A confidence assessment method and system for dynamic skipping of capacity-division processes

Abstract

This invention relates to the field of battery manufacturing technology, and in particular to a confidence assessment method and system for dynamically skipping the capacity grading process. This invention constructs a parameter-capacity mapping matrix by analyzing the correlation between process parameters, material characteristic parameters, and historical battery capacity to address the problem that traditional methods cannot assess parameter interactions. Then, it combines a process capability index to generate a confidence benchmark interval and a design threshold, achieving adaptive adjustment of the assessment standard. Next, it combines the "static historical mapping" with "dynamic electrical performance" through electrical characteristic parameters to obtain the confidence assessment interval. Simultaneously, it calculates the input disturbance coefficient and quality risk index to achieve decoupled discrimination between "data reliability" and "real risk," overcoming the limitation of traditional methods in distinguishing the root causes of fluctuations. Finally, it integrates various indicators and uses a nonlinear algorithm to calculate the dynamic skip confidence, enabling flexible handling of critical-state batteries, thereby improving the precision control of the lithium-ion battery capacity grading process.

Description

Technical Field

[0001] This invention relates to the field of battery manufacturing technology, and in particular to a confidence assessment method and system for dynamically skipping the capacity testing process. Background Technology

[0002] In the lithium-ion battery production process, the "capacity grading process" is a crucial testing step to ensure product quality. Its core function is to accurately measure the actual capacity of the battery and group them according to the capacity value—by screening out batteries with similar parameters such as capacity, internal resistance, and voltage, the consistency of the battery packs after assembly is effectively improved. The "dynamic skip" technology is a revolutionary breakthrough in the field of intelligent manufacturing of lithium-ion batteries. Based on data-driven real-time decision-making, it can selectively omit the traditional capacity grading test while ensuring product quality, thereby significantly improving production efficiency and reducing manufacturing costs. The core technology for achieving this intelligent decision-making is "confidence assessment." This technology constructs a precise mapping relationship between the parameters of the preceding battery processes and capacity characteristics, enabling real-time calculation of the confidence level of each battery to skip the capacity grading test: when the confidence level reaches a preset threshold, the system automatically marks the battery as exempt from capacity grading and allows it to directly proceed to subsequent processes; if the threshold is not reached, the battery still needs to enter the traditional capacity grading process. This shift from relying on "experience-based judgment" to being driven by "data" not only reflects the upgrading and iteration of the industry, but also highlights the key leap of the lithium-ion battery manufacturing industry in the process of intelligent and digital transformation, providing an important practical paradigm for the industry to build an efficient and precise intelligent manufacturing system.

[0003] Existing confidence assessment methods for "dynamic skipping of capacity testing" are mainly based on multi-source data fusion and statistical modeling techniques. These methods collect key process parameters from upstream battery processes and combine them with historical production data to construct a capacity prediction model, calculating the probability of battery capacity meeting standards as the confidence level for skipping capacity testing. However, this assessment method has a significant problem: it lacks reasonable differentiation and flexible handling of critical states. When the confidence level for skipping capacity testing is within a critical range, it cannot distinguish whether the fluctuation originates from model input errors or actual battery quality risks. It mistakenly treats predicted fluctuations under boundary conditions as performance differences, leading to extreme measures of "forced capacity testing" and "direct skipping" for batteries with actual capacity differences less than 0.3%. Even minor parameter fluctuations trigger a step change in decision-making, resulting in inaccurate battery confidence assessments and ultimately leading to both over-testing and under-testing. Summary of the Invention

[0004] The main objective of this invention is to provide a confidence assessment method and system for dynamically skipping capacity-dividing processes, aiming to solve the technical problems in the prior art.

[0005] This invention proposes a confidence assessment method for dynamically skipping capacity-division processes, comprising:

[0006] Acquire multidimensional manufacturing feature data and historical reference data of the battery, wherein the multidimensional manufacturing feature data includes process parameters, material property parameters, electrical property parameters and environmental equipment data;

[0007] The parameter-capacity mapping matrix of the battery is obtained based on the process parameters, material property parameters, and historical reference data.

[0008] The design threshold and confidence baseline range of the battery are obtained based on the multidimensional manufacturing feature data and the parameter-capacity mapping matrix.

[0009] The confidence evaluation range of the battery is obtained based on the parameter-capacity mapping matrix, electrical characteristic parameters, and historical reference data.

[0010] The input disturbance coefficient and quality risk index of the battery are obtained based on the process parameters, environmental equipment data and electrical characteristic parameters.

[0011] The dynamic skip confidence level of the battery is obtained based on the confidence level assessment interval, the confidence level benchmark interval, the input perturbation coefficient, and the quality risk index.

[0012] Determine whether the dynamic skip confidence level is greater than the design threshold;

[0013] If the dynamic skip confidence level is greater than the design threshold, then the battery is determined to be allowed to skip the capacity testing process;

[0014] If the dynamic skip confidence level is not greater than the design threshold, then the battery is determined to be not allowed to skip the capacity testing process.

[0015] Preferably, the step of obtaining the battery parameter-capacity mapping matrix based on the process parameters, material property parameters, and historical reference data includes:

[0016] The core processing parameters are obtained based on the process parameters, including coating surface density, electrode thickness after rolling, and baking temperature. The core material parameters are obtained based on the material characteristic parameters, including positive electrode slurry viscosity, active material median diameter, and membrane porosity.

[0017] Historical battery capacity and historical confidence intervals are obtained based on the historical reference data;

[0018] The capacity influence weight of each parameter is obtained based on the historical confidence interval, core processing parameters, and material property parameters.

[0019] Multiple core processing parameters and multiple material property parameters are arranged and combined to obtain an interaction parameter group, and an interaction effect heat map is obtained based on the interaction parameter group.

[0020] Based on the capacity influence weight and the interaction effect heatmap, multiple core processing parameters and multiple material characteristic parameters are screened to obtain key dominant parameters, and a parameter-capacity mapping matrix is ​​obtained based on the key dominant parameters and the historical battery capacity.

[0021] Preferably, the step of obtaining the battery design threshold and confidence baseline interval based on the multidimensional manufacturing feature data and the parameter-capacity mapping matrix includes:

[0022] Reference sample data is obtained based on historical reference data, and the initial confidence interval of the battery is obtained based on the reference sample data and the parameter-capacity mapping matrix;

[0023] Material fluctuation data and process fluctuation data are obtained based on the multidimensional manufacturing feature data, and the overall consistency coefficient of the battery is obtained based on the material characteristic parameters and the material fluctuation data.

[0024] The historical minimum consistency coefficient and median consistency coefficient are obtained based on the reference sample data, and the design threshold is obtained based on the historical minimum consistency coefficient, median consistency coefficient, overall consistency coefficient, and the initial confidence interval.

[0025] The overall process capability index of the battery is obtained based on the process parameters and the process fluctuation data, and the confidence benchmark interval is obtained based on the initial confidence interval, the process fluctuation data, and the overall process capability index.

[0026] Preferably, the step of obtaining the confidence evaluation interval of the battery based on the parameter-capacity mapping matrix, electrical characteristic parameters, and historical reference data includes:

[0027] The initial confidence assessment interval is obtained based on the parameter-capacity mapping matrix, process parameters, and material property parameters.

[0028] The real-time charge integral, real-time static voltage attenuation rate, and real-time first charge / discharge efficiency are obtained based on the electrical characteristic parameters.

[0029] Multiple complete datasets are obtained based on the historical reference data, and a performance impact weighting set and sensitivity coefficient set are obtained based on the complete datasets, real-time charge integral, real-time static voltage attenuation rate, and real-time first charge and discharge efficiency.

[0030] Based on the multidimensional manufacturing feature data, electrical performance fluctuation data is obtained, and based on the sensitivity coefficient group and the electrical performance fluctuation data, charge correction coefficient, on-state voltage attenuation correction coefficient and efficiency correction coefficient are obtained.

[0031] A comprehensive correction coefficient is obtained based on the charge correction coefficient, the on-state voltage attenuation correction coefficient, the efficiency correction coefficient, and the performance impact weighting. A confidence assessment interval is then obtained based on the comprehensive correction coefficient and the initial confidence assessment interval.

[0032] Preferably, the step of obtaining the battery's input disturbance coefficient and quality risk index based on the process parameters, environmental equipment data, and electrical characteristic parameters includes:

[0033] The process setting values ​​are obtained based on the process parameters, including coating speed, roller gap and slitting tension. Real-time monitoring values ​​are also obtained based on the environmental equipment data, including oven temperature fluctuation variance, hydraulic pressure deviation and vibration spectrum energy.

[0034] The process stability feature vector is obtained based on the process setpoint and the real-time monitoring value, and the quality risk feature vector is obtained based on the electrical characteristic parameters.

[0035] Multiple environmental equipment noise components are obtained based on the process stability feature vector and the quality risk feature vector, and the lithium-ion intercalation entropy change and static open-voltage relaxation constant are obtained based on the quality risk feature vector.

[0036] The input disturbance coefficient is obtained based on the noise components of the multiple environmental devices and the static open-voltage relaxation constant, and the quality risk index is obtained based on the lithium-ion intercalation entropy change and the static open-voltage relaxation constant.

[0037] Preferably, the step of obtaining the dynamic skip confidence level of the battery based on the confidence level assessment interval, the confidence level benchmark interval, the input perturbation coefficient, and the quality risk index includes:

[0038] The boundary proximity determination threshold is obtained based on the historical reference data, and the interval relationship identifier of the battery is obtained based on the confidence evaluation interval, the confidence benchmark interval, and the boundary proximity determination threshold.

[0039] The evaluation interval analysis strategy is obtained based on the interval relationship identifier, and the correction coefficient is obtained based on the evaluation interval analysis strategy, the confidence evaluation interval, and the confidence benchmark interval, wherein the correction coefficient includes the overlap ratio correction coefficient and the boundary distance correction coefficient.

[0040] Obtain influencing factor data, and obtain noise suppression parameters and risk sensitivity parameters based on the influencing factor data;

[0041] A baseline confidence level is obtained based on the confidence level assessment interval, and a dynamic skip confidence level is obtained based on the baseline confidence level, correction coefficient, noise suppression parameter, and risk sensitivity parameter.

[0042] This application also provides a confidence assessment system for dynamically skipping capacity-limiting processes, including:

[0043] The data acquisition module is used to acquire multi-dimensional manufacturing characteristic data and historical reference data of the battery. The multi-dimensional manufacturing characteristic data includes process parameters, material property parameters, electrical property parameters and environmental equipment data.

[0044] The matrix construction module is used to obtain the parameter-capacity mapping matrix of the battery based on the process parameters, material property parameters and historical reference data;

[0045] The benchmark analysis module is used to obtain the battery's design threshold and confidence benchmark interval based on the multidimensional manufacturing feature data and the parameter-capacity mapping matrix;

[0046] The dynamic evaluation module is used to obtain the confidence evaluation range of the battery based on the parameter-capacity mapping matrix, electrical characteristic parameters and historical reference data;

[0047] The risk diagnosis module is used to obtain the battery's input disturbance coefficient and quality risk index based on the process parameters, environmental equipment data, and electrical characteristic parameters.

[0048] The comprehensive evaluation module is used to obtain the dynamic skip confidence level of the battery based on the confidence level evaluation interval, the confidence level benchmark interval, the input perturbation coefficient, and the quality risk index.

[0049] The capacity sizing decision module is used to determine whether the dynamic skip confidence level is greater than the design threshold.

[0050] If the dynamic skip confidence level is greater than the design threshold, then the battery is determined to be allowed to skip the capacity testing process;

[0051] If the dynamic skip confidence level is not greater than the design threshold, then the battery is determined to be not allowed to skip the capacity testing process.

[0052] Preferably, the risk diagnosis module includes:

[0053] The data acquisition unit is used to acquire process setting values ​​based on the process parameters, wherein the process setting values ​​include coating speed, roller gap and slitting tension, and to acquire real-time monitoring values ​​based on the environmental equipment data, wherein the real-time monitoring values ​​include oven temperature fluctuation variance, hydraulic pressure deviation and vibration spectrum energy.

[0054] The feature extraction unit is used to obtain a process stability feature vector based on the process setpoint and the real-time monitoring value, and to obtain a quality risk feature vector based on the electrical characteristic parameters.

[0055] The noise analysis unit is used to obtain multiple environmental equipment noise components based on the process stability feature vector and the quality risk feature vector, and to obtain the lithium-ion intercalation entropy change and the static open-voltage relaxation constant based on the quality risk feature vector.

[0056] The risk quantification unit is used to obtain the input disturbance coefficient based on the noise components of multiple environmental devices and the static open-voltage relaxation constant, and to obtain the quality risk index based on the lithium-ion intercalation entropy change and the static open-voltage relaxation constant.

[0057] The present invention also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the confidence assessment method for dynamically skipping the capacity-sharing process described above.

[0058] The present invention also provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of the confidence assessment method for dynamically skipping the capacity-sharing process described above.

[0059] The beneficial effects of this invention are as follows: This invention uses a multi-source sensor network deployed on the production line to collect multi-dimensional manufacturing characteristic data of batteries in real time, while simultaneously acquiring historical reference data, thereby providing historical experience for current battery evaluation. Then, by analyzing the correlation between process parameters, material characteristic parameters, and historical battery capacity, a parameter-capacity mapping matrix is ​​constructed. This method of quantifying the comprehensive impact of parameter interactions on capacity solves the problems in existing technologies where parameter synergy cannot be evaluated, leading to parameter fluctuations at critical states causing evaluation distortion. Next, based on multi-dimensional manufacturing characteristic data and the parameter-capacity mapping matrix, combined with dynamic adjustment of the process capability index, design thresholds and confidence benchmark intervals are dynamically generated. This method, which allows evaluation standards to adaptively adjust with material fluctuations and process stability, overcomes the limitations of fixed thresholds in existing technologies and solves the problem of "one-size-fits-all standards leading to over-testing or under-testing." Finally, based on the parameter-capacity mapping matrix and historical data... Based on historical reference data, and through real-time correction intervals of electrical characteristic parameters, a confidence assessment interval is obtained by combining "static historical mapping" with "dynamic electrical performance." Then, the input disturbance coefficient is calculated using environmental equipment data, and a quality risk index is obtained by combining electrical characteristic analysis. This achieves decoupling and discrimination between "data reliability" and "real risk," overcoming the limitation of traditional methods that cannot distinguish the root cause of fluctuations. This method provides a dual verification basis for accurate decision-making under critical conditions, ensuring that "the assessment considers not only probability but also the credibility of the probability." Finally, by comprehensively considering the confidence interval, benchmark interval, input disturbance coefficient, and quality risk index, a dynamic skip confidence level is calculated. By employing a nonlinear correction mechanism, batteries with confidence levels at critical states are reasonably distinguished and flexibly processed. This addresses the problem of making extreme decisions for batteries with actual capacity differences of less than 0.3% in existing technologies, as well as the problem of "over-detection and under-detection coexisting," which is conducive to improving the precision control of the lithium-ion battery capacity grading process. Attached Figure Description

[0060] Figure 1 This is a schematic diagram of a method flow according to an embodiment of the present invention.

[0061] Figure 2 This is a schematic diagram of the system structure according to an embodiment of the present invention.

[0062] Figure 3 This is a schematic diagram of the internal structure of a computer device according to an embodiment of this application.

[0063] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0064] It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0065] like Figure 1 As shown, this application provides a confidence assessment method for dynamically skipping capacity-limiting processes, including:

[0066] S1. Obtain multi-dimensional manufacturing feature data and historical reference data of the battery, wherein the multi-dimensional manufacturing feature data includes process parameters, material property parameters, electrical property parameters and environmental equipment data;

[0067] S2. Obtain the battery parameter-capacity mapping matrix based on the process parameters, material property parameters and historical reference data;

[0068] S3. Obtain the battery design threshold and confidence baseline interval based on the multidimensional manufacturing feature data and the parameter-capacity mapping matrix;

[0069] S4. Obtain the confidence evaluation interval of the battery based on the parameter-capacity mapping matrix, electrical characteristic parameters and historical reference data;

[0070] S5. Obtain the input disturbance coefficient and quality risk index of the battery based on the process parameters, environmental equipment data and electrical characteristic parameters;

[0071] S6. Obtain the dynamic skip confidence of the battery based on the confidence assessment interval, the confidence benchmark interval, the input disturbance coefficient, and the quality risk index;

[0072] S7. Determine whether the dynamic skip confidence level is greater than the design threshold;

[0073] If the dynamic skip confidence level is greater than the design threshold, then the battery is determined to be allowed to skip the capacity testing process;

[0074] If the dynamic skip confidence level is not greater than the design threshold, then the battery is determined to be not allowed to skip the capacity testing process.

[0075] As described in steps S1-S7 above, this invention uses a multi-source sensor network deployed on the production line to collect multi-dimensional manufacturing characteristic data of the battery in real time. This multi-dimensional manufacturing characteristic data refers to a set of key data covering the entire battery production process, including process parameters, material property parameters, electrical property parameters, and environmental equipment data. Process parameters refer to the technological parameters of the battery manufacturing process; material property parameters reflect the physicochemical properties of raw materials; electrical property parameters characterize the electrochemical performance of the battery; and environmental equipment data reflects the production environment and equipment status. This data comprehensively depicts the battery's production status and performance characteristics, providing a data foundation for confidence assessment, while also acquiring historical data. Reference data, specifically historical reference data, refers to a collection of process parameters, material properties, electrical properties, environmental equipment data, and corresponding capacity testing results (such as actual capacity and whether the capacity testing process was skipped) for similar batteries produced under the same technological conditions in the past. This provides historical experience for current battery evaluation, enabling the construction of mapping relationships and dynamic adjustment of evaluation standards. Then, by analyzing the correlation between process parameters, material properties, and historical battery capacities, a parameter-capacity mapping matrix is ​​constructed. This matrix quantifies the relationship between parameter combinations and capacity. This method, which transforms the dispersed relationship between parameters and capacity into a structured matrix and quantifies parameter interaction effects, solves the current... There are technical limitations in quantifying the combined impact of parameter combinations on capacity, leading to distorted assessments due to parameter fluctuations at critical states (e.g., the inability to determine whether a parameter fluctuation acts alone or synergistically with other parameters). To address this, a design threshold and confidence benchmark interval are dynamically generated based on multi-dimensional manufacturing feature data and a parameter-capacity mapping matrix. The design threshold determines whether a battery is allowed to skip capacity testing, while the confidence benchmark interval reflects the normal probability range of battery capacity achievement in historical production. This is dynamically adjusted using a process capability index (e.g., the interval shrinks when process stability is high and expands when stability is low) to serve as the benchmark interval for evaluating the current battery. This approach allows the evaluation standard to adaptively adjust to material fluctuations and process stability. This method addresses the limitations of fixed thresholds in existing technologies, resolving the issue of "over-testing or under-testing due to a one-size-fits-all standard" (e.g., using a fixed threshold for batches with large material fluctuations leads to over-testing or under-testing). Then, based on the parameter-capacity mapping matrix and historical reference data, it combines "static historical mapping" with "dynamic electrical performance" by real-time adjusting the range of electrical characteristic parameters to obtain a confidence assessment range. This confidence assessment range refers to the probability range of capacity compliance calculated based on the current battery parameter combination and real-time electrical characteristic parameters; its width dynamically changes with electrical performance fluctuations. Finally, based on process parameters, environmental equipment data, and electrical characteristic parameters, the battery's input disturbance coefficient and quality risk index are obtained.The input disturbance coefficient is an indicator that quantifies the degree of interference of environmental equipment noise (such as vibration and electromagnetic interference) on the evaluation input data. The quality risk index is an indicator that assesses the quality risk of the battery itself. By quantifying data reliability and battery risk separately using these two data points, "noise interference" and "real quality risk" are decoupled. This overcomes the limitation of ignoring the source of risk in existing technologies and solves the core problem of "inability to distinguish the cause of fluctuations." This method of separately quantifying data reliability and battery risk allows the system to clearly determine whether a fluctuation in a parameter is a "false fluctuation" caused by environmental noise or a "real risk" caused by a quality problem in the battery itself. It achieves "evaluation based not only on probability but also on the credibility of probability," thus providing a dual verification basis for accurate decision-making under critical conditions. Finally, by comprehensively considering the confidence interval, benchmark interval, input disturbance coefficient, and quality risk index, the system can determine the true risk of battery quality. The system calculates a risk index and a dynamic skip confidence level. The dynamic skip confidence level is a quantitative indicator that comprehensively assesses the feasibility of skipping the capacity testing process for batteries. By employing a non-linear correction mechanism, the confidence level near the critical value smoothly changes with parameter fluctuations. This allows for the reasonable differentiation and flexible handling of batteries with critical confidence levels, addressing the problem of extreme decisions made for batteries with actual capacity differences less than 0.3% in existing technologies. It avoids drastic switching between "forced capacity testing" and "direct skipping" caused by small fluctuations. Finally, by comparing the dynamic skip confidence level with the design threshold, a capacity testing decision is obtained. This ensures that high-risk batteries enter capacity testing 100% while allowing low-risk batteries to reliably skip it, breaking through the "black and white" decision-making mode of existing technologies and solving the problem of "over-testing and under-testing coexisting." This is beneficial for improving the precision control of the lithium-ion battery capacity testing process.

[0076] In one embodiment, step S2, which involves obtaining the battery's parameter-capacity mapping matrix based on the process parameters, material property parameters, and historical reference data, includes:

[0077] S21. Obtain core processing parameters based on the process parameters, wherein the core processing parameters include coating surface density, electrode thickness after rolling and baking temperature, and obtain core material parameters based on the material characteristic parameters, wherein the core material parameters include positive electrode slurry viscosity, active material median diameter and membrane porosity.

[0078] S22. Obtain the historical battery capacity and historical confidence interval based on the historical reference data;

[0079] S23. Obtain the capacity influence weight of each parameter based on the historical confidence interval, core processing parameters, and material property parameters;

[0080] S24. Arrange and combine multiple core processing parameters and multiple material property parameters to obtain an interaction parameter group, and obtain an interaction effect heat map based on the interaction parameter group;

[0081] S25. Based on the capacity influence weight and the interaction effect heatmap, multiple core processing parameters and multiple material characteristic parameters are screened to obtain key dominant parameters, and a parameter-capacity mapping matrix is ​​obtained based on the key dominant parameters and the historical battery capacity.

[0082] As described in steps S21-S25 above, this invention extracts core processing parameters from process parameters. These core processing parameters are key process parameters that play a decisive role in battery performance and quality during battery production, including coating density, electrode thickness after rolling, and baking temperature. Simultaneously, core material parameters are extracted from material characteristic parameters. These core material parameters are key characteristic parameters of battery materials that have a core impact on battery performance, including positive electrode slurry viscosity, active material diameter, and separator porosity. Existing technologies often include all parameters equally in the model, leading to significant noise interference and dilution of the influence of key parameters. This solution focuses on core parameters, improving the model's sensitivity to key variables and solving the problem of insufficient prediction accuracy caused by parameter redundancy in existing technologies. Next, historical battery capacity and historical confidence intervals corresponding to the above parameters in a time series are extracted from historical reference data. The historical battery capacity refers to… In the past production, testing, or use processes, the recorded actual battery capacity data, historical confidence intervals refer to the statistical analysis of the confidence range of a certain batch of batteries calculated based on historical data. This provides a reference for subsequent analysis of the impact of parameter fluctuations on prediction results. Subsequently, these core processing parameters and core material parameters are used as input features, and historical battery capacity is used as the output label. The XGBoost machine learning feature selection algorithm is used to train the model, and the model outputs the capacity influence weight of each parameter on capacity prediction. Among them, the capacity influence weight refers to the index used to measure the degree of influence of each parameter (such as core processing parameters, core material parameters, etc.) on battery capacity, thereby clarifying the differences in the degree of influence of different parameters on battery capacity. Existing technologies lack quantitative differentiation of parameter influence, leading to the interference of secondary parameter fluctuations in decision-making. This solution solves the problem of misjudgment of critical state caused by the ambiguity of parameter influence in existing technologies by quantifying the importance of parameters.

[0083] Next, from all the acquired core processing parameters and core material parameters, interactive parameter sets that may have mutual influence are generated (such as combinations of all pairs of parameters, including "coating surface density and positive electrode slurry viscosity", "baking temperature and active material median diameter", and "rolled electrode thickness and separator porosity"). An interactive parameter set refers to a set of two parameters that interact with each other. Then, for each interactive parameter set, based on the trained XGBoost model, the SHAP value (i.e., contribution value) of each parameter set to the prediction result is calculated. Simultaneously, SHAP interaction values ​​are calculated for all combinations to obtain the joint interaction contribution value of any two parameters (A and B) in each combination. Finally, a graph is plotted for each interactive parameter set. Construct a SHAP dependency plot: Plot parameter A on the x-axis and the SHAP value of parameter A on the y-axis. The plot uses color depth to indicate the continuous values ​​of parameter B (red indicates high B value, blue indicates low B value). Based on the plot characteristics, determine the interaction relationship between the two parameters: If, when B is high, the SHAP value of A increases rapidly as A increases, it indicates that the positive impact of both A and B on the output is enhanced, and there is a synergistic effect; if, when B is high, the SHAP value of A decreases as A increases (or even turns negative), it indicates that B weakens the positive impact of A, and there is a counteracting effect. Then, for each pair of parameters, calculate its interaction effect strength value: This value is the "joint interaction contribution value of the two parameters" among all samples. The average of the absolute differences between the sums of the SHAP values ​​of individual parameters is used to quantify the strength of nonlinear interactions between parameters that exceed linear superposition: a positive value indicates a synergistic effect (the combined effect is stronger than the sum of individual effects); a negative value indicates a canceling effect (the combined effect is weaker than the sum of individual effects). Then, the interaction effect strength values ​​of all parameter pairs are integrated into a symmetric matrix. The off-diagonal elements in the matrix correspond to the interaction effect strength values ​​of the row and column parameters (positive values ​​indicate synergistic effects, negative values ​​indicate canceling effects, and the absolute value reflects the strength of the interaction). Finally, a matrix is ​​formed using all core feature parameters as row and column labels, with off-diagonal cells representing the interaction effect strength of the corresponding row and column parameters. This scheme... By quantifying interaction effects using SHAP values, the problem of inaccurate confidence assessment caused by ignoring parameter interactions is solved. This enables the model to identify the impact of "parameter combination fluctuations" rather than single parameter fluctuations on capacity. This addresses the issue of existing technologies that only consider the independent effects of individual parameters and ignore the interaction between parameters (e.g., an increase in one parameter is beneficial to capacity, but an increase in another parameter together may be detrimental), leading to prediction bias at critical states. A two-color gradient mapping rule is used for visualization encoding: red (dark red to light pink) represents positive synergistic effects (the higher the intensity, the darker the color), blue (dark blue to light blue) represents negative offsetting effects (the lower the intensity, the darker the color), and white represents no significant interaction (absolute value of intensity less than 0).05), thereby constructing an interaction effect heatmap. This heatmap clearly quantifies whether different parameters have synergistic or counteracting relationships, providing a basis for subsequent optimization of process parameters.

[0084] Then, core processing parameters and core material parameters with significant capacity impact weights, as well as parameters contained in interaction parameter groups with strong interaction strength, are selected as key dominant parameters. Fluctuation data corresponding to the core processing parameters and core material parameters are obtained based on multi-dimensional manufacturing feature data. Next, multiple key dominant parameter groups are obtained based on the key dominant parameters and material data through permutation and combination methods. These key dominant parameter groups (e.g., "coating surface density = 46 mg / cm² + rolling thickness = 80 μm + active material median diameter = 16 μm") are used as row dimensions, and the column dimensions are divided into continuous battery capacity ranges (e.g., 3.5 Ah~3.6 Ah, 3.6 Ah~3.7 Ah, etc.). Similarly, these key dominant parameters are used as input features, and the corresponding historical battery capacity is used as the output label. The XGBoost machine learning feature selection algorithm is used to train the model, enabling the model to learn the nonlinear relationships between parameters (including the nonlinear relationships between individual parameters). (Dynamic influence and interaction effect of interactive parameter groups); After the model training is completed, a large number of artificially constructed key dominant parameters (covering the actual value range of each parameter) are input, and the capacity prediction value and confidence interval corresponding to each key dominant parameter group are output. These are transformed into parameter-capacity interval probability values, which are used as matrix values. The parameter-capacity interval probability value refers to the probability that the battery capacity falls into the corresponding interval under a specific parameter or interactive parameter group. The matrix values ​​are filled into the matrix to finally form the parameter-capacity mapping matrix. Existing technologies rely on a single threshold judgment, which cannot handle the continuity and uncertainty of parameter fluctuations. This solution transforms the parameter combination into the probability distribution of the capacity interval through the mapping matrix, realizing the transformation from "black and white" threshold decision to "probabilistic interval" decision. It solves the problem of critical state step decision-making caused by the lack of parameter-capacity quantitative mapping in existing technologies (such as actual capacity difference <0.3% being classified as different processing methods).

[0085] In one embodiment, step S3, which involves obtaining the battery's design threshold and confidence baseline interval based on the multidimensional manufacturing feature data and the parameter-capacity mapping matrix, includes:

[0086] S31. Obtain reference sample data based on historical reference data, and obtain the initial confidence interval of the battery based on the reference sample data and the parameter-capacity mapping matrix;

[0087] S32. Obtain material fluctuation data and process fluctuation data based on the multi-dimensional manufacturing feature data, and obtain the overall consistency coefficient of the battery based on the material characteristic parameters and the material fluctuation data;

[0088] S33. Obtain the historical minimum consistency coefficient and the median consistency coefficient based on the reference sample data, and obtain the design threshold based on the historical minimum consistency coefficient, the median consistency coefficient, the overall consistency coefficient, and the initial confidence interval.

[0089] S34. Obtain the overall process capability index of the battery based on the process parameters and the process fluctuation data, and obtain the confidence benchmark interval based on the initial confidence interval, the process fluctuation data and the overall process capability index.

[0090] As described in steps S31-S34 above, this invention obtains reference sample data through historical reference data. The reference sample data refers to battery samples that historically "skipped capacity testing and subsequently passed verification," along with their corresponding processing parameters, material parameters, and confirmed compliant actual capacity values. For each battery sample's historical processing parameters and historical material parameters, the corresponding key dominant parameter groups are matched from the parameter-capacity mapping matrix, and the capacity compliance probabilities corresponding to these key dominant parameter groups in the matrix are extracted. The extracted capacity compliance probabilities (e.g., 0.78) are then used to calculate the compliance probability. The values ​​(0.78, 0.81, 0.83, ..., 0.95, 0.97) are summarized to form a "capacity compliance probability distribution". Then, a 95% confidence interval for this distribution is calculated using statistical methods to obtain a confidence reference interval. Specifically, the summarized capacity compliance probability values ​​are sorted from smallest to largest, and the extreme values ​​at both ends (2.5%) are removed. The minimum and maximum values ​​of the remaining data form the confidence reference interval (for example, if the sorted data is 0.78, 0.81, 0.83, ..., 0.95, 0.97, after removing the extreme values ​​at both ends, the interval is [0.81, 0.97, 0.95 ...

[94] Then, based on the degree of influence of the interaction effect of key dominant parameters on the actual battery capacity, the confidence reference interval is calibrated in combination with the interaction effect strength value: when there is a strong synergistic effect between parameters (interaction strength value 0.6~0.8), the difference between the upper and lower limits of the interval is reduced by 0.05; when the interaction strength value > 0.8, the difference between the upper and lower limits of the interval is reduced by 0.08; if there is a significant cancellation effect between parameters (interaction strength value -0.3~-0.5), the difference between the upper and lower limits of the interval is increased by 0.05; when the interaction strength value < -0.5, the difference between the upper and lower limits of the interval is increased. The lower limit difference is increased by 0.08 to obtain the initial confidence interval. This method of calibrating the interval based on the intensity of the interaction effect is a deep understanding and application of the complex relationship between parameters in the battery production process. Existing technologies usually do not fully consider this interaction between parameters, while this invention helps to improve the reliability of the confidence interval. The initial confidence interval refers to the capacity compliance probability range initially determined by matching the parameter-capacity mapping matrix, which provides an initial probability range basis for subsequently determining the design threshold and the confidence benchmark interval.

[0091] Next, material fluctuation data and process fluctuation data are obtained based on multi-dimensional manufacturing characteristic data. Material fluctuation data reflects the degree of deviation of the characteristic parameters of the materials used in the current batch of battery production from standard values ​​during actual production. Process fluctuation data reflects the degree of deviation of the processing parameters during the manufacturing process of the current batch of batteries from set standard values ​​during actual production. Based on this, the standard deviation and average value of each material characteristic parameter in the current batch of materials are calculated individually. Then, the ratio of three times the standard deviation to the average value of that parameter is calculated, and the difference between "1" and this ratio is calculated to obtain the consistency coefficient corresponding to that parameter. Finally, the average of the consistency coefficients corresponding to multiple parameters is taken as the overall consistency coefficient of the battery material characteristics for that batch. The consistency coefficient is an indicator that measures the overall consistency of a batch of battery materials. Based on historical reference data, historical consistency coefficients of multiple different batches of battery materials are obtained. The median and minimum historical consistency coefficients are then calculated from these historical consistency coefficients. The median consistency coefficient is the value in the middle after sorting the consistency coefficients of multiple historically produced batches of battery materials by size. The minimum historical consistency coefficient is the smallest value selected from the consistency coefficients of multiple historically produced batches of battery materials. When the overall consistency coefficient is greater than the median consistency coefficient, the median of the initial confidence interval is used as the design threshold. When the overall consistency coefficient is not greater than the median consistency coefficient, the design threshold is obtained based on the degree of deviation, using the formula… , calculate the design threshold, where T represents the design threshold, M represents the median of the initial confidence interval, and K represents the overall consistency coefficient, K min K represents the minimum historical consistency coefficient. median The overall consistency coefficient represents the median. The essence of this formula is to dynamically allocate the stringency of the threshold based on the degree of deviation in material consistency. A higher overall consistency coefficient (closer to 1) indicates smaller fluctuations in material parameters and higher product quality stability. In this case, the threshold requirement can be relatively lenient (close to 90% of the initial confidence interval median). Conversely, a lower overall consistency coefficient (farther from the median) indicates greater material fluctuations and higher quality risk, requiring stricter screening with a higher threshold (close to 70% of the initial confidence interval median) to prevent defective products from being released. When the overall consistency coefficient is between the historical minimum and the median, the design threshold needs to be "smoothly adjusted" with changes in the overall consistency coefficient, rather than jumping in steps. The formula uses the term "... The essence of the "deviation" part is to normalize the overall consistency coefficient to the range of 0-1 (i.e., "deviation degree": 0 when the overall consistency coefficient equals the minimum consistency coefficient, and 1 when the overall consistency coefficient equals the median consistency coefficient), and then multiply it by a floating ratio of 0.2. Finally, it achieves a linear increase in the design threshold from 70% of the median of the initial confidence interval to 90% of the median of the initial confidence interval. Existing technologies usually do not fully consider the impact of fluctuations in material characteristic parameters on battery quality. Therefore, it is difficult to guarantee the stability of product quality when materials fluctuate. However, this solution, which dynamically adjusts the design threshold according to the material consistency coefficient and flexibly changes the stringency of the threshold according to the material fluctuation, reflects the refined management of battery production quality control. Instead of using a fixed threshold, it can adaptively adjust according to the actual material conditions, making the production standards more in line with the actual production situation and helping to improve the stability of product quality.

[0092] Then, based on the process fluctuation data, obtain the upper limit, lower limit, mean, and standard deviation of each processing parameter for the current batch of materials. Then, use the formula " "Calculate the process capability index corresponding to each processing parameter, where CPK represents the process capability index, USL represents the upper limit of the processing parameter, μ represents the mean of the processing parameter, LSL represents the lower limit of the processing parameter, and σ represents the standard deviation of the processing parameter. This formula assesses process stability and the ability to meet quality requirements by quantifying the degree of matching between the fluctuation of processing parameters in the actual production process and the preset acceptable range. The distance between the mean of the processing parameter and the upper and lower limits of the processing parameter is calculated using the "USL-μ" and "μ-LSL" terms, respectively. Then, these distances are divided by 3 times the standard deviation of the processing parameter, and the smaller of the two values ​​is taken as the process capability index. This takes into account whether the processing parameter deviates from the center and whether the fluctuation is too large. The larger the process capability index value, the stronger the process capability and the more stably products that meet the standards can be produced. Then, the average of the process capability indices corresponding to multiple processing parameters is taken." The mean value serves as the overall process capability index for this batch of battery materials. The overall process capability index measures the overall manufacturing process capability of a batch of batteries. When the overall process capability index is greater than 1.33, the upper and lower limits of the initial confidence interval are simultaneously contracted inward by 5%-10%; when the overall process capability index is less than 1.0, the upper and lower limits of the initial confidence interval are simultaneously expanded outward by 5%-15%, thus forming an adaptive confidence benchmark interval. This method, which assesses the stability of the battery production process from a process perspective, quantifies the process level through the process capability index, and then adjusts the initial confidence interval based on the process level to form the final confidence benchmark interval, allows the interval to more accurately reflect the confidence level of battery capacity under current process conditions. This addresses the problem of insufficient quantitative analysis of the relationship between process and battery quality in existing technologies.

[0093] In one embodiment, step S4, which involves obtaining the confidence evaluation interval of the battery based on the parameter-capacity mapping matrix, electrical characteristic parameters, and historical reference data, includes:

[0094] S41. Obtain the initial confidence assessment interval based on the parameter-capacity mapping matrix, process parameters, and material property parameters;

[0095] S42. Obtain the real-time charge integral, real-time static voltage attenuation rate, and real-time first charge / discharge efficiency based on the electrical characteristic parameters.

[0096] S43. Obtain multiple complete datasets based on the historical reference data, and obtain a performance impact weighting set and sensitivity coefficient set based on the complete datasets, real-time charge integral, real-time static voltage attenuation rate and real-time first charge and discharge efficiency.

[0097] S44. Obtain electrical performance fluctuation data based on the multidimensional manufacturing feature data, and obtain charge correction coefficient, on-state voltage attenuation correction coefficient and efficiency correction coefficient based on the sensitivity coefficient group and the electrical performance fluctuation data.

[0098] S45. Obtain a comprehensive correction coefficient based on the charge correction coefficient, the on-state voltage attenuation correction coefficient, the efficiency correction coefficient, and the performance impact weighting, and obtain a confidence assessment interval based on the comprehensive correction coefficient and the initial confidence assessment interval.

[0099] As described in steps S41-S45 above, this invention matches the current battery's process parameters and material characteristic parameters with a pre-constructed parameter-capacity mapping matrix. Based on the probability distribution of the current parameters in the matrix, an initial confidence assessment interval is obtained. This initial confidence assessment interval refers to the range of capacity achievement probabilities initially determined through statistical analysis by matching the key dominant parameter groups using the parameter-capacity mapping matrix. Next, real-time charge integral, real-time static voltage attenuation rate, and real-time first charge / discharge efficiency are obtained from the battery's electrical characteristic parameters. Then, a large amount of complete dataset is collected from historical reference data. This complete dataset refers to a collection of various data covering the entire battery manufacturing process, including but not limited to processing parameters (such as temperature and pressure), material parameters (such as purity and particle size), historical capacity data, process fluctuation records, and material consistency coefficients. Corresponding historical charge integral, historical static voltage attenuation rate, and historical first charge / discharge efficiency are obtained from the historical electrical characteristic parameters. Finally, these indicators are Z-score standardized (based on historical data). The influence of dimensions is eliminated by using the mean and standard deviation of historical data, and outliers exceeding ±3 times the standard deviation are removed (to ensure data reliability). This yields a preprocessed dataset of historical electrical characteristic indicators. For real-time charge integral, real-time stationary voltage decay rate, and real-time first charge / discharge efficiency, Z-score standardization is performed according to the standardization rules of historical data (i.e., using the mean and standard deviation of historical data). This results in standardized charge integral, standardized stationary voltage decay rate, and standardized first charge / discharge efficiency, which are used as model inputs. Z-score standardization eliminates the dimensional differences between different electrical characteristic indicators (such as charge integral and voltage decay rate) and removes extreme outliers, ensuring the comparability of historical and real-time data. This unifies electrical parameters with different physical meanings to the same analytical dimension, avoiding weight allocation bias caused by differences in units or numerical ranges. This addresses the problem in existing technologies where parameters with high numerical ranges (such as charge integral) excessively dominate the evaluation results when directly mixing parameters with different dimensions for calculation. This lays the data foundation for subsequent weight calculation and correction coefficient generation.

[0100] Since the charge integral directly reflects the degree of lithiation of the active material, it has the highest weight in terms of capacity. The static open-circuit voltage decay rate characterizes the stability of the SEI film and has the second highest impact on the first-cycle capacity. The initial efficiency reflects irreversible lithium loss and mainly affects the initial capacity retention rate, so its weight is relatively the lowest. This method, which combines physical mechanism analysis to determine the priority of each parameter's impact on capacity, ensures that the weight allocation conforms to the internal electrochemical laws of the battery (e.g., the charge integral is directly related to the utilization rate of the active material and is a core determinant of capacity). It avoids counterintuitive weights caused by sample bias in data-driven models (e.g., assigning excessively high weights to secondary parameters, followed by...). The processed historical electrical characteristic index dataset is used as the independent variable, and the corresponding historical measured capacity is used as the dependent variable. A multiple linear regression equation is constructed, and the regression coefficients are calculated using the least squares method. These coefficients are then standardized (divided by their respective standard deviations) to obtain standardized regression coefficients that reflect the influence of each parameter. The standardized regression coefficients are then normalized (making the sum of the weights equal to 1) to obtain the initial weight reassembly. Business constraint data is then acquired, and business constraints are applied to the initial weight reassembly based on this data: if the weight of charge is lower than 0.4 (e.g., 0.35), it is forcibly increased to 0.4, and the weights of other parameters (e.g., static voltage attenuation) are reduced proportionally. The initial efficiency is reduced from 0.3 to 0.25, while the initial efficiency remains at 0.2, ensuring a total weight of 1. When the actual impact of the settling open-circuit voltage decay rate of a certain batch exceeds the model prediction (e.g., historically, 1% decay rate corresponds to 0.3% capacity loss, but this batch actually corresponds to 0.5%), a deviation compensation value (e.g., +0.2%) is calculated, and the weight of the settling open-circuit voltage decay rate is increased from 0.3 to 0.35 (i.e., 0.3 + 0.05). The final performance impact weight reassembly is obtained, where the performance impact weight reassembly refers to a combination of values ​​reflecting the dominance of multiple parameters on performance, including the charge effect weight, the settling open-circuit voltage decay effect weight, and the initial efficiency weight. Efficiency influences weights. This correction method combines data-driven regression results with industrial practice experience, ensuring that weight allocation conforms to physical mechanisms while dynamically adapting to parameter fluctuations in the production process. Before each batch of production, the regression coefficients are recalculated and the weights are updated using the most recent 500 sets of data, and a threshold of no more than 10% is set for a single weight adjustment (to avoid drastic fluctuations). This method, which dynamically adjusts weights through business constraints and compensates for weights based on actual batch deviations, ensures that weight allocation conforms to theoretical laws and adapts to production fluctuations. It can overcome the limitations of pure algorithmic regression, incorporate industrial practice experience, and achieve "dynamic self-calibration" of weights.

[0101] Subsequently, the model was trained using standardized charge, standardized static open-circuit voltage decay rate, and standardized first-time efficiency from historical data as inputs, and capacity correction coefficient as output. HuberLoss was used as the loss function (to balance the impact of outliers and normal data). The training and test sets were divided according to the production time sequence. Time cross-validation was used to simulate the scenario of "predicting the future with past data" in real production, ensuring that the correction coefficient of the model output can make the corrected capacity error ≤1.5% (i.e., the absolute value of the difference between the model's predicted capacity and the measured capacity ≤1.5%), and the misclassification rate of critical batteries (batteries with capacity close to the qualified threshold) <0.5% (to avoid classifying unqualified batteries as qualified or vice versa). By strongly binding the model training with the production time sequence, the evaluation distortion caused by "future data leakage" is avoided.

[0102] Then, the trained model is deployed to the production line's real-time system (such as a manufacturing execution system), and automated operation is achieved through an interface with the database: During real-time calculation, the charge integral, open-circuit voltage decay rate, and initial charge / discharge efficiency of the current batch of batteries are loaded from the database. Standardized charge, open-circuit voltage decay rate, and initial efficiency are calculated according to the standardization rules of historical data (i.e., using the mean and standard deviation of historical data). Based on this trained and corrected model, the standardized charge, open-circuit voltage decay rate, and initial efficiency are input into the model, and a sensitivity coefficient set is output. This sensitivity coefficient set describes the degree of influence of fluctuations in various parameters (or parameter groups) on battery performance (such as capacity), including charge sensitivity coefficient, open-circuit voltage sensitivity coefficient, and charge / discharge efficiency sensitivity coefficient. Next, electrical performance fluctuation data is obtained based on multi-dimensional manufacturing characteristic data. This electrical performance fluctuation data refers to the actual fluctuation records of electrical performance indicators during battery manufacturing or use, including average charge integral, standard deviation of charge integral, maximum allowable open-circuit voltage decay rate, maximum efficiency, and minimum efficiency, etc., and is then calculated according to the formula… "Calculate the charge correction factor, where α represents the charge correction factor, θ1 represents the charge sensitivity factor, and Q..." int Q represents the integral of the current battery charge. avg Q represents the integral of the average charge of batteries in the same batch. std This represents the integral standard deviation of the charge of batteries in the same batch, expressed as " "The integral of the current battery charge is normalized to the standard deviation relative to the batch mean. The normalized result is multiplied by the charge sensitivity coefficient to convert it into a correction ratio for capacity; according to the formula..." "Calculate the static on-state voltage attenuation correction coefficient, where β represents the static on-state voltage attenuation correction coefficient, θ2 represents the static on-state voltage sensitivity coefficient, and V..." decay This indicates the battery's 24-hour static open-voltage degradation rate, V.max This indicates the preset maximum allowable static voltage attenuation rate, expressed through " Mapping the attenuation rate to the [0,1] interval, the larger the attenuation rate, the smaller the static open-circuit attenuation correction coefficient (minimum 0.9), indicating a greater capacity loss; according to the formula… "Calculate the efficiency correction factor, where γ represents the efficiency correction factor, θ3 represents the efficiency sensitivity factor, and ε represents the initial charge / discharge efficiency of the current battery." max ε represents the maximum efficiency. min This formula represents the minimum efficiency, achieved through " "By mapping efficiency to the [0,1] interval, the higher the efficiency, the larger the efficiency correction coefficient (up to 1.15), which indicates a higher reversible capacity. The normalized efficiency is converted into a correction ratio for the capacity. This conversion of the absolute value of the electrical characteristic parameter into a correction ratio relative to the batch average realizes the "interpretable calculation" of the correction coefficient. It solves the problem that the correction of electrical characteristic fluctuations in the existing technology relies on empirical judgment and lacks mathematical basis, thus making the correction under critical state more objective."

[0103] Finally, a weighted summation method is used to calculate the comprehensive correction coefficient based on the charge correction coefficient, the static on-state voltage attenuation correction coefficient, and the efficiency correction coefficient, as well as the corresponding weights for the influence of charge, static on-state voltage attenuation, and initial efficiency. This comprehensive correction coefficient is then calculated according to the formula… Calculate the lower limit of the confidence assessment interval, where E L E represents the lower limit of the confidence assessment interval. init-L This represents the lower limit of the initial confidence assessment interval, and C represents the comprehensive correction coefficient, according to the formula " "The upper limit of the confidence assessment interval is calculated, where E..." U E represents the upper limit of the confidence assessment interval. init-UHere, C represents the upper limit of the initial confidence assessment interval, and C represents the comprehensive correction coefficient. This calculation formula uses the initial confidence assessment interval as a benchmark framework and dynamically shrinks or expands the upper and lower limits of the interval through the comprehensive correction coefficient. The comprehensive correction coefficient is calculated from the sensitivity coefficient set of electrical characteristic parameters and reflects the degree of deviation between the current battery electrical characteristics and the historical benchmark. When the electrical characteristics are better than the historical benchmark (e.g., higher charge, lower decay rate), the comprehensive correction coefficient shifts the lower limit of the interval upward and the upper limit downward (interval shrinkage), improving the assessment accuracy; conversely, it shifts the lower limit of the interval downward and the upper limit upward (interval expansion), reducing the risk of misjudgment. At the same time, the interval adjusted by the max and min functions never exceeds the initial interval range, ensuring that the assessment results reflect both the electrical characteristics and the historical benchmark. The correction of characteristic parameters avoids physical irrationality caused by over-correction (such as the lower limit of the interval being higher than the upper limit or exceeding the theoretical limit of the material), thus achieving adaptive optimization of the evaluation interval. Finally, the calculated lower limit and upper limit of the interval together constitute the final evaluation interval, thereby forming a confidence evaluation interval that is both corrected and conforms to the basic theoretical constraints. This method, which optimizes the evaluation interval using real-time electrical characteristic parameters and avoids over-correction through theoretical constraints, achieves a balance between "local correction + global constraints." It is beneficial to solve the problem in the existing technology where fluctuations in a single parameter under critical conditions cause drastic changes in the interval and trigger decision-making abrupt changes. It makes the confidence evaluation interval present a "smooth transition" within the critical range, reducing extreme decisions caused by small fluctuations.

[0104] In one embodiment, step S5, which involves obtaining the battery's input disturbance coefficient and quality risk index based on the process parameters, environmental equipment data, and electrical characteristic parameters, includes:

[0105] S51. Obtain process setting values ​​based on the process parameters, wherein the process setting values ​​include coating speed, roller gap and slitting tension, and obtain real-time monitoring values ​​based on the environmental equipment data, wherein the real-time monitoring values ​​include oven temperature fluctuation variance, hydraulic pressure deviation and vibration spectrum energy.

[0106] S52. Obtain a process stability feature vector based on the process setpoint and the real-time monitoring value, and obtain a quality risk feature vector based on the electrical characteristic parameters;

[0107] S53. Obtain multiple environmental equipment noise components based on the process stability feature vector and the quality risk feature vector, and obtain the lithium-ion intercalation entropy change and the static open-voltage relaxation constant based on the quality risk feature vector;

[0108] S54. Obtain the input disturbance coefficient based on the multiple environmental equipment noise components and the static open-voltage relaxation constant, and obtain the quality risk index based on the lithium-ion intercalation entropy change and the static open-voltage relaxation constant.

[0109] As described in steps S51-S54 above, this invention extracts coating speed, roller gap, and slitting tension from process parameters. These process settings refer to the standard values ​​of key process parameters set manually during battery production, directly reflecting controllable variables in the production process. Their stability plays a decisive role in the physical properties of the electrode sheets (such as coating uniformity and compaction density). Simultaneously, it obtains oven temperature fluctuation variance, hydraulic pressure deviation, and vibration spectrum energy from environmental equipment data. These real-time monitoring values ​​are indicators collected in real-time from environmental equipment data, reflecting the equipment operating status and workshop environmental fluctuations. They are direct manifestations of the equipment operating status and workshop environmental fluctuations and can serve as real-time "health indicators" of process stability. Through these data, the three-dimensional data system of "input-state-output" of the production process can be integrated, breaking through the limitations of existing technologies that only focus on data from a single link. Then, based on the battery production sequence, the process settings and real-time monitoring values ​​are synchronized at the millisecond level. For example, the coating speed is synchronized with the corresponding time... The oven temperature is precisely matched to eliminate errors caused by data acquisition delays. Then, a multi-head attention mechanism dynamically assigns weights to each parameter of the synchronized multi-source data. When a parameter becomes abnormal (e.g., abnormal roller gap), its weight increases accordingly (e.g., from 0.3 to 0.7). This ultimately generates a 128-dimensional process stability feature vector. The 128 dimensions of this vector correspond to different aspects of process stability during production. For example, dimensions 1-20 reflect the stability of the coating process (including coating speed fluctuations and coating thickness uniformity), dimensions 21-40 reflect the stability of the rolling process (covering roller gap stability and pressure uniformity), dimensions 41-60 correspond to the stability of the slitting process (involving slitting tension fluctuations and cut smoothness), and the remaining dimensions comprehensively reflect the impact of environmental factors on the process. The value of each dimension represents the level of stability in that aspect. Its mathematical expression is: P = Attention(W q X, W k X, W v X), where P represents the process stability feature vector, X represents the aligned multi-source data matrix, and W q X, W k X, W v X represents the learnable parameter. By combining "time synchronization" with "dynamic weights", the feature vector is focused on the parameters that truly affect stability, thus solving the problem of "abnormal signals being submerged" caused by fixed parameter weights in existing technologies. In particular, it can highlight the fluctuation impact of key parameters in critical states.

[0110] The real-time charge integral, real-time static voltage decay rate, and real-time first charge / discharge efficiency extracted from the electrical characteristic parameters directly reflect the quality of the battery's internal electrochemical behavior, such as the SEI film formation quality and interface stability. These parameters are then input into a Generative Adversarial Network (GAN). The encoder extracts latent features (i.e., abstract information extracted from the electrical characteristic parameters that cannot be directly observed but reflects the battery's essential characteristics, such as features reflecting the internal chemical reaction patterns and material performance). The discriminator forces these features to be in the same distribution as the process stability feature vector (i.e., the aforementioned 128-dimensional process stability feature vector) (ensuring that both maintain consistent data distribution patterns for effective comparison and correlation analysis). Finally, a 128-dimensional quality risk feature vector is output. In this process, feature decoupling is achieved by adding specific constraints during model training to control quality risk. The first 64 dimensions of the feature vector are strongly correlated with material property parameters to reflect the intrinsic properties of the material, while the last 64 dimensions are strongly correlated with process parameters to reflect process defects. Simultaneously, when the static voltage attenuation rate is >0.5mV / h, L1 regularization constraints (a constraint mechanism) are applied to the abstract data space where the encoder extracts latent features. This makes the anomaly-related features in the latent space more prominent, enhancing the sensitivity to anomaly data. This method, which forces the distribution of process features and electrical property features to be consistent through the GAN network, achieves effective cross-dimensional data correlation, distinguishes material and process risks through feature decoupling, and enhances anomaly sensitivity through L1 regularization, breaks through the bottleneck of "separation of process and electrical property parameters" in existing technologies. It achieves accurate location of risk sources and solves the problem of misjudging critical states due to the inability to distinguish between "intrinsic material risks" and "process defect risks" in existing technologies.

[0111] Next, a process-electrical characteristic correlation matrix is ​​constructed. This is achieved by transposing the 128-dimensional process stability feature vector (rows to columns) and performing matrix multiplication with the 128-dimensional quality risk feature vector, resulting in a 128×128 matrix. Each element in this matrix represents the correlation strength between process stability and quality risk features. Then, singular value decomposition is used to extract irrelevant components (a part of the matrix decomposition representing information unrelated to both process and quality risk features, primarily mechanical vibrations during equipment operation and electromagnetic interference in the environment). The environmental equipment noise component for each process is then calculated. This component, unrelated to both process stability and quality risk, originates mainly from mechanical vibrations and electromagnetic interference. It quantifies the degree of noise interference with the process signal. For example, the environmental equipment noise component for the coating process is calculated by dividing the L2 norm of the noise component corresponding to the coating process by the L2 norm of the process stability feature vector corresponding to the coating process and 10. -6The sum and the ratio obtained are the noise energy ratio of the coating process. Essentially, it calculates the proportion of noise in the process signal. The higher the proportion, the greater the impact of noise on the process. Add a minimum value of 10 to the denominator. -6 This method is designed to prevent division by zero errors when the process signal strength is zero. It transforms abstract noise into a calculable "energy percentage", making the impact of noise "unknown" "quantifiable". This solves the problem in existing technologies where "parameter fluctuations caused by noise" are misjudged as "real quality differences" due to ignoring noise interference.

[0112] Then, the differential entropy change of the charge-discharge curve is analyzed from the risk feature vector. Specifically, features related to the charge-discharge curve in the risk feature vector (these features are learned and extracted by the model from the charge-discharge curves contained in the electrical characteristic parameters) are used. Combined with thermodynamic principles, the variation law of the charge-discharge curve is first extracted, and then the differential entropy change of the curve is calculated to obtain the lithium-ion intercalation entropy change. The lithium-ion intercalation entropy change refers to the change in the degree of disorder of the system when lithium ions are intercalated into the electrode material, and the normal range is 0.2~0.5 J / mol·K. Then, the static voltage decay curve is fitted by the risk feature vector, that is, the feature related to the static voltage decay curve in the mass risk feature vector is used. The voltage decay-related features (also learned and extracted by the model from the resting voltage decay curve included in the electrical characteristic parameters) are matched with the actual data of voltage decay over time during rest using mathematical curves (such as exponential curves). By adjusting the curve parameters to make the curve as close to the actual data as possible, the resting voltage relaxation constant is calculated based on the fitted curve equation. The resting voltage relaxation constant reflects the rate of battery voltage decay during rest. This method of converting the electrical characteristic curve into thermodynamic parameters (entropy change) and kinetic parameters (relaxation constant) achieves a leap from "observing the curve shape" to "quantifying the mechanism parameters," and then uses the formula… Calculate the input disturbance coefficient, where K noise Represents the input disturbance coefficient, max(φ) noise The maximum noise energy ratio in each process is represented by , t represents the process memory period (default 8h), and τ represents the static pressure relaxation constant. As time progresses, the noise's impact gradually weakens. Multiplying the maximum noise impact by the time decay result gives the actual disturbance level of the noise on the model input at the current moment, expressed by the formula "". "Calculate the quality risk index, where R..." riskThe formula represents the quality risk index, where ΔS represents the entropy change of lithium-ion intercalation, t represents the process memory period (default 8h), and τ represents the static voltage relaxation constant. This formula comprehensively considers the disorder of the SEI film and the voltage relaxation rate to assess the battery quality risk. The larger the entropy change of lithium-ion intercalation (indicating a more disordered SEI film) and the smaller the static voltage relaxation constant (indicating faster voltage decay), the closer the overall score is to 1, indicating a higher intrinsic quality risk. The formula quantifies the noise impact (input disturbance coefficient) and intrinsic risk (quality risk index), achieving a two-dimensional assessment of "data reliability - quality risk" to solve the problem of over-testing or under-testing caused by fixed evaluation standards in existing technologies.

[0113] In one embodiment, step S6, which involves obtaining the dynamic skip confidence level of the battery based on the confidence level assessment interval, the confidence level benchmark interval, the input perturbation coefficient, and the quality risk index, includes:

[0114] S61. Obtain the boundary proximity determination threshold based on the historical reference data, and obtain the interval relationship identifier of the battery based on the confidence evaluation interval, the confidence benchmark interval and the boundary proximity determination threshold;

[0115] S62. Obtain the evaluation interval analysis strategy based on the interval relationship identifier, and obtain the correction coefficient based on the evaluation interval analysis strategy, the confidence evaluation interval and the confidence benchmark interval, wherein the correction coefficient includes the overlap ratio correction coefficient and the boundary distance correction coefficient.

[0116] S63. Obtain influencing factor data, and obtain noise suppression parameters and risk sensitivity parameters based on the influencing factor data;

[0117] S64. Obtain the baseline confidence level based on the confidence level assessment interval, and obtain the dynamic skip confidence level based on the baseline confidence level, correction coefficient, noise suppression parameter and risk sensitivity parameter.

[0118] As described in steps S61-S64 above, this invention obtains the boundary proximity determination threshold based on historical reference data. The boundary proximity determination threshold is a critical value used to define the "boundary proximity" range. The confidence assessment interval and the confidence benchmark interval are compared. Simultaneously, using the boundary proximity determination threshold as a reference, the interval relationship identifier of the battery is obtained. The interval relationship identifier is a classification label used to characterize the positional relationship between the confidence assessment interval and the confidence benchmark interval. The interval relationship identifier is divided into the following four categories: Complete Coverage (Category A): This indicates that the confidence assessment interval is completely covered by the confidence benchmark interval, that is, the lower limit of the confidence assessment interval is greater than or equal to the lower limit of the confidence benchmark interval. The confidence assessment interval has a lower limit, and the upper limit of the confidence assessment interval is less than or equal to the upper limit of the confidence benchmark interval. In this case, the two intervals have an inclusive relationship, and the overall confidence assessment interval is within the range of the confidence benchmark interval. Partial overlap (Type B): The confidence assessment interval and the confidence benchmark interval partially overlap but do not completely contain each other. That is, part of the confidence assessment interval falls within the confidence benchmark interval, and another part falls outside the confidence benchmark interval. The two intervals have overlapping widths, but also each has non-overlapping parts, showing a state of cross-over overlap. Complete separation (Type C): The confidence assessment interval and the confidence benchmark interval have no overlapping parts and are completely independent of each other. The upper limit of the confidence assessment interval is less than the lower limit of the confidence benchmark interval (the confidence assessment interval is to the left of the confidence benchmark interval), or the lower limit of the confidence assessment interval is greater than the upper limit of the confidence benchmark interval (the confidence assessment interval is to the right of the confidence benchmark interval). The two intervals are separate, have no overlapping areas, and the distance between the two intervals is greater than the boundary proximity threshold. Boundary proximity (Class D): The confidence assessment interval and the confidence benchmark interval have no overlapping parts and are completely independent of each other, that is, the upper limit of the confidence assessment interval is less than the lower limit of the confidence benchmark interval (the confidence assessment interval is to the left of the confidence benchmark interval), or the lower limit of the confidence assessment interval is greater than the upper limit of the confidence benchmark interval. The confidence assessment interval (located to the right of the confidence benchmark interval) is defined as follows: the two intervals are separated from each other and have no overlapping area, but the distance between the two intervals is less than the boundary proximity judgment threshold. That is, the edges of the two intervals are close to each other, although they do not overlap but are close in distance. By clarifying four types of interval relationships, the positional relationship between the confidence assessment interval and the confidence benchmark interval is systematically classified, which breaks through the limitation of the existing technology of making a "black and white" judgment on the confidence assessment results. It also uses clear mathematical definitions to distinguish states such as "complete inclusion" and "partial overlap", especially accurately defining the difference between "boundary proximity" and "complete separation", so as to provide a clear basis for subsequent flexible processing of critical states.

[0119] The evaluation interval analysis strategy is obtained based on the interval relationship identifier. The specific evaluation interval analysis strategy is as follows: For Class C batteries, their baseline confidence level is directly set to 0, so their dynamic skip confidence level is also 0, and this type of battery will be forced to enter the capacity testing process; For Class A batteries, the median of their confidence evaluation interval is taken as the dynamic skip confidence level. The processing decision for Class A and Class C batteries achieves "efficient processing under clear conditions" and avoids over-computation.

[0120] For Class B batteries, the median of their confidence assessment interval is taken as the baseline confidence level. Preset weighting parameters are obtained based on the matching data between historical Class B battery capacity assessment results and skip decision data. Then, the formula "" is used to calculate the weighting parameters. "Calculate the overlap ratio correction factor, where ω1 represents the overlap ratio correction factor, O..." (ECI，RCI) W represents the width of the overlap between the confidence assessment interval (ECI) and the confidence baseline interval (RCI). ECI The formula represents the total width of the confidence assessment interval (ECI), where μ represents the preset weight parameter, typically ranging from 0 ≤ μ ≤ 1. "This is used to calculate the overlap ratio, which measures the degree of overlap between the confidence assessment interval and the confidence benchmark interval. The value is 1 when there is complete overlap and 0 when there is no overlap. The overlap ratio is non-linearly adjusted by the weight parameter μ: when μ=1, the correction coefficient is directly equal to the overlap ratio, and the system relies entirely on the interval overlap judgment; when μ=0, the correction coefficient is always 1, and the system completely ignores the interval overlap. This design allows the system to dynamically adjust the confidence in battery quality according to the degree of interval overlap, and achieve accurate decision-making in the capacity grading process: the more overlap there is, the larger the correction coefficient is, and the more likely the battery is to skip the process, thereby optimizing production efficiency while ensuring quality. This decision-making for B-class batteries considers both interval matching degree and historical experience, and uses non-linear adjustment to balance "quality assurance" and "production efficiency", avoiding extreme decisions for partially overlapping batteries, so as to solve the problem of over-testing or under-testing caused by the one-size-fits-all strategy of existing technology, especially to achieve a smooth transition in the critical overlap state."

[0121] For Class D batteries, the median of their confidence assessment interval is taken as the baseline confidence level. The baseline confidence level refers to the starting point for calculating the dynamic skip confidence level, reflecting the initial level of confidence that the battery can skip the capacity testing process without considering factors such as interval relationship correction, noise, and risk. This is expressed by the formula "...". Calculate the boundary distance correction factor, where ω2 represents the boundary distance correction factor, and L eval L represents the lower limit of the confidence assessment interval (ECI). base This represents the lower limit of the confidence baseline interval (RCI). This represents the boundary proximity threshold, used to define the range of "boundary proximity." The formula quantifies the impact of boundary proximity on battery quality assessment by measuring the distance between the lower limit of the confidence assessment interval and the lower limit of the confidence benchmark interval. When the absolute difference between the two lower limits is "...", the boundary proximity threshold is determined. "The smaller the value, the closer the confidence assessment interval is to the confidence benchmark interval on the left boundary. In this case, the closer the boundary distance correction coefficient is to 1, indicating a more reliable battery quality and a higher confidence level in skipping the capacity testing process. Conversely, if the difference exceeds the preset boundary proximity threshold, the boundary distance correction coefficient begins to decrease significantly. When the difference equals the preset boundary proximity threshold, the boundary distance correction coefficient drops to 0, meaning the battery quality risk is high and it must be forced into the capacity testing process. The preset boundary proximity threshold acts as a 'sensitivity regulator': a larger threshold makes the system more tolerant of boundary distances, allowing more batteries to skip capacity testing; a smaller threshold..." A small boundary proximity threshold makes the system more sensitive, ensuring that batteries near the quality boundary undergo rigorous testing. This design mechanism mathematically quantifies the fuzzy concept of "boundary proximity," transforming the positional relationship of intervals into a calculable confidence correction factor. This enables refined decision control over the battery grading process. This processing decision for Class D batteries transforms the fuzzy concept of "boundary proximity" into a calculable correction coefficient. By dynamically adjusting the confidence level through the comparison of distance and threshold, batteries that are "close to the baseline interval but do not overlap" are reasonably evaluated. This solves the coarse decision-making problem in existing technologies where batteries near the boundary are either completely passed or completely blocked.

[0122] For Class B and Class D batteries, noise suppression parameters and risk sensitivity parameters are first obtained based on influencing factor data. Influencing factor data refers to various relevant data affecting battery quality assessment and capacity grading decisions, including the noise level of the model input, the quality risk characteristics of the battery itself, and fluctuations in the production environment. Noise suppression parameters are used to reduce the impact of input disturbance coefficients on the dynamic skip confidence level. Risk sensitivity parameters are used to amplify the impact of the quality risk index on the dynamic skip confidence level. Finally, the parameters are obtained using the formula… "Calculate the dynamic skip confidence level, where C" skip Indicates dynamically skipping confidence levels, C base K represents the baseline confidence level, ω represents the correction coefficient, calculated from the interval relationship (B / D class), and K... noise R represents the input disturbance coefficient. risk This represents the quality risk index, and m represents the noise suppression parameter. This represents the m-th power of the input perturbation coefficient, where k represents the risk-sensitive parameter. This formula, representing the k-th power of the quality risk index, integrates four factors—benchmark confidence, interval relationship correction, model noise, and ontological risk—through a product mechanism to achieve conservative decision quantification under multidimensional constraints. Using the benchmark confidence as the base value, the initial confidence is adjusted using the interval relationship correction coefficient. To address the uncertainty of the model input, the noise coefficient is attenuated by exponential operation using a noise suppression parameter, ensuring a reduction in skip confidence when data is unreliable. Simultaneously, the quality risk index is amplified by exponential operation, significantly reducing the skip confidence of high-risk batteries. Finally, through the product of the four factors, any adverse factor in any dimension (high noise or high risk) will lead to a decrease in the overall skip confidence, thereby ensuring the battery's performance. Optimizing the efficiency of the capacity testing process under the premise of quality, and achieving a scientific balance between data reliability, interval matching degree and quality risk, a high-risk mandatory interception step is then executed. This method systematically correlates scattered influencing factors, so that minor parameter fluctuations only cause a gradual change in confidence level, avoiding decision jumps. In addition, if the quality risk index is greater than the preset risk threshold, the dynamic skip confidence level of the battery is forcibly set to 0, forming a "last line of defense". This method of adding an absolute risk interception mechanism in addition to comprehensive evaluation prioritizes quality requirements and solves the problem that comprehensive calculation may cover up high risks and lead to missed detection, especially for batteries that are close to qualified but have extremely high inherent risks, achieving effective interception.

[0123] This application also provides a confidence assessment system for dynamically skipping capacity-limiting processes, including:

[0124] The data acquisition module is used to acquire multi-dimensional manufacturing characteristic data and historical reference data of the battery. The multi-dimensional manufacturing characteristic data includes process parameters, material property parameters, electrical property parameters and environmental equipment data.

[0125] The matrix construction module is used to obtain the parameter-capacity mapping matrix of the battery based on the process parameters, material property parameters and historical reference data;

[0126] The benchmark analysis module is used to obtain the battery's design threshold and confidence benchmark interval based on the multidimensional manufacturing feature data and the parameter-capacity mapping matrix;

[0127] The dynamic evaluation module is used to obtain the confidence evaluation range of the battery based on the parameter-capacity mapping matrix, electrical characteristic parameters and historical reference data;

[0128] The risk diagnosis module is used to obtain the battery's input disturbance coefficient and quality risk index based on the process parameters, environmental equipment data, and electrical characteristic parameters.

[0129] The comprehensive evaluation module is used to obtain the dynamic skip confidence level of the battery based on the confidence level evaluation interval, the confidence level benchmark interval, the input perturbation coefficient, and the quality risk index.

[0130] The capacity sizing decision module is used to determine whether the dynamic skip confidence level is greater than the design threshold.

[0131] If the dynamic skip confidence level is greater than the design threshold, then the battery is determined to be allowed to skip the capacity testing process;

[0132] If the dynamic skip confidence level is not greater than the design threshold, then the battery is determined to be not allowed to skip the capacity testing process.

[0133] In one embodiment, the risk diagnosis module includes:

[0134] The data acquisition unit is used to acquire process setting values ​​based on the process parameters, wherein the process setting values ​​include coating speed, roller gap and slitting tension, and to acquire real-time monitoring values ​​based on the environmental equipment data, wherein the real-time monitoring values ​​include oven temperature fluctuation variance, hydraulic pressure deviation and vibration spectrum energy.

[0135] The feature extraction unit is used to obtain a process stability feature vector based on the process setpoint and the real-time monitoring value, and to obtain a quality risk feature vector based on the electrical characteristic parameters.

[0136] The noise analysis unit is used to obtain multiple environmental equipment noise components based on the process stability feature vector and the quality risk feature vector, and to obtain the lithium-ion intercalation entropy change and the static open-voltage relaxation constant based on the quality risk feature vector.

[0137] The risk quantification unit is used to obtain the input disturbance coefficient based on the noise components of multiple environmental devices and the static open-voltage relaxation constant, and to obtain the quality risk index based on the lithium-ion intercalation entropy change and the static open-voltage relaxation constant.

[0138] The present invention also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the confidence assessment method for dynamically skipping the capacity-sharing process described above.

[0139] The present invention also provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of the confidence assessment method for dynamically skipping the capacity-sharing process described above.

[0140] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in this application and in the embodiments can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual-speed SDRAM (SSRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0141] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, apparatus, article, or method that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, apparatus, article, or method. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, apparatus, article, or method that includes that element.

[0142] The above description is merely a preferred embodiment of the present invention and does not limit the patent scope of the present invention. Any equivalent structural or procedural transformations made based on the content of the present invention's specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of the present invention.

Claims

1. A confidence assessment method for dynamic skip of a filling procedure, characterized in that, The method comprises the following steps: acquiring multi-dimensional manufacturing feature data and historical reference data of the battery, wherein the multi-dimensional manufacturing feature data comprises process processing parameters, material characteristic parameters, electrical characteristic parameters and environmental equipment data; acquiring a parameter-capacity mapping matrix of the battery according to the process processing parameters, the material characteristic parameters and the historical reference data; acquiring a design threshold and a confidence benchmark interval of the battery according to the multi-dimensional manufacturing feature data and the parameter-capacity mapping matrix; acquiring a confidence evaluation interval of the battery according to the parameter-capacity mapping matrix, the electrical characteristic parameters and the historical reference data; acquiring an input disturbance coefficient and a quality risk index of the battery according to the process processing parameters, the environmental equipment data and the electrical characteristic parameters; acquiring a dynamic skip confidence of the battery according to the confidence evaluation interval, the confidence benchmark interval, the input disturbance coefficient and the quality risk index; judging whether the dynamic skip confidence is greater than the design threshold; if the dynamic skip confidence is greater than the design threshold, determining that the battery is allowed to skip the capacity grading process; if the dynamic skip confidence is not greater than the design threshold, determining that the battery is not allowed to skip the capacity grading process; the step of acquiring the design threshold and the confidence benchmark interval of the battery according to the multi-dimensional manufacturing feature data and the parameter-capacity mapping matrix comprises: acquiring reference sample data according to the historical reference data, and acquiring a confidence initial interval of the battery according to the reference sample data and the parameter-capacity mapping matrix; acquiring material fluctuation data and process fluctuation data according to the multi-dimensional manufacturing feature data, and acquiring an overall consistency coefficient of the battery according to the material characteristic parameters and the material fluctuation data; acquiring a historical minimum consistency coefficient and a consistency coefficient median according to the reference sample data, and acquiring the design threshold according to the historical minimum consistency coefficient, the consistency coefficient median, the overall consistency coefficient and the confidence initial interval; acquiring an overall process capability index of the battery according to the process processing parameters and the process fluctuation data, and acquiring the confidence benchmark interval according to the confidence initial interval, the process fluctuation data and the overall process capability index.

2. The confidence evaluation method of dynamically skipping the sub- capacity process according to claim 1, characterized in that, the step of acquiring the parameter-capacity mapping matrix of the battery according to the process processing parameters, the material characteristic parameters and the historical reference data comprises: acquiring core processing parameters according to the process processing parameters, wherein the core processing parameters comprise a coating surface density, a post-rolling electrode thickness and a baking temperature, and acquiring core material parameters according to the material characteristic parameters, wherein the core material parameters comprise a positive electrode slurry viscosity, an active material median diameter and a separator porosity; acquiring historical battery capacity and a historical confidence interval according to the historical reference data; acquiring a capacity influence weight of each parameter according to the historical confidence interval, the core processing parameters and the material characteristic parameters; arranging and combining a plurality of the core processing parameters and a plurality of the material characteristic parameters to obtain an interaction parameter group, and acquiring an interaction effect heat map according to the interaction parameter group; According to the capacity influence weight and the interaction effect heat map, a plurality of the core processing parameters and a plurality of the material characteristic parameters are screened to obtain key dominant parameters, and a parameter-capacity mapping matrix is obtained according to the key dominant parameters and the historical battery capacity.

3. The confidence evaluation method of dynamically skipping the binning process according to claim 1, wherein, The step of obtaining a confidence evaluation interval of the battery according to the parameter-capacity mapping matrix, the electrical characteristic parameter and the historical reference data comprises: According to the parameter-capacity mapping matrix, the process processing parameter and the material characteristic parameter, an initial confidence evaluation interval is obtained. According to the electrical characteristic parameter, a real-time charge amount integral, a real-time open-circuit voltage decay rate and a real-time first charge-discharge efficiency are obtained. According to the historical reference data, a plurality of complete data sets are obtained, and a performance influence weight group and a sensitivity coefficient group are obtained according to the complete data sets, the real-time charge amount integral, the real-time open-circuit voltage decay rate and the real-time first charge-discharge efficiency. According to the multi-dimensional manufacturing feature data, electrical performance fluctuation data are obtained, and a charge amount correction coefficient, an open-circuit voltage decay correction coefficient and an efficiency correction coefficient are obtained according to the sensitivity coefficient group and the electrical performance fluctuation data. According to the charge amount correction coefficient, the open-circuit voltage decay correction coefficient, the efficiency correction coefficient and the performance influence weight group, a comprehensive correction coefficient is obtained, and a confidence evaluation interval is obtained according to the comprehensive correction coefficient and the initial confidence evaluation interval.

4. The confidence evaluation method of dynamically skipping the sub- capacity process according to claim 1, wherein, The step of obtaining an input disturbance coefficient and a quality risk index of the battery according to the process processing parameter, the environmental equipment data and the electrical characteristic parameter comprises: According to the process processing parameter, process setting values are obtained, wherein the process setting values comprise coating speed, roll gap and slitting tension, and according to the environmental equipment data, real-time monitoring values are obtained, wherein the real-time monitoring values comprise oven temperature fluctuation variance, hydraulic pressure deviation and vibration frequency spectrum energy. According to the process setting values and the real-time monitoring values, a process stability feature vector is obtained, and according to the electrical characteristic parameter, a quality risk feature vector is obtained. According to the process stability feature vector and the quality risk feature vector, a plurality of environmental equipment noise components are obtained, and according to the quality risk feature vector, a lithium ion intercalation entropy change and an open-circuit voltage relaxation constant are obtained. According to the plurality of environmental equipment noise components and the open-circuit voltage relaxation constant, an input disturbance coefficient is obtained, and according to the lithium ion intercalation entropy change and the open-circuit voltage relaxation constant, a quality risk index is obtained.

5. The confidence evaluation method of dynamically skipping the sub- capacity process according to claim 1, wherein, The step of obtaining a dynamic skip confidence of the battery according to the confidence evaluation interval, a confidence benchmark interval, the input disturbance coefficient and the quality risk index comprises: According to the historical reference data, a boundary proximity judgment threshold is obtained, and according to the confidence evaluation interval, the confidence benchmark interval and the boundary proximity judgment threshold, an interval relationship identifier of the battery is obtained. According to the interval relationship identifier, an evaluation interval analysis strategy is obtained, and according to the evaluation interval analysis strategy, the confidence evaluation interval and the confidence benchmark interval, a correction coefficient is obtained, wherein the correction coefficient comprises an overlap proportion correction coefficient and a boundary distance correction coefficient. acquire influence factor data, and acquire noise suppression parameters and risk sensitivity parameters according to the influence factor data; acquire a reference confidence according to the confidence evaluation interval, and acquire a dynamic skip confidence according to the reference confidence, the correction coefficient, the noise suppression parameters, and the risk sensitivity parameters.

6. A confidence assessment system for dynamic skip of a filling procedure, characterized in that The method comprises the following steps: a data acquisition module is configured to acquire multi-dimensional manufacturing feature data and historical reference data of the battery, wherein the multi-dimensional manufacturing feature data comprises process processing parameters, material characteristic parameters, electrical characteristic parameters, and environmental equipment data; a matrix construction module is configured to acquire a parameter-capacity mapping matrix of the battery according to the process processing parameters, the material characteristic parameters, and the historical reference data; a reference analysis module is configured to acquire a design threshold and a confidence reference interval of the battery according to the multi-dimensional manufacturing feature data and the parameter-capacity mapping matrix; a dynamic evaluation module is configured to acquire a confidence evaluation interval of the battery according to the parameter-capacity mapping matrix, the electrical characteristic parameters, and the historical reference data; a risk diagnosis module is configured to acquire an input disturbance coefficient and a quality risk index of the battery according to the process processing parameters, the environmental equipment data, and the electrical characteristic parameters; a comprehensive evaluation module is configured to acquire a dynamic skip confidence of the battery according to the confidence evaluation interval, the confidence reference interval, the input disturbance coefficient, and the quality risk index; a capacity decision module is configured to determine whether the dynamic skip confidence is greater than the design threshold; if the dynamic skip confidence is greater than the design threshold, it is determined that the battery is allowed to skip the capacity sorting process; if the dynamic skip confidence is not greater than the design threshold, it is determined that the battery is not allowed to skip the capacity sorting process.

7. The confidence evaluation system of dynamically skipping the binning procedure according to claim 6, characterized in that, The risk diagnosis module comprises: a data acquisition unit is configured to acquire process setting values according to the process processing parameters, wherein the process setting values comprise coating speed, roll gap, and slitting tension, and to acquire real-time monitoring values according to the environmental equipment data, wherein the real-time monitoring values comprise oven temperature fluctuation variance, hydraulic pressure deviation, and vibration frequency spectrum energy; a feature extraction unit is configured to acquire a process stability feature vector according to the process setting values and the real-time monitoring values, and to acquire a quality risk feature vector according to the electrical characteristic parameters; a noise analysis unit is configured to acquire a plurality of environmental equipment noise components according to the process stability feature vector and the quality risk feature vector, and to acquire lithium ion intercalation entropy change and open-circuit voltage relaxation constant according to the quality risk feature vector; a risk quantification unit is configured to acquire an input disturbance coefficient according to the plurality of environmental equipment noise components and the open-circuit voltage relaxation constant, and to acquire a quality risk index according to the lithium ion intercalation entropy change and the open-circuit voltage relaxation constant.

8. A computer device comprising a memory and a processor, the memory storing a computer program, characterized in that, The processor executes the computer program to realize the steps of the method of any one of claims 1 to 5.

9. A computer-readable storage medium having stored thereon a computer program, characterized in that, The computer program is executed by the processor to realize the steps of the method of any one of claims 1 to 5.

Images