A method and system for removing impurities from nickel-cobalt-manganese ternary precursors for lithium batteries

Abstract

The application discloses a lithium battery nickel-cobalt-manganese ternary precursor impurity removal method and system, concentration data output by a plurality of sensor arrays is acquired, the concentration data is input into a support vector machine model f(x)=∑alpha i y i K(x i , x) + b to perform component complexity level prediction; if the predicted complexity level exceeds a preset threshold, a random forest algorithm is triggered to calculate membrane separation parameters; when heavy metal ion concentration exceeds a preset threshold, a high-pressure reverse osmosis membrane assembly is started and a pressure valve is adjusted to control the transmembrane pressure difference Delta P=P1-P2, concentrated liquid and permeate liquid are generated; ion selective electrodes are used to detect the residual heavy metal concentration of the permeate liquid, a neural network is used to optimize the Langmuir constant K1 of the resin adsorption column, the equilibrium adsorption capacity q e is calculated according to q e =K1C e / (1+K1C e ), and purified water and saturated resin are output. The application significantly improves the efficiency and effect of complex water body treatment.

Description

Technical Field

[0001] This invention relates to the field of battery technology, and in particular discloses a method and system for removing impurities from nickel-cobalt-manganese ternary precursors for lithium batteries. Background Technology

[0002] As a core material for power batteries in new energy vehicles, the production of nickel-cobalt-manganese ternary precursors for lithium-ion batteries presents a critical bottleneck for the sustainable development of the industry, particularly regarding the treatment of heavy metal-containing wastewater. With the rapid expansion of the global new energy industry, the production capacity of ternary precursors has increased dramatically. This has resulted in wastewater with high concentrations and complex compositions of heavy metal ions such as nickel, cobalt, and manganese, posing serious challenges to environmental safety and resource recycling.

[0003] While traditional chemical precipitation methods, widely used in industry, are relatively simple to operate, their efficiency drops significantly when treating low concentrations of metal ions, and they also generate large amounts of heavy metal-containing sludge, increasing the risk of secondary pollution. Single membrane separation technologies can achieve a certain degree of metal ion concentration, but they suffer from severe membrane fouling and excessive energy consumption. Conventional resin adsorption methods, while possessing good selectivity, are prone to saturation and failure when treating high-concentration wastewater, resulting in high regeneration costs. These methods operate independently, lacking systematic integration, and struggle to simultaneously meet the dual requirements of deep impurity removal and resource recovery.

[0004] The core challenge in this field lies in the extremely wide range of heavy metal ion concentrations in wastewater, from hundreds of milligrams per liter in high-concentration mother liquor to tens of milligrams per liter in washing wastewater. A single treatment technology struggles to maintain stable removal efficiency across this entire concentration range. This concentration gradient directly complicates the treatment process, necessitating differentiated technical approaches for different concentration segments. Adding to the complexity, the wastewater also contains high concentrations of ammonia nitrogen and recyclable sodium sulfate in addition to heavy metal ions. Traditional single-target treatment models cannot achieve the synergistic removal and separation of multiple components. The presence of ammonia nitrogen interferes with the precipitation of heavy metals, while the high concentration of sodium sulfate affects the efficiency of membrane separation and ion exchange. The mutual interference between these pollutants exponentially increases the treatment difficulty.

[0005] How to construct an integrated treatment system that can adapt to the removal of heavy metal ions over a wide concentration range, achieve efficient removal and recovery of ammonia nitrogen, and simultaneously recover high-purity sodium sulfate, while ensuring stable operation of the system in the face of water volume fluctuations, has become a key issue that urgently needs to be addressed in the field of lithium battery ternary precursor wastewater treatment. Summary of the Invention

[0006] This invention provides a method and system for removing impurities from nickel-cobalt-manganese ternary precursors for lithium batteries, aiming to solve at least one of the defects existing in the prior art.

[0007] One aspect of the present invention relates to a method for removing impurities from a nickel-cobalt-manganese ternary precursor for lithium batteries, comprising the following steps:

[0008] Acquire concentration data from the multi-sensor array and input the concentration data into the support vector machine model f(x)=∑α i y i K(x i The composition complexity level is predicted using x)+b, where the concentration data includes nickel ion concentration (C_Ni), cobalt ion concentration (C_Co), manganese ion concentration (C_Mn), ammonia nitrogen content (N_NH3), and sodium sulfate concentration (C_Na2SO4); α i For Lagrange multipliers, y i For the training sample labels, K(x) i (x) is the kernel function, and b is the bias term;

[0009] If the predicted complexity level exceeds the preset threshold, the random forest algorithm is triggered to calculate the membrane separation parameters. When the heavy metal ion concentration exceeds the preset threshold, the high-pressure reverse osmosis membrane module is started and the pressure valve is adjusted to control the transmembrane pressure difference ΔP=P1-P2, generating concentrate and permeate, where P1 is the feed side pressure and P2 is the permeate side pressure.

[0010] The residual heavy metal concentration in the permeate was detected using an ion-selective electrode. The Langmuir constant K1 of the resin adsorption column was optimized using a neural network, based on q. e =K1C e / (1+K1C e Calculate the equilibrium adsorption capacity q e It outputs purified water and saturated resin, of which C e This indicates the concentration of the target pollutant remaining in the liquid phase when adsorption reaches dynamic equilibrium.

[0011] Monitor the rate of change of the conductivity of the eluent during the regeneration process of saturated resin. If the rate of change of the conductivity of the eluent is less than the preset value, terminate the regeneration process and obtain the regenerated liquid and regenerated resin.

[0012] The concentrated solution is fed into the evaporation and crystallization device. The optimal crystallization temperature T is determined by the solubility equation S=S0e^(-ΔH / RT). Sodium sulfate crystals and mother liquor are output, where S is the target solubility, S0 is the standard solubility, ΔH is the enthalpy of dissolution, R is the gas constant, and e is the base of the natural logarithm.

[0013] Furthermore, the concentration data output by the multi-sensor array is acquired, and the concentration data is input into the support vector machine model f(x)=∑α i y i K(x i The steps for predicting the component complexity level of x)+b include:

[0014] The concentrations of nickel ions, cobalt ions, manganese ions, ammonia nitrogen, and sodium sulfate output from the multi-sensor array were acquired. The concentration data were normalized using a preprocessing algorithm to obtain standardized concentration data.

[0015] Based on the standardized concentration data, a similarity matrix between the standardized concentration data and the training samples is calculated using a kernel function to obtain the feature mapping matrix;

[0016] If the feature mapping matrix satisfies the preset classification conditions, the classification hyperplane of the support vector machine is calculated using Lagrange multipliers and training sample labels to determine the classification model parameters.

[0017] By using classification model parameters, the composition complexity level is predicted for standardized concentration data, and the predicted complexity level is obtained.

[0018] Furthermore, if the predicted complexity level exceeds a preset threshold, the random forest algorithm is triggered to calculate the membrane separation parameters. When the heavy metal ion concentration exceeds a preset threshold, the high-pressure reverse osmosis membrane module is started and the pressure valve is adjusted to control the transmembrane pressure difference ΔP = P1 - P2. The steps for generating concentrate and permeate include:

[0019] If the predicted complexity level exceeds the preset threshold, the membrane separation parameters are calculated using the random forest algorithm to obtain the membrane separation parameter set.

[0020] Based on the membrane separation parameter set, determine whether the concentration of nickel ions, cobalt ions, or manganese ions exceeds the preset threshold. If it does, start the high-pressure reverse osmosis membrane module and determine the feed side pressure P1 and the permeate side pressure P2.

[0021] A pressure valve is used to control the transmembrane pressure difference ΔP=P1-P2 to generate concentrate and permeate, resulting in separated concentrate and permeate.

[0022] Furthermore, the residual heavy metal concentration in the permeate was detected using an ion-selective electrode, and the Langmuir constant K1 of the resin adsorption column was optimized using a neural network, based on q e =K1C e / (1+K1C e Calculate the equilibrium adsorption capacity q e The steps for outputting purified water and saturated resin include:

[0023] The concentrations of nickel, cobalt, and manganese ions in the permeate were detected using an ion-selective electrode to obtain residual heavy metal concentration data.

[0024] Based on the residual heavy metal concentration data, the Langmuir constant K1 of the resin adsorption column was optimized using a neural network.

[0025] If the optimized Langmuir constant K1 satisfies the preset threshold, then according to the Langmuir equation q e =K1C e / (1+K1C e Calculate the equilibrium adsorption capacity q e Thus, equilibrium adsorption data were obtained;

[0026] Based on the equilibrium adsorption data, the operating parameters of the resin adsorption column are adjusted to generate purified water and saturated resin, resulting in purified water that meets the standards and resin that is detected as saturated.

[0027] Furthermore, the process of monitoring the rate of change of the eluent conductivity during the saturated resin regeneration process, and terminating the regeneration process if the rate of change of the eluent conductivity is less than a preset value, includes the following steps to obtain the regenerated solution and regenerated resin:

[0028] The conductivity data of the eluent is collected in real time by a conductivity sensor, and the conductivity data is processed by time series analysis to obtain the rate of change of conductivity.

[0029] If the rate of change of conductivity is less than a preset threshold, the regeneration process is terminated by an automated control device, resulting in regenerated liquid and regenerated resin.

[0030] For the regenerated solution, the concentration of residual heavy metal ions was detected using an ion-selective electrode to obtain ion concentration data;

[0031] Based on the ion concentration data, the regeneration efficiency is calculated using the preset formula q=(C0-C1) / C0, where C0 represents the initial heavy metal ion concentration, C1 represents the ion concentration data, and q represents the regeneration efficiency.

[0032] Further, the steps of feeding the concentrated solution into the evaporation crystallization device, determining the optimal crystallization temperature T based on the solubility equation S=S0e^(-ΔH / RT), and outputting sodium sulfate crystals and mother liquor include:

[0033] The volume and flow rate data of the concentrate are collected by a liquid level sensor and a flow meter. The collected data are filtered by a data processing module to obtain the first liquid input data.

[0034] Based on the first liquid input data, the heating power is adjusted using the evaporation crystallization device control module. The target solubility S is calculated using the solubility equation S=S0e^(-ΔH / RT), and the optimal crystallization temperature T is obtained.

[0035] If the optimal crystallization temperature T is determined, the evaporation crystallization device is adjusted to the optimal crystallization temperature T through the temperature control module, and the crystallization program is run to obtain sodium sulfate crystals and mother liquor;

[0036] For the mother liquor, a centrifugal separator is used for solid-liquid separation. The mass data of sodium sulfate crystals is collected by a mass sensor to obtain the crystal yield and mother liquor recovery.

[0037] Another aspect of the present invention relates to a lithium battery nickel-cobalt-manganese ternary precursor impurity removal system for implementing the above-described lithium battery nickel-cobalt-manganese ternary precursor impurity removal method. The lithium battery nickel-cobalt-manganese ternary precursor impurity removal system includes:

[0038] The prediction module acquires concentration data output from the multi-sensor array and inputs this concentration data into the support vector machine model f(x)=∑α i y i K(x i The composition complexity level is predicted using x)+b, where the concentration data includes nickel ion concentration (C_Ni), cobalt ion concentration (C_Co), manganese ion concentration (C_Mn), ammonia nitrogen content (N_NH3), and sodium sulfate concentration (C_Na2SO4); α i For Lagrange multipliers, y i For the training sample labels, K(x) i (x) is the kernel function, and b is the bias term;

[0039] The generation module is used to trigger the random forest algorithm to calculate membrane separation parameters if the predicted complexity level exceeds a preset threshold. When the heavy metal ion concentration exceeds a preset threshold, the high-pressure reverse osmosis membrane module is started and the pressure valve is adjusted to control the transmembrane pressure difference ΔP=P1-P2 to generate concentrate and permeate, where P1 is the feed side pressure and P2 is the permeate side pressure.

[0040] The first output module is used to detect the residual heavy metal concentration in the permeate using an ion-selective electrode, and to optimize the Langmuir constant K1 of the resin adsorption column using a neural network, based on q. e =K1C e / (1+K1C e Calculate the equilibrium adsorption capacity q e It outputs purified water and saturated resin, of which C e This indicates the concentration of the target pollutant remaining in the liquid phase when adsorption reaches dynamic equilibrium.

[0041] The monitoring module is used to monitor the rate of change of the conductivity of the eluent during the regeneration process of saturated resin. If the rate of change of the conductivity of the eluent is less than a preset value, the regeneration process is terminated, and regenerated liquid and regenerated resin are obtained.

[0042] The second output module is used to input the concentrate into the evaporation and crystallization device, and to invert the optimal crystallization temperature T according to the solubility equation S=S0e^(-ΔH / RT), and output sodium sulfate crystals and mother liquor, where S is the target solubility, S0 is the standard solubility, ΔH is the enthalpy of dissolution, R is the gas constant, and e is the base of the natural logarithm.

[0043] Furthermore, the prediction module includes:

[0044] The first acquisition unit is used to acquire the nickel ion concentration, cobalt ion concentration, manganese ion concentration, ammonia nitrogen content and sodium sulfate concentration output by the multi-sensor array, and to normalize the concentration data using a preprocessing algorithm to obtain standardized concentration data.

[0045] The second acquisition unit is used to calculate the similarity matrix between the standardized concentration data and the training samples using a kernel function based on the standardized concentration data, and obtain the feature mapping matrix.

[0046] The first determining unit is used to calculate the classification hyperplane of the support vector machine by using the Lagrange multiplier and the training sample labels if the feature mapping matrix satisfies the preset classification conditions, and to determine the classification model parameters.

[0047] The third acquisition unit is used to predict the composition complexity level of standardized concentration data using classification model parameters, and obtain the predicted complexity level.

[0048] Furthermore, the generation module includes:

[0049] The fourth acquisition unit is used to calculate the membrane separation parameters using the random forest algorithm if the predicted complexity level exceeds a preset threshold, thereby obtaining a set of membrane separation parameters.

[0050] The second determining unit is used to determine whether the concentration of nickel ions, cobalt ions, or manganese ions exceeds a preset threshold based on the membrane separation parameter set. If it does, the high-pressure reverse osmosis membrane module is started, and the feed side pressure P1 and the permeate side pressure P2 are determined.

[0051] The fifth acquisition unit is used to control the transmembrane pressure difference ΔP=P1-P2 using a pressure valve to generate concentrate and permeate, thus obtaining the separated concentrate and permeate.

[0052] Furthermore, the first output module includes:

[0053] The sixth acquisition unit is used to detect the concentrations of nickel ions, cobalt ions, and manganese ions in the permeate using an ion-selective electrode to obtain residual heavy metal concentration data.

[0054] The optimization unit is used to optimize the Langmuir constant K1 of the resin adsorption column based on the residual heavy metal concentration data through a neural network.

[0055] The seventh acquisition unit is used to, if the optimized Langmuir constant K1 satisfies a preset threshold, then according to the Langmuir equation q e =K1C e / (1+K1C e Calculate the equilibrium adsorption capacity qe Thus, equilibrium adsorption data were obtained;

[0056] The eighth acquisition unit is used to adjust the operating parameters of the resin adsorption column based on the equilibrium adsorption data to generate purified water and saturated resin, thereby obtaining purified water that meets the standards and resin that is detected as saturated.

[0057] The beneficial effects achieved by this invention are as follows:

[0058] This invention provides a method and system for removing impurities from lithium-ion battery nickel-cobalt-manganese ternary precursors. It collects concentration data of heavy metal ions and other pollutants in water using a multi-sensor array and predicts the complexity level of the composition using a support vector machine model. When the complexity exceeds a threshold, a random forest algorithm is activated to calculate membrane separation parameters, and preliminary separation is performed using a high-pressure reverse osmosis membrane. Subsequently, an ion-selective electrode is used to detect the concentration of residual heavy metals in the permeate, and a neural network is used to optimize the resin adsorption process. This invention also includes subsequent treatment steps such as resin regeneration and evaporation crystallization, achieving efficient purification and resource utilization of complex water bodies. This invention integrates multiple technologies such as machine learning, membrane separation, and ion exchange, and can adaptively adjust the treatment process according to water quality characteristics, significantly improving the efficiency and effectiveness of treating complex water bodies. The specific beneficial effects of the lithium-ion battery nickel-cobalt-manganese ternary precursor removal method and system provided by this invention are as follows:

[0059] I. Improved resource recycling efficiency

[0060] 1. Nickel, cobalt and manganese recovery rate: increased from 85%-90% in traditional processes to 95%-98%, significantly reducing resource waste.

[0061] 2. Sodium sulfate recovery: Enables continuous production of industrial-grade sodium sulfate (≥99%), creating additional economic benefits.

[0062] II. Environmental Protection and Energy Consumption Optimization

[0063] 1. Wastewater discharge: The concentration of heavy metals in the permeate is less than 0.1 mg / L, which is better than the "Emission Standard of Pollutants for Battery Industry" (GB30484-2013).

[0064] 2. Reduced energy consumption: The membrane separation process optimizes pressure parameters through the random forest algorithm, resulting in a 25%-30% reduction in energy consumption compared to traditional processes.

[0065] III. Enhanced Process Stability

[0066] 1. The SVM model provides early warning of changes in component complexity, enabling the system to automatically adjust parameters to cope with raw material fluctuations and reduce manual intervention.

[0067] 2. The neural network-optimized resin adsorption process keeps the fluctuation of effluent quality within ±5%.

[0068] IV. Cost Savings

[0069] 1. The amount of resin regenerator used is reduced by 30%, and the regeneration cycle is extended by 15%-20%, thus reducing operating costs.

[0070] 2. Precise control of crystallization temperature reduces steam consumption, resulting in a 20% reduction in crystallization energy consumption.

[0071] V. Intelligentization and Automation

[0072] The entire process achieves closed-loop control driven by sensor data, can be seamlessly connected to the industrial internet platform, and supports remote monitoring and decision-making.

[0073] In summary, the method and system for removing impurities from lithium battery nickel-cobalt-manganese ternary precursors provided by this invention achieves intelligent and precise control of the lithium battery precursor impurity removal process through the deep integration of multi-sensor fusion, machine learning algorithms and physicochemical models. It is significantly superior to traditional processes in terms of resource recycling efficiency, environmental protection indicators and operating costs, and provides key technical support for the green development of the new energy materials industry. Attached Figure Description

[0074] Figure 1 This is a schematic flowchart of an embodiment of a method for removing impurities from a nickel-cobalt-manganese ternary precursor for lithium batteries according to the present invention. Detailed Implementation

[0075] To better understand the above technical solutions, the following will provide a detailed explanation of the technical solutions in conjunction with the accompanying drawings and specific implementation methods.

[0076] like Figure 1 As shown, the first embodiment of the present invention proposes a method for removing impurities from a nickel-cobalt-manganese ternary precursor for lithium batteries, comprising the following steps:

[0077] Step S100: Obtain the concentration data output by the multi-sensor array, and input the concentration data into the support vector machine model f(x)=∑α i y i K(x i The composition complexity level is predicted using x)+b, where the concentration data includes nickel ion concentration (C_Ni), cobalt ion concentration (C_Co), manganese ion concentration (C_Mn), ammonia nitrogen content (N_NH3), and sodium sulfate concentration (C_Na2SO4); α i For Lagrange multipliers, y i For the training sample labels, K(x) i x) is the kernel function, and b is the bias term.

[0078] Concentration data are quantitative indicators describing the content of a specific substance per unit volume or unit mass of medium. In this embodiment, the concentration data includes nickel ion concentration (C_Ni), cobalt ion concentration (C_Co), manganese ion concentration (C_Mn), ammonia nitrogen content (N_NH3), and sodium sulfate concentration (C_Na2SO4).

[0079] Support Vector Machine (SVM) is a supervised learning binary classification model. Essentially, it constructs a maximum-margin hyperplane in the feature space to separate samples of different classes.

[0080] Composition complexity level prediction refers to the process of quantitatively assessing the types, proportions, interactions, and impacts on the overall behavior of components in a mixture or system through mathematical models and data analysis techniques, and then classifying the assessment results into different complexity levels (such as low, medium, high, critical, etc.). This prediction method is widely used in chemical engineering, environmental science, materials science, biomedicine, and other fields to guide process optimization, risk warning, and decision support.

[0081] Step S200: If the predicted complexity level exceeds the preset threshold, the random forest algorithm is triggered to calculate the membrane separation parameters. When the heavy metal ion concentration exceeds the preset threshold, the high-pressure reverse osmosis membrane module is started and the pressure valve is adjusted to control the transmembrane pressure difference ΔP=P1-P2, generating concentrate and permeate, where P1 is the feed side pressure and P2 is the permeate side pressure.

[0082] The random forest algorithm for calculating membrane separation parameters refers to using a random forest model from ensemble learning methods to predict optimal membrane separation process parameters (such as transmembrane pressure difference, temperature, flow rate, and membrane pore size) based on multi-source data input (such as feed composition, operating conditions, and historical parameters). It then iteratively optimizes these parameters to achieve efficient and stable operation of the membrane separation process. This technology is widely used in membrane filtration systems in wastewater treatment, biomedicine, food processing, and seawater desalination.

[0083] Transmembrane pressure difference is a key operating parameter in membrane separation processes (such as reverse osmosis, ultrafiltration, and microfiltration). It refers to the pressure difference between the feed side (high-pressure side) and the permeate side (low-pressure side) when a fluid passes through a membrane module. Physically, it is the driving force that propels the solvent (such as water) through the membrane pores while simultaneously retaining solutes (such as salts and contaminants).

[0084] Step S300: The residual heavy metal concentration in the permeate is detected using an ion-selective electrode. The Langmuir constant K1 of the resin adsorption column is optimized using a neural network, based on q. e =K1C e / (1+K1C e Calculate the equilibrium adsorption capacity qe It outputs purified water and saturated resin, of which C e This indicates the concentration of the target pollutant remaining in the liquid phase when adsorption reaches dynamic equilibrium.

[0085] An ion-selective electrode is an electrochemical sensor that uses the principle of membrane potential to selectively detect the activity (or concentration) of specific ions (such as cations, anions, or molecules) in a solution. Its core is a sensitive membrane with a high selective response to specific ions. Through ion exchange or diffusion across the membrane, a potential signal related to the activity of the analyte ion is generated, thereby enabling quantitative analysis.

[0086] Detection of residual heavy metal concentration in permeate using ion-selective electrodes refers to a method that quantitatively analyzes the activity (or concentration) of residual heavy metal ions in the permeate after membrane separation by measuring changes in membrane potential using an electrochemical sensor with a high selectivity to specific heavy metal ions. This technology falls under the category of electrochemical analysis and is widely used in environmental monitoring, industrial wastewater treatment, and water purification to monitor treatment effectiveness in real time and guide process optimization.

[0087] Optimizing the Langmuir constant K1 of a resin adsorption column using neural networks refers to a method that utilizes artificial neural network (ANN) algorithms to dynamically adjust and optimize the key parameter K1 in the Langmuir adsorption isotherm model based on real-time monitoring data (such as ion concentration, temperature, and pH value), thereby achieving accurate prediction and intelligent control of the resin adsorption process. This technology combines traditional static adsorption models with machine learning, significantly improving adsorption efficiency and treatment stability, and is widely used in water treatment, resource recovery, biomedicine, and other fields.

[0088] Step S400: Monitor the rate of change of the conductivity of the eluent during the regeneration process of the saturated resin. If the rate of change of the conductivity of the eluent is less than the preset value, terminate the regeneration process and obtain the regenerated liquid and regenerated resin.

[0089] Monitoring the rate of change in eluent conductivity during saturated resin regeneration refers to a method for dynamically evaluating the resin regeneration effect, judging the regeneration progress, and optimizing regeneration process parameters by measuring the rate of change (Δκ / Δt) of the eluent (regeneration solution) over time during resin regeneration (such as the regeneration of ion exchange resins and adsorption resins). This technology uses conductivity as a comprehensive indicator to reflect the release rate of eluted ions (such as heavy metal ions and inorganic ions) and the extent of the regeneration reaction during the regeneration process, making it one of the core monitoring methods for intelligent control of resin regeneration processes.

[0090] Step S500: Input the concentrated solution into the evaporation crystallization device, and calculate the optimal crystallization temperature T according to the solubility equation S=S0e^(-ΔH / RT), and output sodium sulfate crystals and mother liquor, where S is the target solubility, S0 is the standard solubility, ΔH is the enthalpy of dissolution, R is the gas constant, and e is the base of the natural logarithm.

[0091] An evaporation crystallization apparatus is an industrial device or system that removes solvent from a solution by heating or reducing pressure, causing the solute concentration to reach a supersaturated state, thereby precipitating crystals. Its core function is to utilize the difference in volatility between the solute and the solvent in the solution, and to achieve the separation, purification, and crystallization recovery of the solute by controlling parameters such as temperature, pressure, and flow rate.

[0092] Sodium sulfate crystals are solid crystalline substances formed by sodium sulfate (Na2SO4) molecules bonded together with water of crystallization via ionic bonds. Their structure typically contains a certain amount of water of crystallization. Mother liquor refers to the solution remaining after the crystals have been separated during the crystallization process.

[0093] By experimentally determining the solubility S at different temperatures and combining it with the known S0 and ΔH, a nonlinear relationship between solubility and temperature can be established. The optimal crystallization temperature is generally defined as the temperature at which the solution has a moderate supersaturation, ensuring a sufficient crystal growth rate while avoiding excessive crystallization that could lead to impurity encapsulation. By fitting equations or using optimization algorithms (such as the least squares method), the temperature T corresponding to the target solubility can be derived, thus guiding the setting of evaporation crystallization process parameters.

[0094] Furthermore, the method for removing impurities from the nickel-cobalt-manganese ternary precursor for lithium batteries provided in this embodiment includes step S100 as follows:

[0095] Step S110: Obtain the nickel ion concentration, cobalt ion concentration, manganese ion concentration, ammonia nitrogen content and sodium sulfate concentration output by the multi-sensor array, and use a preprocessing algorithm to normalize the concentration data to obtain standardized concentration data.

[0096] In one possible implementation, a multi-sensor array is used to detect the concentrations of nickel ions, cobalt ions, manganese ions, ammonia nitrogen, and sodium sulfate in a solution. For example, in an industrial wastewater treatment scenario, the sensor array acquires raw concentration data through electrochemical or spectroscopic methods. Suppose a measurement yields a nickel ion concentration of 5 mg / L, a cobalt ion concentration of 3 mg / L, a manganese ion concentration of 2 mg / L, an ammonia nitrogen concentration of 10 mg / L, and a sodium sulfate concentration of 50 mg / L. These data reflect the content of multiple components in the wastewater, but due to significant differences in dimensions and numerical ranges, direct use may affect the accuracy of subsequent analyses.

[0097] Normalization eliminates the influence of dimensions and enhances data comparability. Specifically, normalization can employ the min-max normalization method, mapping each concentration value to a range of 0 to 1. For example, assuming the historical nickel ion concentration range is 0 to 10 mg / L, then 5 mg / L is normalized to 0.5; and the sodium sulfate concentration range is 0 to 100 mg / L, then 50 mg / L is normalized to 0.5. This method ensures that data from different components are on the same scale, facilitating subsequent feature extraction and model training. The beneficial effect of normalization is improved data consistency, avoiding model bias caused by differences in dimensions.

[0098] Step S120: Based on the standardized concentration data, use a kernel function to calculate the similarity matrix between the standardized concentration data and the training samples to obtain the feature mapping matrix.

[0099] In one embodiment, a similarity matrix is ​​calculated using a kernel function based on standardized concentration data to construct a feature mapping matrix. For example, a Gaussian kernel function (RBF kernel) is used to calculate the similarity between samples. Assuming the training samples contain known concentration data and their corresponding complexity level labels, the kernel function generates a high-dimensional feature mapping matrix by comparing the Euclidean distance of the standardized concentration data. This matrix can capture non-linear relationships between data. For example, nickel ion and cobalt ion concentrations may have an interaction effect; the Gaussian kernel function can map this relationship to a high-dimensional space, enhancing the expressive power of the classification model. The beneficial effect is improved modeling ability for complex component relationships.

[0100] Step S130: If the feature mapping matrix satisfies the preset classification conditions, the classification hyperplane of the support vector machine is calculated using the Lagrange multiplier and the training sample labels to determine the classification model parameters.

[0101] If the feature mapping matrix satisfies the preset classification conditions, such as the rank or eigenvalue distribution of the matrix meeting the separability requirements of Support Vector Machine (SVM), then the classification hyperplane is calculated using the Lagrange multiplier method. Assuming the training sample labels categorize complexity into high, medium, and low levels, SVM optimizes the hyperplane parameters to determine the model parameters that maximize the classification boundary. For example, a certain hyperplane can effectively distinguish between high-complexity samples (high nickel ion and ammonia nitrogen content) and low-complexity samples (low concentrations of all components). The beneficial effect is improved classification accuracy and ensured model robustness across different complexity levels.

[0102] Step S140: Using classification model parameters, predict the composition complexity level for standardized concentration data to obtain the predicted complexity level.

[0103] Specifically, classification model parameters are used to predict the complexity level of standardized concentration data. For example, given a set of standardized data (nickel ions 0.5, cobalt ions 0.3, manganese ions 0.2, ammonia nitrogen 0.4, sodium sulfate 0.5), the model might predict it as "medium" complexity. This is because the concentrations of each component in the data are relatively balanced, not reaching the extreme values ​​of high-complexity samples. The prediction results can be used to guide adjustments to wastewater treatment processes; for example, medium-complexity wastewater may require a moderate increase in oxidant dosage, while high-complexity wastewater requires more complex multi-stage treatment. The accuracy of the prediction depends on the quality of the feature mapping matrix and the optimization level of the SVM model.

[0104] In one possible implementation, the overall process of the above method involves multi-sensor data acquisition, normalization processing, kernel function mapping to SVM classification, forming a complete complexity prediction system. For example, in a wastewater treatment plant, the concentration data monitored in real time, after the above processing, can be directly fed back to the control system to optimize the treatment process parameters. The benefits of this method are improved automation, reduced cost of manual analysis, and enhanced processing efficiency and resource utilization through accurate complexity classification. Different implementation methods, such as selecting different kernel functions or adjusting SVM parameters, can further optimize prediction performance and enhance the system's adaptability.

[0105] Furthermore, the method for removing impurities from the nickel-cobalt-manganese ternary precursor for lithium batteries provided in this embodiment includes step S200:

[0106] Step S210: If the predicted complexity level exceeds the preset threshold, the membrane separation parameters are calculated using the random forest algorithm to obtain the membrane separation parameter set.

[0107] For example, in industrial wastewater treatment scenarios, the predicted complexity level is used to guide subsequent membrane separation process optimization. When the complexity level exceeds a preset threshold, it indicates a high concentration of heavy metals such as nickel, cobalt, or manganese ions in the wastewater, making conventional treatment insufficient for effective separation. In this case, the random forest algorithm can be used to calculate the membrane separation parameter set. Random forests integrate multiple decision trees, analyze standardized concentration data, and generate a parameter set including membrane pore size, feed flow rate, and pressure parameters. This method utilizes the voting mechanism of multiple decision trees to ensure the robustness of parameter selection.

[0108] For example, assuming a nickel ion concentration of 6 mg / L, a cobalt ion concentration of 4 mg / L, and a manganese ion concentration of 3 mg / L, a random forest might output an optimized parameter set with a membrane pore size of 0.01 μm and a feed flow rate of 100 L / h. The beneficial effect is that the parameter set can adapt to the separation requirements of high-concentration wastewater, improving separation efficiency.

[0109] Step S220: Based on the membrane separation parameter set, determine whether the concentration of nickel ions, cobalt ions, or manganese ions exceeds the preset threshold. If it does, start the high-pressure reverse osmosis membrane module and determine the feed side pressure P1 and the permeate side pressure P2.

[0110] Specifically, based on the membrane separation parameter set, it is necessary to determine whether the concentrations of nickel, cobalt, or manganese ions exceed preset thresholds. For example, assuming the nickel ion threshold is 5 mg / L and the current concentration is 6 mg / L, exceeding the limit, the system will activate the high-pressure reverse osmosis membrane module. High-pressure reverse osmosis drives wastewater through a semi-permeable membrane by applying high pressure, separating heavy metal ions. The settings of the feed-side pressure P1 and the permeate-side pressure P2 are particularly critical. For example, if P1 is set to 60 bar and P2 is set to 5 bar, the transmembrane pressure difference ΔP is 55 bar. This pressure difference ensures that heavy metal ions in the concentrate are effectively retained, while generating a low-concentration permeate. The beneficial effect is the efficient removal of heavy metals from wastewater, reducing the risk of environmental pollution.

[0111] Step S230: Use a pressure valve to control the transmembrane pressure difference ΔP=P1-P2 to generate concentrate and permeate, and obtain the separated concentrate and permeate.

[0112] In one possible implementation, a pressure valve is used to precisely control the transmembrane pressure differential ΔP. For example, by automatically adjusting the pressure valve, ΔP is kept stable at 55 bar, preventing membrane module damage or decreased separation efficiency due to pressure fluctuations. The concentrate may contain high concentrations of nickel and cobalt ions, while the ion concentration in the permeate is significantly reduced. For example, the nickel ion concentration in the permeate is reduced to below 0.5 mg / L, meeting emission standards. The beneficial effect is that precise pressure control extends the service life of the membrane module while ensuring the stability of the separation process.

[0113] The separated concentrate and permeate can be sent to subsequent treatment units. The concentrate can be further processed using chemical precipitation to recover heavy metals, while the permeate can be directly discharged or recycled. For example, nickel ions in the concentrate can be precipitated by adding sodium hydroxide, achieving a recovery rate of over 90%. The permeate, due to its low ion concentration, can be used in the plant's cooling water system. The benefits include resource utilization, reduced treatment costs, and a decrease in the environmental impact of wastewater discharge.

[0114] In one possible implementation, the optimization of the membrane separation parameter set can also be combined with real-time monitoring data. For example, sensors can detect changes in nickel ion concentration on the feed side in real time. If the concentration suddenly increases to 8 mg / L, the random forest algorithm can dynamically adjust P1 to 65 bar to adapt to higher separation requirements. This dynamic adjustment mechanism ensures the system's adaptability to fluctuations in wastewater concentration. The benefits include improved system automation, reduced manual intervention, and increased treatment efficiency.

[0115] Furthermore, in the method for removing impurities from the nickel-cobalt-manganese ternary precursor for lithium batteries provided in this embodiment, step S300 includes:

[0116] Step S310: Use an ion-selective electrode to detect the concentrations of nickel ions, cobalt ions, and manganese ions in the permeate to obtain residual heavy metal concentration data.

[0117] In the field of industrial wastewater treatment, the use of ion-selective electrodes to detect the concentration of heavy metal ions in the permeate is a key technology for ensuring the quality of purified water. Ion-selective electrodes can accurately detect the residual concentration of specific ions, such as nickel, cobalt, and manganese ions, by measuring the potential changes. For example, in the permeate of a wastewater treatment plant, the concentrations of nickel ions were detected as 0.4 mg / L, cobalt ions as 0.2 mg / L, and manganese ions as 0.1 mg / L. These data provide a basis for the subsequent optimization of the resin adsorption column. It should be noted that the high sensitivity of ion-selective electrodes ensures accurate detection even at low concentrations, facilitating timely adjustments to the treatment process.

[0118] Step S320: Based on the residual heavy metal concentration data, optimize the Langmuir constant K1 of the resin adsorption column using a neural network.

[0119] Based on residual heavy metal concentration data, a neural network was used to optimize the Langmuir constant K1 of the resin adsorption column. The neural network predicted the optimal K1 value by analyzing historical concentration data and adsorption performance. For example, assuming an initial K1 value of 0.05 L / mg, it was adjusted to 0.08 L / mg after neural network optimization. This optimization process considered the interactive effects of ion concentration, resin type, and wastewater flow rate.

[0120] Preferably, the training data for the neural network includes adsorption experimental results under various operating conditions to ensure that the K1 value adapts to the actual operating environment. The optimized K1 value increases the resin's affinity for heavy metal ions, thereby improving adsorption efficiency.

[0121] Step S330: If the optimized Langmuir constant K1 satisfies the preset threshold, then according to the Langmuir equation q e =K1C e / (1+K1C e Calculate the equilibrium adsorption capacity q e The equilibrium adsorption data were obtained.

[0122] In one embodiment, the equilibrium adsorption amount q is calculated using the Langmuir equation based on the optimized K1 value. e For example, when the equilibrium concentration of nickel ions in the permeate is C e When the concentration of K1 is 0.4 mg / L and K1 is 0.08 L / mg, the qe value can be obtained, which reflects the adsorption capacity of the resin.

[0123] It should be noted that the Langmuir equation assumes monolayer adsorption and is suitable for describing the adsorption behavior of heavy metal ions by resins. (The last sentence appears to be incomplete and possibly refers to a different topic.) e The system can determine whether the resin is close to saturation based on this value. For example, the calculation result shows q. e A concentration of 10 mg / g indicates that the resin still has an adsorption capacity. This method ensures precise control of the adsorption process and avoids premature resin saturation.

[0124] Step S340: Based on the equilibrium adsorption data, adjust the operating parameters of the resin adsorption column to generate purified water and saturated resin, thereby obtaining purified water that meets the standards and resin that is detected as saturated.

[0125] Based on the equilibrium adsorption data, adjust the operating parameters of the resin adsorption column, such as the feed flow rate and adsorption time, to generate purified water and saturated resin. For example, if q e The value indicates that the resin is close to saturation. The feed flow rate can be reduced from 50L / h to 30L / h, and the adsorption time can be extended to 6 hours to ensure that heavy metal ions are fully adsorbed.

[0126] Preferably, the nickel ion concentration in the purified water can be reduced to below 0.05 mg / L, meeting stringent emission standards. The saturated resin is regenerated through acid washing, for example using a 2 mol / L hydrochloric acid solution, achieving a recovery rate of over 85%. This regeneration method not only extends the resin's lifespan but also reduces operating costs.

[0127] In one possible implementation, the real-time monitoring system is linked to the resin adsorption column. For example, when a sudden increase in nickel ion concentration to 0.6 mg / L is detected, the system automatically reduces the flow rate to 25 L / h and extends the adsorption time. This dynamic adjustment mechanism improves the system's adaptability to concentration fluctuations.

[0128] Preferably, the purified water can be directly used in the plant's internal circulation system, such as for cooling water replenishment, while the regeneration treatment of saturated resin achieves the resource recovery of heavy metals. The advantage of this method lies in its ability to significantly improve the stability and resource utilization efficiency of wastewater treatment through precise detection and optimization.

[0129] Furthermore, in the method for removing impurities from the nickel-cobalt-manganese ternary precursor for lithium batteries provided in this embodiment, step S400 includes:

[0130] Step S410: Collect the conductivity data of the eluent in real time using a conductivity sensor, process the conductivity data using a time series analysis method, and obtain the conductivity change rate.

[0131] In the field of industrial wastewater treatment, real-time acquisition of eluent conductivity data using conductivity sensors is a crucial method for monitoring the resin regeneration process. Conductivity sensors can rapidly reflect changes in ion concentration in the eluent, thus indirectly characterizing the elution of heavy metal ions. In principle, conductivity is related to the total amount of conductive ions in the solution; when heavy metal ions are eluted from saturated resin by acid washing solution, the conductivity changes significantly. Real-time acquisition of this data provides a foundation for subsequent analysis. For example, in a wastewater treatment plant, a conductivity sensor records the eluent conductivity value once per second, with an initial value of 10 mS / cm. As the regeneration process progresses, the conductivity gradually decreases, reflecting the gradual elution of heavy metal ions.

[0132] In one possible implementation, time series analysis is used to process conductivity data to obtain the rate of change. Time series analysis calculates the trend of conductivity over time by performing differential calculations on conductivity data at consecutive time points. For example, the system calculates the rate of change of conductivity every minute; the initial rate of change might be 0.5 mS / cm·min, decreasing to 0.05 mS / cm·min as the regeneration process nears its end. This method can capture the dynamic changes in ion concentration in the eluent, providing a basis for determining the regeneration endpoint.

[0133] It should be noted that time series analysis can also be combined with smoothing algorithms to filter out noise interference and ensure the accuracy of the rate of change.

[0134] Step S420: If the rate of change of conductivity is less than a preset threshold, the regeneration process is terminated by an automated control device to obtain regenerated liquid and regenerated resin.

[0135] When the rate of change in conductivity falls below a preset threshold, such as 0.02 mS / cm·min, it indicates that heavy metal ions in the eluent have been largely eluted, and the system terminates the regeneration process via an automated control device. Specifically, upon receiving the rate of change data, the automated control device shuts off the pickling solution pump, stops the eluent circulation, and yields regenerated solution and regenerated resin. For example, in a regeneration operation, the system detects that the rate of change drops to 0.015 mS / cm·min, automatically stops regeneration, and generates approximately 100 L of regenerated solution and 200 kg of regenerated resin. This automated control improves operational efficiency and reduces waste of pickling solution.

[0136] Step S430: For the regenerated solution, the concentration of residual heavy metal ions is detected using an ion-selective electrode to obtain ion concentration data.

[0137] In one possible implementation, an ion-selective electrode is used to detect the concentration of residual heavy metal ions in the regenerated solution. The ion-selective electrode accurately detects the concentrations of nickel ions, copper ions, etc., in the regenerated solution by measuring the potential changes of specific ions. For example, the detection results show that the concentration of nickel ions in the regenerated solution is 0.3 mg / L, and the concentration of copper ions is 0.15 mg / L. These data provide a basis for evaluating the regeneration effect. It should be noted that the high selectivity of the ion-selective electrode ensures accurate measurement of the target ions and avoids interference from other ions.

[0138] Step S440: Based on the ion concentration data, calculate the regeneration efficiency using the preset formula q=(C0-C1) / C0 to obtain the regeneration efficiency data, where C0 represents the initial heavy metal ion concentration, C1 represents the ion concentration data, and q represents the regeneration efficiency.

[0139] For example, based on ion concentration data, the regeneration efficiency is calculated using the formula q=(C0-C1) / C0, where C0 is the initial concentration and C1 is the detected concentration. Assuming an initial nickel ion concentration of 10 mg / L and a detected concentration of 0.3 mg / L, the regeneration efficiency is 97%. This result indicates that the resin regenerates well, restoring most of its adsorption capacity. It should be noted that high regeneration efficiency means the resin can be reused, reducing operating costs. Furthermore, heavy metal ions in the regenerated solution can be further recovered and converted into resources. This method, through real-time monitoring and data analysis, significantly improves the stability and resource utilization level of the wastewater treatment system.

[0140] Furthermore, in the method for removing impurities from the nickel-cobalt-manganese ternary precursor for lithium batteries provided in this embodiment, step S500 includes:

[0141] Step S510: Collect volume and flow rate data of the concentrate using a level sensor and a flow meter, and filter the collected data using a data processing module to obtain the first liquid input data.

[0142] In the field of industrial wastewater treatment, collecting volume and flow rate data of the concentrate using level sensors and flow meters is a key method for optimizing the evaporation and crystallization process. For example, level sensors measure the height of the concentrate using ultrasound or pressure difference, and calculate the volume by combining this with the cross-sectional area of ​​the tank, while flow meters record the liquid flow rate using electromagnetic or turbine principles.

[0143] For example, in a wastewater treatment plant, a level sensor records the height of the concentrate every minute, with an initial value of 1.2 meters. The tank cross-sectional area is 2 square meters, resulting in a calculated volume of 2.4 cubic meters. A flow meter records a flow rate of 0.5 cubic meters per minute. This data provides the basis for subsequent processing. Specifically, the data processing module filters the collected data to eliminate noise interference.

[0144] In one possible implementation, a moving average filtering algorithm is used to average the liquid level and flow rate data over five consecutive seconds to obtain smooth initial liquid input data. For example, if the initial flow rate fluctuation is between 0.48 and 0.52 cubic meters per minute, it will stabilize at 0.5 cubic meters per minute after filtering. This method ensures data reliability and provides an accurate basis for subsequent control.

[0145] Step S520: Based on the first liquid input data, adjust the heating power using the evaporation crystallization device control module, calculate the target solubility S using the solubility equation S=S0e^(-ΔH / RT), and obtain the optimal crystallization temperature T.

[0146] For the control module of the evaporation crystallization apparatus, adjusting the heating power based on the initial liquid input data is a core step. For example, the system determines the input rate of the concentrate based on volume and flow rate data, and then adjusts the heater power accordingly. When the input rate is high, the system automatically increases the power to 200 kW to accelerate evaporation, while reducing it to 150 kW to save energy when the rate decreases. The system then calculates the target solubility using the solubility equation to determine the optimal crystallization temperature. For instance, the system estimates the solubility of sodium sulfate using preset parameters and determines the optimal crystallization temperature to be 85 degrees Celsius.

[0147] Step S530: If the optimal crystallization temperature T is determined, the evaporation crystallization device is adjusted to the optimal crystallization temperature T through the temperature control module, and the crystallization program is run to obtain sodium sulfate crystals and mother liquor.

[0148] The temperature control module then adjusts the device to the specified temperature and runs the crystallization process to generate sodium sulfate crystals and mother liquor. In one possible implementation, for solid-liquid separation of the mother liquor, a centrifugal separator separates the crystals from the mother liquor by high-speed rotation.

[0149] Step S540: For the mother liquor, a centrifugal separation device is used for solid-liquid separation. The mass data of sodium sulfate crystals is collected by a mass sensor to obtain the crystal output and mother liquor recovery.

[0150] A quality sensor collects crystal quality data in real time. For example, in one operation, 50 kg of sodium sulfate crystals were separated, and 1.8 cubic meters of mother liquor were recovered. This data is used to evaluate crystal production efficiency and the subsequent processing requirements of the mother liquor.

[0151] For example, high-quality crystals can be directly used in industrial production, while the mother liquor can be further recycled to recover residual substances. It should be noted that the coordinated operation of the level sensor and flow meter improves the accuracy of data acquisition, filtering ensures data stability, dynamic adjustment of heating power optimizes energy consumption, and centrifugal separation and mass sensors enhance the controllability of crystal production. This multi-stage collaborative approach significantly improves the production efficiency of sodium sulfate crystals in the wastewater treatment system while reducing mother liquor waste, thus ensuring resource utilization.

[0152] This invention relates to a purification system for lithium-ion battery nickel-cobalt-manganese ternary precursors, used to implement the aforementioned purification method for lithium-ion battery nickel-cobalt-manganese ternary precursors. The purification system for lithium-ion battery nickel-cobalt-manganese ternary precursors includes a prediction module, a generation module, a first output module, a monitoring module, and a second output module. The prediction module acquires concentration data output from a multi-sensor array and inputs the concentration data into a support vector machine model f(x)=∑α. i y i K(x i The composition complexity level is predicted using x)+b, where the concentration data includes nickel ion concentration (C_Ni), cobalt ion concentration (C_Co), manganese ion concentration (C_Mn), ammonia nitrogen content (N_NH3), and sodium sulfate concentration (C_Na2SO4); α i For Lagrange multipliers, y i For the training sample labels, K(x) i x) is the kernel function, b is the bias term; the generation module is used to trigger the random forest algorithm to calculate membrane separation parameters if the predicted complexity level exceeds a preset threshold. When the heavy metal ion concentration exceeds a preset threshold, the high-pressure reverse osmosis membrane module is started and the pressure valve is adjusted to control the transmembrane pressure difference ΔP=P1-P2, generating concentrate and permeate, where P1 is the feed side pressure and P2 is the permeate side pressure; the first output module is used to detect the residual heavy metal concentration in the permeate using an ion-selective electrode, optimize the Langmuir constant K1 of the resin adsorption column through a neural network, and according to q e =K1C e / (1+K1C e Calculate the equilibrium adsorption capacity q e It outputs purified water and saturated resin, of which C eThe first module represents the concentration of the target pollutant remaining in the liquid phase when adsorption reaches dynamic equilibrium; the second module monitors the rate of change of the conductivity of the eluent during the regeneration process of the saturated resin. If the rate of change of the conductivity of the eluent is less than a preset value, the regeneration process is terminated, and regenerated liquid and regenerated resin are obtained; the third module inputs the concentrated liquid into the evaporation crystallization device, calculates the optimal crystallization temperature T according to the solubility equation S=S0e^(-ΔH / RT), and outputs sodium sulfate crystals and mother liquor, where S is the target solubility, S0 is the standard solubility, ΔH is the enthalpy of dissolution, R is the gas constant, and e is the base of the natural logarithm.

[0153] Furthermore, the lithium battery nickel-cobalt-manganese ternary precursor impurity removal system provided in this embodiment includes a prediction module comprising a first acquisition unit, a second acquisition unit, a first determination unit, and a third acquisition unit. The first acquisition unit acquires the nickel ion concentration, cobalt ion concentration, manganese ion concentration, ammonia nitrogen content, and sodium sulfate concentration output by a multi-sensor array, and normalizes the concentration data using a preprocessing algorithm to obtain standardized concentration data. The second acquisition unit calculates the similarity matrix between the standardized concentration data and training samples using a kernel function based on the standardized concentration data to obtain a feature mapping matrix. The first determination unit, if the feature mapping matrix satisfies preset classification conditions, calculates the classification hyperplane of the support vector machine using Lagrange multipliers and training sample labels to determine the classification model parameters. The third acquisition unit uses the classification model parameters to predict the component complexity level of the standardized concentration data to obtain the predicted complexity level.

[0154] Furthermore, the lithium battery nickel-cobalt-manganese ternary precursor impurity removal system provided in this embodiment includes a generation module comprising a fourth acquisition unit, a second determination unit, and a fifth acquisition unit. The fourth acquisition unit is used to calculate membrane separation parameters using a random forest algorithm to obtain a membrane separation parameter set if the predicted complexity level exceeds a preset threshold. The second determination unit is used to determine whether the nickel ion concentration, cobalt ion concentration, or manganese ion concentration exceeds a preset threshold based on the membrane separation parameter set. If it does, the high-pressure reverse osmosis membrane module is activated to determine the feed-side pressure P1 and the permeate-side pressure P2. The fifth acquisition unit is used to control the transmembrane pressure difference ΔP = P1 - P2 using a pressure valve to generate concentrate and permeate, thus obtaining the separated concentrate and permeate.

[0155] Preferably, the lithium battery nickel-cobalt-manganese ternary precursor impurity removal system provided in this embodiment includes a first output module comprising a sixth acquisition unit, an optimization unit, a seventh acquisition unit, and an eighth acquisition unit. The sixth acquisition unit is used to detect the concentrations of nickel ions, cobalt ions, and manganese ions in the permeate using an ion-selective electrode to obtain residual heavy metal concentration data. The optimization unit is used to optimize the Langmuir constant K1 of the resin adsorption column using a neural network based on the residual heavy metal concentration data. The seventh acquisition unit is used to, if the optimized Langmuir constant K1 meets a preset threshold, then according to the Langmuir equation q... e =K1C e / (1+K1C e Calculate the equilibrium adsorption capacity q e The eighth acquisition unit is used to adjust the operating parameters of the resin adsorption column based on the equilibrium adsorption data, generate purified water and saturated resin, and obtain purified water that meets the standards and resin that is detected as saturated.

[0156] The method and system for removing impurities from nickel-cobalt-manganese ternary precursors for lithium batteries provided by this invention have the following specific advantages compared with the prior art:

[0157] I. Improved resource recycling efficiency

[0158] 1. Nickel, cobalt and manganese recovery rate: increased from 85%-90% in traditional processes to 95%-98%, significantly reducing resource waste.

[0159] 2. Sodium sulfate recovery: Enables continuous production of industrial-grade sodium sulfate (≥99%), creating additional economic benefits.

[0160] II. Environmental Protection and Energy Consumption Optimization

[0161] 1. Wastewater discharge: The concentration of heavy metals in the permeate is less than 0.1 mg / L, which is better than the "Emission Standard of Pollutants for Battery Industry" (GB 30484-2013).

[0162] 2. Reduced energy consumption: The membrane separation process optimizes pressure parameters through the random forest algorithm, resulting in a 25%-30% reduction in energy consumption compared to traditional processes.

[0163] III. Enhanced Process Stability

[0164] 1. The SVM model provides early warning of changes in component complexity, enabling the system to automatically adjust parameters to cope with raw material fluctuations and reduce manual intervention.

[0165] 2. The neural network-optimized resin adsorption process keeps the fluctuation of effluent quality within ±5%.

[0166] IV. Cost Savings

[0167] 1. The amount of resin regenerator used is reduced by 30%, and the regeneration cycle is extended by 15%-20%, thus reducing operating costs.

[0168] 2. Precise control of crystallization temperature reduces steam consumption, resulting in a 20% reduction in crystallization energy consumption.

[0169] V. Intelligentization and Automation

[0170] The entire process achieves closed-loop control driven by sensor data, can be seamlessly connected to the industrial internet platform, and supports remote monitoring and decision-making.

[0171] In summary, the method and system for removing impurities from lithium battery nickel-cobalt-manganese ternary precursors provided in this embodiment achieves intelligent and precise control of the lithium battery precursor impurity removal process through the deep integration of multi-sensor fusion, machine learning algorithms and physicochemical models. It is significantly superior to traditional processes in terms of resource recycling efficiency, environmental indicators and operating costs, providing key technical support for the green development of the new energy materials industry.

[0172] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including both the preferred embodiments and all changes and modifications falling within the scope of the invention. Clearly, those skilled in the art can make various alterations and modifications to the invention without departing from its spirit and scope. Thus, if these modifications and modifications of the invention fall within the scope of the claims and their equivalents, the invention is also intended to include these modifications and modifications.

Claims

1. A method for impurity removal of lithium battery nickel-cobalt-manganese ternary precursor, characterized in that, The method comprises the following steps: Concentration data of a multi-sensor array output is acquired, and the concentration data is input into a support vector machine model f(x)=∑α i y i K(x i , x)+b to perform component complexity level prediction, wherein the concentration data includes nickel ion concentration C_Ni, cobalt ion concentration C_Co, manganese ion concentration C_Mn, ammonia nitrogen content N_NH3, and sodium sulfate concentration C_Na2SO4; α i is a Lagrange multiplier, y i is a training sample label, K(x i , x) is a kernel function, and b is a bias term; If the predicted complexity level exceeds the preset threshold, a random forest algorithm is triggered to calculate membrane separation parameters, and when the concentration of heavy metal ions exceeds the preset threshold, a high-pressure reverse osmosis membrane assembly is started and a pressure valve is adjusted to control the transmembrane pressure difference ΔP=P1-P2, to generate concentrated liquid and permeate liquid, wherein P1 is the feed side pressure and P2 is the permeation side pressure; The residual heavy metal concentration in the permeate was detected using an ion-selective electrode. The Langmuir constant K1 of the resin adsorption column was optimized using a neural network, based on q. e =K1C e / (1+K1C e Calculate the equilibrium adsorption capacity q e It outputs purified water and saturated resin, of which C e This indicates the concentration of the target pollutant remaining in the liquid phase when adsorption reaches dynamic equilibrium. The conductivity change rate of the eluent in the saturated resin regeneration process is monitored, and if the conductivity change rate of the eluent is less than a preset value, the regeneration process is terminated to obtain regenerated liquid and regenerated resin; The concentrated liquid is input into an evaporation crystallization device, the optimal crystallization temperature T is obtained according to the solubility equation S=S0e^(-ΔH / RT), the sodium sulfate crystal and the mother liquor are output, wherein S is the target solubility, S0 is the standard solubility, ΔH is the dissolution enthalpy, R is the gas constant, and e is the natural logarithm base.

2. The lithium battery nickel cobalt manganese ternary precursor impurity removal method of claim 1, wherein, The concentration data of the multi-sensor array output is acquired, and the concentration data is input into a support vector machine model f(x) = ∑α i y i K(x i , x) + b to perform the component complexity level prediction step. The nickel ion concentration, cobalt ion concentration, manganese ion concentration, ammonia nitrogen content and sodium sulfate concentration output by the multi-sensor array are obtained, and a pretreatment algorithm is used to normalize the concentration data to obtain standardized concentration data; According to the standardized concentration data, a kernel function is used to calculate the similarity matrix between the standardized concentration data and the training samples to obtain a feature mapping matrix; If the feature mapping matrix meets the preset classification condition, a classification hyperplane of a support vector machine is calculated through the Lagrange multiplier and the training sample labels to determine the classification model parameters; The classification model parameters are used to predict the component complexity level for the standardized concentration data to obtain a predicted complexity level.

3. The method for removing impurities of lithium battery nickel-cobalt-manganese ternary precursor according to claim 1, characterized in that, The step of if the predicted complexity level exceeds the preset threshold, triggering a random forest algorithm to calculate membrane separation parameters, and when the concentration of heavy metal ions exceeds the preset threshold, starting a high-pressure reverse osmosis membrane assembly and adjusting a pressure valve to control the transmembrane pressure difference ΔP=P1-P2, to generate concentrated liquid and permeate liquid, wherein P1 is the feed side pressure and P2 is the permeation side pressure, comprises: If the predicted complexity level exceeds the preset threshold, the membrane separation parameters are calculated through the random forest algorithm to obtain a set of membrane separation parameters; According to the set of membrane separation parameters, it is determined whether the nickel ion concentration, cobalt ion concentration or manganese ion concentration exceeds the preset threshold, and if so, a high-pressure reverse osmosis membrane assembly is started to determine the feed side pressure P1 and the permeation side pressure P2; The transmembrane pressure difference ΔP=P1-P2 is controlled by a pressure valve to generate concentrated liquid and permeate liquid to obtain separated concentrated liquid and permeate liquid.

4. The method for removing impurities of lithium battery nickel-cobalt-manganese ternary precursor according to claim 1, characterized in that, The step of detecting the residual heavy metal concentration of the permeate liquid by using the ion selective electrode, optimizing the Langmuir constant K1 of the resin adsorption column by the neural network, and calculating the equilibrium adsorption capacity q according to q e =K1C e / (1+K1C e ) of the heavy metal ions in the permeate liquid is provided. e The step of outputting purified water and saturated resin includes: The nickel ion concentration, cobalt ion concentration and manganese ion concentration in the permeate liquid are detected by an ion selective electrode to obtain residual heavy metal concentration data; According to the residual heavy metal concentration data, the Langmuir constant K1 of the resin adsorption column is optimized through a neural network; If the optimized Langmuir constant K1 meets a preset threshold, the equilibrium adsorption quantity q e is calculated according to the Langmuir equation q e =K1C e / (1+K1C e ), and equilibrium adsorption quantity data is obtained. According to the equilibrium adsorption capacity data, the operating parameters of the resin adsorption column are adjusted to generate purified water and saturated resin to obtain purified water meeting the standard and saturated resin.

5. The method for removing impurities of lithium battery nickel-cobalt-manganese ternary precursor according to claim 1, characterized in that, The step of monitoring the conductivity change rate of the eluent in the saturated resin regeneration process, and if the conductivity change rate of the eluent is less than a preset value, terminating the regeneration process to obtain regenerated liquid and regenerated resin, comprises: The conductivity data of the eluent is collected in real time by a conductivity sensor, and the conductivity data is processed by a time series analysis method to obtain a conductivity change rate; If the conductivity change rate is less than a preset threshold, the regeneration process is terminated by an automatic control device to obtain a regenerated liquid and regenerated resin; For the regenerated liquid, an ion selective electrode is used to detect the residual heavy metal ion concentration therein to obtain ion concentration data; According to the ion concentration data, a preset formula q=(C0-C1) / C0 is used to calculate the regeneration efficiency to obtain regeneration efficiency data, wherein C0 represents the initial heavy metal ion concentration, C1 represents the ion concentration data, and q represents the regeneration efficiency.

6. The method for removing impurities of lithium battery nickel-cobalt-manganese ternary precursor according to claim 1, characterized in that, The concentrated liquid is input into an evaporation crystallization device, and the optimal crystallization temperature T is obtained according to the solubility equation S=S0e^(-ΔH / RT), and the steps of outputting sodium sulfate crystals and mother liquor include: The volume and flow rate data of the concentrated liquid are collected by a liquid level sensor and a flow meter, and the collected data is filtered by a data processing module to obtain first liquid input data; According to the first liquid input data, a heating power is adjusted by an evaporation crystallization device control module, and the target solubility S is calculated by combining the solubility equation S=S0e^(-ΔH / RT) to obtain the optimal crystallization temperature T; If the optimal crystallization temperature T is determined, the evaporation crystallization device is adjusted to the optimal crystallization temperature T by a temperature control module, and a crystallization program is run to obtain sodium sulfate crystals and mother liquor; For the mother liquor, a solid-liquid separation is performed by a centrifugal separation device, and the mass data of the sodium sulfate crystals are collected by a mass sensor to obtain the crystal output and the mother liquor recovery.

7. A lithium battery nickel-cobalt-manganese ternary precursor impurity removal system for implementing the lithium battery nickel-cobalt-manganese ternary precursor impurity removal method according to any one of claims 1 to 6, characterized in that, The lithium battery nickel-cobalt-manganese ternary precursor impurity removal system comprises: A prediction module is configured to acquire concentration data output by the multi-sensor array, input the concentration data into a support vector machine model f(x)=∑α i y i K(x i , x)+b to perform component complexity level prediction, wherein the concentration data includes nickel ion concentration C_Ni, cobalt ion concentration C_Co, manganese ion concentration C_Mn, ammonia nitrogen content N_NH3, and sodium sulfate concentration C_Na2SO4; α i is a Lagrange multiplier, y i is a training sample label, K(x i , x) is a kernel function, and b is a bias term. A generation module is configured to trigger a random forest algorithm to calculate membrane separation parameters if the predicted complexity level exceeds a preset threshold, and to start a high-pressure reverse osmosis membrane assembly and adjust a pressure valve to control a transmembrane pressure difference ΔP=P1-P2 when the heavy metal ion concentration exceeds a preset threshold, to generate concentrated liquid and permeate liquid, wherein P1 is the feed side pressure and P2 is the permeation side pressure; The first output module is used to detect the residual heavy metal concentration in the permeate using an ion-selective electrode, and to optimize the Langmuir constant K1 of the resin adsorption column using a neural network, based on q. e =K1C e / (1+K1C e Calculate the equilibrium adsorption capacity q e It outputs purified water and saturated resin, of which C e This indicates the concentration of the target pollutant remaining in the liquid phase when adsorption reaches dynamic equilibrium. A monitoring module is configured to monitor the conductivity change rate of the eluent during the saturated resin regeneration process, and to terminate the regeneration process if the conductivity change rate of the eluent is less than a preset value, to obtain regenerated liquid and regenerated resin; A second output module is configured to input the concentrated liquid into an evaporation crystallization device, to obtain the optimal crystallization temperature T according to the solubility equation S=S0e^(-ΔH / RT), and to output sodium sulfate crystals and mother liquor, wherein S is the target solubility, S0 is the standard solubility, ΔH is the solubility enthalpy, R is the gas constant, and e is the natural logarithm base.

8. The lithium battery nickel-cobalt-manganese ternary precursor impurity removal system of claim 7, wherein, The prediction module comprises: A first acquisition unit is configured to acquire nickel ion concentration, cobalt ion concentration, manganese ion concentration, ammonia nitrogen content, and sodium sulfate concentration output by a multi-sensor array, and to perform normalization processing on the concentration data by a preprocessing algorithm to obtain standardized concentration data; A second acquisition unit is configured to calculate a similarity matrix between the standardized concentration data and training samples by a kernel function according to the standardized concentration data to obtain a feature mapping matrix. The first determining unit is configured to calculate a classification hyperplane of a support vector machine by using a Lagrange multiplier and the training sample label if the feature mapping matrix satisfies a preset classification condition, and determine a classification model parameter; The third obtaining unit is configured to perform component complexity level prediction on the standardized concentration data by using the classification model parameter, and obtain a predicted complexity level.

9. The lithium battery nickel-cobalt-manganese ternary precursor impurity removal system of claim 7, wherein, The generating module comprises: The fourth obtaining unit is configured to calculate a membrane separation parameter by using a random forest algorithm if the predicted complexity level exceeds a preset threshold, and obtain a set of membrane separation parameters; The second determining unit is configured to determine whether the nickel ion concentration, the cobalt ion concentration, or the manganese ion concentration exceeds a preset threshold according to the set of membrane separation parameters, and start a high-pressure reverse osmosis membrane assembly if the nickel ion concentration, the cobalt ion concentration, or the manganese ion concentration exceeds the preset threshold, and determine a feed side pressure P1 and a permeation side pressure P2; The fifth obtaining unit is configured to control the transmembrane pressure difference ΔP = P1-P2 by using a pressure valve, generate concentrated liquid and permeated liquid, and obtain separated concentrated liquid and permeated liquid.

10. The lithium battery nickel cobalt manganese ternary precursor impurity removal system of claim 7, wherein, The first output module comprises: The sixth obtaining unit is configured to detect the nickel ion concentration, the cobalt ion concentration, and the manganese ion concentration in the permeated liquid by using an ion selective electrode, and obtain residual heavy metal concentration data; The optimization unit is configured to optimize a Langmuir constant K1 of a resin adsorption column by using a neural network according to the residual heavy metal concentration data; The seventh acquisition unit is used to, if the optimized Langmuir constant K1 satisfies a preset threshold, then according to the Langmuir equation q e =K1C e / (1+K1C e Calculate the equilibrium adsorption capacity q e Thus, equilibrium adsorption data were obtained; The eighth obtaining unit is configured to adjust an operating parameter of the resin adsorption column according to the equilibrium adsorption amount data, generate purified water and saturated resin, and obtain purified water meeting a standard and saturated resin.

Images