Lithium iron phosphate battery material separation and purification system and method under intelligent control

Abstract

The present invention discloses a system and method for separating and purifying lithium iron phosphate battery materials under intelligent control, and relates to the field of intelligent control technology. The method comprises: collecting initial data of the first separation of lithium iron phosphate batteries; inputting the collected initial data into a pre-constructed first machine learning model, and outputting preliminary separation parameters; generating crushing instructions according to the preliminary separation parameters; collecting pH value, ion concentration and temperature of lithium iron phosphate battery recovery liquid; inputting the collected pH value, ion concentration and temperature of lithium iron phosphate battery recovery liquid into a pre-constructed second machine learning model, and outputting an optimal adjustment strategy; generating adjustment instructions according to the optimal adjustment strategy; the present invention can realize more accurate preliminary separation parameter setting and subsequent adjustment strategy selection through intelligent analysis and planning of machine learning models, thereby improving the extraction efficiency and recovery rate of separated materials.

Description

Technical Field

[0001] The present invention relates to the field of intelligent control technology, and more specifically, to a system and method for separating and purifying lithium iron phosphate battery materials under intelligent control. Background Art

[0002] With the rapid development of electric vehicles and energy storage systems, the demand for lithium iron phosphate batteries continues to increase, which also brings a large number of waste lithium iron phosphate batteries. In order to achieve sustainable utilization of resources and environmental protection, the recycling of lithium iron phosphate battery materials has become an important research direction.

[0003] Intelligent control can achieve efficient recycling of waste lithium iron phosphate battery materials, reducing resource waste and environmental pollution. This is of great significance for promoting sustainable energy development and realizing a circular economy.

[0004] The Chinese patent with the existing authorization announcement number CN113097591 B discloses a method for recycling positive electrode materials of lithium iron phosphate batteries, including: ultrasonic alkali washing of waste positive electrode sheets of waste lithium iron phosphate batteries, filtering to obtain active materials of the electrode sheets; dissolving the active materials of the electrode sheets in a first acid solution, adding an oxidant, controlling the pH of the reaction solution to 2.0-2.3 at the end of the reaction, filtering and separating to obtain a first filtrate to be recovered and phosphorus iron slag; adjusting the pH of the first filtrate to be recovered to be greater than or equal to 5 with an alkali solution, filtering and removing the precipitate, adding phosphate to react, controlling the pH of the reaction solution to 10-11 at the end of the reaction, filtering and removing the filtrate and drying to recover lithium phosphate; dissolving the phosphorus iron slag in a second acid solution, filtering and removing the filter residue, adding a complexing agent, lithium hydroxide or lithium carbonate and a carbon source to react, separating to obtain a colloidal precipitate and crushing and roasting, and recovering lithium iron phosphate. Through the above method, the recovery process is simplified, the process is short and the energy consumption is low, and the synthesized material has a high compaction density and good electrical properties.

[0005] The separation and purification of lithium iron phosphate battery materials has broad application prospects in recycling and utilization, but the current mainstream separation and purification of lithium iron phosphate battery materials is still mainly based on manual operation of machines. This is because the separation and purification steps of lithium iron phosphate battery materials are complicated, and the recycling schemes of various types of lithium iron phosphate batteries are also different. In a complex environment, both the initial crushing and separation stage and the later purification stage require manual operation of the machine for screening, classification and adjustment, but manual operation will inevitably cause problems as the working time goes by. At present, the shell stripping work in the crushing stage of lithium iron phosphate batteries is either directly concentrated crushing without classification and then separated and purified in the later stage, or manual sorting and classification. Centralized crushing will increase the burden of later separation and purification operations and reduce the fault tolerance rate, while manual sorting is inefficient and increases labor costs. After the battery has been crushed, immersed in liquid and other procedures, it will enter the recovery liquid purification stage. At this time, the pH value of the liquid will affect the recovery rate, but there are fewer neutral substances in nature, so in an extraction environment that is not completely vacuum, it is inevitable that acidic or alkaline substances will enter the recovery liquid and affect the pH value of the liquid. At this time, it is necessary to monitor and adjust the pH in real time to maximize the recovery rate. The ambient temperature also affects the recovery rate of the recovery liquid. The temperature also needs to be adjusted to maximize the recovery rate and increase the profit of recycling. Summary of the invention

[0006] In order to solve the above problems, the present invention provides a system and method for separating and purifying lithium iron phosphate battery materials under intelligent control. The method for separating and purifying lithium iron phosphate battery materials under intelligent control includes:

[0007] Collecting initial data of the first separation of the lithium iron phosphate battery, and sending the initial data to the first analysis module;

[0008] Inputting the collected initial data into a pre-built first machine learning model, outputting preliminary separation parameters, and sending the preliminary separation parameters to a first execution module;

[0009] Generate a crushing instruction based on the preliminary separation parameters and send it to the machine end;

[0010] Collect the pH value, ion concentration and temperature of the lithium iron phosphate battery recovery liquid, and send the pH value, ion concentration and temperature to the second analysis module;

[0011] Input the collected pH value, ion concentration and temperature of the lithium iron phosphate battery recovery liquid into the pre-built second machine learning model, output the optimal adjustment strategy, and send the optimal adjustment strategy to the second execution module;

[0012] Generate adjustment instructions based on the optimal adjustment strategy and send them to the refining machine.

[0013] Furthermore, the initial data is the initial data of the lithium iron phosphate battery before it is crushed and sent to a recycling plant during the separation and purification process of the lithium iron phosphate battery; the initial data includes the volume and packaging material type of the lithium iron phosphate battery.

[0014] Furthermore, the preliminary separation parameters include parameter 1: applicable crusher model; parameter 2: applicable screen type; parameter 3: recommended preliminary crushing parameters, the recommended preliminary crushing parameters include material particle size and crushing speed; parameter 4: recommended preliminary screening parameters, the recommended preliminary screening parameters include screen mesh and screening frequency.

[0015] Furthermore, the method for constructing the first machine learning model includes:

[0016] Initialize the first machine learning model structure. The first machine learning model structure adopts an MLP type multi-layer forward network structure with one input layer, two hidden layers, and four output layers. The input layer is the first input layer. The number of nodes in the first input layer of the first input layer is 2, corresponding to the volume and packaging material type of the lithium iron phosphate battery respectively. The hidden layer includes the first hidden layer and the second hidden layer. The number of nodes in the first hidden layer of the first hidden layer is 128, and the Relu function is used as the activation function. The number of nodes in the second hidden layer of the second hidden layer is 64, and the Relu function is used as the activation function. The output layer includes the first output layer. Layer, second output layer, third output layer and fourth output layer, the number of nodes in the first output layer of the first output layer is [1, n], predicting the applicable crusher model, n is the number of crusher types; the number of nodes in the second output layer of the second output layer is [1, m], predicting the applicable screen type, m is the number of screen types; the third output layer nodes of the third output layer include nodes 1 and 2, node 1 is the predicted material particle size, node 2 is the predicted crushing speed; the fourth output layer nodes of the fourth output layer include nodes 11 and 12, node 11 is the predicted screen mesh, node 12 is the predicted screening frequency;

[0017] After initializing the first machine learning model structure, the initial data and the preliminary separation parameters corresponding to the initial data are used to train the first machine learning model. The initial data and the preliminary separation parameters corresponding to the initial data used to train the first machine learning model are obtained from a database. The database records each lithium iron phosphate battery crushing record for optimizing the model. The optimizer is Adam, the loss function is the MSE loss function, the batch size is set to 32, the number of iterations is 200 rounds, and the training is terminated when the loss function converges.

[0018] Furthermore, the first machine learning model training method includes:

[0019] Convert the initial data into a corresponding set of feature vectors;

[0020] The initial data is used as the input of the machine learning model, the machine learning model takes a set of preliminary separation parameters corresponding to each set of initial data as output, the preliminary separation parameters actually corresponding to each set of initial data are used as prediction targets, and minimizing the loss function value of the machine learning model is used as the training target; training is stopped when the loss function value of the machine learning model is less than or equal to the preset target loss value.

[0021] Furthermore, the second machine learning model is a reinforcement learning model, using a Q network structure, and the construction steps are as follows:

[0022] Step 1: Design state representation dimensions:

[0023] The design state representation dimension includes design PH state discreteness and design temperature state discreteness;

[0024] The PH state is discrete as follows: (0-3] PH_iron: (3-5] PH_copper: (5-7] (7-14], using One-hot encoding; the temperature state is discrete as follows: (-15℃—5℃] (-5℃—5℃] (5℃—15℃], using One-hot encoding; the state representation dimension is One-hot encoding.

[0025] Step 2: Design the action representation dimension:

[0026] The design action representation dimensions include a design pH adjustment action range and a design temperature adjustment action range;

[0027] The pH adjustment range is: -0.5, -0.2, 0, 0.2, 0.5, using One-hot encoding;

[0028] The temperature adjustment action range is: -5°C, -2°C, 0°C, 2°C, 5°C, using One-hot encoding; the action representation dimension is One-hot encoding.

[0029] Step 3: Build the Q network:

[0030] The Q network has 1 input layer, 2 hidden layers, and 1 output layer. The input layer is the second input layer, and the number of nodes is 4+3=7. The hidden layer includes the third hidden layer and the fourth hidden layer. The activation function of the third hidden layer is the Relu function, and the number of nodes is 14. The structure of the fourth hidden layer is the same as the third hidden layer. The output layer is the fifth output layer, and the number of nodes in the fifth output layer is 10, which is linearly activated.

[0031] Step 4. Set the reward function:

[0032] The reward function formula is:

[0033] Q target=R+γ×maxQ target (s',a');

[0034] In the formula, Q target is the target Q value, R is the reward value, R = 1-(current ion concentration-center of ion concentration target value range), where the center of ion concentration target value range is a preset value, which is determined by technicians in this field based on a large number of experiments, s' is the next state, a' is the optimal action under the next state s', γ is the discount rate, and the initial value is set to 0.9; maxQ target (s',a') is the maximum target Q value of the optimal action for the next state.

[0035] Step 5: Set the optimized execution plan:

[0036] The optimizer is Adam, the learning rate is 0.001, the experience replay pool capacity is 10w, the Q network is updated every 100 steps, and the ε-greedy strategy is used to select actions.

[0037] Furthermore, the method also includes training and optimizing the second machine learning model, including:

[0038] Call c recycling liquid adjustment records in the database as c groups of historical training data, the recycling liquid adjustment records include pH value, temperature and optimal adjustment strategy, and divide c groups of historical training data into training set and test set, with a ratio of 4:1; initialize the experience replay pool, start from a random initial state, adopt ε-greedy strategy to select action, execute action, observe reward and next state, store experience in the replay pool, every δ steps, δ is a preset value, determined by technicians in this field according to data fitting, randomly sample experience samples from the replay pool for learning, and the learning method includes: sampling a (s, a, a', s'), s is the current state, a is the current action, and calculate Q target =R+γ×maxQ target (s',a'), calculate the loss function, the loss function loss formula is:

[0039] loss=(Q target -Q(s, a, θ) 2 ;

[0040] Where θ is the Q network parameter, which is initially 0; Q(s, a, θ) is the actual preset reward value obtained by using the current action according to the Q network parameter in the current state;

[0041] The loss is back-propagated to optimize the Q network parameters. The Adam optimizer is used, and ε gradually decays from 1 to the minimum value. The Q network parameters obtained by training are enhanced. The Q network effect is tested every δ steps. The training is continued until the loss function converges. The Q network parameters are output and the control effect is evaluated on the test set. When the reward value is higher than the preset threshold 0.9 for N consecutive evaluation cycles, N is the preset value, the training can be terminated.

[0042] Lithium iron phosphate battery material separation and purification system under intelligent control, the system includes:

[0043] A first collection module, used for collecting initial data of the first separation of the lithium iron phosphate battery, and sending the initial data to a first analysis module;

[0044] a first analysis module, configured to input the collected initial data into a pre-built first machine learning model, output preliminary separation parameters, and send the preliminary separation parameters to a first execution module;

[0045] A first execution module is used to generate a crushing instruction according to the preliminary separation parameters and send it to the crushing machine end;

[0046] The second collection module collects the pH value, ion concentration and temperature of the lithium iron phosphate battery recovery liquid, and sends the pH value, ion concentration and temperature to the second analysis module;

[0047] The second analysis module is used to input the collected pH value, ion concentration and temperature of the lithium iron phosphate battery recovery liquid into a pre-built second machine learning model, output an optimal adjustment strategy, and send the optimal adjustment strategy to the second execution module;

[0048] The second execution module is used to generate an adjustment instruction according to the optimal adjustment strategy and send it to the refining machine end.

[0049] An electronic device comprises a memory, a processor and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, a method for separating and purifying lithium iron phosphate battery materials under intelligent control is implemented.

[0050] A computer-readable storage medium stores a computer program, and when the computer program is executed, a method for separating and purifying lithium iron phosphate battery materials under intelligent control is implemented.

[0051] The technical effects and advantages of the lithium iron phosphate battery material separation and purification system and method under intelligent control of the present invention are as follows:

[0052] Through intelligent analysis and planning of machine learning models, more accurate initial separation parameter settings and subsequent adjustment strategy selection can be achieved, thereby improving the extraction efficiency and recovery rate of separated materials; closed-loop data transmission and instruction issuance are realized between system modules, and the main processes are automatically executed, greatly reducing human manipulation and improving the automation level of the separation and purification process; the model learns historical experience knowledge and can select the optimal solution for different conditions, making the quality of separation and purification products more stable and controllable; automation improves production efficiency and saves labor costs; optimizing parameters reduces repetitive operations, reduces material loss, and reduces production input costs; the method ideas and constituent modules designed by the system have a certain degree of universality and can be applied to the intelligent control of other similar separation processes; the system has good openness, and the models and algorithms can be continuously improved, and new knowledge obtained from practical applications is absorbed by each other to achieve continuous improvement of the intelligence level. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] Figure 1 This is a schematic diagram of the structural principle of the lithium iron phosphate battery material separation and purification system under intelligent control in Example 1 of the present invention;

[0054] Figure 2 This is a flow chart of the method for separating and purifying lithium iron phosphate battery materials under intelligent control of the present invention;

[0055] Figure 3 is a schematic diagram of an electronic device of the present invention;

[0056] Figure 4 Schematic diagram of the storage medium of the present invention. DETAILED DESCRIPTION

[0057] 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.

[0058] Example 1

[0059] like Figure 1 As shown, this embodiment discloses a lithium iron phosphate battery material separation and purification system under intelligent control, the system comprising:

[0060] The first collection module is used to collect initial data of the first separation of the lithium iron phosphate battery and send the initial data to the first analysis module.

[0061] Among them, the initial data is the data of the lithium iron phosphate battery before it is crushed and sent to the recycling plant during the separation and purification process of the lithium iron phosphate battery; the initial data includes the volume of the lithium iron phosphate battery and the type of packaging material. The volume of the lithium iron phosphate battery is obtained by a laser rangefinder. The volume of the lithium iron phosphate battery will affect the selection of the machine for processing the lithium iron phosphate battery. The lithium iron phosphate battery with a larger volume needs a shearing machine to shear and then send it to the separation machine; the type of lithium iron phosphate battery packaging material is obtained by an infrared spectrometer. The type of lithium iron phosphate battery packaging material will affect the first separation method. For lithium iron phosphate batteries with metal shells, mechanical separation methods can be adopted, such as using a shearing machine, a crusher or a crusher to peel or crush the metal shell and separate the metal part from the internal lithium iron phosphate battery components. The separated metal shell can be recycled and reused for metal; for lithium iron phosphate batteries with plastic shells, physical separation methods can be adopted, such as using technologies such as vibrating screens and airflow separation to separate the plastic shell from the internal lithium iron phosphate battery components. The separated plastic shell can be recycled and reused for plastic.

[0062] The first analysis module is used to input the collected initial data into a pre-built first machine learning model, output preliminary separation parameters, and send the preliminary separation parameters to the first execution module.

[0063] Among them, the preliminary separation parameters include parameter 1: applicable crusher model, such as JC100 shear crusher, ZM200 impact crusher and SM400 axial flow electromagnetic crusher; parameter 2: applicable screen type; parameter 3: recommended preliminary crushing parameters, the recommended preliminary crushing parameters include material particle size and crushing speed; parameter 4: recommended preliminary screening parameters, the recommended preliminary screening parameters include screen mesh and screening frequency.

[0064] The method for constructing the first machine learning model includes:

[0065] Initialize the first machine learning model structure. The first machine learning model structure adopts an MLP type multi-layer forward network structure with one input layer, two hidden layers, and four output layers. The input layer is the first input layer. The number of nodes in the first input layer of the first input layer is 2, corresponding to the volume and packaging material type of the lithium iron phosphate battery respectively. The hidden layer includes the first hidden layer and the second hidden layer. The number of nodes in the first hidden layer of the first hidden layer is 128, and the Relu function is used as the activation function. The number of nodes in the second hidden layer of the second hidden layer is 64, and the Relu function is used as the activation function. The output layer includes the first output layer. The first output layer, the second output layer, the third output layer and the fourth output layer, the first output layer node number of the first output layer is [1, n], predicting the applicable crusher model, n is the number of crusher types; the second output layer node number of the second output layer is [1, m], predicting the applicable screen type, m is the number of screen types; the third output layer nodes of the third output layer include nodes 1 and 2, node 1 is the predicted material particle size, node 2 is the predicted crushing speed; the fourth output layer nodes of the fourth output layer include nodes 11 and 12, node 11 is the predicted screen mesh, node 12 is the predicted screening frequency.

[0066] After initializing the first machine learning model structure, the initial data and the preliminary separation parameters corresponding to the initial data are used to train the first machine learning model. The initial data and the preliminary separation parameters corresponding to the initial data are obtained from a database. The database records each lithium iron phosphate battery crushing record for optimizing the model. The optimizer is Adam, the loss function is the MSE loss function, the batch size is set to 32, the number of iterations is 200 rounds, and the training ends when the loss function converges, indicating that the training is completed.

[0067] It should be noted that the Adam optimizer is one of the five major optimizers commonly used in machine learning, and its full name is Adaptive Moment Estimation.

[0068] The first machine learning model training method comprises:

[0069] Convert the initial data into a corresponding set of feature vectors.

[0070] The initial data is used as the input of the machine learning model, the machine learning model takes a set of preliminary separation parameters corresponding to each set of initial data as output, the preliminary separation parameters actually corresponding to each set of initial data are used as prediction targets, and minimizing the loss function value of the machine learning model is used as the training target; training is stopped when the loss function value of the machine learning model is less than or equal to the preset target loss value.

[0071] The loss function value of the first machine learning model is the mean square error;

[0072] Mean square error is one of the commonly used loss functions. By transforming the loss function formula Minimization is used to train the model so that the machine learning model can better fit the data, thereby improving the performance and accuracy of the model.

[0073] In the loss function, MSE is the loss function value of the first machine learning model, x is the feature vector group number; e is the number of feature vector groups; y is the loss function value of the first machine learning model, x is the number of feature vector groups ... x is a set of preliminary separation parameters output by the first machine learning model corresponding to the x-th set of feature vectors, is the initial separation parameter actually corresponding to the xth group of eigenvectors.

[0074] It should be noted that since it is strategic data training, the optimal decision control group of the initial sample of the preliminary separation parameters corresponding to the initial data is obtained by manually making the optimal decision and recording it after manual experiments. At this stage, the initial stage of decision-making training data requires manual decision-making. Machine learning is the result of logical judgment and classification summary based on manual decision-making. The most typical example is the manually labeled data involved in the big data feeding of GPT.

[0075] The first execution module is used to generate a crushing instruction according to the preliminary separation parameters and send it to the crushing machine end.

[0076] The second collection module is used to collect the pH value, ion concentration and temperature of the lithium iron phosphate battery recovery liquid, and send the pH value, ion concentration and temperature to the second analysis module.

[0077] Among them, the pH value of lithium iron phosphate battery recovery liquid determines the acidity and alkalinity of the solution. Acidic or alkaline environments affect the solubility of metal ions. The concentration of most metal ions increases with the increase of pH value. For example, the peak value of Cu2+ concentration appears under near-neutral conditions. The concentration of some metal ions is very sensitive to pH value. For example, the concentration of Fe3+ drops rapidly in solutions with pH>3. The increase in temperature will increase the kinetic energy of molecular bonding in the solution, thereby increasing the solubility of metal ions, but too high a temperature may also cause the ions to precipitate into a solid state, resulting in a decrease in concentration. This critical temperature is different for different metals. For the same metal ion, its concentration-temperature curve usually presents a peak, indicating that there is a temperature that maximizes the concentration. In addition, temperature will affect the rate of chemical reactions and the dynamic equilibrium of ion formation or decomposition. The ion concentration is obtained by the ion detector, which is closely related to both pH value and temperature, and can be used as a criterion for process optimization.

[0078] The second analysis module is used to input the collected pH value, ion concentration and temperature of the lithium iron phosphate battery recovery liquid into a pre-built second machine learning model, output the optimal adjustment strategy, and send the optimal adjustment strategy to the second execution module.

[0079] The second machine learning model is a reinforcement learning model, using a Q network structure, and the construction steps are as follows:

[0080] Step 1: Design state representation dimensions:

[0081] The design state representation dimension includes design PH state discreteness and design temperature state discreteness;

[0082] The PH state is discrete as follows: (0-3] PH_iron: (3-5] PH_copper: (5-7] (7-14], using One-hot encoding; the temperature state is discrete as follows: (-15℃-5℃] (-5℃-5℃] (5℃-15℃], using One-hot encoding; the state representation dimension is One-hot encoding, and is the sum of the One-hot dimension sum of the PH state and the One-hot dimension sum of the temperature state.

[0083] Step 2: Design the action representation dimension:

[0084] The design action representation dimensions include a design pH adjustment action range and a design temperature adjustment action range;

[0085] The pH adjustment range is: -0.5, -0.2, 0, 0.2, 0.5, using One-hot encoding;

[0086] The temperature adjustment action range is: -5°C, -2°C, 0°C, 2°C, 5°C, using One-hot encoding; the action representation dimension is One-hot encoding, and is the sum of the PH action One-hot dimension sum and the temperature action One-hot dimension sum.

[0087] It is particularly noted that the state representation dimension and the action representation dimension are not the state and action under normal conditions, but are improvements based on the particularity of the state and action of the present invention. PH, ion concentration and temperature data cannot be universal, but when designing states and actions, ion concentration and PH are highly bound together, and although ion concentration is not used as a strategy output in actions, the mutual influence of ion concentration and PH still needs to be considered. It is precisely under this complex state and action design that One-hot encoding expression is required to facilitate subsequent calculations without considering unit issues and to comprehensively consider the setting of reward values; in the form of One-hot encoding, the state representation dimension refers to the total feature space dimension occupied by each environmental state variable after One-hot encoding, which plays the role of the feature vector dimension representing the environmental state. For example, after the PH state is one-hot encoded, each state is represented by a one-hot vector, such as: (0-3] corresponds to the vector [1, 0, 0, 0], (3-5] corresponds to the vector [0, 1, 0, 0], (5-7] corresponds to the vector [0, 0, 1, 0], (7-14] corresponds to the vector [0, 0, 0, 1]; the temperature state is also represented by the same regular one-hot vector after one-hot encoding. The state representation dimension refers to the sum of the dimensions of each state vector after one-hot encoding of the PH state, plus the sum of the dimensions of each state vector after one-hot encoding of the temperature state. For example, there are 4 PH states, each of which is a 4-dimensional one-hot vector; there are 3 temperature states, each of which is a 3-dimensional one-hot vector, then the state representation dimension is:

[0088] The sum of the one-hot dimensions of the PH state = 4 × 4 = 16;

[0089] Temperature state One-hot dimension sum = 3 × 3 = 9;

[0090] The state representation dimension = the sum of the one-hot dimensions of the PH state + the sum of the one-hot dimensions of the temperature state = 16 + 9 = 25, and the same is true for the action representation dimension.

[0091] Step 3: Build the Q network:

[0092] The Q network includes 1 input layer, 2 hidden layers and 1 output layer. The input layer is the second input layer, and the number of nodes is 4+3=7. The hidden layer includes the third hidden layer and the fourth hidden layer. The activation function of the third hidden layer is the Relu function, and the number of nodes is 14. The structure of the fourth hidden layer is the same as the third hidden layer. The output layer is the fifth output layer, and the number of nodes in the fifth output layer is 10, which is linearly activated.

[0093] Step 4. Set the reward function:

[0094] The reward function formula is:

[0095] Qtarget =R+γ×maxQ target (s',a');

[0096] In the formula, Q target is the target Q value, R = 1-(current ion concentration-center of ion concentration target value range), where the center of ion concentration target value range is a preset value, which is determined by technicians in this field based on a large number of experiments, s' is the next state, a' is the optimal action under the next state s', γ is the discount rate, the initial value is set to 0.9, maxQ target (s',a') is the maximum target Q value of the optimal action for the next state.

[0097] Step 5: Set the optimized execution plan:

[0098] The optimizer is Adam, the learning rate is 0.001, the experience replay pool capacity is 10w, the Q network is updated every 100 steps, and the ε-greedy strategy is used to select actions.

[0099] The training optimization method of the second machine learning model includes:

[0100] Call c recycling liquid adjustment records in the database as c groups of historical training data, the recycling liquid adjustment records include pH value, temperature and optimal adjustment strategy, and divide c groups of historical training data into training set and test set, with a ratio of 4:1; initialize the experience replay pool, start from a random initial state, adopt ε-greedy strategy to select action, execute action, observe reward and next state, store experience in the replay pool, every δ steps, δ is a preset value, determined by technicians in this field according to data fitting, randomly sample experience samples from the replay pool for learning, and the learning method includes: sampling a (s, a, a', s'), s is the current state, a is the current action, and calculate Q target =R+γ×maxQ target (s',a'), calculate the loss function, the loss function loss formula is:

[0101] loss=(Q target -Q(s, a, θ) 2 ;

[0102] Where θ is the Q network parameter, which is initially 0; Q(s, a, θ) is the actual preset reward value obtained by using the current action in the current state according to the current Q network parameters.

[0103] It is particularly noted that Q(s, a, θ) in the training phase is determined by a person skilled in the art as a value guide to provide a training direction so that the direction of model change turns to the development direction of pre-solving the problem of the present invention.

[0104] The loss is back-propagated to optimize the Q network parameters. The Adam optimizer is used, and ε gradually decays from 1 to the minimum value. The Q network parameters obtained by training are enhanced. The Q network effect is tested every δ steps. The training is continued until the loss function converges. The Q network parameters are output and the control effect is evaluated on the test set. When the average reward value is higher than the preset threshold of 0.9 for N consecutive evaluation cycles, the training can be terminated.

[0105] The second execution module is used to generate an adjustment instruction according to the optimal adjustment strategy and send it to the refining machine end.

[0106] It is particularly noted that the lithium iron phosphate battery material separation and purification system under intelligent control connects the crushing machine end and the refining machine end in a wireless or wired manner.

[0107] Example 2

[0108] like Figure 2 As shown, this embodiment discloses a method for separating and purifying lithium iron phosphate battery materials under intelligent control, the method comprising:

[0109] Collect initial data for the first separation of lithium iron phosphate batteries.

[0110] The collected initial data is input into the pre-built first machine learning model to output preliminary separation parameters.

[0111] Generate crushing instructions based on preliminary separation parameters.

[0112] Collect the pH value, ion concentration and temperature of lithium iron phosphate battery recovery liquid.

[0113] The collected pH value, ion concentration and temperature of the lithium iron phosphate battery recovery liquid are input into the pre-built second machine learning model to output the optimal adjustment strategy.

[0114] Generate adjustment instructions based on the optimal adjustment strategy.

[0115] Example 3

[0116] See also Figure 3 As shown, the present application also provides an electronic device 500. The electronic device 500 may include one or more processors and one or more memories. The memories store computer-readable codes, and when the computer-readable codes are run by one or more processors, the method for separating and purifying lithium iron phosphate battery materials under intelligent control as described above may be executed.

[0117] The method or system according to the embodiment of the present application can also be used by Figure 3 The electronic device architecture shown in FIG. Figure 3As shown, the electronic device 500 may include a bus 501, one or more CPUs 502, ROM 503, RAM 504, a communication port 505 connected to a network, an input / output 506, a hard disk 507, etc. The storage device in the electronic device 500, such as ROM 503 or hard disk 507, may store the method for separating and purifying lithium iron phosphate battery materials under intelligent control provided in the present application. Furthermore, the electronic device 500 may also include a user interface 508. Of course, Figure 3 The architecture shown is only exemplary and can be omitted according to actual needs when implementing different devices. Figure 3 One or more components of an electronic device are shown.

[0118] Example 4

[0119] See also Figure 4 As shown, the present application also discloses a computer storage medium. Computer-readable instructions are stored on the computer storage medium. When the computer-readable instructions are executed by the processor, the method for separating and purifying lithium iron phosphate battery materials under intelligent control according to the embodiment of the present application described with reference to the above figures can be executed. The storage medium includes, but is not limited to, for example, volatile memory and / or non-volatile memory. Volatile memory may include, for example, random access memory (RAM) and cache memory (cache), etc. Non-volatile memory may include, for example, read-only memory (ROM), hard disk, flash memory, etc.

[0120] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

[0121] Finally: The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for separating and purifying lithium iron phosphate battery materials under intelligent control, characterized in that: Methods include: Collect initial data of the first separation of lithium iron phosphate batteries; Inputting the collected initial data into a pre-built first machine learning model to output preliminary separation parameters; The method for constructing the first machine learning model includes: Initialize the first machine learning model structure. The first machine learning model structure adopts an MLP type multi-layer forward network structure. The MLP type multi-layer forward network structure includes an input layer, two hidden layers and four output layers. The input layer is the first input layer. The number of nodes in the first input layer of the first input layer is 2, corresponding to the volume and packaging material type of the lithium iron phosphate battery respectively. The hidden layer includes the first hidden layer and the second hidden layer. The number of nodes in the first hidden layer of the first hidden layer is 128, and the Relu function is used as the activation function. The number of nodes in the second hidden layer of the second hidden layer is 64, and the Relu function is used as the activation function. The output layer includes the first output layer, the second output layer, the third output layer and the fourth output layer. The number of nodes in the first output layer of the first output layer is [1, n], and the applicable crusher model is predicted, where n is the number of crusher types; the number of nodes in the second output layer of the second output layer is [1, m], and the applicable screen type is predicted, where m is the number of screen types; the third output layer nodes of the third output layer include nodes 1 and 2, where node 1 is the predicted material particle size, and node 2 is the predicted crushing speed; the fourth output layer nodes of the fourth output layer include nodes 11 and 12, where node 11 is the predicted screen mesh, and node 12 is the predicted screening frequency; After initializing the first machine learning model structure, the first machine learning model is trained using the initial data and the preliminary separation parameters corresponding to the initial data. The optimizer is Adam, the loss function is the MSE loss function, the batch size is set to 32, the number of iterations is 200, and the training is terminated when the loss function converges. The first machine learning model training method comprises: Convert the initial data into a corresponding set of feature vectors; The initial data is used as the input of the machine learning model, the machine learning model takes a set of preliminary separation parameters corresponding to each set of initial data as output, takes the preliminary separation parameters actually corresponding to each set of initial data as the prediction target, and takes minimizing the loss function value of the machine learning model as the training target; when the loss function value of the machine learning model is less than or equal to the preset target loss value, the training is stopped; Generate crushing instructions according to preliminary separation parameters; Collect the pH value, ion concentration and temperature of lithium iron phosphate battery recovery liquid; Input the collected pH value, ion concentration and temperature of the lithium iron phosphate battery recovery liquid into the pre-built second machine learning model to output the optimal adjustment strategy; The second machine learning model is a reinforcement learning model, using a Q network structure, and the construction steps are as follows: Step 1: Design state representation dimensions: The design state representation dimension includes design PH state discreteness and design temperature state discreteness; The PH state is discrete as follows: (0-3] PH_iron: (3-5] PH_copper: (5-7] (7-14], using One-hot encoding; the temperature state is discrete as follows: (-15°C-5°C] (-5°C-5°C] (5°C-15°C], using One-hot encoding; the state representation dimension is One-hot encoding; Step 2: Design the action representation dimension: The design action representation dimensions include a design pH adjustment action range and a design temperature adjustment action range; The pH adjustment range is: -0.5, -0.2, 0, 0.2, 0.5, using One-hot encoding; The temperature adjustment action range is: -5°C, -2°C, 0°C, 2°C, 5°C, using One-hot encoding; the action representation dimension is One-hot encoding; Step 3: Build the Q network: The Q network includes 1 input layer, 2 hidden layers and 1 output layer. The input layer is the second input layer with 7 nodes. The hidden layer includes the third hidden layer and the fourth hidden layer. The activation function of the third hidden layer is the Relu function with 14 nodes. The structure of the fourth hidden layer is the same as the third hidden layer. The output layer is the fifth output layer with 10 nodes and linear activation. Step 4. Set the reward function: The reward function formula is: Q target =R+γ×maxQ target (s',a'); In the formula, Q target is the target Q value, R is the reward value, R = 1-(current ion concentration-center of ion concentration target value range), where the center of ion concentration target value range is the preset value, s' is the next state, a' is the optimal action under the next state s', γ is the discount rate, and the initial value is set to 0.9; maxQ target (s', a') is the maximum target Q value of the optimal action for the next state; Step 5: Set the optimized execution plan: The optimizer is Adam, the learning rate is 0.001, the experience replay pool capacity is 10w, the Q network is updated every 100 steps, and the ε-greedy strategy is used to select actions; The method also includes training and optimizing the second machine learning model, including: Call c recycling liquid adjustment records in the database as c groups of historical training data. The recycling liquid adjustment records include pH value, temperature and optimal adjustment strategy. The c groups of historical training data are divided into training set and test set with a ratio of 4:

1. Initialize the experience replay pool, start from a random initial state, use the ε-greedy strategy to select actions, execute actions, observe rewards and the next state, store experience in the replay pool, and randomly sample experience samples from the replay pool for learning every δ steps, δ is a preset value. The learning method includes: sampling a (s, a, a', s'), s is the current state, a is the current action, and calculate Q target =R+γ×maxQ target (s',a'), calculate the loss function, the loss function loss formula is: loss=(Q target -Q(s,a,θ)) 2 ; Where θ is the Q network parameter, which is initially 0; Q(s, a, θ) is the actual preset reward value obtained by using the current action according to the Q network parameter in the current state; The loss is back-propagated to optimize the Q network parameters. The Adam optimizer is used. ε gradually decays from 1 to the minimum value. The Q network parameters obtained by training are enhanced. The Q network effect is tested every δ steps. The training is continued until the loss function converges. The Q network parameters are output and the control effect is evaluated on the test set. When the reward value is higher than the preset threshold 0.9 for N consecutive evaluation cycles, N is the preset value, the training is terminated. Generate adjustment instructions based on the optimal adjustment strategy.

2. The method for separating and purifying lithium iron phosphate battery materials under intelligent control according to claim 1, characterized in that: The initial data is the initial data of the lithium iron phosphate battery before it is crushed and sent to the recycling plant during the separation and purification process of the lithium iron phosphate battery; the initial data includes the volume and packaging material type of the lithium iron phosphate battery.

3. The method for separating and purifying lithium iron phosphate battery materials under intelligent control according to claim 2, characterized in that: The preliminary separation parameters include parameter 1: applicable crusher model; parameter 2: applicable screen type; parameter 3: recommended preliminary crushing parameters, including material particle size and crushing speed; parameter 4: recommended preliminary screening parameters, including screen mesh and screening frequency.

4. The lithium iron phosphate battery material separation and purification system under intelligent control is characterized by: The method for separating and purifying lithium iron phosphate battery materials under intelligent control according to any one of claims 1 to 3 is implemented, and the system comprises: A first collection module, used for collecting initial data of the first separation of the lithium iron phosphate battery, and sending the initial data to a first analysis module; a first analysis module, configured to input the collected initial data into a pre-built first machine learning model, output preliminary separation parameters, and send the preliminary separation parameters to a first execution module; A first execution module is used to generate a crushing instruction according to the preliminary separation parameters and send it to the crushing machine end; The second collection module collects the pH value, ion concentration and temperature of the lithium iron phosphate battery recovery liquid, and sends the pH value, ion concentration and temperature to the second analysis module; The second analysis module is used to input the collected pH value, ion concentration and temperature of the lithium iron phosphate battery recovery liquid into a pre-built second machine learning model, output an optimal adjustment strategy, and send the optimal adjustment strategy to the second execution module; The second execution module is used to generate an adjustment instruction according to the optimal adjustment strategy and send it to the refining machine end.

5. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the method for separating and purifying lithium iron phosphate battery materials under intelligent control as described in any one of claims 1 to 3 is implemented.

6. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed, the method for separating and purifying lithium iron phosphate battery materials under intelligent control as described in any one of claims 1 to 3 is implemented.

Images