A control method for lithium battery electrode rolling equipment

Abstract

This invention provides a control method for lithium battery electrode rolling equipment. It is based on a BP neural network to construct a parameter prediction model for the rolling equipment, and then optimizes the model using a GA genetic algorithm to obtain the optimal parameters. During processing, the equipment parameters of the lithium battery electrode rolling equipment are collected in real time and input into the trained parameter prediction model. The model then outputs the corresponding predicted rolling force. This method firstly reduces the workload of debugging and experimentation before formal production, thus lowering the overall cost. Furthermore, during processing, the parameter prediction model can determine the most suitable rolling force based on the current equipment parameters to complete the rolling of the electrode sheets, effectively improving the processing quality of the electrode sheets.

Description

Technical Field

[0001] This invention relates to the field of battery processing technology, specifically to a control method for lithium battery electrode rolling equipment. Background Technology

[0002] In the electrode sheet processing of lithium-ion batteries, the rolling process is the final step, determining the final compaction density and porosity of the positive and negative electrodes, which has a decisive impact on the capacity and degradation of the lithium-ion battery. The rolling process compacts the slurry particles on the current collector after coating and drying processes, primarily to increase its density and improve the adhesion between the current collector and the slurry particles. A reasonable compaction density can increase the battery's capacity and lifespan. When subjected to pressure, the slurry particles on the current collector move randomly, such as... Figure 1 As shown in the schematic diagram of the electrode pressing roller, the particle density gradually increases as rolling progresses. Typically, coated and dried electrode sheets are stored in a wound form. Before rolling, the electrode sheets are unwound and fed under a certain tension between the upper and lower rollers of the rolling mill. The upper roller remains stationary, while the lower roller, under the action of a hydraulic cylinder, slowly moves upward, working together with the upper roller to press the electrode sheet to the pre-designed thickness.

[0003] In practice, the main quality problems of battery electrode rolling equipment manifest in the uneven thickness of the electrode after rolling. Inconsistent thickness leads to inconsistent compaction density, a key factor affecting battery performance consistency. Other factors significantly impacting electrode rolling quality include tension control devices, alignment devices, slicing devices, and dust removal devices. During rolling, the electrode requires a certain tension; insufficient tension leads to wrinkles, while excessive tension causes breakage. Dust removal devices prevent surface defects caused by impurities during rolling. Alignment and slicing devices primarily affect the dimensional accuracy of the electrode cutting.

[0004] In the existing technology, before the formal production of electrode sheets, a very complicated experimental and debugging process is required to obtain suitable rolling equipment parameters for production, which greatly increases the production cost of electrode sheets. Summary of the Invention

[0005] To address the problem of lengthy and complex debugging processes before processing lithium-ion battery electrode sheets in existing technologies, this invention provides a control method for lithium battery electrode sheet rolling equipment. This method can quickly and accurately determine suitable control parameters for the rolling equipment, thereby reducing the processing cost of lithium battery electrode sheets.

[0006] The technical solution of this invention is as follows: a control method for a lithium battery electrode rolling equipment, characterized in that it includes the following steps: S1: Sample data collected based on lithium battery electrode rolling equipment; The sample data includes: equipment parameters and quality parameters; The equipment parameters include: electrode tension, correction amount, electrode preset thickness, roller feed amount, and roller gap width; The quality parameters include: rolling force and electrode thickness; S2: Based on the sample data, construct a training dataset and a test dataset; the data ratio of the training dataset to the test dataset is 7:3; S3: Constructing a parameter prediction model for roller pressing equipment based on BP neural network; S4: The connection weights and thresholds between the input layer and the hidden layer, and between the hidden layer and the output layer in the parameter prediction model of the roller press are randomly initialized so that the parameter prediction model of the roller press has the most basic prediction conditions. S5: Optimize the parameter prediction model of the roller press based on the principle of GA algorithm to obtain the initial weights and thresholds of the parameter prediction model of the roller press. S6: Based on the optimized roller press equipment parameter prediction model, the training dataset is used to train the roller press equipment parameter prediction model to obtain the trained model. S7: During the processing, the equipment parameters of the lithium battery electrode rolling equipment are collected in real time and recorded as: real-time parameters; S8: Input the real-time parameters into the trained rolling mill parameter prediction model, and the rolling mill parameter prediction model outputs the corresponding predicted rolling force; S9: Based on the predicted rolling force, control the lithium battery electrode rolling equipment to perform processing.

[0007] Its further features are: Step S5 specifically includes the following steps: a1: Initialization; Set the population size, current generation, crossover probability, mutation probability, and initial population size; Set the maximum number of generations for the genetic algorithm and the maximum number of iterations for the roller press parameter prediction model; The connection values ​​of a BP neural network are represented by codes representing the internal values ​​of individuals within the population. a2: Calculate the fitness value of each individual in the population in turn, and perform selection, crossover and mutation operations based on the fitness value; a3: Repeat the linear step a2 until the GA algorithm reaches the maximum number of iterations or the error accuracy reaches the limit value, then stop the genetic operation and repeat the linear step a4; a4: After obtaining the optimal value, it is automatically input into the parameter prediction model of the roller press equipment as the initial weight and threshold of the parameter prediction model of the roller press equipment. Step S6 includes the following steps: b1: Based on the initial weights and thresholds, run the parameter prediction model for the roller pressing equipment; b2: Compare the error between the expected value and the output value in real time. If the final error meets the preset accuracy, the training stops and proceeds to step b3. Otherwise, return to step a2, re-optimize the weights and thresholds, and update the optimal value; b3: Optimization ends. Output the fitness changes as coordinate axis curves and output the corresponding optimal optimization parameters as a matrix.

[0008] This invention provides a control method for lithium battery electrode rolling equipment. It is based on a BP neural network to construct a parameter prediction model for the rolling equipment, and then optimizes the model using a GA genetic algorithm to obtain the optimal parameters. During processing, the equipment parameters of the lithium battery electrode rolling equipment are collected in real time and input into the trained parameter prediction model. The model then outputs the corresponding predicted rolling force. This method firstly reduces the workload of debugging and experimentation before formal production, thus lowering the overall cost. Furthermore, during processing, the parameter prediction model can determine the most suitable rolling force based on the current equipment parameters to complete the rolling of the electrode sheets, effectively improving the processing quality of the electrode sheets. Attached Figure Description

[0009] Figure 1 This is a schematic diagram of the pressure roller for the electrode sheet. Figure 2 Schematic diagram of the calculation principle for electrode pressure rollers; Figure 3 This is a force analysis diagram of the electrode sheet during the pressure roller process; Figure 4 This is a schematic diagram of the structure of the battery electrode. Figure 5 This is a flowchart of the control method for the lithium battery electrode rolling equipment in this paper; Figure 6 The graph shows a comparison between the system response performance after applying this method and existing methods. Detailed Implementation

[0010] like Figure 2As shown, during the rolling process, the area that comes into contact with the rolls and undergoes compression deformation is called the deformation zone (ABCD). The upper roll is stationary, while the lower roll, driven by the drive components, moves slowly upwards, working together with the upper roll to press the electrode sheet to the pre-designed thickness. Typically, a relatively large roll pressure is required. The reaction force of this roll pressure acts on the pressure plate and is transmitted to the overall structure of the equipment through the bearings of the pressure roll. Therefore, the pressure roll can be viewed as a support beam.

[0011] The formula for calculating the feed amount of the electrode pressure roller is: ∆H=Hh=2R(1-cosα); Taking the roll view below as an example, let's first analyze the situation where the roll pressure load is symmetrically distributed in the middle of the pressure roll. The stress situation is shown in the figure below. Figure 3 .

[0012] The electrode width is b. During the rolling process, the centerline of the electrode coincides with the centerline of the pressure roller. The distance from the edge of the electrode to the edge of the pressure roller on both the left and right sides is a. The total length of the pressure roller is I. The reaction force of rolling the electrode is p, which acts uniformly on the lower pressure roller in a vertically downward direction. The maximum deflection is f. max The calculation formula is as follows: ; In the formula: f max ρ is the maximum deflection (mm); p is the uniformly distributed load (N / m); E is the material's elastic modulus (GPa); I is the moment of inertia of the roller (kg.m). 2 ).

[0013] Electrode thickness uniformity includes transverse thickness uniformity and longitudinal thickness uniformity, such as... Figure 4 As shown, the causes of transverse and longitudinal thickness non-uniformity are different. The main influencing factors of transverse thickness non-uniformity of the electrode sheet are the bending deformation of the rolls, the rigidity of the mill base, the elastic deformation of the main load-bearing components, the roll pressure, and the width of the electrode sheet. During the operation of the rolling mill, the roll pressure causes deformation of the rolls and the mill base, which ultimately manifests as the deflection deformation of the rolls, resulting in the electrode sheet being thicker in the middle and thinner at the edges in the transverse direction. The main influencing factors of longitudinal thickness non-uniformity of the electrode sheet are the machining accuracy and installation accuracy of the rolls, bearings, bearing seats, etc. Machining errors of key components will cause the roll pressure acting on the electrode sheet to fluctuate periodically when the rolls rotate, resulting in uneven compaction thickness of the electrode sheet in the longitudinal direction.

[0014] To achieve more accurate predictions, this application provides a control method for lithium battery electrode rolling equipment. This method constructs a parameter prediction model for the rolling equipment based on a BP neural network, and optimizes the BP neural network using a genetic algorithm. The specific GA-BP algorithm flow is shown in the attached figure. Figure 5 This method specifically includes the following steps.

[0015] S1: Sample data collected based on lithium battery electrode rolling equipment; Sample data includes: equipment parameters and quality parameters; Equipment parameters include: electrode tension, correction amount, electrode preset thickness, roller feed amount, and roller gap width; Quality parameters include: rolling force and electrode thickness.

[0016] S2: Based on the sample data, construct the training dataset and the test dataset; the ratio of the training dataset to the test dataset is 7:3.

[0017] S3: Construct a parameter prediction model for roller press equipment based on BP neural network.

[0018] S4: The connection weights and thresholds between the input layer and the hidden layer, and between the hidden layer and the output layer in the parameter prediction model of the roller press are randomly initialized to ensure that the parameter prediction model of the roller press has the most basic prediction conditions.

[0019] S5: Define the variable names of the input and output variables in the training set as global variables, input the global variable data into the GA genetic algorithm, optimize the parameter prediction model of the roller press based on the principle of the GA algorithm, and obtain the initial weights and thresholds of the parameter prediction model of the roller press.

[0020] Step S5 specifically includes the following steps: a1: Initialization.

[0021] Set the population size, current generation number (1), crossover probability, mutation probability, and initial population size; Set the maximum number of generations for the genetic algorithm and the maximum number of iterations for the roller press parameter prediction model; The connection values ​​of a BP neural network are represented by codes representing the internal values ​​of individuals within the population. a2: Calculate the fitness value of each individual in the population in turn, and perform selection, crossover and mutation operations based on the fitness value; a3: Repeat the linear step a2 until the GA algorithm reaches the maximum number of iterations or the error accuracy reaches the limit value, then stop the genetic operation and repeat the linear step a4; a4: After obtaining the optimal value, it is automatically input into the parameter prediction model of the roller press equipment as the initial weight and threshold of the parameter prediction model of the roller press equipment.

[0022] In genetic algorithms, selection refers to choosing individuals with higher fitness from the population to serve as parents for the next generation, thus achieving natural selection in the process of biological evolution. In genetic algorithms, individuals with higher fitness have a greater probability of becoming parents of the next generation. In this method, the selection operation involves using the selected set as a reference; if a certain set of parameters in the system happens to satisfy the requirement of producing an appropriate or predetermined electrode thickness, this selected individual is highly likely to be the parent of the next generation.

[0023] In genetic algorithms, crossover involves pairing the chromosomes of two parent individuals and exchanging information at corresponding gene loci to produce new offspring. Crossover increases population diversity, thereby improving global search capabilities. In this method, it involves replacing one or more parameter values ​​in the selected parent samples after the selection operation, calculating new samples to choose from.

[0024] In genetic algorithms, mutation refers to subjecting offspring to a certain degree of random mutation after crossover, thereby increasing population diversity. Mutation prevents the population from getting trapped in local optima. In this method, it involves modifying a variable (within a specified range) in a sample and calculating a new sample.

[0025] An individual's fitness measures its advantage in the population, distinguishing it from "good" or "bad." Fitness is calculated using a fitness function, also called an evaluation function, which primarily determines an individual's fitness based on its characteristics.

[0026] The topology of a neural network includes the number of layers, the number of neurons in each layer, and the way these neurons are connected. From a topological perspective, artificial neural network models can be categorized into hierarchical and interconnected types. A hierarchical model divides the neural network into an input layer, hidden layers, and an output layer, with these layers connected sequentially. The input layer neurons receive input information from the outside world and transmit it to the hidden layer neurons. The hidden layers are responsible for information processing and transformation within the neural network. Typically, depending on the transformation requirements, one or more hidden layers are designed.

[0027] S6: Based on the optimized roller press parameter prediction model, train the model using the training dataset to obtain the trained roller press parameter prediction model.

[0028] In practical applications, the genetic algorithm is performed based on the gaot toolbox.

[0029] Step S6 includes the following steps: b1: Determine the topology of the parameter prediction model for the roller press equipment. The specific model structure should be selected and set according to actual needs. Then, based on the initial weights and thresholds, the parameter prediction model for the roller press is run; b2: Compare the error between the expected value and the predicted output value in real time. If the final error meets the preset accuracy, the training stops and proceeds to step b3. Otherwise, return to step a2, re-optimize the weights and thresholds, and update the optimal value; In the specific implementation of step b2, existing error comparison methods can be applied. For example, the sum of squares of the differences between the predicted output value and the expected value of the electrode thickness can be constructed as the objective function. The minimization process of the objective function is regarded as the backpropagation process of the parameter prediction model of the rolling mill. According to the quality requirements of the product, a target output value and an allowable error range are preset as the quality accuracy target. The difference between the value of the objective function and the target accuracy is compared in real time. As the number of iterations increases, this difference will become smaller and smaller until the difference meets the allowable error range, at which point training stops. b3: Optimization ends. Output the fitness changes as a coordinate axis curve and output the corresponding optimal optimization parameters as a matrix. In practical applications, the data of the parameter prediction model of the roller pressing equipment can be processed based on the neural network toolbox built into MATLAB software to obtain curves, and other functions in MATLAB software can be used to complete the process.

[0030] S7: During the processing, the equipment parameters of the lithium battery electrode rolling equipment are collected in real time and recorded as: real-time parameters.

[0031] S8: Input the real-time parameters into the trained rolling mill parameter prediction model, and the rolling mill parameter prediction model outputs the corresponding predicted rolling force.

[0032] S9: Based on the predicted rolling force, control the lithium battery electrode rolling equipment for processing.

[0033] This method is based on the GA-BP algorithm for the pressure roller equipment. Through real-time acquisition equipment, parameters affecting the quality of the electrode, such as electrode tension, correction amount, electrode preset thickness, roller feed amount, and roller gap, are used as input samples, as shown in Table 1 below. After rapid real-time calculation, the rolling force is obtained, and the output is controlled by the equipment to obtain battery electrodes with uniform thickness.

[0034] Table 1 Partial Sample Data like Figure 6The figure shows a comparison of the system response performance after applying this method with existing methods. The horizontal axis represents the response time in seconds, and the vertical axis represents the displacement, indicating the output position of the control component, which is the position of the hydraulic cylinder lifting, used to control the roll gap, in meters. It can be seen that compared to traditional electrode rolling mills, the GA-BP algorithm in this method improves the response time by 40% compared to traditional PID control or analog control, with almost zero steady-state error, higher system dynamic response, and smaller overshoot.

Claims

1. A control method for a lithium battery electrode rolling mill, characterized in that, It includes the following steps: S1: Sample data collected based on lithium battery electrode rolling equipment; The sample data includes: equipment parameters and quality parameters; The equipment parameters include: electrode tension, correction amount, electrode preset thickness, roller feed amount, and roller gap width; The quality parameters include: rolling force and electrode thickness; S2: Based on the sample data, construct a training dataset and a test dataset; the data ratio of the training dataset to the test dataset is 7:3; S3: Constructing a parameter prediction model for roller press equipment based on BP neural network; S4: The connection weights and thresholds between the input layer and the hidden layer, and between the hidden layer and the output layer in the parameter prediction model of the roller press are randomly initialized so that the parameter prediction model of the roller press has the most basic prediction conditions. S5: Optimize the parameter prediction model of the roller press based on the principle of GA algorithm to obtain the initial weights and thresholds of the parameter prediction model of the roller press. S6: Based on the optimized roller press equipment parameter prediction model, the training dataset is used to train the roller press equipment parameter prediction model to obtain the trained model. S7: During the processing, the equipment parameters of the lithium battery electrode rolling equipment are collected in real time and recorded as: real-time parameters; S8: Input the real-time parameters into the trained rolling mill parameter prediction model, and the rolling mill parameter prediction model outputs the corresponding predicted rolling force; S9: Based on the predicted rolling force, control the lithium battery electrode rolling equipment to perform processing; Step S5 specifically includes the following steps: a1: Initialization; Set the population size, current generation, crossover probability, mutation probability, and initial population size; Set the maximum number of generations for the genetic algorithm and the maximum number of iterations for the roller press parameter prediction model; The connection values ​​of a BP neural network are represented by codes representing the internal values ​​of individuals within the population. a2: Calculate the fitness value of each individual in the population in turn, and perform selection, crossover and mutation operations based on the fitness value; a3: Repeat step a2 until the GA algorithm reaches the maximum number of iterations or the error accuracy reaches the limit value, then stop the genetic operation and execute step a4; a4: After obtaining the optimal value, it is automatically input into the parameter prediction model of the roller press equipment as the initial weight and threshold of the parameter prediction model of the roller press equipment.

2. The control method for a lithium battery electrode rolling equipment according to claim 1, characterized in that: Step S6 includes the following steps: b1: Based on the initial weights and thresholds, run the parameter prediction model for the roller pressing equipment; b2: Compare the error between the expected value and the output value in real time. If the final error meets the preset accuracy, the training stops and step b3 is executed. Otherwise, return to step a2, re-optimize the weights and thresholds, and update the optimal value; b3: Optimization ends. Output the fitness changes as a coordinate axis curve and output the corresponding optimal optimization parameters as a matrix.

Images