CN-122020064-A - High-purity lithium salt production process on-line monitoring system

CN122020064ACN 122020064 ACN122020064 ACN 122020064ACN-122020064-A

Abstract

The embodiment of the invention provides an online monitoring system for a high-purity lithium salt production process, and relates to the technical field of online monitoring technology for lithium salt production. The system comprises a data acquisition module, a quality prediction module and an alarm generation module, wherein the data acquisition module is used for acquiring process variables in real time and constructing a time sequence feature matrix based on the process variables, the quality prediction module is used for inputting the time sequence feature matrix into a pre-trained deep time sequence prediction model to obtain a predicted value of a target product quality index, and the alarm generation module is used for comparing the predicted value with a preset quality specification and generating and outputting alarm information when the predicted value meets an early warning condition. The invention solves the problem of low monitoring precision in lithium salt production, and further achieves the effect of improving the monitoring precision of lithium salt production quality.

Inventors

HU JIAN
LIU LIN
FANG CAIBAO
LIU ZHENGWEI
ZHAN JIANG
LIU XINGHUA
ZHU DEXIANG
SUN JIN
CHENG SAI

Assignees

浙江司太立制药股份有限公司
浙江健立化学有限公司

Dates

Publication Date: 20260512
Application Date: 20260408

Claims (10)

1. An on-line monitoring system for a high purity lithium salt production process, comprising: The data acquisition module is used for acquiring process variables in real time and constructing a time sequence feature matrix based on the process variables; The quality prediction module is used for inputting the time sequence feature matrix into a pre-trained depth time sequence prediction model so as to obtain a predicted value of a target product quality index; and the alarm generation module is used for comparing the predicted value with a preset quality specification, and generating and outputting alarm information when the predicted value meets the early warning condition.
2. The system of claim 1, wherein the real-time acquisition of process variables comprises: Performing a health assessment on the real-time data of the process variable to generate a health tag; when the health tag determines to be abnormal, a sensor maintenance alarm is triggered.
3. The system of claim 2, wherein the performing a health assessment on real-time data of the process variable comprises: performing flatness verification on the real-time data, wherein the health degree tag is judged to be abnormal when the numerical variation of the real-time data in a preset verification time is smaller than a preset flatness threshold value; And/or the number of the groups of groups, And performing change rate verification on the real-time data, wherein the health degree label is judged to be abnormal when the instantaneous change rate of the real-time data is larger than a preset change rate threshold value.
4. The system of claim 1, wherein after generating and outputting the alert information, the steps of: acquiring an actual assay value of the target product quality index; Calculating a multi-dimensional prediction residual vector based on the predicted value and the actual assay value; normalizing the multidimensional prediction residual vector according to an inverse matrix of the historical residual covariance matrix; performing exponential weighted moving average calculation on a first distance sequence within a preset time period to obtain smooth control statistics, wherein the first distance is obtained through a normalization process; and when the smooth control statistic is larger than the preset mismatch control limit, judging that the model mismatch condition is met, and triggering a model updating instruction.
5. An on-line monitoring method for a high-purity lithium salt production process is characterized by comprising the following steps: Collecting process variables in real time; constructing a time sequence feature matrix based on the process variable; Inputting the time sequence feature matrix into a pre-trained depth time sequence prediction model to obtain a predicted value of a target product quality index; And comparing the predicted value with a preset quality specification, and generating and outputting alarm information when the predicted value meets the early warning condition.
6. The method of claim 5, wherein the real-time acquisition of process variables comprises: Performing a health assessment on the real-time data of the process variable to generate a health tag; when the health tag determines to be abnormal, a sensor maintenance alarm is triggered.
7. The method of claim 6, wherein the performing a health assessment on real-time data of the process variable comprises: performing flatness verification on the real-time data, wherein the health degree tag is judged to be abnormal when the numerical variation of the real-time data in a preset verification time is smaller than a preset flatness threshold value; And/or the number of the groups of groups, And performing change rate verification on the real-time data, wherein the health degree label is judged to be abnormal when the instantaneous change rate of the real-time data is larger than a preset change rate threshold value.
8. The method of claim 5, wherein after the generating and outputting the alert information, the method further comprises: acquiring an actual assay value of the target product quality index; Calculating a multi-dimensional prediction residual vector based on the predicted value and the actual assay value; normalizing the multidimensional prediction residual vector according to an inverse matrix of the historical residual covariance matrix; performing exponential weighted moving average calculation on a first distance sequence within a preset time period to obtain smooth control statistics, wherein the first distance is obtained through a normalization process; and when the smooth control statistic is larger than the preset mismatch control limit, judging that the model mismatch condition is met, and triggering a model updating instruction.
9. A computer readable storage medium, characterized in that the computer readable storage medium has stored therein a computer program, wherein the computer program is arranged to execute the method of any of the claims 5 to 8 when run.
10. An electronic device comprising a memory and a processor, characterized in that the memory has stored therein a computer program, the processor being arranged to run the computer program to perform the method of any of the claims 5 to 8.

Description

High-purity lithium salt production process on-line monitoring system Technical Field The embodiment of the invention relates to the technical field of online monitoring of lithium salt production, in particular to an online monitoring system for a high-purity lithium salt production process. Background The levels of trace impurities (particularly moisture content and free metal ion concentration) within high purity lithium hexafluorophosphate constitute key limiting factors that limit the electrochemical window, cycle life, and critical temperature for thermal runaway of the battery. In current industrial practice, verification of lithium hexafluorophosphate product quality relies primarily on off-line detection in a central laboratory, including chemical titration and mass spectrometry, a mode with sampling and detection delays of several hours. The information feedback causes hysteresis due to the continuity of crystallization and drying processes, and when an analytical instrument detects that the quality is over-limit, a large amount of irreversible inferior products are produced and packaged on the production line, so that quality monitoring cannot be realized in a complex nonlinear coupling chemical scene. Disclosure of Invention The embodiment of the invention provides an online monitoring system for a high-purity lithium salt production process, which at least solves the problem of low quality monitoring precision of lithium salt in the related technology. According to one embodiment of the present invention, there is provided an on-line monitoring system for a high purity lithium salt production process, comprising: The data acquisition module is used for acquiring process variables in real time and constructing a time sequence feature matrix based on the process variables; The quality prediction module is used for inputting the time sequence feature matrix into a pre-trained depth time sequence prediction model so as to obtain a predicted value of a target product quality index; and the alarm generation module is used for comparing the predicted value with a preset quality specification, and generating and outputting alarm information when the predicted value meets the early warning condition. In one exemplary embodiment, the real-time acquisition process variable includes: Performing a health assessment on the real-time data of the process variable to generate a health tag; when the health tag determines to be abnormal, a sensor maintenance alarm is triggered. In one exemplary embodiment, the performing a health assessment on real-time data of the process variable includes: performing flatness verification on the real-time data, wherein the health degree tag is judged to be abnormal when the numerical variation of the real-time data in a preset verification time is smaller than a preset flatness threshold value; And/or the number of the groups of groups, And performing change rate verification on the real-time data, wherein the health degree label is judged to be abnormal when the instantaneous change rate of the real-time data is larger than a preset change rate threshold value. In one exemplary embodiment, after generating and outputting the alert information, the following steps are also performed: acquiring an actual assay value of the target product quality index; Calculating a multi-dimensional prediction residual vector based on the predicted value and the actual assay value; normalizing the multidimensional prediction residual vector according to an inverse matrix of the historical residual covariance matrix; performing exponential weighted moving average calculation on a first distance sequence within a preset time period to obtain smooth control statistics, wherein the first distance is obtained through a normalization process; and when the smooth control statistic is larger than the preset mismatch control limit, judging that the model mismatch condition is met, and triggering a model updating instruction. According to another embodiment of the present invention, there is provided an on-line monitoring method for a high purity lithium salt production process, including: Collecting process variables in real time; constructing a time sequence feature matrix based on the process variable; Inputting the time sequence feature matrix into a pre-trained depth time sequence prediction model to obtain a predicted value of a target product quality index; And comparing the predicted value with a preset quality specification, and generating and outputting alarm information when the predicted value meets the early warning condition. In one exemplary embodiment, the real-time acquisition process variable includes: Performing a health assessment on the real-time data of the process variable to generate a health tag; when the health tag determines to be abnormal, a sensor maintenance alarm is triggered. In one exemplary embodiment, the performing a health assessment on real-time data of the process v