Method and equipment for predicting leaching rate of various metals in waste lithium battery and storage medium

Abstract

The invention provides a method and equipment for predicting the leaching rate of various metals in a waste lithium battery and a storage medium, and the method comprises the following steps: generating a descriptor of an organic acid by using RDKit, and obtaining a basic data set of multi-task learning; the basic data set is preprocessed, and a feature subset most representative of the leaching rate is obtained; defining leaching tasks of various metals to be leached and constructing a basic framework of multi-task learning; performing division and index evaluation on the preprocessed data set to obtain a multi-task learning optimization framework; performing hyper-parameter optimization to obtain an optimal multi-task learning hyper-parameter; a multi-task learning model is trained based on the optimal multi-task learning hyper-parameters, and a multi-task learning-based multi-metal leaching rate prediction model is obtained; unknown waste lithium battery leaching data are input into the multiple metal leaching rate prediction model, and meanwhile the leaching rates of multiple metals are obtained; according to the method, the leaching rates of various metals in the waste lithium battery can be rapidly, reliably and simultaneously predicted.

Description

Technical Field

[0001] The present invention relates to the field of environmental protection technology, and in particular to a method, device and storage medium for predicting leaching rates of multiple metals in waste lithium batteries based on multi-task learning modeling. Background Art

[0002] Waste power batteries contain a large amount of metal resources. Taking ternary materials as an example, for every 100g of ternary power batteries recycled, 12g of nickel, 5g of cobalt, 1.2g of lithium and 7g of manganese can be recovered. The non-ferrous metal content is far higher than that of primary ore. In traditional selective recovery, the simultaneous leaching process is a very important step. The leaching process needs to consider multiple factors such as the type of leaching agent, leaching agent concentration, temperature, pH, time, etc. These conditions and factors often have complex nonlinear relationships. The mutual influence between the conditions and factors greatly increases the difficulty of the experiment. Traditional experiments can only carry out refined experiments on specific samples, and it is impossible to extract common laws from scattered research to guide the process.

[0003] Traditional machine learning models can usually only predict the leaching rate of a single metal, while ignoring the correlation between the four metals and lacking the necessary data leakage management steps, making it impossible to reliably guide the simultaneous leaching process of the four metals.

[0004] Therefore, there is an urgent need for a method, device and storage medium for predicting the leaching rates of multiple metals in waste lithium batteries, which can quickly and reliably predict the leaching rates of multiple metals in waste lithium batteries. Summary of the invention

[0005] The purpose of the present invention is to provide a method, device and storage medium for predicting the leaching rates of multiple metals in waste lithium batteries, aiming to solve the technical problems that traditional machine learning models can usually only predict the leaching rate of a single metal and the reliability of the prediction results is poor.

[0006] To achieve the above objectives, in a first aspect, the present invention provides a method for predicting the leaching rates of multiple metals in waste lithium batteries, the steps of which include:

[0007] S1: Collect relevant data, generate descriptors of organic acids using RDKit, and obtain the basic dataset for multi-task learning;

[0008] S2: preprocessing the basic data set obtained in step S1 to obtain the feature subset that is most representative of the leaching rate; defining leaching tasks for multiple metals to be leached, and constructing a basic framework for multi-task learning;

[0009] S3: Divide the preprocessed data set and perform indicator evaluation on the divided test set; determine the appropriate data division range based on the evaluation indicators, and adjust the basic framework of multi-task learning to obtain the optimization framework of multi-task learning;

[0010] S4: Optimizing the hyperparameters of the multi-task learning optimization framework to obtain optimal multi-task learning hyperparameters;

[0011] S5: training a multi-task learning model based on the optimal multi-task learning hyperparameters to obtain a multi-metal leaching rate prediction model based on multi-task learning;

[0012] The unknown waste lithium battery leaching data is input into the multi-metal leaching rate prediction model based on multi-task learning to obtain the leaching rates of the multiple metals at the same time.

[0013] As a further improvement of the above scheme, in step S1, the method steps for obtaining the basic data set for multi-task learning are specifically as follows:

[0014] S11: Obtain literature information on the data to be collected;

[0015] S12: using Origin-image digitization tool to obtain leaching rate data of multiple metals in literature, and record the relevant waste lithium battery properties, organic acid types and experimental conditions in the literature; preferably, obtaining leaching rate data of four metals: lithium, cobalt, manganese and nickel;

[0016] S13: Organic acids were converted into Morgan fingerprints using RDKit as descriptors of organic acids.

[0017] As a further improvement of the above solution, in step S2, the method steps for preprocessing the basic data set obtained in step S1 are specifically as follows:

[0018] S31: Classify the single variable experimental data into a group of data, remove duplicate values ​​and delete outliers from the processed data;

[0019] S32: PCA dimensionality reduction is performed on the organic acid descriptors in the processed data, and the principal components of 95% of the explanatory variables are selected as the features after dimensionality reduction, and correlation tests are performed to obtain the feature subset that is most representative of the leaching rate.

[0020] As a further improvement of the above scheme, in step S2, when constructing the basic architecture of multi-task learning, the shared layers and task-specific layers of the multi-task learning model are set, and the number of layers of the shared layers, the number of neurons in each layer, and the type of activation function are initialized;

[0021] The characteristic task layer sets the output layer corresponding to the various metals to be leached;

[0022] Preferably, the characteristic task layer is provided with four output layers, which respectively correspond to the lithium leaching rate, the cobalt leaching rate, the manganese leaching rate and the nickel leaching rate.

[0023] As a further improvement of the above scheme, in step S2, the mean absolute error (MAE), mean square error (MSE), root mean square error (RMSE), coefficient of determination (R 2 ) or OverallAccuracy Index (OAI) to evaluate the model effect of the basic architecture of the multi-task learning constructed;

[0024] Preferably, the Overall Accuracy Index (OAI) is used to evaluate the model effect of the constructed basic architecture of the multi-task learning.

[0025] As a further improvement of the above scheme, the calculation formula of the mean absolute error is shown as follows:

[0026]

[0027] The calculation formula of the mean square error is as follows:

[0028]

[0029] The calculation formula of the root mean square error is as follows:

[0030]

[0031] The calculation formula of the determination coefficient is shown as follows:

[0032]

[0033] The calculation formula of the full-precision index is as follows:

[0034]

[0035] Where N is the number of samples, y is the predicted value of the output, is the true value of the output, y i and is the true value and predicted value of the i-th sample.

[0036] As a further improvement of the above solution, in step S3, the method steps for obtaining the optimization framework of multi-task learning are specifically as follows:

[0037] S31: Dividing the feature subset data into groups of training set, validation set and test set according to a preset ratio, preferably, the preset ratio is 6:2:2;

[0038] S32: Select a preset number of random seeds that meet the proportional division for screening, set the training set and the validation set as the total training set, perform Overall Accuracy Index (OAI) evaluation on the test set, and select random seeds at the median level of OAI for subsequent data division; preferably, the preset number is 100;

[0039] S33: Evaluate different multi-task learning frameworks according to step S31 and step S32, evaluate the number of shared layers and the number of neurons, and determine the optimized framework of multi-task learning according to the optimal value of the full-precision indicator.

[0040] As a further improvement of the above solution, in step S4, hyperparameter optimization is performed using the total training set on the obtained optimization framework of the multi-task learning;

[0041] The hyperparameter optimization method uses any of the grid search, random search, and particle swarm optimization methods, and performs five-fold cross validation to obtain the best multi-task learning hyperparameters.

[0042] In a second aspect, the present invention further provides a device comprising a memory for storing computer program instructions and a processor for executing computer program instructions, wherein, when the computer program instructions are executed by the processor, the device is triggered to execute a method for predicting the leaching rates of multiple metals in waste lithium batteries provided in the first aspect.

[0043] In a third aspect, the present invention further provides a storage medium having a computer program stored thereon, wherein when the program is run, the device where the storage medium is located is controlled to execute a method for predicting the leaching rates of multiple metals in waste lithium batteries provided in the first aspect.

[0044] Since the present invention adopts the above technical solution, the beneficial effects of this application are:

[0045] The present invention provides a method for predicting the leaching rates of multiple metals in waste lithium batteries, comprising: using RDKit to generate descriptors of organic acids to obtain a basic data set for multi-task learning; preprocessing the basic data set to obtain a feature subset that is most representative of the leaching rate; defining leaching tasks of multiple metals to be leached and constructing a basic framework for multi-task learning; dividing and evaluating the preprocessed data set to obtain an optimization framework for multi-task learning; performing hyperparameter optimization to obtain optimal multi-task learning hyperparameters; training a multi-task learning model based on the optimal multi-task learning hyperparameters to obtain a multi-metal leaching rate prediction model based on multi-task learning. The invention discloses a method for predicting the leaching rate of multiple metals in waste lithium batteries by using a multi-task learning-based prediction model for leaching rates of multiple metals; the method comprises the following steps: inputting unknown leaching data of waste lithium batteries into a multi-metal leaching rate prediction model based on multi-task learning to simultaneously obtain the leaching rates of multiple metals; the invention can quickly and reliably predict the leaching rates of multiple metals in waste lithium batteries at the same time; based on the data of simultaneous leaching of waste lithium batteries by organic acids, the invention uses a multi-metal leaching rate prediction model based on multi-task learning to simultaneously predict the leaching rates of multiple metals; compared with traditional experimental chemical analysis tests, the prediction method provided by the invention can save a lot of manpower and material resources, has the advantages of small workload, low cost, high accuracy, high reliability, etc., and can simultaneously obtain the leaching rates of multiple metals;

[0046] In addition, the present invention combines data leakage management modeling and multi-task learning model algorithms together to establish a more reliable and convenient system than single modeling and random data division, and has better robustness. The results of predicting the types of waste lithium batteries and organic acids with unknown raw materials are more reasonable and stable. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] 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 the structures shown in these drawings without paying any creative work.

[0048] Figure 1 A schematic diagram of a process for predicting the leaching rates of multiple metals in waste lithium batteries disclosed in the present invention;

[0049] Figure 2 It is the PCA dimensionality reduction graph of the organic acid Morgan fingerprint disclosed in the present invention;

[0050] Figure 3 A graph showing changes in evaluation indicators versus the number of iterations during the hyperparameter optimization process disclosed in the present invention;

[0051] Figure 4 It is a comparison chart of the leaching rates R2 of four metals on the training set and the test set disclosed in the present invention.

[0052] The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with the implementation methods and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0053] 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.

[0054] Furthermore, the technical solutions between the various embodiments of the present invention may be combined with each other, but this must be based on the fact that they can be implemented by ordinary technicians in the field. When the combination of technical solutions is mutually contradictory or cannot be implemented, it should be deemed that such combination of technical solutions does not exist and is not within the scope of protection required by the present invention.

[0055] Embodiment 1:

[0056] See also Figure 1-Figure 4 The present invention provides a method for predicting the leaching rates of multiple metals in waste lithium batteries, the steps of which include:

[0057] S1: Collect relevant data, generate descriptors of organic acids using RDKit, and obtain the basic dataset for multi-task learning;

[0058] Specifically, the steps for obtaining the basic data set for multi-task learning are as follows:

[0059] S11: Download relevant literature data from Web of Science and China National Knowledge Infrastructure to obtain literature data to be collected;

[0060] S12: using Origin-image digitization tool to obtain leaching rate data of multiple metals in literature, and record relevant information such as properties of waste lithium batteries, types of organic acids, leaching reagents, experimental conditions and leaching rates in the literature; in this embodiment, leaching rate related data of four metals, lithium, cobalt, manganese and nickel, are obtained;

[0061] S13: Organic acids were converted into Morgan fingerprints using RDKit as descriptors of organic acids;

[0062] S2: preprocessing the basic data set obtained in step S1 to obtain the feature subset that is most representative of the leaching rate; defining leaching tasks for multiple metals to be leached, and constructing a basic framework for multi-task learning;

[0063] S3: Divide the preprocessed data set and perform indicator evaluation on the divided test set; determine the appropriate data division range based on the evaluation indicators, and adjust the basic framework of multi-task learning to obtain the optimization framework of multi-task learning;

[0064] S4: Optimizing the hyperparameters of the multi-task learning optimization framework to obtain optimal multi-task learning hyperparameters;

[0065] S5: training a multi-task learning model based on the optimal multi-task learning hyperparameters to obtain a multi-metal leaching rate prediction model based on multi-task learning;

[0066] Inputting unknown waste lithium battery leaching data into the multi-metal leaching rate prediction model based on multi-task learning to simultaneously obtain the leaching rates of the multiple metals;

[0067] Specifically, when predicting the leaching rates of various metals in unknown waste lithium batteries, the types of organic acids and experimental conditions are input into the model. The results are as follows: Figure 4 As shown in Figure 2, the R ratios between the predicted lithium leaching rate, cobalt leaching rate, nickel leaching rate and manganese leaching rate of the multi-metal leaching rate prediction model based on multi-task learning and the experimental values ​​on the test set are as follows: 2 They are 0.75, 0.8, 0.8 and 0.7 respectively. The predicted values ​​are very close to the true values. This shows that it is feasible to use the multi-metal leaching rate prediction model based on multi-task learning to simultaneously predict the leaching rates of four metals using the leaching data of waste lithium batteries under the consideration of data leakage management, and it has good generalization ability.

[0068] In the present invention, based on the data of simultaneous leaching of waste lithium batteries with organic acid, a multi-metal leaching rate prediction model based on multi-task learning is constructed to simultaneously predict the leaching rates of multiple metals; compared with traditional experimental chemical analysis tests, the prediction method provided by the present invention can save a lot of manpower and material resources, has the advantages of small workload, low cost, high accuracy, high reliability, etc., and can simultaneously obtain the leaching rates of multiple metals;

[0069] In addition, the present invention combines data leakage management modeling and multi-task learning model algorithms together to establish a more reliable and convenient system than single modeling and random data division, and has better robustness. The results of predicting the types of waste lithium batteries and organic acids with unknown raw materials are more reasonable and stable.

[0070] As a preferred embodiment, in step S2, the method steps for preprocessing the basic data set obtained in step S1 are specifically as follows:

[0071] S31: Classify the single variable experimental data into a group of data, remove duplicate values ​​and delete outliers from the processed data;

[0072] Specifically, first, according to the experimental conditions, the single variable experimental data are classified into a group of data to prevent the same group of data from appearing in the training set and the test set at the same time, causing data leakage. The processed data are duplicated to prevent the same features from corresponding to different outputs. Outliers are deleted based on the 3σ principle, assuming that more than 99% of the data are concentrated within the range of 3 standard deviations above and below the mean. The maximum or minimum values ​​exceeding this range are identified as outliers and deleted. Missing value processing is based on the mean of the existing training set to fill in.

[0073] S32: See Figure 2 , PCA dimension reduction is performed on the organic acid descriptors in the processed data, 95% of the principal components of the explanatory variables are selected as the features after dimension reduction, and a correlation test is performed to obtain the most representative feature subset with the leaching rate; specifically, in this embodiment, Figure 2 As shown in the figure, the first eight principal components of 1024 columns of features are selected as input features according to the explanatory variable rate of 95%; the linear correlation and monotonic relationship between the features are tested by Pearson correlation coefficient and Spearman correlation coefficient respectively, and finally 17 features are selected as the feature subset for predicting the leaching rates of the four metals.

[0074] As a preferred embodiment, in step S2, when constructing the basic architecture of multi-task learning, the shared layers and task-specific layers of the multi-task learning model are set, and the number of layers of the shared layers, the number of neurons in each layer, and the type of activation function are initialized;

[0075] The characteristic task layer sets the output layer corresponding to the various metals to be leached;

[0076] Specifically, the multi-task learning model with hard parameter sharing needs to set up a model sharing layer and a task feature layer. The number of layers of the sharing layer and the number of neurons in each layer largely determine the effect of model prediction. In this embodiment, the number of layers of the sharing layer is set to range from 3 to 6, and the number of neurons is set to range from 50 to 200. The feature task layer is set to have 4 output layers, which respectively represent "lithium leaching rate", "cobalt leaching rate", "manganese leaching rate" and "nickel leaching rate"; the activation functions between different layers are all selected as "relu" functions, which can significantly increase the nonlinear ability of the model and maximize the screening ability of neurons, and has the advantages of simple function but high computational efficiency;

[0077] At the same time, the mean absolute error (MAE), mean square error (MSE), root mean square error (RMSE), and coefficient of determination (R 2 ) or OverallAccuracy Index (OAI) to evaluate the model effect of the basic architecture of the multi-task learning constructed;

[0078] Specifically, the calculation formula of the mean absolute error is as follows:

[0079]

[0080] The calculation formula of the mean square error is as follows:

[0081]

[0082] The calculation formula of the root mean square error is as follows:

[0083]

[0084] The calculation formula of the determination coefficient is shown as follows:

[0085]

[0086] The calculation formula of the full-precision index is as follows:

[0087]

[0088] Where N is the number of samples, y is the predicted value of the output, is the true value of the output, y i and is the true value and predicted value of the i-th sample.

[0089] In this embodiment, the Overall Accuracy Index (OAI) is used to evaluate the model effect of the constructed basic architecture of the multi-task learning.

[0090] As a preferred embodiment, in step S3, the method steps for obtaining the optimization framework of multi-task learning are specifically as follows:

[0091] S31: Divide the feature subset data into groups according to a preset ratio into a training set, a validation set, and a test set;

[0092] Specifically, the feature subset data is divided into training set, validation set and test set according to the ratio of 6:2:2 without disrupting the groups;

[0093] S32: Select a preset number of random seeds that meet the proportional division for screening, set the training set and the validation set as the total training set, perform Overall Accuracy Index (OAI) evaluation on the test set, and select a random seed at the median level of OAI for subsequent data division; specifically, consider 100 random seeds that meet the data division, that is, fully try different data distribution forms to avoid fluctuations in model prediction accuracy caused by over-optimal or over-poor data distribution;

[0094] S33: Evaluate different multi-task learning frameworks according to step S31 and step S32, evaluate the number of shared layers and the number of neurons, and determine the optimized framework of multi-task learning according to the optimal value of the full-precision index; specifically, according to step S31 and step S32, evaluate the prediction effect of different random seeds, number of shared layers, and number of neurons on the test set, and merge the data of the training set and the validation set as the data of the training model; finally, the number of shared layers is selected to be 4, and the number of neurons from outside to inside are 200, 100, 50 and 25 respectively, and the number of random seeds is selected as 1288 according to the median of OAI of 100 random seeds. At the same time, the R2 of the four metals on the test set are 0.6, 0.7, 0.6 and 0.6 respectively.

[0095] As a preferred embodiment, in step S4, hyperparameter optimization is performed using the total training set on the obtained optimization framework of the multi-task learning;

[0096] The hyperparameter optimization method uses any of the grid search, random search, and particle swarm optimization methods, and performs five-fold cross validation to obtain the best multi-task learning hyperparameters;

[0097] Specifically, the hyperparameter optimization step is performed on the total training set, and the five-fold cross validation is performed without disrupting the grouping. The hyperparameters that need to be optimized are the learning rate of the multi-task learning model (optimization range is 0.001-0.01), the regularization parameter (optimization range is 10-50%), and the number of sample training (optimization range is 16, 32, and 64). The MSE of the model is used as the loss function, as shown in Figure 3 As shown, the model iterates 100 times to find the best hyperparameter combination of learning rate 0.003, regularization parameter 15%, and sample training number 16. At this time, the average RMSE of the multi-task learning model drops to a minimum of 15.8.

[0098] The present invention collects relevant subject data from public literature, considers risk management processes such as random seed screening and data leakage management, groups and divides the literature according to different experimental conditions, selects a multi-task learning model for hyperparameter optimization, reduces the risk of overfitting of the model, and finally obtains a reliable multi-metal leaching rate prediction model based on multi-task learning. Compared with traditional chemical experiments to determine metal leaching rates and single regression models, the present invention considers a more rigorous modeling process to predict the simultaneous leaching rate of multiple metals, which is simpler, greener, more reliable and stable.

[0099] Embodiment 2:

[0100] The present invention also provides a device, comprising a memory for storing computer program instructions and a processor for executing the computer program instructions, wherein when the computer program instructions are executed by the processor, the device is triggered to execute some or all of the steps in Embodiment 1;

[0101] The processor may include one or more processing units, for example, the processor may include an application processor (AP), a modem processor, a graphics processor (GPU), an image signal processor (ISP), a controller, a video codec, a digital signal processor (DSP), a baseband processor, and / or a neural-network processing unit (NPU), etc. Different processing units may be independent devices or integrated into one or more processors.

[0102] The controller can be the nerve center and command center of the electronic device. The controller can generate operation control signals according to the instruction operation code and timing signal to complete the control of fetching and executing instructions.

[0103] A memory may also be provided in the processor for storing instructions and data. In some embodiments, the memory in the processor is a cache memory. The memory may store instructions or data that the processor has just used or is cyclically used. If the processor needs to use the instruction or data again, it may be directly called from the memory. This avoids repeated access, reduces the waiting time of the processor, and thus improves the efficiency of the system.

[0104] Embodiment 3:

[0105] The present invention also provides a storage medium having a computer program stored thereon, wherein when the program is executed, the device where the storage medium is located is controlled to execute part or all of the steps in Example 1.

[0106] The storage medium may include a high-speed RAM memory, and may also include a nonvolatile memory (nonvolatile memory), such as at least one disk memory. It is understandable that the storage medium can be a random access memory (Random Access Memory, RAM), a disk, a hard disk, a solid state disk (solid state disk, SSD) or a nonvolatile memory, etc., which can store program codes and other machine-readable media.

[0107] Those skilled in the art will appreciate that the embodiments of the present invention may be provided as methods or storage media. Therefore, the embodiments of the present invention may take the form of complete hardware embodiments, complete software embodiments, or embodiments combining software and hardware. Moreover, the embodiments of the present invention may take the form of a computer program product implemented in one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes.

[0108] The above are only preferred embodiments of the present invention, and are not intended to limit the patent scope of the present invention. All equivalent structural changes made using the contents of the present invention's specification and drawings, or directly / indirectly applied in other related technical fields, are included in the patent protection scope of the present invention.

Claims

1. A method for predicting the leaching rate of multiple metals in waste lithium batteries, characterized in that: The steps include: S1: Collect relevant data, generate descriptors of organic acids using RDKit, and obtain the basic dataset for multi-task learning; S2: preprocessing the basic data set obtained in step S1 to obtain the feature subset that is most representative of the leaching rate; defining leaching tasks for multiple metals to be leached, and constructing a basic framework for multi-task learning; S3: Divide the preprocessed data set and perform indicator evaluation on the divided test set; determine the data division range based on the evaluation indicators, and adjust the basic framework of multi-task learning to obtain the optimization framework of multi-task learning; S4: Optimizing the hyperparameters of the multi-task learning optimization framework to obtain optimal multi-task learning hyperparameters; S5: training a multi-task learning model based on the optimal multi-task learning hyperparameters to obtain a multi-metal leaching rate prediction model based on multi-task learning; The unknown waste lithium battery leaching data is input into the multi-metal leaching rate prediction model based on multi-task learning to obtain the leaching rates of the multiple metals at the same time.

2. The method for predicting the leaching rate of multiple metals in waste lithium batteries according to claim 1, characterized in that: In step S1, the method steps for obtaining the basic data set for multi-task learning are as follows: S11: Obtain literature information on the data to be collected; S12: Use Origin-image digitization tool to obtain the leaching rate data of various metals in the literature, and record the relevant information on the properties of waste lithium batteries, types of organic acids and experimental conditions in the literature; S13: Organic acids were converted into Morgan fingerprints using RDKit as descriptors of organic acids.

3. The method for predicting the leaching rate of multiple metals in waste lithium batteries according to claim 1 or 2, characterized in that: In step S2, the method steps for preprocessing the basic data set obtained in step S1 are specifically as follows: S31: Classify the single variable experimental data into a group of data, remove duplicate values ​​and delete outliers from the processed data; S32: PCA dimensionality reduction is performed on the organic acid descriptors in the processed data, and the principal components of 95% of the explanatory variables are selected as the features after dimensionality reduction, and correlation tests are performed to obtain the feature subset that is most representative of the leaching rate.

4. The method for predicting the leaching rate of multiple metals in waste lithium batteries according to claim 3, characterized in that: In step S2, when constructing the basic architecture of multi-task learning, the shared layers and task-specific layers of the multi-task learning model are set, and the number of layers of the shared layers, the number of neurons in each layer, and the type of activation function are initialized; The characteristic task layer is provided with output layers corresponding to the various metals to be leached.

5. The method for predicting the leaching rate of multiple metals in waste lithium batteries according to claim 1 or 2, characterized in that: In step S2, one of the mean absolute error, mean square error, root mean square error, coefficient of determination or full precision index is used to evaluate the model effect of the constructed basic architecture of the multi-task learning.

6. The method for predicting the leaching rate of multiple metals in waste lithium batteries according to claim 5, characterized in that: The calculation formula of the mean absolute error is as follows: The calculation formula of the mean square error is as follows: The calculation formula of the root mean square error is as follows: The calculation formula of the determination coefficient is shown as follows: The calculation formula of the full-precision index is as follows: Where N is the number of samples, y is the predicted value of the output, is the true value of the output, y i and is the true value and predicted value of the i-th sample.

7. The method for predicting the leaching rate of multiple metals in waste lithium batteries according to claim 1 or 2, characterized in that: In step S3, the method steps for obtaining the optimization framework of multi-task learning are specifically as follows: S31: Divide the feature subset data into groups according to a preset ratio into a training set, a validation set, and a test set; S32: Select a preset number of random seeds that meet the proportional division for screening, set the training set and the validation set as the total training set, perform full-precision index evaluation on the test set, and select a random seed at the OAI median level for subsequent data division; S33: Evaluate different multi-task learning frameworks according to step S31 and step S32, evaluate the number of shared layers and the number of neurons, and determine the optimized framework of multi-task learning according to the optimal value of the full-precision indicator.

8. The method for predicting the leaching rate of multiple metals in waste lithium batteries according to claim 7, characterized in that: In step S4, hyperparameter optimization is performed using the total training set on the obtained optimization framework of the multi-task learning; The hyperparameter optimization method uses any of the grid search, random search, and particle swarm optimization methods, and performs five-fold cross validation to obtain the best multi-task learning hyperparameters.

9. A device, characterized in that: It comprises a memory for storing computer program instructions and a processor for executing the computer program instructions, wherein when the computer program instructions are executed by the processor, the device is triggered to execute a method for predicting the leaching rates of multiple metals in waste lithium batteries as described in any one of claims 1 to 8.

10. A storage medium, characterized in that: A computer program is stored thereon, wherein when the computer program is running, the device where the storage medium is located is controlled to execute a method for predicting the leaching rates of multiple metals in waste lithium batteries as described in any one of claims 1-8.

Images