A method for active remanufacturing time domain based on power battery cycle aging degradation decision
The method uses deep residual shrinkage and radial basis networks with twin data fusion to address the challenge of determining power battery remanufacturing time, enhancing recycling accuracy and extending battery life.
Patent Information
- Application Number
- CN202311594059.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2023-11-27
- Publication Date
- 2024-12-24
- Estimated Expiration
- 2043-11-27
AI Technical Summary
Conventional methods struggle to accurately determine the life cycle and remanufacturing time domain of power batteries due to complex interactions among electrochemical, temperature, and stress fields, leading to difficulties in analyzing cyclic aging and degradation mechanisms, especially for large-scale retired batteries.
A method involving quantitative characterization of power battery performance degradation using deep residual shrinkage networks and radial basis networks, coupled with twin data fusion and gradient descent algorithms, to determine the upper and lower limits of the active remanufacturing time domain.
This approach provides accurate decision-making for power battery remanufacturing, extending the service life and predicting the remanufacturing time domain by analyzing cell inconsistencies and mapping performance degradation states, thus improving recycling efficiency.
Smart Images
Figure CN117907833B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of power batteries, and in particular to a method for actively remanufacturing time domain based on power battery cycle aging degradation decision. Background Art
[0002] With the popularization of new energy vehicles, the application of power batteries has also become popular; power batteries are mainly lithium batteries, and their life cycle is limited. When the power battery life cycle reaches the end, the power battery needs to be recycled. Conventional recycling methods include cascade utilization, recycling and remanufacturing.
[0003] Among them, cascade utilization is mainly used for power batteries in the early stage of their life cycle, while recycling is used for power batteries entering the end stage of their life cycle, and remanufacturing is used for power batteries between the early and late stages of their life cycle. In order to more accurately grasp the life cycle of power batteries, improve the level of power battery recycling, effectively solve the time domain decision-making difficulties of active remanufacturing of power batteries, and improve the level of power battery recycling, it is necessary to accurately judge the relationship between the life cycle of power batteries and their service cycle.
[0004] In the existing technology, conventional detection methods are unable to cope with large-scale retired power batteries, and the aging and degradation process of power batteries is affected by multiple physical field parameters such as electrochemical field, temperature field and stress field. The coupling relationship between the multiple physical field parameters dynamically evolves under the cyclic alternation of three operating conditions: dynamic charging and discharging, constant charging and static. As a result, the power battery capacitance degradation rate presents a staged sudden change, making it difficult to analyze the cyclic aging and degradation mechanism of power batteries.
[0005] Therefore, to address the above issues, a method for active remanufacturing time domain based on power battery cycle aging degradation decision is proposed. Summary of the invention
[0006] In order to make up for the deficiencies of the prior art and solve at least one of the above problems, the present invention proposes a method for actively remanufacturing time domain based on power battery cycle aging degradation decision.
[0007] A method for active remanufacturing time domain based on power battery cycle aging degradation decision, the method comprising the following steps:
[0008] S1: Quantitative characterization of power battery performance degradation state based on cell inconsistency;
[0009] S2: Decision-making in the active remanufacturing time domain of power batteries based on the fusion of analytical model and twin data;
[0010] Among them, the quantitative characterization of the power battery performance degradation state based on the inconsistency of the cells includes building a deep residual shrinkage network to extract the power battery twin data, then passing the data to the quantization system for quantization, and finally quantifying the key parameters in the power battery twin data through the gradient descent algorithm;
[0011] Among them, the decision-making of the active remanufacturing time domain of power batteries by integrating analytical models and twin data includes constructing a radial basis network, extracting twin data, and characterizing the quantitative key parameters through analytical models to jointly decide the upper and lower limits of the active remanufacturing time domain.
[0012] Preferably, in S1, the deep residual shrinkage network includes a residual module, a soft threshold function and an attention mechanism, wherein the soft threshold function is used to perform noise reduction on the data; the attention mechanism is used to generate a weight coefficient and is used for weighting each feature channel.
[0013] Preferably, in S1, the gradient descent algorithm is expressed as follows:
[0014]
[0015] In the formula, f(x k +αd k ) is the function value at the current position, f(x k ) is the function value of the previous position, and the new position is smaller than the original position α is the learning rate. The number of iterations can be reduced by adjusting the preset expected value c.
[0016] Preferably, in S2, the decision in the time domain of active remanufacturing of power batteries by fusion of analytical model and twin data can quantify the performance degradation curve of power batteries based on the quantitative characterization of the performance degradation state of power batteries, and record the life cycle of power batteries as stable period, sharp period and scrap period according to the turning points of the curve.
[0017] Preferably, in S2, the upper limit of the active remanufacturing time of the power battery is calculated based on the service value per unit time. For measurement purposes, The parsing model is as follows:
[0018]
[0019]
[0020] Where P(t) is the analytical formula of the performance degradation curve, T is the new life cycle started after active remanufacturing, P′(t) is the analytical formula of the performance degradation curve under the new life cycle, and C(t) is the other value consumed by active remanufacturing, including inspection, disassembly and reassembly.
[0021] Preferably, in S2, the radial basis network is to find a surface that best matches the problem to be solved in the high-dimensional space, and then use the resulting surface to process the test data, so that the original linearly inseparable problem becomes linearly separable; the radial basis function is an important module of the radial basis network, and the output result of the network hidden layer can be expressed as:
[0022]
[0023] In the formula, x i is the monomer performance parameter vector; x k is the current monomer performance parameter vector; s is the surface area; the middle layer of the radial basis network adopts multiple hidden layers, and the activation function takes the distance between the input vector and the weight vector as the independent variable, thereby forming a reasonable mapping mechanism between the input vector and the service time, and finally predicts the lower limit of the active remanufacturing time domain of the power battery based on the time series twin data.
[0024] The present invention is beneficial in that:
[0025] 1. The present invention analyzes the influencing factors of the active remanufacturing time domain decision-making for power batteries, and reveals the influence of the inconsistency of single cells under the multi-scale structure of power batteries on their overall performance degradation state; on the basis of extracting the performance parameter data of power battery single cells, by using the Sklar theory to map it into the overall performance degradation state of the power battery, the cycle aging degradation mechanism and the twin data fusion method are further used to make decisions on the upper and lower limits of the active remanufacturing time domain of the power battery, thereby providing an important basis for extending the service life of the power battery and dynamically predicting the active remanufacturing time domain of the power battery. BRIEF DESCRIPTION OF THE DRAWINGS
[0026] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.
[0027] Figure 1 A flowchart of an embodiment of the present invention;
[0028] Figure 2 A structural diagram of a deep residual shrinkage network in an embodiment of the present invention;
[0029] Figure 3 This is a power battery performance degradation curve diagram in an embodiment of the present invention. DETAILED DESCRIPTION
[0030] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0031] See also Figure 1-Figure 3 As shown, a method for active remanufacturing time domain based on power battery cycle aging degradation decision-making includes the following steps:
[0032] S1: Quantitative characterization of power battery performance degradation state based on cell inconsistency;
[0033] S2: Decision-making in the active remanufacturing time domain of power batteries based on the fusion of analytical model and twin data;
[0034] Among them, the quantitative characterization of the power battery performance degradation state based on the inconsistency of the cells includes building a deep residual shrinkage network to extract the power battery twin data, then passing the data to the quantization system for quantization, and finally quantifying the key parameters in the power battery twin data through the gradient descent algorithm;
[0035] Among them, the decision-making of the active remanufacturing time domain of power batteries by integrating analytical models and twin data includes constructing a radial basis network, extracting twin data, and characterizing the quantitative key parameters through analytical models to jointly decide the upper and lower limits of the active remanufacturing time domain.
[0036] As an implementation mode of the present invention, in S1, the deep residual shrinkage network includes a residual module, a soft threshold function and an attention mechanism, wherein the soft threshold function is used to perform noise reduction on the data; the attention mechanism is used to generate a weight coefficient and is used for weighting each feature channel; as shown in the attached Figure 2 As shown in a, the cuboid represents a feature map with a channel number of C, a width of W, and a height of 1; BN, ReLU, Conv, and Identity shortcut represent batch normalization, rectified linear unit activation function, convolution layer, and identity path, respectively; K represents the number of convolution kernels in the convolution layer, which is equal to the number of channels C of the feature map; the soft threshold function and its derivative are shown in Figure 2 b, the derivative can only be 0 or 1, which can reduce the noise of the data and help prevent "gradient disappearance" and "gradient explosion"; the attention mechanism is to learn a set of weight coefficients through a small network, which is used to weight each feature channel, such as Figure 2 As shown in c.
[0037] As an embodiment of the present invention, in S1, in order to accurately describe the inconsistency of the performance degradation process between cells under the multi-scale structure of the power battery and the hierarchical conduction relationship between its overall performance degradation state, the Sklar theorem is applied and the multivariate Copula function is introduced to couple the performance degradation processes of different cells of the power battery:
[0038] F(x1,x2,…,x n )=C(F1(x1),F2(x2),…,F n (x n );p)
[0039] In the formula, F l (x i ) is the power battery cell performance l about the performance parameter vector x i The marginal probability distribution function, F(x1, x2, ..., x n ) is the joint probability distribution function of the overall performance of the power battery with respect to the performance of n monomers, C(*) is the multivariate Gaussian Copula function, and ρ is the correlation coefficient to be estimated. The analytical expression of ρ can be obtained through maximum likelihood estimation. In order to efficiently obtain the accurate result of ρ, a gradient descent algorithm based on the Armijo criterion is designed to calculate a large amount of time series twin data. The gradient descent algorithm is expressed in the following formula:
[0040]
[0041] In the formula, f(x k +αd k ) is the function value at the current position, f(x k ) is the function value of the previous position, and the new position is smaller than the original position α is the learning rate. The number of iterations can be reduced by adjusting the preset expected value c.
[0042] As an embodiment of the present invention, in S2, the decision of the active remanufacturing time domain of the power battery by fusion of the analytical model and the twin data can quantify the power battery performance degradation curve based on the quantitative characterization of the power battery performance degradation state, and record the life cycle of the power battery as a stable period, a sharp period and a scrap period according to the turning point of the curve, as shown in the attached Figure 3 shown.
[0043] As an embodiment of the present invention, in S2, the upper limit of the active remanufacturing time domain of the power battery is calculated based on the service value per unit time. For measurement purposes, The parsing model is as follows:
[0044]
[0045]
[0046] Wherein, P(t) is the analytical expression of the performance degradation curve, T is the new life cycle started after active remanufacturing, P′(t) is the analytical expression of the performance degradation curve under the new life cycle, and C(t) is the other value consumed by active remanufacturing, including inspection, disassembly and reorganization. The prediction of the lower limit of the active remanufacturing time domain of power batteries depends on the effective estimation of the duration of the sharp period of performance degradation. Since the performance degradation of power batteries during the sharp period is nonlinear, a radial basis function neural network will be designed and combined with the power battery twin data and the quantitative representation of the performance degradation state for prediction.
[0047] As an implementation mode of the present invention, in S2, the radial basis network is to find a surface that best matches the problem to be solved in the high-dimensional space, and then use the resulting surface to process the test data, so that the originally linearly inseparable problem becomes linearly separable; the radial basis function is an important module of the radial basis network. Using the Gaussian function, the output result of the hidden layer of the network can be expressed as:
[0048]
[0049] In the formula, x i is the monomer performance parameter vector; x k is the current monomer performance parameter vector; s is the surface area; the middle layer of the radial basis network adopts multiple hidden layers, and the activation function takes the distance between the input vector and the weight vector as the independent variable, thereby forming a reasonable mapping mechanism between the input vector and the service time, and finally predicts the lower limit of the active remanufacturing time domain of the power battery based on the time series twin data.
[0050] Working principle: By analyzing the influencing factors of the active remanufacturing time domain decision of power batteries, the influence of the inconsistency of single cells under the multi-scale structure of power batteries on their overall performance degradation state is revealed; on the basis of extracting the performance parameter data of power battery single cells, by using the Sklar theory to map it to the overall performance degradation state of the power battery, the cycle aging degradation mechanism and twin data fusion method are further used to make decisions on the upper and lower limits of the active remanufacturing time domain of power batteries, thereby providing an important basis for extending the service life of power batteries and dynamically predicting the active remanufacturing time domain of power batteries.
[0051] In the description of this specification, the description with reference to the terms "one embodiment", "example", "specific example", etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in a suitable manner.
[0052] The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments, and the above embodiments and descriptions are only for explaining the principles of the present invention. Without departing from the spirit and scope of the present invention, the present invention may have various changes and improvements, and these changes and improvements all fall within the scope of the present invention to be protected.
Claims
1. A method for active remanufacturing time domain based on power battery cycle aging and degradation decision making, characterized by: The method comprises the following steps: S1: Construct a multi-scale digital model of power batteries; S2: Superimpose the data-driven model into the multi-scale digital model to build a multi-scale digital twin model of the power battery; S3: Based on the multi-scale digital twin model, a multi-scale equivalent circuit model of the power battery is constructed; S4: Construct a deep residual shrinkage network to extract the performance parameter data of power battery cells; S5: Based on S4, a quantitative characterization of the power battery performance degradation state with inconsistent cells is obtained; S6: Based on S5, quantify the power battery performance degradation curve, analyze the model and twin data fusion to make decisions in the active remanufacturing time domain of power batteries; Among them, the quantitative characterization of the power battery performance degradation state based on the inconsistency of the cells includes constructing a deep residual shrinkage network, extracting the power battery twin data, and then transferring the data to the quantification system for quantification. Finally, the key parameters in the power battery twin data are quantitatively characterized by the gradient descent algorithm. In order to accurately characterize the inconsistency of the performance degradation process between cells under the multi-scale structure of the power battery and the hierarchical conduction relationship between its overall performance degradation state, the Sklar theorem is applied and the multivariate Copula function is introduced to couple the performance degradation processes of different cells of the power battery: ； In the formula, For power battery monomer performance About the performance parameter vector The marginal probability distribution function of is the joint probability distribution function of the overall performance of the power battery with respect to the performance of n monomers, C(*) is the multivariate Gaussian Copula function, is the correlation coefficient to be estimated; it can be obtained by maximum likelihood estimation The analytical expression of is to efficiently obtain To obtain accurate results, we will design a gradient descent algorithm based on the Armijo criterion to calculate a large amount of time series twin data. The gradient descent algorithm is expressed in the following formula: ； In the formula is the function value at the current position, is the function value of the previous position, and the new position is smaller than the original position" ”, is the learning rate. The number of iterations can be reduced by adjusting the preset expected value c; Among them, the decision-making of the active remanufacturing time domain of power batteries by integrating analytical models and twin data includes constructing a radial basis network, extracting twin data, and characterizing the quantitative key parameters through analytical models to jointly decide the upper and lower limits of the active remanufacturing time domain.
2. The method for active remanufacturing time domain based on power battery cycle aging and degradation decision according to claim 1 is characterized by: The deep residual shrinkage network includes a residual module, a soft threshold function and an attention mechanism, wherein the soft threshold function is used to perform noise reduction on the data; the attention mechanism is used to generate a weight coefficient and is used for weighting each feature channel.
3. The method for active remanufacturing time domain based on power battery cycle aging and degradation decision according to claim 2 is characterized by: The decision-making in the time domain of active remanufacturing of power batteries by fusing the analytical model with twin data can quantify the power battery performance degradation curve based on the quantitative characterization of the power battery performance degradation state. According to the turning points of the curve, the life cycle of the power battery can be recorded as a stable period, a sharp period and a scrap period.
4. The method for active remanufacturing time domain based on power battery cycle aging and degradation decision according to claim 3 is characterized by: The upper limit of the active remanufacturing time of the power battery is expressed in terms of service value per unit time. For measurement purposes, The parsing model is as follows: ； In the formula, is the analytical expression of the performance degradation curve, T is the new life cycle started after active remanufacturing, is the analytical expression of the performance degradation curve under the new life cycle, and C(t) is the other value consumed by active remanufacturing, including inspection, disassembly and reassembly.
5. The method for active remanufacturing time domain based on power battery cycle aging degradation decision according to claim 4 is characterized in that: The radial basis network is to find a surface that best matches the problem to be solved in a high-dimensional space, and then use the resulting surface to process the test data, so that the original linearly inseparable problem becomes linearly separable; the radial basis function is an important module of the radial basis network. Using the Gaussian function, the output result of the hidden layer of the network can be expressed as: ； In the formula, is the monomer performance parameter vector; is the current monomer performance parameter vector; s is the surface area; the middle layer of the radial basis network adopts multiple hidden layers, and the activation function takes the distance between the input vector and the weight vector as the independent variable, thereby forming a reasonable mapping mechanism between the input vector and the service time, and finally predicts the lower limit of the active remanufacturing time domain of the power battery based on the time series twin data.