Window-based electrodialysis plant anomaly detection method and apparatus
By setting time series models for input and prediction windows and combining them with deep learning models, the problem of anomaly detection in bipolar membrane electrodialysis equipment was solved, enabling real-time anomaly detection and automated control of the equipment, thereby improving production efficiency and product quality.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- POSCO HLDG INC
- Filing Date
- 2024-11-21
- Publication Date
- 2026-07-14
AI Technical Summary
Existing technologies are insufficient to effectively detect abnormal operating conditions of bipolar membrane electrodialysis equipment, which affects production efficiency and product quality.
By setting appropriate input and prediction windows, time series models are applied to preprocess and detect anomalies in the control variables of electrodialysis equipment, and deep learning models such as Informer and DLinear models are used to predict and compare management variables.
It enables real-time anomaly detection of electrodialysis equipment, improving the accuracy of automated control in the production process and the stability of product quality.
Smart Images
Figure CN122396661A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to a window-based method and apparatus for detecting anomalies in electrodialysis equipment. More specifically, this invention relates to a window-based method and apparatus for detecting anomalies in electrodialysis equipment, used for the optimal automated operation of BPED (bipolar membrane electrodialysis) processes. Background Technology
[0002] Bipolar electrodialysis (BPED) is an electro-separation process used to separate or concentrate specific ions from a solution using ion exchange materials. BPED has attracted considerable attention as a particularly important water separation and regeneration technology.
[0003] BPED operational data can include various data collected during the operation of the technology. Examples of BPED operational data include current and voltage data, substance concentration data, temperature and pressure data, and time data. Summary of the Invention
[0004] (a) Technical problems to be solved One embodiment of the present invention aims to provide a window-based method and apparatus for detecting anomalies in electrodialysis equipment. As a preprocessing process specifically for operating data, it sets appropriate input and prediction window sizes and ratios, converts continuous segment unit data, and applies a time series model to predict management variables at each stage of bipolar electrodialysis (BPED).
[0005] (II) Technical Solution In some embodiments, a window-based method for detecting anomalies in an electrodialysis device includes: setting an input window range for control variables of the electrodialysis device; converting time-series data of the control variables into continuous segment unit data corresponding to the input window range; predicting a management variable corresponding to the control variables using an artificial intelligence model trained based on the continuous segment unit data; and comparing the predicted management variable with a benchmark value to detect whether the electrodialysis device is abnormal.
[0006] The control variables may include the solution inflow rate, circulation flow rate, and rectifier voltage at each stage of the electrodialysis equipment, and the management variables may include the conductivity at each stage and the stack voltage of the rectifier.
[0007] The step of setting the input window range may include: setting the input window range by reflecting the propagation time of the solution introduced into each stage of the electrodialysis equipment after ion exchange into the next stage.
[0008] The step of setting the input window range may include: controlling the propagation time by adjusting the flow rates of the first-level salt, the third-level acid, and the third-level base in the control variables.
[0009] It may further include the step of setting a prediction window range for the management variable based on the set input window range, the step of setting the prediction window range may include: setting the prediction window range such that the ratio of the prediction window range to the input window range is 1:1.
[0010] The step of converting to continuous segment unit data may include: performing data cleaning and data interpolation preprocessing on the continuous segment data within the input window range.
[0011] The steps for predicting the management variables may include: calculating the goodness of fit of the trained artificial intelligence model, and resetting the input window range and the prediction window range based on the calculated goodness of fit.
[0012] The step of resetting the input window range and the prediction window range may include: controlling the propagation time by adjusting the flow rate of the first-stage salt, the flow rate of the third-stage acid, and the flow rate of the third-stage alkali.
[0013] The steps for predicting the management variable may include: aligning the target time interval between the target time of the prediction window range and the starting point of the input window range with the input window range.
[0014] The artificial intelligence model may include a deep learning model for time series data processing, and the deep learning model includes the Informer model and the DLinear model.
[0015] In some embodiments, a window-based anomaly detection device for an electrodialysis device includes: a preprocessing unit for setting an input window range for control variables of the electrodialysis device; a data conversion unit for converting time-series data of the control variables into continuous segment unit data corresponding to the input window range; a prediction unit for predicting management variables corresponding to the control variables using an artificial intelligence model trained based on the continuous segment unit data; and an anomaly detection unit for comparing the predicted management variables with a benchmark value to detect whether the electrodialysis device is abnormal.
[0016] The control variables may include the solution inflow rate, circulation flow rate, and rectifier voltage at each stage of the electrodialysis equipment, and the management variables may include the conductivity at each stage and the stack voltage of the rectifier.
[0017] The pretreatment unit may include an input window setting unit, used to set the input window range by reflecting the propagation time of the solution introduced into each stage of the electrodialysis equipment after ion exchange into the next stage.
[0018] The input window setting unit can control the propagation time by adjusting the flow rates of the first-stage salt, the third-stage acid, and the third-stage alkali in the control variables.
[0019] The preprocessing unit may further include a prediction window setting unit, used to set the prediction window range to a ratio of 1:1 between the prediction window range and the input window range based on the input window range.
[0020] The data conversion unit can perform data cleaning and data interpolation preprocessing on continuous segments of data within the input window range.
[0021] The prediction unit can calculate the goodness of fit of the trained artificial intelligence model, and the preprocessing unit can reset the input window range and the prediction window range based on the goodness of fit until the goodness of fit reaches or exceeds the benchmark level.
[0022] The input window setting unit can control the propagation time by adjusting the flow rate of the first-stage salt, the flow rate of the third-stage acid, and the flow rate of the third-stage alkali, and reset the input window range by reflecting the propagation time.
[0023] The prediction unit can adjust the target time interval between the target time of the prediction window range and the starting point of the input window range to be consistent with the input window range.
[0024] The artificial intelligence model may include a deep learning model for time series data processing, and the deep learning model includes the Informer model and the DLinear model.
[0025] (III) Beneficial Effects According to an embodiment of the present invention, a window-based method and apparatus for detecting anomalies in an electrodialysis device, as a preprocessing procedure specifically for operational data, can set appropriate input and prediction window sizes and ratios, and after converting continuous segment unit data, apply a time series model to predict management variables at each stage of bipolar electrodialysis (BPED). Attached Figure Description
[0026] Figure 1 This is a schematic diagram of an electrodialysis device according to an embodiment of the present invention.
[0027] Figure 2 This is a schematic diagram of one end of an electrodialysis device according to an embodiment of the present invention.
[0028] Figure 3 This is a block diagram illustrating the preprocessing steps of a window-based anomaly detection method for electrodialysis equipment according to an embodiment of the present invention.
[0029] Figures 4a to 4d This is a schematic diagram illustrating the input window and prediction window set according to an embodiment and comparative example of the present invention.
[0030] Figure 5 This is a schematic diagram illustrating an example of a window-based anomaly detection method for an electrodialysis device according to an embodiment of the present invention.
[0031] Figure 6 This is a flowchart of a window-based anomaly detection method for electrodialysis equipment according to an embodiment of the present invention.
[0032] Figure 7 This is a block diagram of a window-based electrodialysis equipment anomaly detection device according to an embodiment of the present invention.
[0033] Figures 8a to 8g This is a view showing the prediction results according to an embodiment and comparative example of the present invention.
[0034] Figure 9 This is a view used to describe a computing device according to an embodiment of the present invention. Detailed Implementation
[0035] In the following description, embodiments of the present invention will be described in detail with reference to the accompanying drawings to facilitate implementation by those skilled in the art. However, the present invention can be implemented in many different ways and is not limited to the described embodiments. Furthermore, for the sake of clear description in the drawings, parts irrelevant to the description have been omitted, and similar reference numerals have been used for similar parts throughout the specification.
[0036] Throughout the specification and claims, when a portion is described as "comprising" a particular element, unless specifically stated otherwise, it indicates that other elements may be further included, not that other elements are excluded. Ordinal terms such as "first," "second," etc., may be used to describe various elements, but the elements are not limited by these terms. These terms are used only to distinguish one element from others.
[0037] The terms “...unit”, “...device”, “module”, etc. used in this specification refer to a unit capable of performing at least one function or action described in this specification, which can be implemented by hardware or circuits, software, or a combination of hardware or circuits and software.
[0038] Various embodiments of the present invention will now be described with reference to the accompanying drawings.
[0039] Figure 1 This is a schematic diagram of an electrodialysis device according to an embodiment of the present invention. Figure 2 This is a schematic diagram of one end of an electrodialysis device according to an embodiment of the present invention.
[0040] exist Figure 1 and Figure 2 In this context, the electrodialysis equipment can be a bipolar electrodialysis (BPED) device. That is, the electrodialysis equipment can be a bipolar membrane electrodialysis device. An electrodialysis device (BPED) can be a device that converts an aqueous solution of lithium sulfate into lithium hydroxide and sulfuric acid. Here, lithium sulfate is Li₂SO₄, lithium hydroxide is LiOH, and sulfuric acid is H₂SO₄. An electrodialysis device (BPED) can also be an aqueous solution treatment device that simultaneously performs water splitting / ion separation using an electrodialysis membrane within an electric field.
[0041] Reference Figure 1 and Figure 2 Electrodialysis equipment (BPED) may include cation exchange membranes (CEM), anion exchange membranes (AEM), and bipolar membranes (BPM).
[0042] Cation exchange membranes (CEMs) have internal anionic groups, thus allowing only cations (e.g., Li) to pass through. + Anion exchange membranes (AEMs) allow only anions (e.g., SO42-) to pass through their internal cation groups. 2- The bipolar membrane (BPM) is composed of a cation exchange membrane and an anion exchange membrane, with a water splitting catalyst sandwiched between them. The BPM can split water within an electric field, thereby generating hydrogen ions (H+). + ), hydroxide ions (OH-) - ).
[0043] In other words, a bio-dilation device (BPED) can simultaneously perform water splitting (decomposition into H+) using an electrodialysis membrane (cation exchange membrane, anion exchange membrane, bipolar membrane) within an electric field. + OH - ) / Ion separation (Li + SO4 2- Aqueous solution treatment equipment for ion separation.
[0044] exist Figure 1 In a BPED electrodialysis unit, through a three-stage process (Press) comprising stages 1 to 3, the lithium sulfate (LS) solution can transfer Li and SO4 ions to the lithium hydroxide (LH) solution and the sulfuric acid (H2SO4) solution. For example, in a BPED unit, deionized water (DI water) can be contacted with the LS solution in a counter-flow manner and converted into LH and sulfuric acid solutions. Here, the LS solution is lithium sulfate (Li2SO4), and the LH solution is lithium hydroxide (LiOH).
[0045] Levels 1 through 3 can respectively include a salt chamber, an acid chamber, an alkali chamber, a salt tank, an acid tank, and an alkali tank.
[0046] In each stage, the salt chamber can be supplied with lithium sulfate (Li₂SO₄) to produce desalted water after the reaction. The acid chamber can be supplied with deionized water (DI water) to produce sulfuric acid (H₂SO₄) after the reaction. The alkali chamber can be supplied with deionized water (DI water) to produce lithium hydroxide (LiOH) after the reaction.
[0047] Deionized water produced is stored in a salt tank. Sulfuric acid produced is stored in an acid tank. Lithium hydroxide produced is stored in an alkali tank.
[0048] The production capacity of a BPED (Biodialysis Electrodialysis) unit depends on the discharge flow rates of sulfuric acid and lithium hydroxide. Production capacity is also dependent on the control of the BPED's water (H₂O) inflow rate, lithium sulfate inflow rate, circulation flow rate, circulation pressure, rectifier current and voltage, and the management of pH and conductivity in each chamber. The solution inflow rate, rectifier current, voltage, pH, conductivity, circulation flow rate, and circulation pressure—corresponding to the controlled and managed variables—can be detected by sensors installed in each chamber. The controlled and managed variables can be operational data affecting the concentrations of lithium hydroxide and sulfuric acid produced.
[0049] Figure 3 This is a block diagram illustrating the preprocessing steps of a window-based electrodialysis equipment anomaly detection method according to an embodiment of the present invention. The window-based electrodialysis equipment anomaly detection method can be executed by a window-based electrodialysis equipment anomaly detection device.
[0050] exist Figure 3 In the process, the window-based method for detecting anomalies in electrodialysis equipment may include a preprocessing step 310, a data conversion step 320, and a training, evaluation, and prediction step 330 based on an artificial intelligence model.
[0051] In the preprocessing step 310, the window-based electrodialysis equipment anomaly detection device can select control variables, select management variables, and set the input window range and prediction window range.
[0052] The window-based anomaly detection device for electrodialysis equipment allows for the selection of control and management variables for optimal automated operation of the equipment. In other words, the window-based anomaly detection device can select control and management variables comprised of the main factors affecting the optimal automated operation of the electrodialysis equipment.
[0053] For example, control variables may include the inflow rate, circulation rate, and rectifier voltage of each stage of the electrodialysis unit. Managed variables may change in response to changes in the control variables. Managed variables may include the conductivity, pH, and substack voltage of each stage of the electrodialysis unit.
[0054] Window-based anomaly detection devices for electrodialysis equipment monitor outliers by analyzing control and management variables, including time-series data. When analyzing time-series data, these devices can use appropriate window preprocessing to rationally select the range or size of the input and prediction windows.
[0055] An input window can be a subset of control variables that are continuous over time. An input window can be defined as a unit of control variables that are input into the AI model for processing at once. The input window range can refer to the size of one unit of the control variable. For example, the input window range can be defined as a specific time range from 1 hour to 24 hours in a time series.
[0056] A forecast window can be a subset of continuous managed variables. A forecast window can correspond to a managed variable unit that is linked to a control variable unit. The forecast window range can also be defined as a specific time range from 1 hour to 24 hours. The forecast window range can be determined based on the input window range.
[0057] In data transformation step 320, the window-based electrodialysis equipment anomaly detection device can clean the continuous segment data within the set input window range. The continuous segment data can be a collection of continuous time series data within the input window range. Data cleaning techniques such as removing missing and outlier values and scaling can be used for cleaning the continuous segment data.
[0058] A window-based anomaly detection device for electrodialysis equipment can perform interpolation processing on cleaned continuous segment data. Subsequently, this device can convert the preprocessed time-series data into continuous segment unit data within the input window range. Continuous segment unit data transformation refers to the process of dividing continuous time-series data into segments and converting them into new categories or values based on the corresponding segments. This transformation can be used to simplify data or modify the data structure to meet specific analytical purposes.
[0059] The continuous segment unit can correspond to the input window range. In other words, the continuous segment unit data can include control variable related data within the continuous segment corresponding to the input window range.
[0060] The window-based anomaly detection device for electrodialysis equipment allows for setting the input time interval for an artificial intelligence model. The model input time interval represents the time interval between consecutive data points provided to the model. For example, if daily measurements are used, the input time interval could be one day.
[0061] In step 330 of the training, evaluation, and prediction based on the artificial intelligence model, the window-based electrodialysis equipment anomaly detection device can train the artificial intelligence model based on continuous segment unit data. The window-based electrodialysis equipment anomaly detection device evaluates the trained artificial intelligence model. The window-based electrodialysis equipment anomaly detection device can use the trained artificial intelligence model to predict management variables within the prediction window range. The window-based electrodialysis equipment anomaly detection device can compare the predicted management variables with the actual management variables to determine the goodness of fit of the prediction.
[0062] The window-based electrodialysis equipment anomaly detection device can compare predicted management variables with preset benchmark values to detect whether BPED is abnormal.
[0063] Figures 4a to 4d This is a schematic diagram illustrating the input window and prediction window set according to an embodiment and comparative example of the present invention.
[0064] The range occupied by the input window on the input variable (X) axis is the input window range. The range occupied by the prediction window on the target variable (Y) axis is the prediction window range.
[0065] exist Figures 4a to 4dIn this context, the target time is the starting point of the prediction window. In other words, the target time is the baseline time for the prediction range determined based on the prediction window range. The AI model predicts the managed variable within the prediction range set after the target time. The target time interval is the interval between the start time of the input window and the start time of the prediction window. The target time interval can be the interval between the start point of the input window and the target time.
[0066] Artificial intelligence models use input variables (X) within an input window range to predict target variables (Y) within a prediction window range. When the target time interval matches the input window range, there are no gaps in the prediction of time series data. The input variables are control variables, while the target variable can be a management variable.
[0067] Figure 4a An input window and a prediction window are shown according to one embodiment.
[0068] exist Figure 4a In this model, the size or interval of the input window range and the prediction window range can be the same. That is, the ratio of the input window range to the prediction window range can be 1:1. Within the prediction window range, the target variable (Y) can be predicted using an artificial intelligence model trained with the input variables (X) within the input window range. The time interval between the input window range and the target variable can be the same.
[0069] Figure 4b The input window and prediction window are shown according to the first comparison example. The first comparison example is the case where the size of the prediction window range is larger than the size of the input window range.
[0070] exist Figure 4b In this approach, the input window range and the target time interval can be the same. That is, there is no gap between the target time and the input window range. Therefore, time series data can be predicted in real time.
[0071] However, the size of the input window and the prediction window can differ. For example, the ratio of the input window to the prediction window can be 1:2. When the input window and the prediction window are not in a 1:1 ratio, the prediction fit or predictive power may decrease. For example, if the input window is 2 hours and the prediction window is 4 hours, and if predictions are made based on windows trained in 2 hours for 2 hours (assuming a 90% fit), then predictions are made again for 2 hours, the errors will multiply, leading to a decrease in the fit (90% x 90% = 81%).
[0072] Figure 4cThe input window and prediction window are shown according to the second comparison example. The second comparison example is the case where the size of the prediction window range is smaller than the size of the input window range.
[0073] exist Figure 4c In this scenario, the ratio of the input window range to the prediction window range can be 2:1. If the ratio is not 1:1, the prediction fit or predictive power may decrease. For example, if the input window range is 2 hours and the prediction window range is 1 hour, overfitting may occur, leading to reduced predictive power.
[0074] Figure 4d The input window and prediction window are shown according to the third comparison example. The third comparison example is a case where there is a gap between the target time and the input window range.
[0075] exist Figure 4d In this model, the size or interval of the input window range and the prediction window range can be the same. That is, the ratio of the input window range to the prediction window range can be 1:1. Within the prediction window range, the target variable (Y) can be predicted using an artificial intelligence model trained with the input variables (X) within the input window range.
[0076] At this point, there is a gap between the target time and the input window range. In other words, there is a certain time gap in time series data prediction. This gap hinders real-time prediction. Here, the training window can refer to the set of input variables (Y) trained within the input window range.
[0077] Figure 5 This is a schematic diagram illustrating an example of predicting management variables based on the steps of a window-based method for detecting anomalies in electrodialysis equipment. Figure 5 The input window and prediction window are shown according to one embodiment.
[0078] exist Figure 5 In the window-based electrodialysis equipment anomaly detection device, during the preprocessing step 510 of the time series data used to train the artificial intelligence model, the input window range and the prediction window range can both be set to the same 2-hour period. When setting the input window range, the window-based electrodialysis equipment anomaly detection device can reflect the propagation time of the solution entering each stage of the electrodialysis equipment through ion exchange to the next stage. The window-based electrodialysis equipment anomaly detection device can set a prediction window range consistent with the thus-set input window range. For example, the window-based electrodialysis equipment anomaly detection device can set a 2-hour input window range from 0 to 2 for the control variables X1 to Xn.
[0079] A window-based anomaly detection device for electrodialysis equipment can convert control variables (X) within an input window range into continuous interval unit data. The continuous interval unit can be a 2-hour unit corresponding to the input window range.
[0080] In step 520 of the training process for the artificial intelligence model in the window-based electrodialysis equipment anomaly detection device, a deep learning model for time series processing can be trained based on continuous segment unit data. The deep learning model for time series processing may include an Informer model and a DLinear model.
[0081] The Informer model is one of the deep learning-based time series data prediction models, primarily aiming to provide high-performance predictions for long-series time series data. The DLinear model combines time series decomposition with linear layers. The DLinear model first generates a moving average and removes it, then decomposes it into trend and periodic data for separate training. Then, the DLinear model applies a single linear layer to each component for training and sums the results to calculate the final prediction.
[0082] In the prediction step 530 of the management variable prediction device, the window-based electrodialysis equipment anomaly detection device can predict the management variable (Y) within the prediction window range using a trained time series processing deep learning model. For example, the time series processing deep learning model can predict the management variable (Y) within a prediction window range set to 2 hours. The window-based electrodialysis equipment anomaly detection device can predict the management variable (Y) within the prediction window range between 2 and 4 points. That is, the target time can be 2 points, and the target time interval can be 2 hours. Therefore, in one embodiment, the target time interval can be consistent with the input window range. In other words, the window-based electrodialysis equipment anomaly detection device can predict the management variable in real time when there is no gap between the input window and the prediction window.
[0083] According to one embodiment, the window-based anomaly detection device for electrodialysis equipment can set the ratio of the input window range to the prediction window range to 1:1. Furthermore, the window-based anomaly detection device can make the target time interval consistent with the input window range so that there is no gap between the input window and the prediction window. Therefore, the window-based anomaly detection device for electrodialysis equipment can improve the prediction fit by making real-time predictions of the managed variables.
[0084] Window-based anomaly detection devices for electrodialysis equipment can compare predicted management variables with actual management variables to calculate the goodness of fit, thereby improving the prediction goodness of fit. If the calculated goodness of fit is not higher than a specific benchmark, the input window range and prediction window range can be reset.
[0085] When resetting the input and prediction window ranges, the window-based anomaly detection device for electrodialysis equipment can control the propagation time by adjusting the flow rates of the first-stage salt, third-stage acid, and third-stage base in the control variables. The device can then reset the input window range based on the propagation time and, consequently, the prediction window range.
[0086] Figure 6 This is a flowchart of a window-based anomaly detection method for electrodialysis equipment according to an embodiment of the present invention. Figure 6 The window-based method for detecting anomalies in electrodialysis equipment can be executed by a window-based electrodialysis equipment anomaly detection device.
[0087] exist Figure 6 In this embodiment, the window-based electrodialysis equipment anomaly detection device can set the input window range by reflecting the propagation time of the solution entering the next stage of the electrodialysis equipment after ion exchange (step S100). In one embodiment, the window-based electrodialysis equipment anomaly detection device can control the propagation time by adjusting the flow rates of the first-stage salt, the third-stage acid, and the third-stage base in the control variables. Since the input window range is determined by reflecting the propagation time, the input window range will change when the propagation time changes.
[0088] The window-based electrodialysis equipment anomaly detection device can set the prediction window range for the management variable based on the set input window range (step S200). The window-based electrodialysis equipment anomaly detection device can provide feedback on the window size and ratio based on the goodness of fit of the prediction window.
[0089] For example, a window-based anomaly detection device for electrodialysis equipment can set the prediction window range to a 1:1 ratio between the input window range and the prediction window range. That is, if the input window range is 2 hours, the window-based anomaly detection device for electrodialysis equipment can also set the prediction window range to 2 hours.
[0090] Even when the ratio of the input window range to the prediction window range is 1:1, the window size cannot be trained if it is too large. Therefore, the window size can be reasonably determined based on the goodness of fit of the prediction window.
[0091] Since the prediction window range is determined based on the input window range, it will ultimately change over time. A window-based anomaly detection device for electrodialysis equipment can eliminate the gap in the target time interval between the input window range and the prediction window range. In other words, a window-based anomaly detection device for electrodialysis equipment can ensure that the target time interval, which serves as the interval between the starting point of the input window and the starting point of the prediction window, is consistent with the input window range.
[0092] The window-based anomaly detection device for electrodialysis equipment can convert time-series data of control variables into continuous segment unit data corresponding to continuous segments within the input window range (step S300). Continuous segment unit data is defined as data corresponding to continuous segment units within the input window range. Continuous segment unit data can be obtained by preprocessing and processing the control variables within the input window range.
[0093] The window-based anomaly detection device for electrodialysis equipment can train an artificial intelligence model based on continuous segment unit data, and predict management variables based on control variables using the trained artificial intelligence model (step S400). The artificial intelligence model may include a deep learning model for time series processing. The window-based anomaly detection device for electrodialysis equipment can make the target time, which is the starting point of the prediction window range, coincide with the ending point of the input window. That is, the window-based anomaly detection device for electrodialysis equipment can predict management variables within the prediction window range in real time based on control variables within the input window range. In other words, it can eliminate the gap time between the input window range and the prediction window range.
[0094] Furthermore, the window-based electrodialysis equipment anomaly detection device can compare predicted management variables with preset benchmark values to determine whether the electrodialysis equipment is abnormal (step S400). For example, each stage of the rectifier consists of three sub-stacks. The stack voltage, which is the sub-stack voltage, is a management variable. The limit voltage of the stack voltage can be 90V. For the window-based electrodialysis equipment anomaly detection device, if any of the multiple stack voltages is predicted to be above 90V, an anomaly in the electrodialysis equipment can be detected.
[0095] Figure 7 This is a block diagram of a window-based electrodialysis equipment anomaly detection device according to an embodiment of the present invention.
[0096] Reference Figure 7The window-based electrodialysis equipment anomaly detection device 100 may include a preprocessing unit (110, 120), a data conversion unit 130, a prediction unit 140, and an anomaly detection unit 150. The preprocessing unit may include an input window setting unit 110 and a prediction window setting unit 120.
[0097] The preprocessing unit can preprocess the control variables that serve as input variables for deep learning models used in time series processing. Preprocessing may include data cleaning, data interpolation, and scaling. The preprocessing unit can set the input window range for the control variables and the prediction window range for the management variables.
[0098] The input window setting unit 110 can set the input window range by reflecting the propagation time of the solution entering the next stage after ion exchange in each stage of the electrodialysis equipment. The input window setting unit 110 can control the propagation time by adjusting the flow rates of the first-stage salt, the third-stage acid, and the third-stage alkali in the control variables.
[0099] The prediction window setting unit 120 can set the prediction window range based on the input window range. The prediction window setting unit 120 can set the prediction window range to a ratio of 1:1 to the input window range.
[0100] The data conversion unit 130 can convert the time series data of the control variables into continuous segment unit data corresponding to the input window range. The data conversion unit 130 can perform data cleaning and data interpolation preprocessing on the continuous segment data within the input window range. After preprocessing, the continuous segment data can be converted back into continuous segment unit data.
[0101] The prediction unit 140 can predict the management variable corresponding to the control variable using an artificial intelligence model trained on continuous segment unit data. The prediction unit 140 can calculate the goodness of fit of the trained artificial intelligence model. Based on the goodness of fit calculated by the prediction unit 140, the preprocessing units (110, 120) can reset the input window range and the prediction window range until the goodness of fit reaches above the baseline level. The preprocessing units (110, 120) can reset the input window range by controlling the propagation time, and then reset the prediction window range based on this.
[0102] The prediction unit 140 can adjust the target time interval between the target time of the prediction window range, including the management variables, and the starting point of the input window range through an artificial intelligence model, making it consistent with the input window range. In other words, it can eliminate the gap time between the input window and the prediction window and predict the management variables in real time.
[0103] The anomaly detection unit 150 can compare predicted management variables with reference values to detect whether the electrodialysis equipment is malfunctioning. For example, the reference value may include 90V, which is the limit voltage for stack voltage monitoring.
[0104] Figures 8a to 8g This is a view showing the prediction results according to an embodiment and comparative example of the present invention.
[0105] Figure 8a This is a view showing the conductivity prediction results according to an embodiment of the present invention. Figure 8a It is a graph showing the conductivity when the ratio of the input window range to the prediction window range is set to 1:1.
[0106] Figure 8b and Figure 8c This is a view showing the stack voltage prediction results according to an embodiment of the present invention. Figure 8b and Figure 8c It is a graph showing the stack voltage when the ratio of the input window range to the prediction window range is set to 1:1.
[0107] from Figures 8a to 8c It can be seen that the actual and predicted values of conductivity and stack voltage exhibited by each solution are similar to each other.
[0108] Figure 8d and Figure 8e This is a view showing the prediction results based on the first comparison example. The first comparison example is the case where the size of the prediction window is larger than the size of the input window (training window).
[0109] from Figure 8d It can be seen that the predicted results curves representing lithium production, current efficiency, and grade 1 acid purity differ from the actual results curves.
[0110] from Figure 8e It can be seen that in the curves showing the purity (S / Li ratio) of Grade 1 Base product (LiOH), Grade 2 Base product (LiOH), and Grade 3 Base product (LiOH), the predicted and actual results differ from each other.
[0111] Figure 8f This is a view showing the prediction results based on the second comparison example. Figure 8f The first example is a graph showing the conductivity of the first-order solution. The second comparative example is the case where the prediction window is smaller than the input window (training window).
[0112] from Figure 8f It can be seen that in the various curves for the conductivity of Grade 1 salt, Grade 1 acid, and Grade 1 base, the predicted and actual values do not behave similarly. Although not shown, the predicted and actual values for the conductivity of each solution in Grade 2 and Grade 3 also differed in the second comparative example.
[0113] Figure 8g This is a view showing the prediction results based on another comparison example. Figure 8f The comparison examples show a situation where the ratio of the input window range to the prediction window range is set to 1:1, but their respective sizes are set too large. For example, Figure 8g The results show the input window size and prediction window size set to 24 hours.
[0114] Figure 8g The diagram shows that no training was actually performed in this comparative example. Training cannot proceed when the prediction window size and training time are too large.
[0115] The first curve (LS1) is the loss calculated using the evaluation data from each epoch, and the second curve (LS2) is the loss calculated using the training data from each epoch.
[0116] Since the training of the second curve (LS2) was not performed well, it can be seen that the loss value is the same as 1.06 regardless of the number of epochs.
[0117] Here, an epoch represents the number of times the entire training dataset passes through the model in one go.
[0118] Figure 9 This is a view used to describe a computing device according to an embodiment of the present invention.
[0119] Reference Figure 9 The window-based electrodialysis equipment anomaly detection method and apparatus according to the embodiments can be implemented using a computing device 900.
[0120] 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, which communicate via a bus 920. The computing device 900 may further include a network interface 970 electrically connected to a network 90. The network interface 970 can transmit or receive signals with other entities via the network 90.
[0121] The processor 910 can be implemented as various types such as MCU (Microcontroller Unit), AP (Application Processor), CPU (Central Processing Unit), GPU (Graphics Processing Unit), NPU (Neural Processing Unit), etc., and can be any semiconductor device that executes commands stored in memory 930 or storage device 960. The processor 910 can be configured to implement... Figure 1 The aforementioned functions and methods are related to Figure 8.
[0122] The memory 930 and storage device 960 may include various forms of volatile or non-volatile storage media. For example, the memory may include ROM (Read-Only Memory) 931 and RAM (Random Access Memory) 932. In this embodiment, the memory 930 may be located inside or outside the processor 910, and the memory 930 may be connected to the processor 910 by various known means.
[0123] In some embodiments, at least some components or functions of the window-based electrodialysis equipment anomaly detection method and apparatus according to various embodiments may be implemented by a program or software running in a computing device 900, and the program or software may be stored in a computer-readable medium.
[0124] In some embodiments, at least some components or functions of the window-based electrodialysis equipment anomaly detection method and apparatus according to various embodiments may be implemented by the hardware or circuitry of the computing device 900, or may be implemented by separate hardware or circuitry that can be electrically connected to the computing device 900.
[0125] Although the embodiments of the present invention have been described in detail above, the scope of the present invention is not limited to the above embodiments. Various modifications and improvements made by those skilled in the art using the basic concepts of the present invention as defined in the claims also fall within the scope of the present invention.
[0126] [Explanation of reference numerals in the attached figures] 100: Window-based anomaly detection device for electrodialysis equipment 110: Input window setting unit 120: Prediction Window Setting Unit 130: Data Conversion Unit 140: Prediction Unit 150: Anomaly Detection Unit
Claims
1. A window-based method for detecting anomalies in electrodialysis equipment, comprising: The steps for setting the input window range for control variables of the electrodialysis equipment; The step of converting the time series data of the control variable into continuous segment unit data corresponding to the input window range; The step of predicting the management variable corresponding to the control variable using an artificial intelligence model trained based on the continuous segment unit data; as well as The step of comparing the predicted management variables with the baseline values to detect whether the electrodialysis equipment is malfunctioning.
2. The window-based anomaly detection method for electrodialysis equipment according to claim 1, wherein, The control variables include the solution inflow rate, circulation flow rate, and rectifier voltage at each stage of the electrodialysis equipment, and the management variables include the conductivity at each stage and the stack voltage of the rectifier.
3. The window-based anomaly detection method for electrodialysis equipment according to claim 1, wherein, The step of setting the input window range includes: setting the input window range by reflecting the propagation time of the solution introduced into each stage of the electrodialysis equipment through ion exchange into the next stage.
4. The window-based anomaly detection method for electrodialysis equipment according to claim 3, wherein, The step of setting the input window range includes: controlling the propagation time by adjusting the flow rates of the first-stage salt, the third-stage acid, and the third-stage alkali in the control variables.
5. The window-based anomaly detection method for electrodialysis equipment according to claim 4, further comprising: The step of setting the prediction window range for the managed variable based on the set input window range. The step of setting the prediction window range includes: setting the prediction window range to a ratio of 1:1 between the prediction window range and the input window range.
6. The window-based anomaly detection method for electrodialysis equipment according to claim 5, wherein, The step of converting to continuous segment unit data includes: performing data cleaning and data interpolation preprocessing on the continuous segment data within the input window range.
7. The window-based anomaly detection method for electrodialysis equipment according to claim 6, wherein, The steps for predicting the management variables include: calculating the goodness of fit of the trained artificial intelligence model, and resetting the input window range and the prediction window range based on the calculated goodness of fit.
8. The window-based anomaly detection method for electrodialysis equipment according to claim 7, wherein, The step of resetting the input window range and the prediction window range includes the step of controlling the propagation time by adjusting the flow rate of the first-stage salt, the flow rate of the third-stage acid, and the flow rate of the third-stage alkali.
9. The window-based anomaly detection method for electrodialysis equipment according to claim 6, wherein, The steps for predicting the management variable include: aligning the target time interval between the target time of the prediction window range and the starting point of the input window range with the input window range.
10. The window-based anomaly detection method for electrodialysis equipment according to claim 1, wherein, The artificial intelligence model includes a deep learning model for time series data processing, and the deep learning model includes the Informer model and the DLinear model.
11. A window-based anomaly detection device for electrodialysis equipment, comprising: The preprocessing unit is used to set the input window range for the control variables of the electrodialysis equipment; A data conversion unit is used to convert the time series data of the control variable into continuous segment unit data corresponding to the input window range; The prediction unit is used to predict the management variable corresponding to the control variable using an artificial intelligence model trained based on the continuous segment unit data; as well as An anomaly detection unit is used to compare the predicted management variables with a baseline value to detect whether the electrodialysis equipment is abnormal.
12. The window-based electrodialysis equipment anomaly detection device according to claim 11, wherein, The control variables include the solution inflow rate, circulation flow rate, and rectifier voltage at each stage of the electrodialysis equipment, and the management variables include the conductivity at each stage and the stack voltage of the rectifier.
13. The window-based electrodialysis equipment anomaly detection device according to claim 11, wherein, The pretreatment unit includes an input window setting unit, which is used to set the input window range by reflecting the propagation time of the solution introduced into each stage of the electrodialysis equipment after ion exchange into the next stage.
14. The window-based electrodialysis equipment anomaly detection device according to claim 13, wherein, The input window setting unit controls the propagation time by adjusting the flow rates of the first-stage salt, the third-stage acid, and the third-stage alkali in the control variables.
15. The window-based electrodialysis equipment anomaly detection device according to claim 14, wherein, The preprocessing unit further includes a prediction window setting unit, used to set the prediction window range to a ratio of 1:1 between the prediction window range and the input window range based on the input window range.
16. The window-based electrodialysis equipment anomaly detection device according to claim 15, wherein, The data conversion unit performs data cleaning and data interpolation preprocessing on continuous segments of data within the input window range.
17. The window-based electrodialysis equipment anomaly detection device according to claim 16, wherein, The prediction unit calculates the fit of the trained artificial intelligence model. The preprocessing unit resets the input window range and the prediction window range based on the fit degree until the fit degree reaches or exceeds the baseline level.
18. The window-based electrodialysis equipment anomaly detection device according to claim 17, wherein, The input window setting unit controls the propagation time by adjusting the flow rate of the first-stage salt, the flow rate of the third-stage acid, and the flow rate of the third-stage alkali, and resets the input window range by reflecting the propagation time.
19. The window-based electrodialysis equipment anomaly detection device according to claim 17, wherein, The prediction unit adjusts the target time interval between the target time of the prediction window range and the starting point of the input window range to be consistent with the input window range.
20. The window-based electrodialysis equipment anomaly detection device according to claim 19, wherein, The artificial intelligence model includes a deep learning model for time series data processing, and the deep learning model includes the Informer model and the DLinear model.