Method and device for predicting lithium concentration based on artificial intelligence

Abstract

A method for predicting lithium concentration based on artificial intelligence, according to an embodiment, comprises: a cleaning step of obtaining a plurality of first variables through preprocessing of a plurality of initial variables used in an electrodialysis facility that produces lithium; a preprocessing step of inputting the plurality of first variables to a boosting-based artificial intelligence model to preprocess same, and obtaining final variables; and a step of inputting the final variables to the boosting-based artificial intelligence model to predict the concentration of the lithium.

Description

Artificial intelligence-based lithium concentration prediction method and device

[0001] The present invention relates to an artificial intelligence-based lithium concentration prediction method and device, and more particularly, to an artificial intelligence-based lithium concentration prediction method and device that predicts lithium concentration with a minimum of variables using artificial intelligence.

[0002] Bipolar Electrodialysis (BPED) is an electrochemical separation process that uses ion exchange materials to separate or concentrate specific ions from a solution. BPED is gaining attention as a particularly important water separation and regeneration technology.

[0003] BPED operational data may include various data collected during the operation of the technology. BPED operational data may include, for example, current and voltage data, material concentration data, temperature and pressure data, and time data.

[0004] To improve BPED operations based on experience and enhance lithium (Li) productivity, a model is needed that predicts the lithium concentration produced based on operational data. The prediction results can be used to control and optimize the production process.

[0005] One embodiment of the present invention provides an artificial intelligence-based lithium concentration prediction method and device that effectively predicts lithium concentration with a minimum number of variables by first preprocessing variables based on the importance of each variable according to permutation importance and second preprocessing variables based on mean absolute percentage error (MAPE) by applying a boosting series artificial intelligence model in a preprocessing process.

[0006] Among the embodiments, the artificial intelligence-based lithium concentration prediction method includes a cleaning step of obtaining a plurality of first variables through preprocessing of a plurality of initial variables used in an electrodialysis facility that produces lithium, a preprocessing step of inputting the plurality of first variables into a boosting series artificial intelligence model to preprocess them and obtain final variables, and a step of inputting the final variables into the boosting series artificial intelligence model to predict the lithium concentration.

[0007] The above cleaning step may include a data cleaning step of removing missing values, removing non-identical data, and removing data for processes other than the BPED process.

[0008] The above cleaning step may further include a step of detecting and removing at least one initial variable among the initial variables whose correlation is greater than a reference value through correlation analysis.

[0009] The above boosting series artificial intelligence model can be selected from the group consisting of AdaBoost, Random forest, Catboost, Gradient Boosting Model, Light GBM, and XGBoost.

[0010] The above preprocessing step includes a first preprocessing step of inputting the first variables into a boosting series artificial intelligence model to calculate the variable importance of each of the first variables, and obtaining a plurality of second variables through preprocessing based on the variable importance, and the variable importance may use permutation importance.

[0011] The first preprocessing step may include a step of removing first variables among the first variables whose permutation importance is less than a preset second criterion.

[0012] The above preprocessing step may further include a second preprocessing step of inputting the second variables into the boosting series artificial intelligence model to obtain a plurality of third variables that satisfy a predicted consistency higher than a first criterion with a minimum number of variables, and the second preprocessing step may include a step of calculating the predicted consistency through a mean absolute percentage error (MAPE).

[0013] The second preprocessing step may further include a step of calculating variable importance for each of the acquired third variables.

[0014] The second preprocessing step may further include a step of comparing the degree of matching when only one variable with the highest variable importance among the third variables is input into the boosting series artificial intelligence model with the degree of predicted matching, and if the degree of predicted matching is greater, determining the third variables as final variables.

[0015] The above third variables may each be selected from among conductivity, circulation flow rate, and circulation pressure for any one of lithium sulfate aqueous solution, sulfuric acid, and lithium hydroxide.

[0016] Among the embodiments, the artificial intelligence-based lithium concentration prediction device includes a cleaning unit that obtains a plurality of first variables through preprocessing of a plurality of initial variables used in an electrodialysis facility that produces lithium, a preprocessing unit that inputs the plurality of first variables into a boosting series artificial intelligence model to preprocess them and obtain final variables, and a concentration prediction unit that inputs the final variables into the boosting series artificial intelligence model to predict the concentration of the lithium.

[0017] The above cleaning unit can remove missing values, remove non-identical data, and remove data for processes other than the BPED process.

[0018] The above cleaning unit can detect and remove at least one initial variable among the initial variables whose correlation is greater than a reference value through correlation analysis.

[0019] The above boosting series artificial intelligence model can be selected from the group consisting of AdaBoost, Random forest, Catboost, Gradient Boosting Model, Light GBM, and XGBoost.

[0020] The above preprocessing unit includes a first preprocessing unit that inputs the first variables into a boosting series artificial intelligence model to calculate the variable importance of each of the first variables, and obtains a plurality of second variables through the first preprocessing based on the variable importance, and the variable importance may use permutation importance.

[0021] The first preprocessing unit can remove first variables among the first variables whose permutation importance is less than a preset second criterion.

[0022] The above preprocessing unit further includes a second preprocessing unit that inputs the second variables into the boosting series artificial intelligence model to obtain a plurality of third variables that satisfy a predicted consistency higher than a first criterion with a minimum number of variables, and the second preprocessing unit can calculate the predicted consistency through a mean absolute percentage error (MAPE).

[0023] The above second preprocessing unit can calculate variable importance for each of the acquired third variables.

[0024] The second preprocessing unit can compare the matching degree and the predicted matching degree in the case where only one variable with the highest variable importance among the third variables is input into the boosting series artificial intelligence model, and if the predicted matching degree is greater, the third variables can be confirmed as final variables.

[0025] The above third variables may each be selected from among the conductivity, circulation flow rate, and circulation pressure for any one of the lithium sulfate aqueous solution, sulfuric acid, and lithium hydroxide.

[0026] An artificial intelligence-based lithium concentration prediction method and device according to one embodiment of the present invention applies a boosting series artificial intelligence model in a preprocessing process, thereby first preprocessing variables based on the importance of each variable according to permutation importance, and second preprocessing variables based on mean absolute percentage error (MAPE), thereby effectively predicting lithium concentration with a minimum number of variables.

[0027] FIG. 1 is a drawing showing the configuration of an electrodialysis facility according to one embodiment of the present invention.

[0028] FIG. 2 is a drawing showing one end of an electrodialysis facility according to one embodiment of the present invention.

[0029] FIG. 3 is a block diagram of an artificial intelligence-based lithium concentration prediction device according to one embodiment of the present invention.

[0030] FIG. 4 and FIG. 5 are flowcharts showing an artificial intelligence-based lithium concentration prediction method according to one embodiment of the present invention.

[0031] FIG. 6 is a diagram showing a preprocessing step of an artificial intelligence-based lithium concentration prediction method according to one embodiment of the present invention.

[0032] Figure 7 is a graph showing the effectiveness of an artificial intelligence-based lithium concentration prediction method according to one embodiment of the present invention.

[0033] FIG. 8 is a drawing for explaining a computing device according to one embodiment of the present invention.

[0034] Below, with reference to the attached drawings, embodiments of the present invention are described in detail so that those skilled in the art can easily implement the present invention. However, the present invention may be implemented in various different forms and is not limited to the embodiments described herein. In addition, in the drawings, parts irrelevant to the description are omitted for clarity of description, and similar parts are designated with similar reference numerals throughout the specification.

[0035] Throughout the specification and claims, whenever a part is referred to as "comprising" a component, this does not exclude other components, but rather includes other components, unless otherwise stated. Terms including ordinal numbers, such as "first," "second," etc., may be used to describe various components, but these components are not limited by these terms. These terms are used solely to distinguish one component from another.

[0036] Terms such as “part,” “unit,” and “module” described in the specification may mean a unit capable of processing at least one function or operation described in the specification, which may be implemented by hardware or a circuit, software, or a combination of hardware or a circuit and software.

[0037] Hereinafter, embodiments of the present invention will be described with reference to the drawings.

[0038] Fig. 1 is a drawing showing the configuration of an electrodialysis facility according to one embodiment of the present invention. Fig. 2 is a drawing showing one end of an electrodialysis facility according to one embodiment of the present invention.

[0039] In FIG. 1 and FIG. 2, the electrodialysis equipment may be a BPED (Bipolar ElectroDialysis) equipment. That is, the electrodialysis equipment may be a bipolar electrodialysis equipment. The electrodialysis equipment (BPED) may be equipment that converts a lithium sulfate aqueous solution into lithium hydroxide and sulfuric acid. Here, lithium sulfate is Li2SO4, lithium hydroxide is LiOH, and sulfuric acid is H2SO4. The electrodialysis equipment (BPED) may be an aqueous solution treatment equipment that simultaneously performs water decomposition / ion separation using an electrodialysis membrane in an electric field.

[0040] Referring to FIGS. 1 and 2, the electrodialysis device (BPED) may include a cation exchange membrane (CEM), an anion exchange membrane (AEM), and a bipolar membrane (BPM).

[0041] A cation exchange membrane (CEM) has an internal anionic group, allowing only cations (e.g., Li+) to pass through. An anion exchange membrane (AEM) allows only anions (e.g., SO42-) to pass through due to its internal cation group. A bipolar membrane (BPM) consists of a cation membrane and an anion membrane overlapping each other with a water-splitting catalyst in between. A bipolar membrane (BPM) can decompose water in an electric field to produce hydrogen ions (H+) and hydroxide ions (OH-).

[0042] That is, the electrodialysis equipment (BPED) may be an aqueous solution treatment equipment that simultaneously performs water decomposition (decomposition into H+, OH-) / ion separation (ion separation of Li+, SO42-) using an electrodialysis membrane (cation dialysis membrane, anion dialysis membrane, bipolar membrane) in an electric field.

[0043] In Fig. 1, in a BPED device, the LS solution can transfer Li and SO4 ions to the LH solution and sulfuric acid (H2SO4) solution through a three-stage process (Press) including the first to third stages. For example, in the BPED device, deionized water (DI water) can be converted into the LH solution and sulfuric acid solution by contacting the LS solution in a counterflow manner. Here, the LS solution is lithium sulfate (Li2SO4), and the LH solution is lithium hydroxide (LiOH).

[0044] The first to third stages may each include a salt room, an acid room, a base room, a salt tank, an acid tank, and a base tank.

[0045] In each stage, the Salt chamber supplies lithium sulfate (Li2SO4) to produce desalted water after the reaction. The Acid chamber supplies deionized water (DI water) to produce sulfuric acid (H2SO4) after the reaction. The Base chamber supplies deionized water (DI water) to produce lithium hydroxide (LiOH) after the reaction.

[0046] The salt tank stores the produced desalinated water. The acid tank stores the produced sulfuric acid. The base tank stores the produced lithium hydroxide.

[0047] The production volume of the BPED (Bioelectrodialysis Device) is determined by the discharge flow rates of sulfuric acid and lithium hydroxide. The production volume can be determined by controlling the input flow rates of water (H2O) and lithium sulfate into the BPED, the current and voltage of the rectifier, and the management of pH, conductivity, circulation flow, and circulation pressure within each chamber. The input flow rates of the solution, the current and voltage of the rectifier, pH, conductivity, circulation flow, and circulation pressure, which correspond to the control and management elements, can be detected through internal sensors installed in each chamber. The control and management elements can correspond to variables that determine the concentration of lithium hydroxide and sulfuric acid produced.

[0048] FIG. 3 is a block diagram of an artificial intelligence-based lithium concentration prediction device according to one embodiment of the present invention.

[0049] An artificial intelligence-based lithium concentration prediction device (100) can predict the production concentration of lithium (Li) through specialized preprocessing of operation data including control elements and management elements of a BPED (Bioelectrodialysis Equipment). The artificial intelligence-based lithium concentration prediction device (100) can reduce noise and computational load, quantify the derivation of major factors, and improve the degree of matching by using an artificial intelligence model of the boosting series in the preprocessing of operation data and prediction of lithium concentration. That is, the artificial intelligence-based lithium concentration prediction device (100) performs an artificial intelligence-based lithium concentration prediction method that predicts the production concentration of lithium (Li) through specialized preprocessing of operation data using an artificial intelligence model of the boosting series. A specific description of the artificial intelligence-based lithium concentration prediction method will be described later with reference to FIGS. 4 to 6.

[0050] Referring to FIG. 3, the artificial intelligence-based lithium concentration prediction device (100) may include a cleaning unit (110), a first preprocessing unit (120), a second preprocessing unit (130), and a concentration prediction unit (140).

[0051] The first preprocessing unit (120) and the second preprocessing unit (130) may be collectively referred to as a preprocessing unit. The preprocessing unit may preprocess a plurality of initial variables and obtain final variables. The cleaning unit (110) may obtain a plurality of first variables through preprocessing a plurality of initial variables used in a lithium-producing electrodialysis device (BPED). The cleaning unit (110) may remove missing values, outliers, duplicate data, non-identical data (non-operating sensor data), and data for processes other than the BPED process through preprocessing of data cleaning from the initial variables including all variables of the BPED process. The cleaning unit (110) may also remove data that moves in the same manner as duplicate data based on correlation analysis for each data.

[0052] The cleaning unit (110) can obtain first variables whose number is reduced compared to the initial variables through data cleaning.

[0053]

[0054] The first preprocessing unit (120) can input the first variables into a boosting series artificial intelligence model and calculate the variable importance of each of the first variables.

[0055] Boosting-type AI models can be AI models that utilize boosting algorithms. Boosting is a type of machine learning algorithm that combines weak learners to create a strong learner. Boosting can be used for classification and regression problems. Boosting algorithms train models sequentially. At each step, the boosting algorithm adds a new model that corrects the errors of previous models. The boosting algorithm combines all models to produce a final prediction.

[0056] The boosting family of AI models can be selected from the group consisting of AdaBoost, Random Forest, Catboost, Gradient Boosting Model, Light GBM, and XGBoost. The boosting family of AI models is preferably XGBoost.

[0057] AdaBoost (Adaptive Boosting) is one of the early boosting algorithms. It learns by assigning weights to each data point and giving more weight to misclassified data points.

[0058] Gradient Boosting Machines (GBM) train models to reduce residual (error) errors. At each step, the residual is calculated using the gradient of the loss function.

[0059] XGBoost (Extreme Gradient Boosting) is an extended version of GBM, which includes features such as regularization, parallel processing, and missing value handling.

[0060] LightGBM is a boosting algorithm specialized for large datasets, featuring fast learning speed and efficient memory usage. Like XGBoost, it includes regularization and parallel processing capabilities.

[0061] CatBoost is a boosting algorithm specialized for categorical data, providing automatic categorical feature transformation and fast learning speed.

[0062] Boosting AI models, for regression problems like predicting production concentration, preprocess data, train models, and analyze performance and feature importance. Once a satisfactory model is obtained, boosting AI models are applied to real-world environments to predict concentration in real time and leverage this knowledge for production control and optimization.

[0063] In this process, the boosting series artificial intelligence model can identify which variables have the greatest influence on predicting production concentration by analyzing the importance of features or variables.

[0064] That is, the first preprocessing unit (120) can analyze the variable importance of the first variables through a boosting series artificial intelligence model and determine which of the first variables has a greater influence on the prediction of lithium production concentration.

[0065] The first preprocessing unit (120) can use permutation importance to determine variable importance. Permutation importance is a methodology for estimating the importance of a feature by determining how much the value of a model performance indicator (accuracy, F1-score, R^2, etc.) decreases when a feature is excluded from an artificial intelligence model.

[0066] The first preprocessing unit (120) can obtain a plurality of second variables through the first preprocessing based on the calculated variable importance. For example, the first preprocessing unit (120) can remove first variables whose permutation importance is lower than a preset specific criterion among the first variables. That is, the first preprocessing unit (120) can assume that the lower the importance of a variable according to the permutation importance of each first variable, the lower the impact on the performance of the artificial intelligence model, and can improve the lithium concentration prediction performance of the artificial intelligence model by reducing data noise by removing variables whose importance is lower than a specific criterion. The second variables can include variables remaining after removing variables whose permutation importance is lower than a specific criterion (the second criterion) among the first variables.

[0067] The second preprocessing unit (130) inputs the second variables into a boosting series artificial intelligence model to obtain a plurality of third variables that satisfy a prediction consistency higher than the first criterion with a minimum number of variables.

[0068]

[0069] Boosting-type AI models improve the predictive performance of models by generating models using preprocessed data and evaluating their performance. In one embodiment, the second preprocessing unit (130) evaluates the model's performance using second variables and, based on the evaluation results, obtains third variables with a predictive accuracy exceeding a certain threshold. For example, the second preprocessing unit (130) may use evaluation metrics such as MAE, RMSE, and MAPE.

[0070] MAE (Mean Absolute Error) is the average of the absolute values ​​of each prediction error. RMSE (Root Mean Square Error) is the square root of the average of the squares of each prediction error. MAPE (Mean Absolute Percentage Error) is the mean absolute error, which is calculated by dividing the absolute value of each prediction error by the actual value and then taking the average of the entire error. In other words, it is the percentage ratio of the difference between the actual value and the predicted value divided by the actual value. The result can be expressed as a percentage. MAPE is a tool for evaluating prediction consistency and is frequently used in regression problems / models. Because it represents the relative error ratio, it allows for intuitive interpretation of the prediction results.

[0071] The second preprocessing unit (130) calculates the predicted consistency of the artificial intelligence model using the second variables using the mean absolute error, and obtains third variables included in a set composed of the smallest number of variables among a plurality of sets of variables that satisfy a predetermined specific criterion (first criterion) for which the calculated predicted consistency is obtained.

[0072] The concentration prediction unit (140) can calculate the variable importance for each of the acquired third variables. The concentration prediction unit (140) can compare the predicted consistency when only one variable with the highest variable importance among the third variables is input into the boosting series artificial intelligence model with the predicted consistency when all of the third variables are input. If the predicted consistency when all of the third variables are input is greater than the predicted consistency when the one variable with the highest variable importance is input, the concentration prediction unit (140) can finalize the third variables as final variables and input them into the boosting series artificial intelligence model to predict the concentration of lithium.

[0073] The third variables may be selected from any one of conductivity, circulation flow rate and circulation pressure for any one of lithium sulfate aqueous solution, sulfuric acid and lithium hydroxide.

[0074] Figures 4 and 5 are flowcharts illustrating an artificial intelligence-based lithium concentration prediction method according to an embodiment of the present invention. Figure 6 is a diagram illustrating a preprocessing step of an artificial intelligence-based lithium concentration prediction method according to an embodiment of the present invention. The artificial intelligence-based lithium concentration prediction method can be performed through an artificial intelligence-based lithium concentration prediction device (100, see Figure 3).

[0075] In FIG. 4, an artificial intelligence-based lithium concentration prediction device (100) can obtain a first variable through preprocessing of initial variables used in a lithium-producing electrodialysis device (BPED) (step S100). The initial variables may include BPED operation data. The BPED operation data may include data on control elements and management elements that determine the production results of the BPED. In an artificial intelligence model, the operation data may include features or variables.

[0076] For example, an artificial intelligence-based lithium concentration prediction device (100) can obtain the first variables of Table 2 below through preprocessing using data cleaning from the initial variables of Table 1 below. Data cleaning may include removing missing values, outliers, duplicate data, non-identical data (non-operating sensor data), and data for processes other than the BPED process.

[0077] Number Variable Number Variable 1 First flow rate value......2 Second flow rate value384 First coolant valve adjustment value3 Third flow rate value385 Third coolant valve adjustment value4 Fourth flow rate value386 Fourth coolant valve adjustment value5 Fifth flow rate value387 Fifth coolant valve adjustment value6 Sixth flow rate value388 Seventh coolant valve adjustment value7 Seventh flow rate value389 Eighth coolant valve adjustment value8 Eighth flow rate value390 Ninth coolant valve adjustment value9 Nineth flow rate value391 Eleventh coolant valve adjustment value10 Tenth flow rate value392 Twelveth coolant valve adjustment value........26 First temperature value517 First stack voltage27 Second temperature value518 Second stack voltage28 Third temperature value519 Third stack voltage29 Fourth temperature value520 Fourth stack voltage30 Fifth temperature value521 Fifth stack voltage31 Sixth temperature value522 Sixth stack voltage32 Seventh temperature value523 Seventh Stack voltage 33rd temperature value 524th temperature value 34th temperature value 525th temperature value 9th stack voltage

[0078] In Table 1, the total number of initial variables is 525. The initial variables include all missing values, outliers, duplicate data, and data with the same correlation. Fig. 6 is a diagram showing how variables are removed through correlation analysis during data cleaning. In Fig. 6, the first coolant valve adjustment value, the third to fifth coolant valve adjustment values, the seventh to ninth coolant valve adjustment values, and the eleventh and twelfth coolant valve adjustment values ​​represent the coolant valve adjustment values ​​of the corresponding plurality of sensors, respectively. The first to twelfth temperature values ​​represent the temperature values ​​of the corresponding first to twelfth sensors, respectively. That is, Fig. 6 shows the correlation between the coolant valve adjustment values ​​and the temperature values. Referring to FIG. 6, the first coolant valve adjustment value, the third to fifth coolant valve adjustment values, the seventh to ninth coolant valve adjustment values, the eleventh and twelfth coolant valve adjustment values, and the first to twelfth temperature values ​​represent temperature values ​​and coolant valve adjustment values ​​measured through corresponding sensors within the same facility, respectively.

[0079] That is, since the temperature value and the coolant valve adjustment value for the same sensor within the same facility have a correlation of 1, they are in fact overlapping variables. For example, the first coolant valve adjustment value and the first temperature value measured by the first sensor have a correlation of 1. Therefore, the artificial intelligence-based lithium concentration prediction device (100) can remove any one of the variables of the first coolant valve adjustment value, the third to fifth coolant valve adjustment values, the seventh to ninth coolant valve adjustment values, the eleventh and twelfth coolant valve adjustment values, or the first temperature value, the third to fifth temperature values, the seventh to ninth temperature values, and the eleventh and twelfth temperature values ​​through preprocessing of data cleaning.

[0080] In addition, the artificial intelligence-based lithium concentration prediction device (100) can eliminate variables related to processes other than the BPED process (e.g., the water level of the tank corresponding to the process following BPED, a total of 86).

[0081] Number Variable 1 First flow rate value 2 Second flow rate value 3 Third flow rate value 4 Fourth flow rate value 5 Fifth flow rate value 6 Sixth flow rate value 7 Seventh flow rate value 8 Eighth flow rate value 9 Ninth flow rate value 10 Tenth flow rate value......394 Second conductivity value 395 Third conductivity value 396 Fourth conductivity value 397 Fifth conductivity value 398 Sixth conductivity value 399 Seventh conductivity value 400 Eighth conductivity value

[0082] In Table 2, the number of first variables is 400 in total. The first variables include the remaining variables after removing some variables through preprocessing of data cleaning from the initial variables. That is, 125 variables can be removed in the cleaning step. The artificial intelligence-based lithium concentration prediction device (100) can obtain second variables by inputting the first variables into a boosting series artificial intelligence model to calculate the variable importance of each of the first variables, and performing preprocessing to remove lower-order factors based on the calculated variable importance (step S200). In one embodiment, the artificial intelligence-based lithium concentration prediction device (100) can calculate the permutation importance for each of the 400 first variables. For example, the artificial intelligence-based lithium concentration prediction device (100) can obtain the second variables of Table 3, which include the top 100 variables based on the permutation importance among the 400 first variables of Table 2. Here, the top 100 can be determined variably by an arbitrary number.

[0083] Number Variable 1 3rd Conductivity Value 2 8th Conductivity Value 3 11th Flow Rate Value 4 3rd Flow Rate Value 5 10th Flow Rate Value......9 4th Flow Rate Value 95 19th Flow Rate Value......100 5th Flow Rate Value

[0084] The artificial intelligence-based lithium concentration prediction device (100) can input the second variables into a boosting series artificial intelligence model to obtain key variables that produce a prediction consistency higher than a specific standard for lithium concentration with a minimum number of variables (step S300). The artificial intelligence-based lithium concentration prediction device (100) can obtain four third variables of Table 4 based on the 100 second variables of Table 2. The third variables are the minimum number of variables that can predict lithium concentration with a prediction consistency higher than a specific standard by inputting them into the artificial intelligence model, and can be referred to as key variables.

[0085] Ranking VariableVariable Importance13rd Conductivity Value1.034210th Flow Rate Value0.42438th Flow Rate Value0.39411th Conductivity Value0.361

[0086] The artificial intelligence-based lithium concentration prediction device (100) inputs 100 second variables into a boosting series artificial intelligence model by starting from them and subtracting them one by one, and can compare the mean absolute error (MAPE) for each number of variables. The mean absolute error may be a standard for judging the degree of agreement. The artificial intelligence-based lithium concentration prediction device (100) can identify variables that satisfy the second criterion of permutation importance when the degree of prediction agreement according to the mean absolute error is higher than the first criterion (e.g., 95% or higher, MAPE standard 5 or lower) and the number of major variables is minimum. For example, the artificial intelligence-based lithium concentration prediction device (100) can determine that the case in which there are finally 4 major variables in Table 4 is the major variable with the minimum number of fixed consistency. The artificial intelligence-based lithium concentration prediction device (100) can compare the learning and evaluation results of the artificial intelligence model based on the 4 selected major variables with the learning and evaluation results of the model based on 1 major variable (e.g., the 3rd conductivity value of the 3rd sensor with the highest importance). The artificial intelligence-based lithium concentration prediction device (100) finally determines the 4 major variables when the evaluation result (prediction consistency) based on the 4 major variables is the best.

[0087] The artificial intelligence-based lithium concentration prediction device (100) can predict the concentration of lithium produced through BPED by inputting the acquired key variables into a boosting series artificial intelligence model (step S400). The artificial intelligence-based lithium concentration prediction device (100) can predict the lithium concentration using the four key variables in Table 4.

[0088] In one embodiment, the third variables (primary variables) may be selected from any one of conductivity, input flow rate, circulation flow rate, and circulation pressure for any one of lithium sulfate aqueous solution, sulfuric acid, and lithium hydroxide in multiple stages.

[0089] For example, the third conductivity value may be the conductivity of lithium hydroxide of the first base measured through the third sensor, the tenth flow rate value may be the circulation flow rate of sulfuric acid of the second acid measured through the tenth sensor, the eighth flow rate value may be the circulation flow rate of the first electrode solution through the eighth sensor, and the eleventh conductivity value may be the conductivity of lithium hydroxide of the third base through the eleventh sensor. Fig. 5 specifically shows one embodiment of Fig. 4. Fig. 5 sequentially shows an example of a preprocessing process of BPED operation data and a lithium concentration prediction process using a boosting series artificial intelligence model.

[0090] In FIG. 5, first, the AI-based lithium concentration prediction method may include a cleaning step of preparing and cleaning a dataset of BPED operation data (step S110). The AI-based lithium concentration prediction method obtains a first variable by cleaning the initial variable (step S120).

[0091] Afterwards, the artificial intelligence-based lithium concentration prediction method can preprocess the first variable using XGBoost, which is a boosting series artificial intelligence model. The artificial intelligence-based lithium concentration prediction method can input the first variable into XGBoost (step S210). The artificial intelligence-based lithium concentration prediction method determines the variable importance of each first variable using permutation importance (step S220). The artificial intelligence-based lithium concentration prediction method obtains the second variable by removing the subfactor of the permutation importance (step S230).

[0092] Afterwards, the AI-based lithium concentration prediction method inputs the second variable into XGBoost to obtain feedback on performance through the degree of matching (step S310). The AI-based lithium concentration prediction method searches for the condition where the degree of matching is 90% or higher and the number of variables satisfying the variable importance is minimal through the degree of matching determination factor such as MAPE (step S320). The AI-based lithium concentration prediction method derives and quantifies the major factors satisfying the above conditions as the minimum variables that affect the process (step S330).

[0093] Thereafter, the AI-based lithium concentration prediction method trains and evaluates an AI model using the derived key variables (step S410). The AI-based lithium concentration prediction method can verify the influence of key factors (key variables) based on the predicted consistency of the evaluation results (step S420). The influence of key factors can be verified through the graph in Figure 7.

[0094] Figure 7 is a graph illustrating the effectiveness of an AI-based lithium concentration prediction method according to one embodiment of the present invention. Figure 7 is a graph showing changes in lithium concentration over an operating period. Figure 7a shows a comparative example in which one key variable is applied. Figure 7b shows an embodiment of the present invention in which all four key variables are applied.

[0095] In Fig. 7a, the actual graph waveform of lithium concentration change (Real) differs from the predicted waveform of lithium concentration change (Pred). While the predicted waveform (Pred) remains constant at approximately 18 g / L, the actual waveform (Real) varies between approximately 16 g / L and approximately 19 g / L throughout the operating period. The mean absolute error (MAPE) is 5.406, indicating that the agreement is below 95%.

[0096] On the other hand, in Fig. 7b, the actual graph waveform of lithium concentration change (Real) is similar to the predicted waveform of lithium concentration change (Pred). The predicted (Pred) and actual (Real) lithium concentrations are similar, varying from about 16 g / L to about 19 g / L during the operation period. Since the mean absolute error (MAPE) is 2.021, the agreement is approximately 98%, exceeding the standard value.

[0097] FIG. 8 is a drawing for explaining a computing device according to one embodiment of the present invention.

[0098] Referring to FIG. 8, the artificial intelligence-based lithium concentration prediction method and device according to the embodiments can be implemented using a computing device (900).

[0099] The computing device (900) may include at least one of a processor (910), a memory (930), a user interface input device (940), a user interface output device (950), and a storage device (560) that communicate via a bus (920). The computing device (900) may also include a network interface (970) electrically connected to a network (90). The network interface (970) may transmit or receive signals to or from other entities via the network (90).

[0100] The processor (910) may be implemented in various types such as an MCU (Micro Controller Unit), an AP (Application Processor), a CPU (Central Processing Unit), a GPU (Graphics Processing Unit), an NPU (Neural Processing Unit), etc., and may be any semiconductor device that executes instructions stored in a memory (930) or a storage device (960). The processor (910) may be configured to implement the functions and methods described above with respect to FIGS. 1 to 7.

[0101] The memory (930) and storage device (960) may include various types of volatile or non-volatile storage media. For example, the memory may include read-only memory (ROM) (931) and random access memory (RAM) (932). In the present embodiment, the memory (930) may be located inside or outside the processor (910), and the memory (930) may be connected to the processor (910) via various known means.

[0102] In some embodiments, at least some components or functions of the artificial intelligence-based lithium concentration prediction method and device according to the embodiments may be implemented as a program or software running on a computing device (900), and the program or software may be stored on a computer-readable medium.

[0103] In some embodiments, at least some components or functions of the artificial intelligence-based lithium concentration prediction method and device according to the embodiments may be implemented using hardware or circuitry of the computing device (900), or may be implemented as separate hardware or circuitry that can be electrically connected to the computing device (900).

[0104] Although the embodiments of the present invention have been described in detail above, the scope of the present invention is not limited thereto, and various modifications and improvements made by a person of ordinary skill in the art to which the present invention pertains using the basic concept of the present invention defined in the following claims also fall within the scope of the present invention.

[0105] [Explanation of symbols]

[0106] 100: AI-based lithium concentration prediction device

[0107] 110: Cleaning Department

[0108] 120: Preprocessing Unit 1

[0109] 130: Second preprocessing unit

[0110] 140: Concentration Prediction Unit

Claims

1. A cleaning step for obtaining a plurality of first variables through preprocessing of a plurality of initial variables used in an electrodialysis facility for producing lithium; A preprocessing step of inputting the above plurality of first variables into a boosting series artificial intelligence model to preprocess them and obtain final variables; and An artificial intelligence-based lithium concentration prediction method comprising a step of inputting the final variables into the boosting series artificial intelligence model to predict the lithium concentration.

2. An artificial intelligence-based lithium concentration prediction method in the first paragraph, wherein the cleaning step includes a data cleaning step of removing missing values, removing non-identical data, and removing data for a process other than an electrodialysis process.

3. In the second paragraph, the cleaning step further includes a step of detecting and removing at least one initial variable among the initial variables whose correlation is greater than a reference value through correlation analysis. An artificial intelligence-based lithium concentration prediction method.

4. In the first paragraph, the boosting series artificial intelligence model is an artificial intelligence-based lithium concentration prediction method selected from the group consisting of AdaBoost, Random forest, Catboost, Gradient Boosting Model, Light GBM, and XGBoost.

5. In the first paragraph, the preprocessing step includes a first preprocessing step of inputting the first variables into a boosting series artificial intelligence model to calculate the variable importance of each of the first variables, and obtaining a plurality of second variables through preprocessing based on the variable importance. The above variable importance is an artificial intelligence-based lithium concentration prediction method using permutation importance.

6. In the fifth paragraph, an artificial intelligence-based lithium concentration prediction method, wherein the first preprocessing step includes a step of removing first variables among the first variables whose permutation importance is less than a preset second criterion.

7. In the 6th paragraph, the preprocessing step further includes a second preprocessing step of inputting the second variables into the boosting series artificial intelligence model to obtain a plurality of third variables that satisfy a prediction consistency higher than the first criterion with a minimum number of variables. An artificial intelligence-based lithium concentration prediction method, wherein the second preprocessing step includes a step of calculating the prediction consistency through the mean absolute percentage error (MAPE).

8. An artificial intelligence-based lithium concentration prediction method in claim 7, wherein the second preprocessing step further includes a step of calculating variable importance for each of the acquired third variables.

9. In the 8th paragraph, the second preprocessing step compares the matching degree and the predicted matching degree when only one variable with the highest variable importance among the third variables is input into the boosting series artificial intelligence model. An artificial intelligence-based lithium concentration prediction method further comprising a step of confirming the third variables as the final variables if the above prediction accuracy is greater.

10. An artificial intelligence-based lithium concentration prediction method in the first paragraph, wherein the final variables are each selected from among conductivity, circulation flow rate, and circulation pressure for any one of lithium sulfate aqueous solution, sulfuric acid, and lithium hydroxide.

11. A cleaning unit for obtaining a plurality of first variables through preprocessing of a plurality of initial variables used in an electrodialysis facility for producing lithium; A preprocessing unit that inputs the above plurality of first variables into a boosting series artificial intelligence model to preprocess them and obtain final variables; and An artificial intelligence-based lithium concentration prediction device including a concentration prediction unit that predicts the concentration of lithium by inputting the final variables into the boosting series artificial intelligence model.

12. In the 11th paragraph, the cleaning unit is an artificial intelligence-based lithium concentration prediction device that removes missing values, non-identical data, and data for processes other than the electrodialysis process.

13. In the 12th paragraph, the cleaning unit is an artificial intelligence-based lithium concentration prediction device that detects and removes at least one initial variable among the initial variables whose correlation is greater than a reference value through correlation analysis.

14. In the 11th paragraph, the boosting series artificial intelligence model is an artificial intelligence-based lithium concentration prediction device selected from the group consisting of AdaBoost, Random forest, Catboost, Gradient Boosting Model, Light GBM, and XGBoost.

15. In the 11th paragraph, the preprocessing unit includes a first preprocessing unit that inputs the first variables into a boosting series artificial intelligence model to calculate the variable importance of each of the first variables, and obtains a plurality of second variables through the first preprocessing based on the variable importance. The above variable importance is an artificial intelligence-based lithium concentration prediction device using permutation importance.

16. In the 15th paragraph, the first preprocessing unit is an artificial intelligence-based lithium concentration prediction device that removes first variables among the first variables whose permutation importance is less than a preset second criterion.

17. In the 16th paragraph, the preprocessing unit further includes a second preprocessing unit that inputs the second variables into the boosting series artificial intelligence model to obtain a plurality of third variables that satisfy a prediction consistency higher than the first criterion with a minimum number of variables. The above second preprocessing unit is an artificial intelligence-based lithium concentration prediction device that calculates the prediction consistency through the mean absolute percentage error (MAPE).

18. In the 17th paragraph, the second preprocessing unit is an artificial intelligence-based lithium concentration prediction device that calculates variable importance for each of the acquired third variables.

19. In the 18th paragraph, the second preprocessing unit compares the matching degree with the predicted matching degree when only one variable with the highest variable importance among the third variables is input into the boosting series artificial intelligence model. An artificial intelligence-based lithium concentration prediction device that determines the third variables as the final variables when the above prediction accuracy is greater.

20. In the 11th paragraph, an artificial intelligence-based lithium concentration prediction device, wherein the final variables are each selected from among conductivity, circulation flow rate, and circulation pressure for any one of a lithium sulfate aqueous solution, sulfuric acid, and lithium hydroxide.

Images