form of digital signal processing (Digital Signal Processing, DSP), field programmable gate array (Field-Programmable Gate Array, FPGA), and programmable logic array (Programmable Logic Array, PLA). The processor 301 can integrate one or a combination of a central processing unit (Central Processing Unit, CPU), a graphics processing unit (Graphics Processing Unit, GPU) and a modem. Among them, the CPU mainly processes the operating system, user interface and application programs; the GPU is responsible for rendering and drawing the content to be displayed on the display screen; the modem is used to process wireless communications. It can be understood that the above-mentioned modem may not be integrated into the processor 301, and it can be implemented separately through a chip.

[0087] Among them, the memory 305 may include a random access memory (Random Access Memory, RAM) and may also include a read-only memory (Read-Only Memory). Optionally, the memory 305 includes a non-transitory computer-readable storage medium. The memory 305 can be used to store instructions, programs, codes, code sets or instruction sets. The memory 305 may include a program storage area and a data storage area, wherein the program storage area may store instructions for implementing an operating system, instructions for at least one function (such as a touch function, a sound playback function, an image playback function, etc.), instructions for implementing the above-mentioned various method embodiments, etc.; the data storage area may store data involved in the above-mentioned various method embodiments, etc. The memory 305 may optionally also be at least one storage device located away from the aforementioned processor 301. Refer to Figure 3 , the memory 305 as a computer storage medium may include an operating system, a network communication module, a user interface module and an application program of a lithium battery monitoring method based on a transfer learning algorithm.

[0088] exist Figure 3In the electronic device 300 shown, the user interface 303 is mainly used to provide an input interface for the user and obtain the data input by the user; and the processor 301 can be used to call the application program storing a lithium battery monitoring method based on a transfer learning algorithm in the memory 305, and when executed by one or more processors 301, the electronic device 300 executes one or more of the methods described in the above embodiments. It should be noted that for the aforementioned method embodiments, for the sake of simple description, they are all expressed as a series of action combinations, but those skilled in the art should be aware that the present application is not limited to the described order of actions, because according to the present application, certain steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also be aware that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily required for the present application.

[0089] In the above embodiments, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0090] In the several embodiments provided in this application, it should be understood that the disclosed devices can be implemented in other ways. For example, the device embodiments described above are only schematic, such as the division of units, which is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some service interfaces, and the indirect coupling or communication connection of devices or units can be electrical or other forms.

[0091] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0092] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0093] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable memory. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product, which is stored in a memory and includes several instructions for a computer device (which can be a personal computer, server or network device, etc.) to execute all or part of the steps of the various embodiments of the present application. The aforementioned memory includes: various media that can store program codes, such as USB flash drives, mobile hard drives, magnetic disks or optical disks.

[0094] The above is only an exemplary embodiment of the present disclosure and cannot be used to limit the scope of the present disclosure. That is, any equivalent changes and modifications made according to the teachings of the present disclosure are still within the scope of the present disclosure. After considering the disclosure of the specification and the truth of practice, those skilled in the art will easily think of other embodiments of the present disclosure.

[0095] This application is intended to cover any variation, use or adaptation of the present disclosure, which follows the general principles of the present disclosure and includes common knowledge or customary technical means in the art not described in the present disclosure. The description and examples are to be regarded as exemplary only, and the scope and spirit of the present disclosure are defined by the claims.

Claims

1. A lithium battery monitoring method based on transfer learning algorithm, characterized in that: The method comprises: Acquire a first environmental parameter of a first area and battery data under the first environmental parameter, and construct a standard prediction model based on the first environmental parameter and the battery data, wherein the standard prediction model includes standard model parameters; Acquire second environmental parameters corresponding to each of the second areas, and determine at least one target area based on each of the second environmental parameters and the first environmental parameters; Based on the transfer learning algorithm and the standard model parameters, a basic prediction model corresponding to the target area is constructed; Acquire historical environmental parameters and historical battery data corresponding to the target area, adjust the basic prediction model based on the historical environmental parameters and the historical battery data, obtain a target prediction model corresponding to the target area, and monitor the lithium batteries in the target area based on the target prediction model.

2. A lithium battery monitoring method based on transfer learning algorithm according to claim 1, characterized in that: The constructing a standard prediction model based on the first environmental parameter and the battery data includes: Perform feature extraction and data preprocessing on the first environmental parameter and the battery data to obtain sample data, wherein the sample data includes battery feature labels and corresponding feature values ​​of the same lithium battery in different time periods; Dividing the sample data into a training set and a test set; Training an initial prediction model based on the training set; The initial prediction model is verified and trained according to the test set until the accuracy of the initial prediction model reaches an accuracy threshold, and the training is stopped to obtain a standard prediction model.

3. A lithium battery monitoring method based on transfer learning algorithm according to claim 1, characterized in that: The determining at least one target area based on each of the second environmental parameters and the first environmental parameter comprises: Respectively calculating the parameter difference between each dimensional parameter in the first environmental parameter and each dimensional parameter in the corresponding second environmental parameter; If there is any abnormal environmental parameter in the second environmental parameter whose parameter difference is greater than a preset difference range, the area corresponding to the abnormal environmental parameter is taken as the target area.

4. A lithium battery monitoring method based on transfer learning algorithm according to claim 1, characterized in that: The step of constructing a basic prediction model corresponding to the target area based on the transfer learning algorithm and the standard model parameters includes: Using the standard model parameters as basic model parameters corresponding to the target area; Based on the transfer learning algorithm, the fixed network layer in the basic prediction model is parameterized according to the basic model parameters, so that the fixed network layer is adjusted according to the basic model parameters to obtain the basic prediction model.

5. A lithium battery monitoring method based on transfer learning algorithm according to claim 4, characterized in that: The adjusting the basic prediction model based on the historical environmental parameters and the historical battery data to obtain a target prediction model corresponding to the target area includes: Using the historical environmental parameters and the historical battery data corresponding to the target area as model adjustment samples, and training the feature processing layer in the basic prediction model based on the model adjustment samples to obtain a trained feature processing layer; The target prediction model is obtained by combining the trained feature processing layer and the fixed network layer after parameter adjustment.

6. A lithium battery monitoring method based on transfer learning algorithm according to claim 1, characterized in that: After monitoring the lithium batteries in the target area based on the target prediction model, the method further includes: Screening out sample batteries in the target area, and acquiring target battery data corresponding to the sample batteries, wherein the target battery data includes first battery data in a first time period and second battery data in a second time period, wherein the first time period is earlier than the second time period; Obtaining a prediction result of the target prediction model on the first battery data; The target prediction model is adjusted based on the prediction result and the second battery data.

7. A lithium battery monitoring method based on transfer learning algorithm according to claim 1, characterized in that: The method further comprises: Performing an accuracy test on the standard prediction model in real time, and when the accuracy of the standard prediction model is lower than an accuracy threshold, re-determining a new target battery in the first area, and acquiring new battery data corresponding to the target battery; The standard prediction model is modified based on the new battery data to obtain a new standard prediction model.

8. A lithium battery monitoring system based on transfer learning algorithm, characterized in that: The system comprises: A standard model building module, used to obtain a first environmental parameter of a first area and battery data under the first environmental parameter, and build a standard prediction model based on the first environmental parameter and the battery data, wherein the standard prediction model includes standard model parameters; a target area determination module, configured to obtain second environmental parameters corresponding to each of the second areas, and determine at least one target area based on each of the second environmental parameters and the first environmental parameters; A parameter migration module, used to construct a basic prediction model corresponding to the target area based on a migration learning algorithm and the standard model parameters; A target model construction module is used to obtain historical environmental parameters and historical battery data corresponding to the target area, adjust the basic prediction model based on the historical environmental parameters and historical battery data, obtain a target prediction model corresponding to the target area, and monitor the lithium batteries in the target area based on the target prediction model.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a plurality of instructions, and the instructions are suitable for being loaded by a processor and executing the method according to any one of claims 1 to 7.

10. An electronic device, characterized in that: It includes a processor, a memory, a user interface and a network interface, the memory is used to store instructions, the user interface and the network interface are used to communicate with other devices, and the processor is used to execute the instructions stored in the memory so that the electronic device executes the method as described in any one of claims 1-7.

Images