Power battery recovery echelon utilization optimization method and system

Abstract

The invention relates to a power battery recycling echelon utilization optimization method and system, and the method comprises the steps: obtaining a multi-dimensional state data set of a target battery, inputting the multi-dimensional state data set into a multi-channel prediction model, and enabling the multi-channel prediction model to generate and output a prediction information set; dividing the target battery to a corresponding echelon utilization performance grade based on the prediction information set and a preset grading strategy; acquiring real-time market data, and respectively executing simulation calculation; based on a simulation calculation result, taking batch residual value maximization as a decision-making target, and generating an optimization scheme corresponding to each echelon utilization performance grade; in conclusion, the prediction model is constructed by fusing the multi-dimensional state data to realize accurate evaluation, simulation calculation is executed in combination with the dynamic market data, and finally an optimization scheme with residual value maximization as a target is generated. The method has the advantages that the battery state evaluation accuracy is improved, dynamic economic decision optimization is realized, potential safety hazards are reduced, and the resource utilization efficiency is improved.

Description

Technical Field

[0001] This application relates to the technical field of battery recycling, and in particular to an optimized method and system for recycling and recycling power batteries. Background Technology

[0002] With the rapid development of the new energy vehicle industry, a large-scale retirement wave of power batteries has arrived, and their green recycling and high-value utilization have become key factors restricting the sustainable development of the industry. Currently, the industry generally adopts a combination of traditional manual experience and single performance testing for the assessment and disposal of retired batteries. Specifically, existing technologies usually rely on simple voltage and internal resistance measurements and time-consuming full-capacity tests to roughly determine the health status of the battery, and then roughly classify it into "usable" or "unusable" based on a few fixed thresholds.

[0003] However, the above-mentioned model has many drawbacks: First, its evaluation dimensions are extremely limited, and it seriously lacks the ability to perceive the deep state of the battery, such as the integrity of the internal physical structure, the evolution and consistency of the chemical composition of key materials. This leads to a large number of potential batteries being misjudged as deteriorated products and directly dismantled, while some batteries with internal defects (such as micro-short circuits and lithium plating) are classified into the category of secondary use, which poses serious safety hazards. Second, the entire evaluation process is highly dependent on the experience of the operators, which is inefficient and difficult to meet the pace requirements of large-scale recycling and processing. Summary of the Invention

[0004] To address the aforementioned shortcomings, this application provides an optimized method and system for the recycling and secondary utilization of power batteries.

[0005] The above-mentioned objective of this application is achieved through the following technical solution: An optimized method for recycling and reusing power batteries includes the following steps: A multidimensional state dataset of the target battery is obtained, which includes physical structure data, chemical composition data, and dynamic electrical property data. The multidimensional state dataset is input into a pre-trained multi-channel prediction model, which generates and outputs a prediction information set, including battery health state prediction, remaining life prediction, consistency index, and safety risk identifier. Based on the predictive information set and the pre-set grading strategy, the target battery is classified into the corresponding tiered utilization performance level; Real-time market data is acquired, and simulation calculations based on preset calculation paths are performed on target batteries at each performance level of utilization. The preset calculation paths include utilization paths and dismantling and recycling paths. Based on the simulation results, with the goal of maximizing batch residual value, optimization schemes are generated for each utilization performance level.

[0006] In a preferred embodiment, this application can be further configured as follows: the multi-channel prediction model includes a multi-channel feature extraction network, a feature fusion layer, and a task decision layer; the step of inputting a multi-dimensional state dataset into a pre-trained multi-channel prediction model, causing the multi-channel prediction model to generate and output a prediction information set, wherein the prediction information set includes battery health state prediction values, remaining life prediction values, consistency index, and safety risk identifiers, includes the following steps: The multi-channel feature extraction network extracts physical structure feature vectors, chemical composition feature vectors, and electrical property feature vectors based on the input physical structure data, chemical composition data, and dynamic electrical property data. The feature fusion layer performs cross-modal fusion on the received physical structure feature vector, chemical composition feature vector, and electrical property feature vector to generate a joint feature vector; Based on the received joint feature vector, the task decision layer calculates the predicted battery health status, predicted remaining life, consistency index, and safety risk identifier, and outputs a set of predicted information.

[0007] In a preferred embodiment, this application can be further configured as follows: the task decision layer calculates the predicted battery health status, predicted remaining life, consistency index, and safety risk identifier based on the received joint feature vector, and outputs the predicted information set, including the following steps: The task decision layer inputs the joint feature vector into its pre-set first branch network, second branch network, third branch network and fourth branch network, so that they can perform the corresponding computation tasks simultaneously, and obtain the initial prediction value of each computation task and its corresponding task confidence score. The task decision layer performs consistency verification on the initial prediction values ​​of each computation task based on a pre-defined decay mechanism rule base, and generates the final prediction values ​​of each computation task based on the consistency verification results and the task confidence score. The task decision layer uses the final prediction values ​​of each computation task as the predicted values ​​of battery health status, remaining life, consistency index, and safety risk identifier, and outputs a set of prediction information.

[0008] In a preferred embodiment, this application can be further configured as follows: the step of classifying the target battery into the corresponding tiered utilization performance level based on the predicted information set and a pre-set grading strategy includes the following steps: The predicted information set is input into a pre-built fuzzy logic reasoning model, which outputs a recommendation membership degree that is associated with the performance level of the tiered utilization. The target battery is classified into the corresponding tiered utilization performance level based on the pre-set grading strategy and recommended membership degree.

[0009] In a preferred embodiment, this application can be further configured such that: the fuzzy logic reasoning model includes a fuzzy processing layer, a rule reasoning layer, and a defuzzification layer; the step of inputting the prediction information set into the pre-constructed fuzzy logic reasoning model to output the recommended membership degree associated with the tiered utilization performance level includes the following steps: The fuzzy processing layer transforms the predicted information set into a comprehensive membership vector of the fuzzy language based on a predefined set of membership functions, and outputs it to the rule inference layer. The rule reasoning layer performs fuzzy implication operations on the comprehensive membership vector based on a preset fuzzy rule base, and outputs the fuzzy implication operation results to the defuzzification layer. The defuzzification layer aggregates the received fuzzy implication calculation results, performs defuzzification calculation based on a predefined defuzzification function, and obtains the recommended membership degree of the target battery.

[0010] In a preferred embodiment, this application can be further configured as follows: the membership function set includes a first membership function and a second membership function; the fuzzy processing layer transforms the predicted information set into a comprehensive membership vector of the fuzzy language based on the predefined membership function set, and outputs it to the rule inference layer, including the following steps: The fuzzy processing layer calculates the membership degree corresponding to the preset first fuzzy subset based on the first membership function, with the battery health state prediction value as input, and generates a state membership vector. The fuzzy processing layer calculates the membership degree corresponding to the preset second fuzzy subset based on the second membership function, with the security risk identifier as input, and generates a risk membership vector. The state membership vector and the risk membership vector are combined to form a comprehensive membership vector of length M+N, where M is the number of the first fuzzy subset and N is the number of the second fuzzy subset. The comprehensive membership vector is output to the rule inference layer.

[0011] In a preferred embodiment, this application can be further configured as follows: the rule inference layer performs fuzzy implication operations on the comprehensive membership vector based on a preset fuzzy rule base, and outputs the fuzzy implication operation results to the defuzzification layer, including the following steps: The rule reasoning layer inputs the comprehensive membership vector into a preset fuzzy rule base, calculates the premise matching degree between each rule and the comprehensive membership vector, and activates all rules with premise matching degree greater than zero. The rule includes a premise part and a consequent part. The rule reasoning layer takes the premise matching degree of the activated rule as input and applies predefined implication operators to calculate the activation strength of the rule for its output fuzzy subset. The rule inference layer generates an inference result output set based on the output fuzzy subset of all activated rules and their corresponding activation intensities, and outputs the inference result output set as the fuzzy implication operation result to the defuzzification layer.

[0012] In a preferred embodiment, this application can be further configured as follows: the step of the defuzzification layer aggregating the received fuzzy implication operation results, performing defuzzification calculation based on a predefined defuzzification function, and obtaining the recommended membership degree of the target battery includes the following steps: The defuzzification layer uses the centroid method to calculate the geometric centroid of the fuzzy implication operation result, and sets the abscissa value of the geometric centroid on the universe of discourse as the recommended value; The defuzzing layer inputs the recommended value into a predefined set of membership functions, and calculates the membership degree of the recommended value to each performance level membership function. The set of membership functions includes membership functions corresponding to each tier of utilization performance level. The defuzzing layer combines the calculated membership degrees into a recommended membership vector, and outputs this recommended membership vector as the recommended membership degree of the target battery.

[0013] In a preferred embodiment, this application can be further configured as follows: the step of acquiring real-time market data and performing simulation calculations based on a preset calculation path for target batteries at each performance level of utilization, wherein the preset calculation path includes a utilization path and a dismantling and recycling path, includes the following steps: A market data vector is constructed based on real-time market data. The market data vector includes a battery product price vector, a target metal price vector, an energy storage transformation cost vector, and a dismantling and recycling cost vector. Using the unit price of second-hand battery products, the cost vector of energy storage transformation, and the predicted value of battery health status as independent variables for the second-hand utilization path, the residual value of second-hand utilization is generated. Using the target metal price vector, dismantling and recycling cost vector, and chemical composition data as independent variables for the dismantling and recycling path, the dismantling and recycling residual value is generated.

[0014] The second objective of this invention is achieved through the following technical solution: A power battery recycling and cascade utilization optimization system includes: The data acquisition module is used to acquire a multidimensional state dataset of the target battery, which includes physical structure data, chemical composition data and dynamic electrical characteristic data. The information prediction module is used to input the multidimensional state dataset into the pre-trained multi-channel prediction model, so that the multi-channel prediction model generates and outputs a prediction information set, which includes battery health state prediction value, remaining life prediction value, consistency index and safety risk identifier. The tiered grading module is used to classify target batteries into corresponding tiered utilization performance levels based on the predicted information set and the pre-set grading strategy. The simulation calculation module is used to acquire real-time market data and perform simulation calculations based on preset calculation paths for target batteries at each performance level of utilization. The preset calculation paths include utilization paths and dismantling and recycling paths. The scheme generation module is used to generate optimization schemes corresponding to each utilization performance level based on the results of simulation calculations, with the goal of maximizing batch residual value.

[0015] In summary, the power battery recycling and cascade utilization optimization method and system provided in this application obtains a multi-dimensional state dataset and generates a prediction information set using a multi-channel prediction model. It combines a fuzzy logic reasoning model and market dynamic data to perform hierarchical decision-making and residual value optimization, which solves the problems of single evaluation dimensions, coarse grading and low resource utilization in traditional methods. It has the advantages of improving evaluation dimensions, achieving accurate hierarchical decision-making, and dynamically optimizing resource utilization efficiency. Attached Figure Description

[0016] Figure 1 This is a flowchart of an embodiment of an optimized method for recycling and reusing power batteries according to this application; Figure 2 This is a flowchart of step S20 in an embodiment of the optimized method for recycling and reusing power batteries according to this application; Figure 3 This is a flowchart of step S30 in an embodiment of the optimized method for recycling and reusing power batteries according to this application. Detailed Implementation

[0017] The following is in conjunction with the appendix Figures 1-3 This application will be described in further detail.

[0018] In one embodiment, such as Figure 1 As shown, this application discloses an optimized method for the recycling and secondary utilization of power batteries, which specifically includes the following steps: S10: Obtain a multidimensional state dataset of the target battery, which includes physical structure data, chemical composition data, and dynamic electrical characteristic data; In this embodiment, the target battery is a retired power battery selected and optimized for secondary utilization in a power battery recycling production line or processing batch. Depending on the specific situation, the target battery may be single or multiple. The multidimensional state dataset is a comprehensive data set covering the battery's physical morphology, material composition, and operating characteristics. Specifically, it can be achieved by using X-ray tomography to obtain casing deformation data, inductively coupled plasma optical emission spectrometry to detect the metal content of the positive and negative electrode materials, and a charge-discharge tester to record dynamic impedance spectra. The multidimensional state dataset provides a foundation for a comprehensive assessment of the battery's state. Among them, the physical structure data is data reflecting the integrity of the battery's physical morphology, obtained through technologies such as industrial CT and 3D scanning. This includes casing deformation, the degree of electrode tab corrosion, and the wrinkling or breakage of internal electrode sheets. Structural data is crucial for determining whether a battery has experienced internal short circuits or mechanical damage. Chemical composition data, obtained through non-destructive or minimal-destructive testing techniques such as inductively coupled plasma spectroscopy (ICP) and X-ray diffraction (XRD), reflects changes in the battery's internal chemical system, including the loss rate of lithium in the cathode material, the dissolution ratio of transition metals (cobalt, nickel, and manganese), and the concentration of electrolyte degradation products. Chemical composition data is the core for assessing the root causes of irreversible capacity decay in batteries. Dynamic electrical characteristic data, collected during charge-discharge tests under specific operating conditions, includes capacity, internal resistance, voltage curve relaxation characteristics at different rates, frequency domain response of electrochemical impedance spectroscopy (EIS), and the shape of the capacity decay curve during charge-discharge cycles. Dynamic electrical characteristic data contains rich information about the battery aging mechanism.

[0019] S20: Input the multidimensional state dataset into the pre-trained multi-channel prediction model, so that the multi-channel prediction model generates and outputs a prediction information set, which includes battery health state prediction value, remaining life prediction value, consistency index and safety risk identifier. In this embodiment, the multi-channel prediction model is a neural network architecture capable of processing heterogeneous data in parallel. Specifically, it can be implemented by extracting physical structure features through convolutional networks, modeling chemical component correlations through graph neural networks, and analyzing the evolution of electrical characteristics through time-series networks. This multi-channel prediction model can effectively capture the nonlinear relationships between multi-source data. The prediction information set is a set of key indicators synchronously output by the multi-channel prediction model to comprehensively evaluate the current state and future performance of the target battery. The battery health status prediction value is a quantified percentage value, representing the ratio of the battery's actual usable capacity at the current moment to its rated capacity when it was in its new state. The remaining life prediction value is a value in terms of cycles or time (years), predicting the duration or number of cycles that the battery can continue to operate under specific conditions from its current state until the end of its life (e.g., SOH drops to 70% or 80%). The consistency index is a quantified indicator used to evaluate the degree of dispersion of key performance parameters (such as capacity, internal resistance, and self-discharge rate) among multiple retired batteries of the same batch and model. The safety risk identifier is a comprehensive risk score or risk level label used to identify the potential possibility of safety accidents such as thermal runaway, short circuit, and leakage in subsequent use of the battery.

[0020] S30: Based on the predicted information set and the pre-set grading strategy, the target battery is classified into the corresponding tiered utilization performance level; In this embodiment, the pre-set grading strategy is a set of decision rules based on fuzzy logic. Specifically, the membership function can be used to quantify the correlation between battery state and safety risk. The performance level is determined by rule reasoning. That is, based on the predicted information set (SOH, RUL, consistency index, safety risk identifier) ​​as input, the target battery is mapped to a certain level in the tiered utilization performance level. This grading strategy is used to solve the rigidity defects of traditional threshold division. The tiered utilization performance level is an application scenario adaptability classification based on the battery's remaining value and technical characteristics.

[0021] For example, the preset performance level for tiered utilization includes: Grade A (High-performance / High-value): Technical standards: High state of health (SOH) (e.g., >80%), long remaining life (RUL), excellent consistency, and a safety risk identifier of "safe".

[0022] Target applications: Scenarios with the highest requirements for performance, lifespan and safety, such as: power grid frequency regulation, backup power for communication base stations, and high-end energy storage systems.

[0023] Economic value: highest; can be used directly as a "high-quality second-hand product" with low renovation investment and high premium.

[0024] Grade B (Balanced Grade / Standard Value Grade): Technical standards: Medium SOH (e.g., 60%-80%), moderate RUL, good consistency, and safety risk identifier is "Warning" but controllable.

[0025] Target applications: Scenarios where performance requirements are not extreme and cost-effectiveness is the priority, such as: low-speed electric vehicles (sightseeing vehicles, sanitation vehicles), wind and solar energy storage systems, and home energy storage.

[0026] Economic value: Medium, requires some investment in system adaptation and monitoring.

[0027] Grade C (Low Performance / Basic Value): Technical standards: Low SOH (e.g., 40%-60%), short RUL, poor consistency, but controllable safety risks.

[0028] Target applications: Scenarios with extremely low performance requirements and only basic charging and discharging functions, such as street light energy storage, billboard power supply, and backup lighting power supply.

[0029] Economic value: Low, typically requires a complex battery management system (BMS) for protection and monitoring.

[0030] Dismantling level (no value for secondary use): Technical standards: extremely low SOH (e.g., <40%), safety risk identifier is "hazardous", or there is serious physical damage.

[0031] Target application: Where there is no possibility of secondary utilization, its value lies in the internal metallic materials (lithium, cobalt, nickel, etc.).

[0032] Economic value: Depends entirely on the value of recycled raw materials.

[0033] S40: Acquire real-time market data and perform simulation calculations based on preset calculation paths for target batteries at each performance level of utilization. The preset calculation paths include utilization paths and dismantling and recycling paths. In this embodiment, real-time market data refers to key economic parameters that dynamically affect battery residual value assessment and are obtained from external markets. This real-time nature is a core feature distinguishing it from traditional static assessment models. The preset calculation path is a mathematical model for residual value assessment. Specifically, it can construct a residual value function for tiered utilization that correlates health status with modification costs, and a dismantling and recycling function that correlates metal content with market prices. This preset calculation path is used to achieve dynamic optimization of economic assessment. The tiered utilization path simulates the path chain for reusing the battery as a whole functional unit. Its calculation model can be set as: Path residual value = (Selling price of the battery as a product) - (Remanufacturing / modification cost). The value of the tiered utilization path lies in the battery's remaining performance (SOH, RUL). The dismantling and recycling path simulates the path chain for recycling materials from the battery as a mineral resource carrier. Its calculation model can be set as: Path residual value = Σ(Value of metal materials) - (Dismantling and recycling cost). The value of the dismantling and recycling path lies in the battery's chemical composition and weight.

[0034] S50: Based on the results of simulation calculations, with the goal of maximizing batch residual value, an optimization scheme corresponding to each utilization performance level is generated.

[0035] In this embodiment, maximizing the batch residual value is the decision objective function of the entire optimization process. It does not pursue the maximum residual value of a single battery, but rather the maximum residual value of the entire recycling batch (i.e., all target batteries currently being processed).

[0036] Specifically, the integrity data of the battery casing welds is collected using industrial CT equipment, and the changes in lithium salt concentration in the electrolyte are detected simultaneously. Voltage relaxation curves under different operating conditions are recorded to construct a multi-dimensional state dataset containing morphological, compositional, and behavioral characteristics. The multi-dimensional state dataset is then input into a trained multi-channel prediction model, which generates and outputs a prediction information set. Based on the prediction information set, the target battery is classified into the corresponding cascade utilization performance level using fuzzy inference rules and other methods. Based on real-time market data, the cascade utilization revenue and dismantling and recycling revenue are calculated separately, and the maximum residual value combination is solved using a linear programming algorithm.

[0037] Through the above technical solutions, this application effectively distinguishes batteries with intact appearance but with lithium plating risk, avoiding their misuse in high-power scenarios; accurately identifies batteries with high cobalt content in the cathode material, prioritizing their dismantling and recycling to maximize the utilization of metal resources; and finds the optimal balance between cascade utilization and dismantling and recycling through real-time optimization algorithms.

[0038] In one embodiment, the multi-channel prediction model includes a multi-channel feature extraction network, a feature fusion layer, and a task decision layer, such as... Figure 2 As shown, step S20 includes: S21: The multi-channel feature extraction network extracts physical structure feature vectors, chemical component feature vectors, and electrical property feature vectors based on the input physical structure data, chemical composition data, and dynamic electrical property data. In this embodiment, the multi-channel feature extraction network is a neural network architecture with parallel processing capabilities. Specifically, it can be implemented using a hybrid structure of convolutional neural networks and graph neural networks. It is used to extract physical structure feature vectors from physical structure data, chemical component feature vectors from chemical component data, and electrical characteristic feature vectors from dynamic electrical characteristic data. The physical structure feature vector is an abstract numerical vector representing a certain structural characteristic inside the battery (such as tab alignment and electrode porosity). The chemical component feature vector is a numerical vector encoding the health status of the internal chemical system of the battery, such as the lattice integrity of the cathode material and the concentration characteristics of electrolyte degradation products. The electrical characteristic feature vector is a numerical vector that condenses key information of the battery's dynamic performance, such as capacity decay rate, internal resistance growth mode, and relaxation characteristics.

[0039] S22: The feature fusion layer performs cross-modal fusion on the received physical structure feature vector, chemical composition feature vector, and electrical property feature vector to generate a joint feature vector; In this embodiment, the feature fusion layer is a computational module for integrating heterogeneous features. Specifically, it can be implemented using an attention mechanism plus a fully connected layer. By establishing correlation weights between cross-modal features, it achieves effective integration of multi-dimensional information. Cross-modal fusion is the process of integrating heterogeneous feature vectors from different channels into a unified, complementary joint representation through specific technical means. The joint feature vector is a unified feature vector generated after cross-modal fusion. This joint feature vector can be regarded as the "health fingerprint" of the entire battery. It retains all the effective information from three different detection methods to the maximum extent, providing a rich data foundation for the final accurate prediction of indicators such as SOH and RUL.

[0040] S23: Based on the received joint feature vector, the task decision layer calculates the predicted battery health status, predicted remaining life, consistency index, and safety risk identifier, and outputs the predicted information set.

[0041] In this embodiment, the task decision layer is a multi-task joint reasoning module, which can be implemented using a multi-branch neural network structure. Each branch network corresponds to a nonlinear mapping relationship of a specific prediction task, and multi-task collaborative optimization is achieved by sharing underlying features.

[0042] Specifically, physical structure data is generated into point cloud data through 3D scanning and then input into the geometric feature extraction network; chemical composition data is analyzed by spectral analysis and then input into the material feature extraction network; and dynamic electrical property data is modeled over time and then input into the electrochemical feature extraction network. The three feature vectors are weighted by the cross-modal attention mechanism of the feature fusion layer to form a joint feature vector containing information related to structure, materials, and electrochemistry. This joint feature vector is then input into four parallel branches of the task decision layer, which respectively predict health status values ​​through a regression network, predict remaining lifespan through a survival analysis model, calculate a consistency index through a clustering algorithm, and generate safety risk identifiers through a classifier.

[0043] Through the above technical solutions, this application solves the problem of misjudgment caused by the single dimension of traditional evaluation methods. It improves the accuracy of health status prediction through multimodal data fusion and enhances the reliability of safety risk identification through joint feature analysis, providing more comprehensive data support for tiered utilization decisions. Specifically, it can identify hidden defects such as tab corrosion and electrolyte drying that cannot be detected by traditional methods, avoiding misjudging batteries with potential safety hazards as usable. It can also quantitatively assess the consistency differences between individual cells within the battery pack, providing data basis for energy storage system reorganization.

[0044] In one embodiment, step S23 includes: S231: The task decision layer inputs the joint feature vector into its pre-set first branch network, second branch network, third branch network and fourth branch network, so that they can perform the corresponding computation tasks simultaneously, and obtain the initial prediction value of each computation task and its corresponding task confidence score. In this embodiment, the branch networks are neural network structures designed independently for different prediction tasks. Specifically, they can be implemented using a combination of fully connected layers and attention mechanisms to process joint feature vectors in parallel to generate initial prediction results. The first branch network is used for SOH prediction and can employ a deep regression network. Its output layer uses the Sigmoid activation function to constrain the predicted value to the [0,1] interval (corresponding to 0%-100% health). The second branch network is used for RUL prediction and can employ a survival analysis model to output the remaining number of cycles or time of the battery. ReLU activation is used to ensure non-negativity. The third branch network is used for consistency index prediction and can employ a statistical learning network to output the coefficient of variation or standard deviation of the batch of batteries. The fourth branch network is used for safety risk prediction and can employ a probability output network to output the risk probability value (0-1). The task confidence score is a quantitative indicator reflecting the reliability of the prediction results of each branch network. Specifically, it can be calculated using a probability output layer or an uncertainty estimation module and is used for weight allocation when multiple prediction results conflict.

[0045] S232: The task decision layer performs consistency verification on the initial prediction values ​​of each computation task based on the preset decay mechanism rule base, and generates the final prediction values ​​of each computation task based on the consistency verification results and the task confidence score. In this embodiment, the degradation mechanism rule base is a database that stores the aging laws and failure modes of battery materials. Specifically, it can be constructed by electrochemical mechanism models and historical degradation data to verify whether the prediction results conform to the battery degradation laws.

[0046] S233: The task decision layer uses the final prediction values ​​of each computation task as the predicted values ​​of battery health status, remaining life, consistency index, and safety risk identifier, and outputs a set of prediction information.

[0047] Specifically, the joint feature vector is simultaneously input into four independent branch networks, each of which outputs initial predicted values ​​for battery health status, remaining lifespan, consistency index, and safety risk identifier. During computation, each branch network synchronously generates a task confidence score, reflecting the degree of matching between the current input data and the model training distribution. Subsequently, battery aging knowledge stored in the decay mechanism rule base is used to physically validate the initial predicted values, such as whether there is a logical contradiction between the remaining lifespan prediction and the health status prediction. For conflicting predictions, dynamic adjustments are made based on the task confidence score; for example, when the confidence score for the health status prediction is higher than that for the remaining lifespan prediction, the health status prediction is prioritized to correct the remaining lifespan calculation. Finally, the consistent predicted values ​​are integrated into a complete prediction information set for output.

[0048] Through the above technical solution, this application solves the problem of lack of physical consistency in the prediction results in the prior art, and effectively reduces the risk of misjudgment of tiered utilization caused by model prediction deviation. Specifically, the verification mechanism based on the decay mechanism rule base can forcibly correct the prediction values ​​that do not conform to the battery aging law, while the task confidence score provides a quantitative basis for resolving conflicts in multi-task prediction results, thereby improving the reliability of the overall prediction system.

[0049] In one embodiment, step S30 includes: S31: Input the predicted information set into the pre-built fuzzy logic reasoning model, so that its output is the recommendation membership degree associated with the performance level of the tiered utilization; S32: Based on the pre-set grading strategy and recommended membership degree, the target battery is classified into the corresponding tiered utilization performance level.

[0050] In this embodiment, the fuzzy logic reasoning model is a computational model that uses fuzzy mathematics theory to process uncertain and fuzzy data. It is used to transform imprecise indicators such as battery health status prediction values ​​and safety risk identifiers into quantifiable membership degrees. The recommended membership degree is a quantified probability value of the target battery belonging to different performance levels of utilization. Specifically, it can be calculated through membership degree functions to reflect the degree of matching between battery performance and each level. The membership degree function set is a set of mathematical functions used to map numerical inputs to fuzzy linguistic variables, such as trigonometric functions or Gaussian functions, to fuzzify battery health status prediction values ​​and safety risk identifiers to eliminate errors caused by single threshold division.

[0051] Through the above technical solutions, this application solves the problem of inaccurate battery tiered utilization classification caused by a single threshold in the prior art, and reduces batch residual value loss caused by differences in human experience; by using a fuzzy logic reasoning model to fuse multi-dimensional prediction information, it improves the rationality and interpretability of battery performance level classification, and provides a reliable basis for subsequent residual value optimization.

[0052] In one embodiment, the fuzzy logic reasoning model includes a fuzzy processing layer, a rule-based reasoning layer, and a defuzzification layer, such as... Figure 3 As shown, step S31 includes: S311: The fuzzy processing layer transforms the predicted information set into a comprehensive membership vector of the fuzzy language based on a predefined set of membership functions, and outputs it to the rule inference layer. In this embodiment, the fuzzy processing layer is a functional module used to transform continuous prediction data into fuzzy linguistic variables. Specifically, it can be implemented using Gaussian membership functions or triangular membership functions. By setting the coverage range of different fuzzy subsets, parameters such as battery health status prediction values ​​and safety risk identifiers are mapped to corresponding membership vectors. The membership function set is a set of mathematical functions used to map numerical inputs to fuzzy linguistic variables, such as triangular functions or Gaussian functions, to fuzzify battery health status prediction values ​​and safety risk identifiers to eliminate errors caused by single threshold division. The comprehensive membership vector is the output of the fuzzy processing layer, which is a structured vector representing the membership degree of the input data to all relevant fuzzy concepts.

[0053] S312: The rule reasoning layer performs fuzzy implication operation on the comprehensive membership vector based on the preset fuzzy rule base, and outputs the fuzzy implication operation result to the defuzzification layer; In this embodiment, the rule reasoning layer is a computational unit for fuzzy rule execution logic reasoning based on expert experience. Specifically, it can be implemented using a Mamdani-type fuzzy reasoning system. By matching the condition strength of the input membership vector with the premise part of the rule, the corresponding output fuzzy subset is activated. The fuzzy rule base is a set of "IF-THEN" rules defined based on the battery decay mechanism and expert knowledge, used to describe the fuzzy logical relationship between the input and the output. The fuzzy implication operation is the process of calculating the premise matching degree of each rule and deriving the activation strength of its conclusion.

[0054] S313: The defuzzification layer aggregates the received fuzzy implication operation results, performs defuzzification calculation based on a predefined defuzzification function, and obtains the recommended membership degree of the target battery.

[0055] In this embodiment, the defuzzification layer is an output module that transforms the fuzzy inference results into deterministic values. Specifically, it can be implemented using the centroid method or the maximum membership method. By calculating the geometric centroid of the fuzzy set or selecting the value range point corresponding to the maximum membership degree, the final recommended membership degree is determined. The predefined defuzzification function is a mathematical method that transforms the fuzzy inference results into specific values. Commonly used methods include the centroid method and the maximum membership method.

[0056] Specifically, parameters such as the predicted battery health status and safety risk identifiers from the prediction information set are first input into the fuzzy processing layer, and then transformed into corresponding fuzzy linguistic variables through a predefined membership function. For example, the predicted battery health status can be divided into three fuzzy subsets: "high," "medium," and "low," each corresponding to a different membership distribution range. Subsequently, the rule reasoning layer performs logical operations based on a preset fuzzy rule library, such as "if the health status is high and the safety risk is low, then the recommended level is Class I tiered utilization." Finally, the defuzzification layer aggregates the outputs of multiple activated fuzzy rules and transforms them into specific recommended membership values ​​through mathematical operations. These values ​​reflect the degree to which the target battery is suitable for each tiered utilization level.

[0057] Through the above technical solutions, this application realizes the fuzzy processing and intelligent reasoning of battery performance parameters, improving the accuracy and interpretability of the classification of tiered utilization levels; by establishing a multi-level fuzzy reasoning mechanism, it can effectively handle the uncertainty of prediction parameters, reduce the classification error caused by unreasonable setting of single parameter thresholds, and at the same time, through the knowledge expression of the rule reasoning layer, the classification decision-making process is traceable, providing a reliable basis for the generation of subsequent optimization schemes.

[0058] In one embodiment, the membership function set includes a first membership function and a second membership function, and step S311 includes: S3111: The fuzzy processing layer calculates the membership degree corresponding to the preset first fuzzy subset based on the first membership function and the battery health state prediction value as input, and generates a state membership vector. In this embodiment, the first membership function refers to a mathematical function used to map the predicted battery health status value to a preset fuzzy subset. Specifically, it can be implemented using a trapezoidal membership function or a Gaussian membership function. By defining the membership distribution of different health status intervals, the conversion from precise numerical values ​​to fuzzy semantics is achieved. The first fuzzy subset is a set of fuzzy concepts predefined to describe the battery health status. The state membership vector is a set of vectors generated by the first membership function that reflects the degree of membership of the battery health status on each fuzzy subset. Its dimension is determined by the preset number of the first fuzzy subset.

[0059] S3112: The fuzzy processing layer calculates the membership degree corresponding to the preset second fuzzy subset based on the second membership function, with the security risk identifier as input, and generates a risk membership vector; In this embodiment, the second membership function refers to the membership relationship used to quantify the risk level corresponding to the safety risk identifier. Specifically, it can be implemented using a triangular membership function or a piecewise linear function. By setting boundary conditions for different risk levels, the discrete safety identifier is transformed into a continuous fuzzy quantity. The second fuzzy subset is a set of fuzzy concepts predefined to describe the battery safety risk situation. The risk membership vector is a set of vectors generated by the second membership function that characterize the degree of membership of the safety risk identifier at each risk level. Its dimension is determined by the preset number of the second fuzzy subset.

[0060] S3113: Combine the state membership vector and the risk membership vector to form a comprehensive membership vector of length M+N, where M is the number of the first fuzzy subset and N is the number of the second fuzzy subset; In this embodiment, the comprehensive membership vector is a multi-dimensional vector formed by concatenating the state membership vector and the risk membership vector, which is used to comprehensively characterize the fuzzy semantic information of battery health status and safety risk.

[0061] S3114: Output the comprehensive membership vector to the rule inference layer.

[0062] Specifically, the fuzzy processing layer first receives the predicted battery health status and the safety risk identifier as input. For the predicted battery health status, a first membership function is used to calculate its membership degree on multiple preset fuzzy subsets of health status. For example, the health status is divided into four fuzzy subsets: "Excellent," "Good," "Medium," and "Poor," with each subset corresponding to different membership function parameters. Simultaneously, for the safety risk identifier, a second membership function is used to calculate its membership degree on preset risk level fuzzy subsets. For example, the risk level is divided into three fuzzy subsets: "Low Risk," "Medium Risk," and "High Risk." By concatenating the two independently generated membership vectors, a comprehensive membership vector with M+N dimensions is formed, where M and N correspond to the number of fuzzy subsets of health status and risk level, respectively. This comprehensive membership vector can fully preserve the fuzzy semantic features of battery performance and safety, providing multi-dimensional input data for subsequent rule reasoning.

[0063] Through the above technical solutions, this application effectively solves the technical defects of existing technologies, such as the single dimension of health status assessment and insufficient quantification of safety risks. By constructing a multi-dimensional membership function system, it achieves a synergistic quantitative assessment of battery performance and safety, improving the accuracy of the classification of tiered utilization levels. At the same time, the construction method of the comprehensive membership vector can retain the multi-dimensional feature information of the original data, providing sufficient data support for subsequent fuzzy inference and reducing the occurrence of misclassification problems caused by information loss in traditional methods. In addition, this technical solution can also adapt to the needs of different battery types and application scenarios, and achieve flexible adaptation of the assessment model by adjusting the number of fuzzy subsets and parameter configuration.

[0064] In one embodiment, step S312 includes: S3121: The rule reasoning layer inputs the comprehensive membership vector into the preset fuzzy rule base, calculates the premise matching degree between each rule and the comprehensive membership vector, and activates all rules with premise matching degree greater than zero. The rule includes a premise part and a consequent part. In this embodiment, the fuzzy rule base is a set of rules that store condition-conclusion relationships derived from expert experience or historical data. Specifically, it can be implemented using a structured rule set in the form of if-then. Each rule contains a premise part consisting of multiple membership conditions and a corresponding conclusion output. The premise matching degree is the degree of conformity between the input data and the rule premise conditions. Specifically, it can be implemented by taking the minimum value of each membership condition or by multiplying them, and is used to filter out effective rules related to the current input. The premise part is the "condition" or "input" part of the fuzzy rule, which describes the fuzzy conditions that must be met for the rule to be activated. The consequent part is the "conclusion" or "output" part of the fuzzy rule, which describes the fuzzy conclusion that should be drawn when the premise conditions are met. In this embodiment, the consequent part points to the output fuzzy subset.

[0065] S3122: The rule reasoning layer takes the premise matching degree of the activated rule as input and applies a predefined implication operator to calculate the activation strength of the rule for its output fuzzy subset. In this embodiment, the implication operator is a mathematical method used to map the premise matching degree to the conclusion output strength. Specifically, it can be implemented using Mamdani minimum implication or Larsen product implication, and is used to quantify the contribution weight of the rule to the final conclusion.

[0066] S3123: The rule inference layer generates an inference result output set based on the output fuzzy subset of all activated rules and their corresponding activation intensities, and outputs the inference result output set as the fuzzy implication operation result to the defuzzification layer.

[0067] In this embodiment, the activation strength is the actual degree of influence of the rule on the conclusion after processing by the implication operator. Specifically, it can be achieved by combining the premise matching degree with the calculation result of the implication operator, which is used for the aggregation calculation of subsequent inference results.

[0068] Specifically, when the comprehensive membership vector is input to the rule inference layer, each rule in the fuzzy rule base is automatically traversed, and valid rules are selected by calculating the degree of matching between the input data and the rule premises. For each activated rule, a preset implication operator is used to convert the premise matching degree into the conclusion output strength, and the conclusion outputs of all activated rules are superimposed to form the inference result output set. Through this processing method, complex multi-dimensional membership data can be transformed into quantifiable rule outputs, realizing an automated inference process based on expert experience.

[0069] Through the above technical solution, this application effectively solves the problems of inconsistent rule application and opaque reasoning process in the evaluation of retired batteries. By using a standardized rule matching mechanism and mathematical strength calculation method, it ensures that the evaluation process of different batches of batteries is repeatable and verifiable, while significantly improving the processing efficiency of rule reasoning.

[0070] In one embodiment, step S313 includes: S3131: The defuzzification layer uses the centroid method to calculate the geometric centroid of the fuzzy implication operation result, and sets the abscissa value of the geometric centroid on the universe of discourse as the recommended value; In this embodiment, the centroid method is a defuzzification method that determines the precise output value by calculating the abscissa of the geometric center of the fuzzy set. Specifically, it can be implemented by integral operation or discrete point weighted average to eliminate the uncertainty of the fuzzy inference result. The universe of discourse is the set of all possible values ​​of the output variable (tiered utilization performance level). The recommended value is the coordinate point of the universe of discourse obtained by the centroid method, which can be implemented by calculating the first moment of the area of ​​the fuzzy set, and serves as the input benchmark for subsequent membership degree calculation.

[0071] S3132: The defuzzification layer inputs the recommended value into a predefined set of membership functions, and calculates the membership degree of the recommended value to each performance level membership function. The set of membership functions includes membership functions corresponding to each tier of utilization performance level. In this embodiment, the membership function set consists of multiple membership distribution functions associated with the tiered utilization performance level. Specifically, Gaussian, triangular, or trapezoidal functions can be used to quantify the degree to which the recommended value belongs to each performance level.

[0072] S3133: The defuzzification layer combines the calculated membership degrees into a recommended membership vector, and outputs this recommended membership vector as the recommended membership degree of the target battery.

[0073] In this embodiment, the recommended membership vector is a vector structure composed of multiple membership values. Specifically, it can be generated by calculating the output value of the recommended value under each membership function, which serves as the quantitative basis for performance level classification.

[0074] Specifically, the results of fuzzy implication operations are aggregated to form a fuzzy set. The geometric centroid abscissa of this set is calculated using the centroid method and used as the recommended value. This value reflects the comprehensive tendency of the fuzzy inference results. The recommended value is then input into a predefined set of membership functions, each function corresponding to a tiered utilization performance level. The membership degree of the recommended value under that level is calculated. For example, when the recommended value is 0.72, the output values ​​of its membership functions under levels A, B, and C can be calculated respectively to form a recommended membership vector [0.85, 0.15, 0]. This recommended membership vector is ultimately used as the recommended membership output, providing multi-dimensional quantitative basis for subsequent hierarchical decision-making.

[0075] Through the above technical solutions, this application solves the problem of insufficient accuracy in the process of converting fuzzy inference results into precise values, effectively improving the accuracy of the performance level classification for tiered utilization; by recommending the multidimensional quantitative expression of membership vectors, it provides an interpretable decision basis for subsequent grading strategies, reduces the phenomenon of battery misclassification caused by membership calculation deviations, and optimizes the selection accuracy of tiered utilization and dismantling and recycling paths.

[0076] In one embodiment, step S40 includes: S41: Construct a market data vector based on real-time market data, the market data vector including a battery product price vector, a target metal price vector, an energy storage transformation cost vector, and a dismantling and recycling cost vector; In this embodiment, the market data vector is a numerical set integrating dynamic market parameters. Specifically, it can be constructed using historical price data and real-time transaction data from a time-series database. For example, by collecting futures prices of metals such as lithium and cobalt and market quotations for energy storage system retrofit costs, a multi-dimensional vector can be formed. The market data vector can reflect the impact of the current market supply and demand relationship on residual value calculation, providing dynamic input for subsequent economic analysis. Among them, the battery product price vector is a structured vector representing the unit price of batteries with different performance levels in the second-hand market; the target metal price vector is a vector representing the real-time spot price of key metals (such as lithium, cobalt, nickel, and manganese) with recycling value in the battery cathode material; the energy storage retrofit cost vector and the dismantling and recycling cost vector are structured vectors representing the unit cost required to retrofit retired batteries (target batteries) into energy storage systems and dismantle and recycle valuable metals, respectively.

[0077] S42: Using the unit price of secondary battery products, the cost vector of energy storage transformation, and the predicted value of battery health status as independent variables for the secondary utilization path, generate the residual value of secondary utilization. In this embodiment, the residual value of tiered utilization is the estimated economic value of the battery after modification for use in scenarios such as energy storage. Specifically, it can be calculated using a cost-benefit model. For example, the unit price of the tiered battery product is multiplied by the predicted health status value, and then the equipment purchase cost and labor cost in the energy storage modification cost vector are deducted. This calculation method combines battery performance with modification cost to quantify the feasibility of the tiered utilization path.

[0078] S43: Using the target metal price vector, dismantling and recycling cost vector, and chemical composition data as independent variables for the dismantling and recycling path, generate the dismantling and recycling residual value.

[0079] In this embodiment, the residual value of dismantling and recycling is the estimated economic value of the metal materials recovered after battery dismantling. Specifically, it can be calculated by multiplying the metal mass by its real-time unit price and then deducting the dismantling cost. For example, the total amount of recyclable metal can be calculated based on the nickel and cobalt content in the chemical composition data, and the total revenue can be calculated by combining the real-time price in the target metal price vector. This calculation method dynamically links the material composition with the market price, improving the accuracy of recycling decisions.

[0080] Specifically, real-time market data is accessed through data acquisition interfaces, such as metal price data from commodity trading platforms and transformation cost data from supply chain management systems; market data vectors are updated using a sliding time window mechanism, such as refreshing price information every 30 minutes; in the calculation of residual value for tiered utilization, the predicted value of battery health status is used as a capacity conversion factor, for example, when the predicted health status value is 80%, the unit price of the tiered battery product is calculated at 80% of the original price; when calculating the residual value of dismantling and recycling, the metal content in the chemical composition data needs to be corrected in conjunction with the metal extraction rate parameter, for example, converting the theoretical content into the actual recyclable amount based on the electrolysis process efficiency.

[0081] In some specific implementations, the battery product price vector can adopt a regionally differentiated pricing strategy, such as adjusting the unit price of energy storage batteries according to the electricity price policy of the target market; the dismantling and recycling cost vector can include environmental treatment costs, such as adding a heavy metal harmless treatment cost item according to local hazardous waste treatment standards; the energy storage transformation cost vector can be further broken down into BMS replacement costs and structural reorganization costs, such as modeling the battery management system upgrade costs and enclosure transformation costs separately.

[0082] Through the above technical solution, this application solves the problems of single residual value assessment dimension and lack of market dynamic factors in the prior art; by simultaneously calculating the economic value of the two paths of cascade utilization and dismantling and recycling, and combining real-time price data and battery performance parameters, the optimal disposal method for batteries of different performance levels can be accurately identified; for example, when the predicted health status value of a batch of batteries is less than 60% but contains a high proportion of cobalt, the dismantling and recycling path will be automatically recommended to maximize the residual value, reducing the economic losses caused by the traditional method of erroneously selecting the cascade utilization path due to ignoring component data.

[0083] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0084] In one embodiment, a power battery recycling and cascade utilization optimization system is provided, which corresponds one-to-one with the power battery recycling and cascade utilization optimization method described in the above embodiment. The power battery recycling and cascade utilization optimization system includes: The data acquisition module is used to acquire a multidimensional state dataset of the target battery, which includes physical structure data, chemical composition data and dynamic electrical characteristic data. The information prediction module is used to input the multidimensional state dataset into the pre-trained multi-channel prediction model, so that the multi-channel prediction model generates and outputs a prediction information set, which includes battery health state prediction value, remaining life prediction value, consistency index and safety risk identifier. The tiered grading module is used to classify target batteries into corresponding tiered utilization performance levels based on the predicted information set and the pre-set grading strategy. The simulation calculation module is used to acquire real-time market data and perform simulation calculations based on preset calculation paths for target batteries at each performance level of utilization. The preset calculation paths include utilization paths and dismantling and recycling paths. The scheme generation module is used to generate optimization schemes corresponding to each utilization performance level based on the results of simulation calculations, with the goal of maximizing batch residual value.

[0085] Specific limitations regarding the optimized system for the recycling and reuse of power batteries can be found in the limitations of the optimized method for the recycling and reuse of power batteries described above, and will not be repeated here. Each module in the aforementioned optimized system for the recycling and reuse of power batteries can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the memory of a computer device as software, so that the processor can call and execute the corresponding operations of each module.

[0086] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application, and should all be included within the protection scope of this application.

Claims

1. A power battery recycling cascade utilization optimization method, characterized in that, The method comprises the steps of: obtaining a multi-dimensional state data set of a target battery, the multi-dimensional state data set comprising physical structure data, chemical component data and dynamic electrical characteristic data; inputting the multi-dimensional state data set into a pre-trained multi-channel prediction model to enable the multi-channel prediction model to generate and output a prediction information set, the prediction information set comprising a battery health state prediction value, a remaining life prediction value, a consistency index and a safety risk identifier; dividing the target battery into a corresponding performance grade of cascade utilization based on the prediction information set and a pre-set grading strategy; obtaining real-time market data and performing simulation calculation on the target battery in each performance grade of cascade utilization based on a pre-designed calculation path, the pre-designed calculation path comprising a cascade utilization path and a disassembly and recycling path; generating an optimized scheme corresponding to each performance grade of cascade utilization based on the results of the simulation calculation, with the decision target being maximization of the residual value of each batch.

2. The method of claim 1, wherein: The multi-channel prediction model comprises a multi-channel feature extraction network, a feature fusion layer and a task decision layer. The step of inputting the multi-dimensional state data set into the pre-trained multi-channel prediction model to enable the multi-channel prediction model to generate and output a prediction information set, the prediction information set comprising a battery health state prediction value, a remaining life prediction value, a consistency index and a safety risk identifier, comprises the steps of: the multi-channel feature extraction network extracts a physical structure feature vector, a chemical component feature vector and an electrical characteristic feature vector based on the input physical structure data, chemical component data and dynamic electrical characteristic data; the feature fusion layer performs cross-modal fusion on the received physical structure feature vector, chemical component feature vector and electrical characteristic feature vector to generate a joint feature vector; the task decision layer calculates the battery health state prediction value, the remaining life prediction value, the consistency index and the safety risk identifier based on the received joint feature vector and outputs the prediction information set.

3. The method of claim 2, wherein: The step of the task decision layer calculating the battery health state prediction value, the remaining life prediction value, the consistency index and the safety risk identifier based on the received joint feature vector and outputting the prediction information set comprises the steps of: the task decision layer inputs the joint feature vector into a first branch network, a second branch network, a third branch network and a fourth branch network preset in the task decision layer to enable the task decision layer to perform corresponding calculation tasks simultaneously to obtain initial prediction values of the calculation tasks and corresponding task confidence scores of the calculation tasks; the task decision layer performs consistency verification on the initial prediction values of the calculation tasks based on a pre-set decay mechanism rule library and generates final prediction values of the calculation tasks based on the consistency verification results and the task confidence scores; the task decision layer takes the final prediction values of the calculation tasks as the battery health state prediction value, the remaining life prediction value, the consistency index and the safety risk identifier and outputs the prediction information set.

4. The method of claim 1, wherein: The step of dividing the target battery into a corresponding performance grade of cascade utilization based on the prediction information set and a pre-set grading strategy comprises the steps of: inputting the prediction information set into a pre-constructed fuzzy logic reasoning model to enable the fuzzy logic reasoning model to output recommended membership degrees associated with the performance grade of cascade utilization; The target battery is classified into a corresponding performance grade of cascade utilization based on a preset classification strategy and a recommended membership.

5. The method of claim 4, wherein: The fuzzy logic reasoning model comprises a fuzzy processing layer, a rule reasoning layer and a defuzzification layer, the step of inputting the prediction information set into the pre-built fuzzy logic reasoning model to make the fuzzy logic reasoning model output the recommended membership associated with the performance grade of cascade utilization comprises the following steps: The fuzzy processing layer converts the prediction information set into a comprehensive membership vector for fuzzy language according to a predefined membership function set and outputs the comprehensive membership vector to the rule reasoning layer; The rule reasoning layer performs fuzzy implication operation on the comprehensive membership vector based on a preset fuzzy rule base and outputs the fuzzy implication operation result to the defuzzification layer; The defuzzification layer aggregates the received fuzzy implication operation result, performs defuzzification calculation based on a predefined defuzzification function and obtains the recommended membership of the target battery.

6. The power battery recycling cascade utilization optimization method according to claim 5, characterized in that: The fuzzy processing layer converts the prediction information set into a comprehensive membership vector for fuzzy language according to a predefined membership function set and outputs the comprehensive membership vector to the rule reasoning layer, which comprises the following steps: The fuzzy processing layer calculates the membership corresponding to the preset first fuzzy subset based on the first membership function and the predicted battery state of health value as input and generates a state membership vector; The fuzzy processing layer calculates the membership corresponding to the preset second fuzzy subset based on the second membership function and the safety risk identifier as input and generates a risk membership vector; The state membership vector and the risk membership vector are combined to form a comprehensive membership vector with a length of M+N, wherein M is the number of the first fuzzy subset and N is the number of the second fuzzy subset; The comprehensive membership vector is output to the rule reasoning layer.

7. The method of claim 5, wherein: The rule reasoning layer performs fuzzy implication operation on the comprehensive membership vector based on a preset fuzzy rule base and outputs the fuzzy implication operation result to the defuzzification layer, which comprises the following steps: The rule reasoning layer inputs the comprehensive membership vector into the preset fuzzy rule base, calculates the premise matching degrees of each rule and the comprehensive membership vector, activates all rules with a premise matching degree greater than zero, and the rules comprise a premise part and a consequent part; The rule reasoning layer takes the premise matching degree of the activated rule as input, applies a predefined implication operator to calculate the activation strength of the rule for its output fuzzy subset; The rule reasoning layer generates a reasoning result output set based on the output fuzzy subsets of all activated rules and their corresponding activation strengths and outputs the reasoning result output set as the fuzzy implication operation result to the defuzzification layer.

8. The power battery recycling cascade utilization optimization method of claim 5, wherein: The defuzzification layer aggregates the received fuzzy implication operation result, performs defuzzification calculation based on a predefined defuzzification function and obtains the recommended membership of the target battery, which comprises the following steps: The defuzzification layer calculates the geometric barycenter of the fuzzy implication operation result by using the barycenter method and sets the horizontal coordinate value of the geometric barycenter on the domain as the recommended value. The de-fuzzification layer inputs the recommended value into a set of predefined membership functions corresponding to each performance grade, respectively calculates the membership of the recommended value to each performance grade membership function, and the set of membership functions includes membership functions corresponding to each ladder utilization performance grade; The de-fuzzification layer combines the calculated membership into a recommended membership vector, and outputs the recommended membership vector as the recommended membership of the target battery.

9. The method of claim 1, wherein: The real-time market data is obtained, and simulation calculation based on a pre-designed calculation path is performed on the target battery in each ladder utilization performance grade, the pre-designed calculation path includes the steps of ladder utilization path and disassembly and recycling path, including steps: Based on the real-time market data, a market data vector is constructed, including a battery product price vector, a target metal price vector, a storage energy conversion cost vector, and a disassembly and recycling cost vector; The single price of the ladder battery product, the storage energy conversion cost vector, and the battery health state prediction value are used as the independent variables of the ladder utilization path to generate the ladder utilization residual value; The target metal price vector, the disassembly and recycling cost vector, and the chemical component data are used as the independent variables of the disassembly and recycling path to generate the disassembly and recycling residual value.

10. A power battery recycling cascade utilization optimization system, characterized in that, It includes: A data acquisition module for acquiring a multi-dimensional state data set of a target battery, the multi-dimensional state data set including physical structure data, chemical component data, and dynamic electrical characteristic data; An information prediction module for inputting the multi-dimensional state data set into a pre-trained multi-channel prediction model to generate and output a prediction information set, the prediction information set including a battery health state prediction value, a remaining life prediction value, a consistency index, and a safety risk identifier; A ladder classification module for classifying the target battery into a corresponding ladder utilization performance grade based on the prediction information set and a pre-set classification strategy; A simulation calculation module for obtaining real-time market data and performing simulation calculation based on a pre-designed calculation path on the target battery in each ladder utilization performance grade, the pre-designed calculation path including the ladder utilization path and the disassembly and recycling path; A scheme generation module for generating an optimized scheme corresponding to each ladder utilization performance grade based on the results of the simulation calculation, with batch residual value maximization as the decision target.

Images