Method and system for combining electrical measurement steps based on prediction algorithm

Abstract

The invention discloses an electrical testing work step merging method and system based on a prediction algorithm, and belongs to the technical field of battery testing. The invention relates to an electrical measurement work step merging method and system based on a prediction algorithm, and the method comprises the steps: S1, adding an initial SOC emptying step in an electrical measurement process, and carrying out the primary electrical measurement and secondary electrical measurement of a battery pack; s2, collecting electrical test data of the tested battery pack, and selecting specific features to construct a training set and a test set; s3, designing a prediction algorithm model; s4, updating parameters in the prediction model through a gradient optimization descent algorithm until the model performance is optimal; s5, deploying the trained prediction model into software, automatically extracting the characteristics required by the battery pack by the software, inputting the characteristics into the prediction model to obtain performance indexes, and automatically judging whether the battery pack is qualified or not; according to the invention, the battery test production efficiency and the battery test accuracy can be improved.

Description

Technical Field

[0001] The present invention relates to the technical field of battery testing, and in particular to a method and system for merging electrical testing steps based on a prediction algorithm. Background Art

[0002] Lithium-ion battery packs for electric vehicles are core components of the power source for new energy vehicles. To ensure the key performance of lithium-ion power batteries in long-term stable energy output, battery pack manufacturers will conduct charge and discharge tests on the battery packs. Currently, traditional battery electrical testing methods, such as fully charging the battery pack (filling the battery pack with energy), then fully discharging the battery pack (discharging the battery pack with energy), and finally charging the battery pack to the shipping energy for electrical testing, pose a risk of excessively high battery pack temperatures during testing. To avoid high temperature risks, the electrical testing process is often divided into a primary electrical test (full charge) and a secondary electrical test (full discharge), which results in additional static and transport work, is not conducive to single-flow production, and is inefficient. In summary, the existing electrical testing process has safety risks, consumes human resources, and is too time-consuming.

[0003] With the development of machine learning, there are currently some battery pack electrical performance prediction methods based on machine learning prediction models, but these methods are often only aimed at predicting the battery pack's SOC (State of Charge) and discharge capacity. Among them, the existing invention patent application document "Battery State Prediction Model Optimization Method, Capacity Prediction Method and Related Device" with publication number CN115563704A includes using the first SOC starting value and the first SOC ending value of the reference charging period to correct and predict the battery pack's SOC state. Although this method uses a prediction model to predict battery pack performance, it only corrects and predicts the SOC and cannot comprehensively judge the overall state of the battery pack. Summary of the Invention

[0004] The purpose of the present invention is to solve the problems in the prior art and to propose a method and system for merging electrical measurement steps based on a prediction algorithm.

[0005] In order to achieve the above object, the present invention adopts the following technical solutions:

[0006] A method for merging electrical measurement steps based on a prediction algorithm includes the following steps:

[0007] S1. Add an initial SOC discharge step during the electrical test process and perform a primary and secondary electrical test on the battery pack.

[0008] S2. Collect the electrical measurement data of the battery pack tested above and select specific features to construct training and test sets;

[0009] S3. Design prediction algorithm models, including single-input first-order prediction models, multi-input first-order prediction models, multi-input second-order prediction models, and multi-layer perceptron models, and train the prediction models using the L1 loss function to predict the full discharge capacity, full discharge energy, full discharge dynamic pressure difference, and full discharge static pressure difference of the battery pack;

[0010] S4. Update the parameters in the prediction model through the gradient optimization descent algorithm until the model performance reaches the optimal level;

[0011] S5. Deploy the trained prediction model to the software. The software automatically extracts the required features of the battery pack and inputs them into the prediction model to obtain performance indicators, and automatically determines whether the battery pack is qualified.

[0012] Preferably, in step S2, the selected characteristics include initial total pressure, initial discharge total pressure, initial discharge static total pressure, initial discharge capacity, initial discharge energy, initial discharge dynamic pressure difference, initial discharge static pressure difference, charging initial temperature, charging end temperature, full charge total pressure, discharge capacity, discharge energy, discharge end dynamic pressure difference, and discharge end static pressure difference.

[0013] Preferably, in step S3, the prediction algorithm model specifically includes: a single-input first-order prediction model y=wx+b, a multi-input first-order prediction model y=WX+b, a multi-input second-order prediction model y=X T W 1 X+W2X-b, multi-layer perceptron model y=σ(W T X+b), where w, W, and b are the parameters of the prediction model;

[0014] Use L1 loss function Train the prediction model and select the most appropriate model battery pack's full discharge capacity, full discharge energy, full discharge dynamic pressure difference, and full discharge static pressure difference for training;

[0015] S-fold cross validation is used when training the prediction model to ensure the generalization ability of the model. Specifically:

[0016] The dataset is randomly divided into S mutually exclusive subsets of equal size. In each iteration, S-1 subsets are selected as training sets, and the remaining 1 subset is used as the test set. This process is repeated S times, and a different subset is selected as the validation set each time to ensure that each subset has a chance to be used as a validation set. Then the best model selection and hyperparameters are selected to avoid overfitting the training data. Finally, the performance of the model is optimized by the relative error.

[0017]

[0018] After the model is trained, it is tested on the test set to finally evaluate the generalization ability of the model.

[0019] Preferably, in step S4, the parameters w and b in the prediction model are updated by a gradient optimization descent algorithm, and the specific process is as follows:

[0020] First, find the gradient of weights w and b, and set y as the true value. is the predicted value, and its formula is:

[0021]

[0022]

[0023] Then use the gradient descent method to update the parameters w and b. Set the learning rate to α and the weight update formula to be:

[0024]

[0025]

[0026] Preferably, after step S5, the step of fine-tuning the model is also included, specifically: extracting battery packs for full charge and discharge tests every day, collecting data and extracting features to input into the model for retraining, so as to fine-tune the model weights and ensure the stability of the prediction model.

[0027] Preferably, during the model fine-tuning process, the loss function used for fine-tuning is first determined, the fine-tuning learning rate is fixed, and then the model parameters are updated using the gradient descent method. The model parameter update formula is as follows:

[0028]

[0029] Among them, θ is the model parameter, η is the learning rate, and J(-) is the loss function The gradient with respect to the parameters θ.

[0030] Preferably, the prediction algorithm model is a multi-layer perceptron model, which can comprehensively consider multiple input features to achieve accurate prediction of the overall state of the battery pack.

[0031] In a more specific technical solution, a system for merging electrical and measuring work steps based on a prediction algorithm includes:

[0032] The data collection module is used to add an initial SOC discharge step during the electrical measurement process, perform primary and secondary electrical measurements on the battery pack, and collect the electrical measurement data of the battery pack;

[0033] Feature selection module, used to select features from the collected electrical measurement data and construct training sets and test sets;

[0034] The model training module is used to design a prediction algorithm model and train the prediction model using the L1 loss function to predict the full discharge capacity, full discharge energy, full discharge dynamic pressure difference, and full discharge static pressure difference of the battery pack;

[0035] The model deployment module is used to deploy the trained prediction model to the software, enabling the software to automatically extract the required features of the battery pack and input the prediction model to obtain performance indicators, and automatically determine whether the battery pack is qualified.

[0036] Preferably, the model training module further includes a cross-validation unit for performing S-fold cross-validation on the prediction model to ensure the generalization ability of the model.

[0037] Preferably, the system also includes a model fine-tuning module for extracting battery packs for full charge and discharge tests every day, collecting data and extracting features to input into the model for retraining, so as to fine-tune the model weights and ensure the stability of the prediction model.

[0038] Compared with the prior art, the present invention provides a method and system for merging electrical measurement steps based on a prediction algorithm, which has the following beneficial effects:

[0039] 1. This high-temperature-stability, high-power film capacitor combines electrical testing steps by adding an initial SOC discharge step during electrical testing and using a deep learning prediction algorithm model to predict key performance indicators of the battery pack. This avoids the additional static and transport work caused by the separate primary electrical testing (full charge) and secondary electrical testing (full discharge) in traditional electrical testing methods, thereby reducing time waste and human resource consumption in the production process.

[0040] 2. This high-temperature-stability, high-power film capacitor uses a deep learning prediction algorithm model to predict key performance indicators of the battery pack, such as full discharge capacity, full discharge energy, full discharge dynamic pressure difference, and full discharge static pressure difference. Compared with traditional methods, it is more accurate and ensures the stability and long-term accuracy of the prediction model.

[0041] The parts not involved in the high-temperature stability and high-power film capacitor are the same as the existing technology or can be implemented using the existing technology. The present invention can improve the battery testing production efficiency and battery testing accuracy, and can also comprehensively judge the overall status of the battery pack. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 This is a schematic diagram of a model training method for an electrical measurement step merging method based on a prediction algorithm and its system proposed by the present invention;

[0043] Figure 2 This is a schematic diagram of a model deployment of a method for merging electrical measurement steps based on a prediction algorithm and its system proposed in the present invention;

[0044] Figure 3 This is a schematic diagram of the primary and secondary electrical measurement steps of a method for merging electrical measurement steps based on a prediction algorithm and a system thereof proposed by the present invention;

[0045] Figure 4 This is a schematic diagram of a method for merging electrical measurement steps based on a prediction algorithm and a system data collection step thereof proposed by the present invention;

[0046] Figure 5 This is a schematic diagram of a method for merging electrical measurement steps based on a prediction algorithm and a system merging steps thereof proposed by the present invention. DETAILED DESCRIPTION

[0047] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments.

[0048] In the description of the present invention, it should be understood that the terms "upper", "lower", "front", "back", "left", "right", "top", "bottom", "inside", "outside", etc., indicating directions or positional relationships, are based on the directions or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific direction, be constructed and operated in a specific direction, and therefore should not be understood as limiting the present invention.

[0049] Reference Figure 1-Figure 5 The present invention proposes a method for merging electrical measurement steps based on a prediction algorithm, which includes the following steps:

[0050] S1. Add an initial SOC discharge step during the electrical test process and perform a primary and secondary electrical test on the battery pack.

[0051] S2. Collect the electrical measurement data of the battery pack tested above and select specific features to construct training and test sets;

[0052] S3. Design prediction algorithm models, including single-input first-order prediction models, multi-input first-order prediction models, multi-input second-order prediction models, and multi-layer perceptron models, and train the prediction models using the L1 loss function to predict the full discharge capacity, full discharge energy, full discharge dynamic pressure difference, and full discharge static pressure difference of the battery pack;

[0053] S4. Update the parameters in the prediction model through the gradient optimization descent algorithm until the model performance reaches the optimal level;

[0054] S5. Deploy the trained prediction model to the software. The software automatically extracts the required features of the battery pack and inputs them into the prediction model to obtain performance indicators, and automatically determines whether the battery pack is qualified.

[0055] In the present invention, by combining the primary and secondary electrical testing steps and introducing an initial SOC venting step, the stationary and transport work of the battery pack during the test process is reduced, thereby improving the production efficiency of the battery test;

[0056] Using deep learning algorithms to build a prediction model that comprehensively considers multiple input features can more accurately predict key performance indicators of battery packs, such as full discharge capacity and full discharge energy, thereby improving the accuracy of battery performance testing and comprehensively judging the overall status of the battery pack;

[0057] Through automated testing and identification, manual intervention is reduced, human resources are saved, and test errors caused by human factors are reduced;

[0058] The initial SOC venting step helps reduce the temperature of the battery pack during testing, thereby reducing safety hazards.

[0059] Reference Figure 1-Figure 5 In step S2, the selected features include initial total pressure, initial discharge total pressure, initial discharge static total pressure, initial discharge capacity, initial discharge energy, initial discharge dynamic pressure difference, initial discharge static pressure difference, charging initial temperature, charging end temperature, full charge total pressure, discharge capacity, discharge energy, discharge end dynamic pressure difference, and discharge end static pressure difference.

[0060] In this embodiment, the model type that best suits the current data set and task requirements is selected to optimize the prediction results.

[0061] Reference Figure 1-Figure 5 In step S3, the prediction algorithm model specifically includes: single-input first-order prediction model y=wx+b, multi-input first-order prediction model y=WX+b, multi-input second-order prediction model y=X T W 1 X+W2X-b, multi-layer perceptron model y=σ(W T X+b), where w, W, and b are the parameters of the prediction model;

[0062] Use L1 loss function Train the prediction model and select the most appropriate model battery pack's full discharge capacity, full discharge energy, full discharge dynamic pressure difference, and full discharge static pressure difference for training;

[0063] S-fold cross validation is used when training the prediction model to ensure the generalization ability of the model. Specifically:

[0064] The dataset is randomly divided into S mutually exclusive subsets of equal size. In each iteration, S-1 subsets are selected as training sets, and the remaining 1 subset is used as the test set. This process is repeated S times, and a different subset is selected as the validation set each time to ensure that each subset has a chance to be used as a validation set. Then the best model selection and hyperparameters are selected to avoid overfitting the training data. Finally, the performance of the model is optimized by the relative error.

[0065]

[0066] After the model is trained, it is tested on the test set to finally evaluate the generalization ability of the model.

[0067] In the present invention, by calculating the gradient, the user can quickly know how to adjust the model parameters to minimize the loss function, thereby improving the prediction accuracy of the model;

[0068] By continuously iteratively updating the model parameters, we can gradually approach the minimum point of the loss function, thereby making the model's performance on the training set better and better;

[0069] In summary, updating the parameters in the prediction model through the gradient optimization descent algorithm ensures the continuous optimization and improvement of the model, and can gradually reduce the prediction error and improve the accuracy and reliability of the model.

[0070] Reference Figure 1-Figure 5 In step S4, the parameters w and b in the prediction model are updated by the gradient optimization descent algorithm. The specific process is as follows:

[0071] First, find the gradient of weights w and b, and set y as the true value. is the predicted value, and its formula is:

[0072]

[0073]

[0074] Then use the gradient descent method to update the parameters w and b. Set the learning rate to α and the weight update formula to be:

[0075]

[0076]

[0077] During the training process of the present invention, attention should be paid to the convergence of the model. If the model fails to converge to the optimal solution for a long time or converges too slowly, it may be necessary to adjust the learning rate, optimization algorithm, or other related parameters. The learning rate α can be customized according to user needs.

[0078] In order to prevent the model from falling into a local optimal solution or other problems during training, parameters need to be initialized in practical applications. Common initialization methods include zero initialization, He initialization, Xavier initialization, and normal distribution initialization.

[0079] The S-fold cross-validation method is used to train the prediction model, which enhances the generalization ability of the model;

[0080] Through multiple iterations and verifications, the best model selection and hyperparameters can be selected to avoid overfitting the training data, thereby ensuring the stability and reliability of the model in practical applications.

[0081] Reference Figure 1-Figure 5 After step S5, the model is fine-tuned. Specifically, the battery pack is sampled daily for full charge and discharge testing, data is collected, and features are extracted and input into the model for retraining to fine-tune the model weights and ensure the stability of the prediction model. During the model fine-tuning process, the loss function used for fine-tuning is first determined, the fine-tuning learning rate is fixed, and then the model parameters are updated using the gradient descent method. The model parameter update formula is as follows:

[0082]

[0083] Among them, θ is the model parameter, η is the learning rate, and J(-) is the loss function The gradient with respect to the parameters θ.

[0084] The prediction algorithm model is a multi-layer perceptron model, which can comprehensively consider multiple input features to achieve accurate prediction of the overall status of the battery pack.

[0085] In this embodiment, the performance of the battery pack may change over time, due to changes in usage conditions or adjustments to the production process. The initial training data set may not fully cover all possible battery pack performance and test conditions. Therefore, the model needs to be regularly fine-tuned to adapt to these changes and ensure the accuracy and reliability of the model. By sampling battery packs for full charge and discharge tests daily and collecting data to fine-tune the model weights, the stability and accuracy of the prediction model can be continuously maintained. This method can adapt to changes in battery pack performance and different test conditions, ensuring the effectiveness and reliability of the model in actual production.

[0086] A certain number of battery packs are sampled daily for full charge and discharge tests, and relevant electrical measurement data is collected. Features relevant to the prediction model, such as initial total pressure, discharge capacity, and discharge energy, are extracted from the collected data. The extracted features are input into the prediction model, and the model is retrained to fine-tune its weights.

[0087] After fine-tuning the model, it is necessary to evaluate the model on the test set to check whether the generalization ability and accuracy of the model have been improved;

[0088] During the model fine-tuning process, parameters such as loss function and learning rate can be customized. In this embodiment, they are comprehensively considered based on specific task requirements, data distribution, and model performance.

[0089] The above model parameter update formula is the core of the gradient descent algorithm. By iteratively updating model parameters, the optimal solution is gradually approached. In practical applications, this formula is implemented in conjunction with specific optimization algorithms (such as stochastic gradient descent and mini-batch gradient descent). By continuously calculating gradients and updating parameters, the model can gradually learn the underlying patterns in the data, thereby improving prediction accuracy.

[0090] In this embodiment, during the model fine-tuning process, by determining the loss function used for fine-tuning and fixing the fine-tuning learning rate, the rationality and effectiveness of the model parameter update are ensured, the model parameters can be gradually optimized, the prediction error can be reduced, and the accuracy and reliability of the model can be improved;

[0091] In addition, the collected data needs to be processed by feature extraction before it can be input into the model for retraining. Feature extraction is a key step in model input, which determines the type and amount of information that the model can learn;

[0092] In this embodiment, the specified prediction algorithm model is a multi-layer perceptron model, which can comprehensively consider multiple input features to achieve accurate prediction of the overall state of the battery pack, thereby improving the accuracy and comprehensiveness of the prediction and helping to better evaluate the performance and quality of the battery pack.

[0093] Reference Figure 1-Figure 5 In this embodiment, a system for merging electrical measurement steps based on a prediction algorithm includes:

[0094] The data collection module is used to add an initial SOC discharge step during the electrical measurement process, perform primary and secondary electrical measurements on the battery pack, and collect the electrical measurement data of the battery pack;

[0095] Feature selection module, used to select features from the collected electrical measurement data and construct training sets and test sets;

[0096] The model training module is used to design a prediction algorithm model and train the prediction model using the L1 loss function to predict the full discharge capacity, full discharge energy, full discharge dynamic pressure difference, and full discharge static pressure difference of the battery pack;

[0097] The model deployment module is used to deploy the trained prediction model to the software, enabling the software to automatically extract the required features of the battery pack and input the prediction model to obtain performance indicators, and automatically determine whether the battery pack is qualified.

[0098] This embodiment proposes a system for merging electrical test steps based on a predictive algorithm, including a data collection module, a feature selection module, a model training module, and a model deployment module. This systematic approach efficiently collects, processes, and analyzes battery test data, as well as builds and deploys predictive models, thereby improving the production efficiency and accuracy of battery testing.

[0099] The data collection module is used to execute step S1.

[0100] The feature selection module is used to execute step S2,

[0101] The model training module is used to perform step S3 and S-fold cross validation.

[0102] The model deployment module is used to execute step S4,

[0103] The model fine-tuning module is used to perform the model fine-tuning process in steps S5 and S6.

[0104] Reference Figure 1-Figure 5 ,The model training module also includes a cross-validation unit, which is used to perform S-fold ,cross-validation on the prediction model to ensure the generalization ability of the ,model.

[0105] In this embodiment, the cross-validation unit in the model training module can further ensure the generalization ability and stability of the prediction model, thereby reducing the prediction error of the model on unknown data and improving the practicality and reliability of the model.

[0106] Reference Figure 1-Figure 5 ,The system also includes a model fine-tuning module that ,extracts battery packs daily for full charge and discharge tests, ,collects data and extracts features to input into the model for ,retraining in order to fine-tune the model weights and ensure the ,stability of the prediction model.

[0107] In this embodiment, by introducing a model fine-tuning module, the performance of the prediction model can be continuously optimized. This method can adapt to changes in battery pack performance and different testing requirements, ensuring the continued effectiveness and reliability of the model in actual production; at the same time, the model fine-tuning module also facilitates subsequent research and improvements.

[0108] The above description is only a preferred specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any technician familiar with the technical field, within the technical scope disclosed by the present invention, who makes equivalent replacements or changes based on the technical solution and inventive concept of the present invention, should be covered by the scope of protection of the present invention.

Claims

1. A method for merging electrical measurement steps based on a prediction algorithm, characterized in that: The following steps are involved: S1. Add an initial SOC discharge step during the electrical test process and perform a primary and secondary electrical test on the battery pack. S2. Collect the electrical measurement data of the battery pack tested above and select specific features to construct training and test sets; S3. Design prediction algorithm models, including single-input first-order prediction models, multi-input first-order prediction models, multi-input second-order prediction models, and multi-layer perceptron models, and train the prediction models using the L1 loss function to predict the full discharge capacity, full discharge energy, full discharge dynamic pressure difference, and full discharge static pressure difference of the battery pack; S4. Update the parameters in the prediction model through the gradient optimization descent algorithm until the model performance reaches the optimal level; S5. Deploy the trained prediction model to the software. The software automatically extracts the required features of the battery pack and inputs them into the prediction model to obtain performance indicators, and automatically determines whether the battery pack is qualified.

2. The method for merging electrical measurement steps based on a prediction algorithm according to claim 1, characterized in that: In step S2, the selected features include initial total pressure, initial discharge total pressure, initial discharge static total pressure, initial discharge capacity, initial discharge energy, initial discharge dynamic pressure difference, initial discharge static pressure difference, charging initial temperature, charging end temperature, full charge total pressure, discharge capacity, discharge energy, discharge end dynamic pressure difference, and discharge end static pressure difference.

3. The method for merging electrical measurement steps based on a prediction algorithm according to claim 1, characterized in that: In step S3, the prediction algorithm model specifically includes: a single-input first-order prediction model y=wx+b, a multi-input first-order prediction model y=WX+b, a multi-input second-order prediction model y=X T W 1 X+W2X-b, multi-layer perceptron model y=σ(W T X+b), where w, W, and b are the parameters of the prediction model; Use L1 loss function Train the prediction model and select the most appropriate model battery pack's full discharge capacity, full discharge energy, full discharge dynamic pressure difference, and full discharge static pressure difference for training; S-fold cross validation is used when training the prediction model to ensure the generalization ability of the model. Specifically: The dataset is randomly divided into S mutually exclusive subsets of equal size. In each iteration, S-1 subsets are selected as training sets, and the remaining 1 subset is used as the test set. This process is repeated S times, and a different subset is selected as the validation set each time to ensure that each subset has a chance to be used as a validation set. Then the best model selection and hyperparameters are selected to avoid overfitting the training data. Finally, the performance of the model is optimized by the relative error. After the model is trained, it is tested on the test set to finally evaluate the generalization ability of the model.

4. The method for merging electrical measurement steps based on a prediction algorithm according to claim 1, characterized in that: In step S4, the parameters w and b in the prediction model are updated by the gradient optimization descent algorithm. The specific process is as follows: First, find the gradient of weights w and b, and set y as the true value. is the predicted value, and its formula is: Then use the gradient descent method to update the parameters w and b. Set the learning rate to α and the weight update formula to be:

5. The method for merging electrical measurement steps based on a prediction algorithm according to claim 1, characterized in that: After step S5, the process also includes a step of fine-tuning the model, specifically: extracting battery packs daily for full charge and discharge testing, collecting data and extracting features to input into the model for retraining, so as to fine-tune the model weights and ensure the stability of the prediction model.

6. The method for merging electrical measurement steps based on a prediction algorithm according to claim 5, characterized in that: During the model fine-tuning process, we first determine the loss function used for fine-tuning, fix the fine-tuning learning rate, and then use the gradient descent method to update the model parameters. The model parameter update formula is as follows: Among them, θ is the model parameter, η is the learning rate, and J(-) is the loss function The gradient with respect to the parameters θ.

7. A method for merging electrical measurement steps based on a prediction algorithm according to any one of claims 1 to 6, characterized in that: The prediction algorithm model is a multi-layer perceptron model, which can comprehensively consider multiple input features to achieve accurate prediction of the overall state of the battery pack.

8. An electrical measurement step merging system based on a prediction algorithm, characterized in that: include: The data collection module is used to add an initial SOC discharge step during the electrical measurement process, perform primary and secondary electrical measurements on the battery pack, and collect the electrical measurement data of the battery pack; Feature selection module, used to select features from the collected electrical measurement data and construct training sets and test sets; The model training module is used to design a prediction algorithm model and train the prediction model using the L1 loss function to predict the full discharge capacity, full discharge energy, full discharge dynamic pressure difference, and full discharge static pressure difference of the battery pack; The model deployment module is used to deploy the trained prediction model to the software, enabling the software to automatically extract the required features of the battery pack and input the prediction model to obtain performance indicators, and automatically determine whether the battery pack is qualified.

9. The electrical measurement step merging system based on prediction algorithm according to claim 8, characterized in that: The model training module also includes a cross-validation unit for performing S-fold cross-validation on the prediction model to ensure the generalization ability of the model.

10. The electrical measurement step merging system based on prediction algorithm according to claim 8, characterized in that: The system also includes a model fine-tuning module, which is used to extract battery packs for full charge and discharge tests every day, collect data and extract features to input into the model for retraining, so as to fine-tune the model weights and ensure the stability of the prediction model.

Images