A method for monitoring waste gas from lithium carbonate production

Abstract

The present invention relates to the technical field of data processing, and more specifically, to a method for monitoring waste gas from lithium carbonate production. The method comprises: obtaining k nearest neighbors and reverse k nearest neighbors of each data point in a waste gas data set, recording different data points in the k nearest neighbors and reverse k nearest neighbors as difference points of the data points, and determining density features of the data points, wherein the density features are used to describe the distribution of the difference points in the data set; calculating the influence of the difference points on the data points; determining the actual influence of the difference points on the data points through the variance of the data points in the k nearest neighbors of the difference points and the Frobenius norm of the k nearest neighbor correlation matrix; recording the accumulated and normalized values ​​of the actual influence of all the difference points on the data points as the importance of the data points; and determining the abnormal scores of the data points according to the importance of the data points in an INFLO model to realize the monitoring of waste gas from lithium carbonate production, thereby effectively improving the accuracy of obtaining the monitoring results of waste gas from lithium carbonate production.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and more specifically, to a method for monitoring waste gas in lithium carbonate production. Background Art

[0002] Lithium carbonate is mainly used in battery manufacturing (such as lithium-ion batteries), glass and ceramic industries, and other chemical applications. Its production process usually releases a variety of waste gases, such as carbon dioxide and volatile organic compounds. The emission of waste gases is harmful to the environment and health, and also involves safety issues in industrial production. Therefore, the development of effective waste gas monitoring methods can help companies monitor the emission of harmful gases in real time, detect and deal with excessive emissions in a timely manner, thereby effectively reducing environmental pollution and protecting air quality.

[0003] In the prior art, the monitoring of waste gas emissions is mostly focused on the calculation of waste gas content. For example, the patent application document with publication number CN110866856A discloses a greenhouse gas emission monitoring system and method for a steel enterprise. The application includes a real-time online monitoring module for carbon dioxide and other greenhouse gases, a carbon emission monitoring module for production processes and carbon fixation products, an energy consumption conversion carbon dioxide emission calculation module, and a comprehensive carbon dioxide emission monitoring platform. The comprehensive carbon dioxide emission monitoring platform receives data transmitted by the real-time online monitoring module for carbon dioxide, other greenhouse gas monitoring modules, and the energy consumption conversion carbon dioxide emission calculation module, and calculates the real-time comprehensive carbon dioxide emissions of the steel enterprise based on the data.

[0004] Although the existing technology can accurately obtain the waste gas content to a certain extent, there is little research on how to monitor the abnormal and excessive waste gas data in the huge amount of waste gas data. Based on this, how to accurately realize the waste gas monitoring of lithium carbonate production is an urgent problem to be solved by technicians in this field. Summary of the invention

[0005] In order to solve the above technical problem of how to accurately realize the waste gas monitoring of lithium carbonate production, the present invention proposes a waste gas monitoring method for lithium carbonate production, which comprises the following steps:

[0006] The k nearest neighbors and reverse k nearest neighbors of each data point in the exhaust gas data set are obtained, the different data points in the k nearest neighbors and reverse k nearest neighbors are recorded as the difference points of the data points, and the density characteristics of the data points are determined, and the density characteristics are used to describe the distribution of the difference points in the data set; the influence of the difference points on the data points is calculated: ;

[0007] In the formula, Indicates the influence of the j-th difference point on the i-th data point, represents the density feature of the j-th difference point, represents the density feature of the i-th data point, represents the distance between the jth difference point and the i-th data point, represents the exponential function with base e, represents a linear normalization function; a first ratio is obtained after normalizing the ratio of the variance of the data point in the k nearest neighbors of the difference point to the Frobenius norm of the k nearest neighbors correlation matrix, and the absolute value of the difference between the first ratio and the influence degree is used as the actual influence degree of the difference point on the data point; the accumulation and normalized value of the actual influence degree of all the difference points of the data point on the data point is recorded as the importance of the data point; in the INFLO model, the abnormal score of the data point is determined according to the importance of the data point to realize the monitoring of lithium carbonate production waste gas.

[0008] The present invention monitors abnormal and excessive waste gas data in a huge amount of waste gas data through an anomaly detection algorithm. In this process, the present invention takes into account that not all data points require the participation of reverse k-nearest neighbors when determining the anomaly score. Therefore, the influence of the abnormal point in the k-nearest neighbor and reverse k-nearest neighbor of the data point on the data point is obtained, thereby determining the influence space of the data point; on this basis, the present invention also takes into account that noise data will affect the accuracy of the influence degree. Therefore, the influence degree is corrected by further obtaining the deviation degree and correlation between the abnormal point and the data point in its k-nearest neighbor, thereby effectively improving the accuracy of determining the abnormal score of the data point, thereby effectively improving the accuracy of obtaining the monitoring results of lithium carbonate production waste gas.

[0009] According to a method for monitoring waste gas from lithium carbonate production provided by the present invention, the k nearest neighbors and reverse k nearest neighbors of each data point in the waste gas data set are obtained, including: deploying gas sensors to collect waste gas data generated in the production of lithium carbonate, and obtaining each data point in the waste gas data set after preprocessing the waste gas data; presetting the number k of neighboring data points of each data point, and taking the k data points with the closest Euclidean distance to the data point as the k nearest neighbors of the data point; obtaining the data point with the data point as the k nearest neighbor, and all the data points with the data point as the k nearest neighbors constitute the reverse k nearest neighbors of the data point.

[0010] According to a lithium carbonate production waste gas monitoring method provided by the present invention, determining the density characteristics of data points includes: obtaining the ratio of the cumulative sum of distances between the k nearest neighbors of the data point and the data point to the number of data points in the k nearest neighbors, which is recorded as the density characteristics of the data point.

[0011] According to a lithium carbonate production waste gas monitoring method provided by the present invention, the method for obtaining the k-nearest neighbor correlation matrix includes: obtaining the correlation between the difference point and each data point in the k-nearest neighbors of the difference point through a DTW algorithm; and recording the matrix composed of the correlation between the difference point and each data point as the k-nearest neighbor correlation matrix of the difference point.

[0012] The present invention takes into account that noise data will affect the accuracy of the data point influence space. Therefore, by obtaining the degree of deviation and correlation between the abnormal point and the data points in its k nearest neighbors and correcting the degree of influence, the accuracy of the data point abnormality score can be effectively improved.

[0013] According to a lithium carbonate production waste gas monitoring method provided by the present invention, the abnormal score of the data point is determined according to the importance of the data point in the INFLO model, including: presetting a threshold; if the importance of the data point is greater than the preset threshold, the k nearest neighbors of the data point are used as the influence space of the data point; if the importance of the data point is not greater than the preset threshold, the k nearest neighbors and the reverse k nearest neighbors of the data point are used as the influence space of the data point; and the influence space of the data point is used in the INFLO model to obtain the abnormal score of the data point.

[0014] The present invention takes into account that not all data points require the participation of reverse k-nearest neighbors when determining anomaly scores. Therefore, the importance of the data point is obtained by calculating the degree of influence of the anomaly point on the data point, and whether the data point needs reverse k-nearest neighbors is determined based on its importance, thereby effectively improving the accuracy of the anomaly score of the data point.

[0015] According to a lithium carbonate production waste gas monitoring method provided by the present invention, the influence space of the data point is used in the INFLO model to obtain the anomaly score of the data point, including: recording the inverse of the distance between the data point and the nearest data point in its influence space as the local density of the data point; recording the ratio of the mean of the local density of all data points in the influence space of the data point to the local density of the data point as the anomaly score of the data point.

[0016] According to a method for monitoring waste gas from lithium carbonate production provided by the present invention, the abnormality score of the data point is determined according to the importance of the data point in the INFLO model to realize the monitoring of waste gas from lithium carbonate production, including: setting an abnormality threshold; if the abnormality score of the data point is greater than the abnormality threshold, the data point is an abnormal data point; if the abnormality score of the data point is not greater than the abnormality threshold, the data point is a normal data point.

[0017] The present invention takes into account that the determination of abnormal exhaust gas needs to be specifically set according to different situations, so by presetting thresholds, abnormal data in various situations can be accurately identified.

[0018] According to a method for monitoring waste gas from lithium carbonate production provided by the present invention, the monitoring of waste gas from lithium carbonate production is achieved, and then the method further includes: marking the abnormal data points and issuing an abnormal warning.

[0019] The present invention takes into account that abnormally excessive exhaust gas may have an impact on the human body and the environment. Therefore, after detecting abnormally excessive exhaust gas data, an early warning is issued in time to remind staff to carry out repairs.

[0020] The present invention has the following beneficial effects:

[0021] Based on the above technical solution, when monitoring waste gas produced by lithium carbonate, the present invention determines the actual required influence space of the data point by obtaining the influence degree of the abnormal points in the k-nearest neighbors and reverse k-nearest neighbors of the data point on the data point, so that the abnormal and excessive waste gas data in the waste gas data with huge data volume can be accurately monitored; on this basis, the present invention further obtains the deviation degree and correlation between the abnormal point and the data point in its k-nearest neighbors to correct the influence degree, reduces the influence of noise data on the actual required influence space of the data point, effectively improves the accuracy of determining the abnormal score of the data point, and thus effectively improves the accuracy of obtaining the monitoring results of waste gas produced by lithium carbonate. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] By reading the following detailed description with reference to the accompanying drawings, the above and other objects, features and advantages of the exemplary embodiments of the present invention will become readily understood. In the accompanying drawings, several embodiments of the present invention are shown in an exemplary and non-limiting manner, and the same or corresponding reference numerals represent the same or corresponding parts, wherein:

[0023] Figure 1 A schematic flow chart of a method for monitoring waste gas from lithium carbonate production provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0024] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.

[0025] The specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0026] It should be noted that in order to accurately detect abnormal and excessive exhaust gas data in a large amount of exhaust gas data, the exhaust gas data set can be processed by an outlier monitoring algorithm, such as the Influenced Outlierness (INFLO) algorithm based on local density. This algorithm can consider the k-nearest neighbor data points and the reverse k-nearest neighbor data points of the data point when calculating the local anomaly factor of the data point, thereby improving the efficiency of anomaly monitoring and avoiding misjudgment of cluster edge data points. However, in practical applications, data points with abnormal distribution only account for a small number of data sets, and most data points are normal data points. Therefore, most data points can be identified as outliers without the calculation of reverse k-nearest neighbors.

[0027] Based on this, an embodiment of the present invention discloses a method for monitoring waste gas from lithium carbonate production. By obtaining different difference points between the k nearest neighbors and the reverse k nearest neighbors of a data point, the degree of influence of the difference point on the data point is determined to assist in judging whether the data point requires the calculation of the reverse k nearest neighbors, and the influence space actually required by the data point is obtained. The abnormal score is obtained through the influence space to realize waste gas monitoring in the lithium carbonate production process.

[0028] For example, see Figure 1 As shown, Figure 1 A flow chart of a method for monitoring waste gas from lithium carbonate production provided in an embodiment of the present invention is provided, and the method specifically comprises the following steps.

[0029] S1: Get the k nearest neighbors and reverse k nearest neighbors of each data point in the exhaust gas data set, record the different data points in the k nearest neighbors and reverse k nearest neighbors as the difference points of the data points, and determine the density characteristics of the data points. The density characteristics are used to describe the distribution of the difference points in the data set.

[0030] It is understandable that the types of waste gas generated during the production of lithium carbonate vary depending on the production process. For example, in the production of lithium carbonate, the use of the natural lithium carbonate method may produce waste gas containing sulfur oxides such as sulfur dioxide and nitrogen oxides during the roasting and acidification of lithium ore; the use of the lithium nitrate method may produce waste gas containing sulfur dioxide, nitrogen oxides, etc.

[0031] Based on this, in the embodiment of the present invention, the type of exhaust gas may be sulfur dioxide, carbon dioxide, etc., and the type of exhaust gas may be specifically set according to actual needs.

[0032] It should be noted that there is an existing Faster Influenced Outlierness (FINFLO) algorithm that improves the INFLO algorithm. It determines whether a data point needs the participation of reverse k-nearest neighbors by the relationship between the number of data points in the reverse k-nearest neighbors of the data point and the average number of reverse k-nearest neighbors of all data points. Although this algorithm can determine whether a data point needs reverse k-nearest neighbors, it ignores the information of the actual differences between data points, resulting in that even if the number of reverse k-nearest neighbors of a data point is large, but it contains data points that are significantly different from the current data point, the influence of such data points is ignored.

[0033] The difference between only considering the k nearest neighbors of a data point and considering both the k nearest neighbors and the reverse k nearest neighbors of a data point is only the different data points in the k nearest neighbors and the reverse k nearest neighbors. Based on this, the embodiment of the present invention obtains the influence of different difference points in the k nearest neighbors and the reverse k nearest neighbors of a data point on the data point, thereby determining whether the reverse k nearest neighbors need to participate in determining the abnormal score of the data point.

[0034] By way of example, in an embodiment of the present invention, the k nearest neighbors and reverse k nearest neighbors of each data point in the exhaust gas data set are obtained, including: deploying gas sensors to collect exhaust gas data generated in the production of lithium carbonate, and obtaining each data point in the exhaust gas data set after preprocessing the exhaust gas data; presetting the number k of neighboring data points of each data point, and taking the k data points with the closest Euclidean distance to the data point as the k nearest neighbors of the data point; obtaining the data point that takes the data point as the k nearest neighbor, and all the data points that take the data point as the k nearest neighbor constitute the reverse k nearest neighbors of the data point.

[0035] The preprocessing method may be data cleaning, data format conversion, etc., which may be set according to actual needs, and the embodiments of the present invention are not described in detail here. The number k of neighboring data points of each data point may be set according to actual needs, and the embodiments of the present invention do not impose too many restrictions here.

[0036] Specifically, when obtaining exhaust gas data, gas sensors can be placed in the exhaust gas pipeline during the lithium carbonate production process. The gas sensors collect exhaust gas data at a fixed frequency, and each collected exhaust gas data is used as a data point. All data points are pre-processed and together form an exhaust gas data set.

[0037] It should be noted that when obtaining the degree of influence of the difference point on the data point, the density feature of the data point can determine the density of the data point in the local area. The data points in the high-density area are more likely to be normal data points, and vice versa, the data points in the low-density area are more likely to be abnormal data points. In addition, the distance between the data point and its difference point can also describe the degree of influence of the difference point on the data point. The larger the distance, the greater the distance between the data point and its difference point in the feature space, and the less the data point is affected by the difference point.

[0038] Based on this, the degree of influence of the outlier point on the data point can be described by obtaining the deviation of the density characteristics between the difference point and the data point and the deviation of the distance between the difference point and the data point. If the deviation is large, it means that the difference point may not be within the normal distribution range of the data point, and its influence on the data point is smaller.

[0039] For example, in an embodiment of the present invention, determining the density feature of a data point includes: obtaining the ratio of the cumulative sum of distances between the data point's k nearest neighbors and the data point to the number of data points in the k nearest neighbors, and recording it as the density feature of the data point.

[0040] For example, the density characteristics of the data points are determined, and the specific relationship can be seen in the following formula:

[0041] ;

[0042] In the formula, Indicates The density characteristics of the data points, Indicates The number of data points in the k nearest neighbors of a data point, Indicates A data point and its k nearest neighbors The distance between data points.

[0043] After obtaining the density characteristics of each data point and the distance between the data points based on the above steps, the following steps may be continued to determine the degree of influence of the difference point on the data point.

[0044] S2: Calculate the influence of the difference point on the data point.

[0045] For example, in the embodiment of the present invention, the influence degree of the difference point on its data point is determined, and the specific details can be referred to the following relationship:

[0046] ;

[0047] In the formula, Indicates the influence of the j-th difference point on the i-th data point, represents the density feature of the j-th difference point, represents the density feature of the i-th data point, represents the distance between the jth difference point and the i-th data point, represents the exponential function with base e, represents the linear normalization function.

[0048] In the above formula, It represents the absolute difference in density characteristics between the ith data point and its jth difference point. The larger the value is, the greater the difference in density characteristics between the ith data point and its jth difference point is, the greater the possibility that the jth difference point is not within the normal distribution range of the ith data point, and the smaller the corresponding impact is.

[0049] It represents the distance between the i-th data point and its j-th difference point. The larger the distance, the less the i-th data point is affected by the j-th difference point, and the smaller the corresponding degree of influence is.

[0050] After obtaining the degree of influence of the difference point between the k nearest neighbors and the reverse k nearest neighbors of a data point on the data point based on the above method, it can be determined whether the reverse k nearest neighbors need to participate in determining the abnormal score of the data point according to the degree of influence of the difference point on the data point.

[0051] S3: Normalize the ratio of the variance of the data point in the k nearest neighbors of the difference point to the Frobenius norm of the k nearest neighbor correlation matrix to obtain a first ratio, and take the absolute value of the difference between the first ratio and the influence degree as the actual influence degree of the difference point on the data point.

[0052] It should be noted that when the gas sensor collects exhaust gas data, it is easily affected by airflow, which leads to interference of noise data in the collected exhaust gas data. This noise data will cause the degree of influence of the difference point obtained based on the above steps on the data point to deviate from the actual degree, affecting the accuracy of determining the influence space of the data point.

[0053] Based on this, the embodiment of the present invention reduces the interference of noise data by obtaining the variance of data points in the k nearest neighbors of the difference point and the Frobenius norm of the k nearest neighbor correlation matrix, and measuring the discreteness of data points in the k nearest neighbors of the difference point and the correlation between data points.

[0054] By way of example, in an embodiment of the present invention, a method for obtaining a k-nearest neighbor correlation matrix includes: obtaining the correlation between a difference point and each data point in the k-nearest neighbors of the difference point through a DTW algorithm; and recording a matrix composed of the correlations between the difference point and each data point as the k-nearest neighbor correlation matrix of the difference point.

[0055] Among them, the specific steps of calculating the correlation between the difference point and each data point in the k nearest neighbors of the difference point through the dynamic time warping (DTW) algorithm can be implemented by the existing technology, and the embodiments of the present invention are not described in detail here. The specific steps of obtaining the Frobenius norm of the matrix can be implemented by the existing technology, and the embodiments of the present invention are not described in detail here.

[0056] For example, in the embodiment of the present invention, the actual influence degree of the difference point on the data point is determined, and the specific details can be referred to the following relationship:

[0057] ;

[0058] In the formula, Indicates the actual impact of the j-th difference point on the i-th data point, Indicates the influence of the j-th difference point on the i-th data point, represents the variance of the data points in the k nearest neighbors of the jth difference point, represents the Frobenius norm of the k-neighbor correlation matrix of the j-th difference point, represents the absolute value symbol, represents the linear normalization function, Represents a hyperparameter.

[0059] In the above formula, set the hyperparameters The purpose is to avoid the situation where the formula is 0, so the specific value of the hyperparameter can be set according to actual needs.

[0060] It represents the first ratio. The larger the variance, the greater the dispersion of data points in the k nearest neighbors of the difference point, the greater the possibility of noise data in the k nearest neighbors, the greater the degree of correction for the degree of influence, and the smaller the corresponding actual degree of influence after correction. The Frobenius norm can evaluate the correlation between the difference point and the data points in the k nearest neighbors. The smaller the value, the smaller the correlation between the difference point and the data points in the k nearest neighbors, the greater the possibility of noise data in the k nearest neighbors, and the smaller the corresponding actual degree of influence after correction.

[0061] In this way, the embodiment of the present invention obtains the discreteness and correlation of the difference points and data points based on the above steps to correct the influence degree, which can effectively reduce the influence of noise data on the actual influence degree, so that the influence space of each data point can be accurately obtained based on this in the subsequent process, that is, execute the following steps.

[0062] S4: The accumulated and normalized value of the actual impact of all differences on the data point is recorded as the importance of the data point; in the INFLO model, the abnormal score of the data point is determined according to the importance of the data point to realize the monitoring of lithium carbonate production waste gas.

[0063] It should be noted that after the collected waste gas data is input into the INFLO model, the INFLO model will determine the influence space of the data point according to the actual influence of each difference point on the data point, so as to accurately obtain the abnormal score of the data point to realize the monitoring of waste gas in lithium carbonate production.

[0064] For example, when determining the importance of a data point, the accumulated and normalized value of the actual influence of all differences of the data point on the data point may be used as the importance of the data point.

[0065] After determining the importance of each data point in the exhaust gas data set, the anomaly score of the data point can be determined based on the importance of each data point.

[0066] By way of example, in an embodiment of the present invention, the anomaly score of a data point is determined according to the importance of the data point in the INFLO model, including: presetting a threshold; if the importance of the data point is greater than the preset threshold, taking the k nearest neighbors of the data point as the influence space of the data point; if the importance of the data point is not greater than the preset threshold, taking the k nearest neighbors and the reverse k nearest neighbors of the data point as the influence space of the data point; and using the influence space of the data point in the INFLO model to obtain the anomaly score of the data point.

[0067] The preset threshold value may be set to 0.7; the value of the preset threshold value may be set specifically according to actual needs, which will not be elaborated in detail in the embodiment of the present invention.

[0068] In this way, the embodiment of the present invention can accurately determine whether a data point needs a reverse k-nearest neighbor when determining its anomaly score through the importance of the data point, so that the data point can be processed more accurately, reducing the amount of data processing while accurately obtaining the anomaly score of the data point.

[0069] For example, in an embodiment of the present invention, the influence space of a data point is used in the INFLO model to obtain an anomaly score of the data point, including: recording the inverse of the distance between the data point and the nearest data point in its influence space as the local density of the data point; recording the ratio of the mean of the local density of all data points in the influence space of the data point to the local density of the data point as the anomaly score of the data point.

[0070] The specific steps of obtaining the abnormal score of the data point through the INFLO model can be implemented through the existing technology, and the embodiments of the present invention will not be described in detail here.

[0071] After obtaining the anomaly score of each data point based on the above steps, the monitoring result of the data point can be determined according to the anomaly score of the data point.

[0072] By way of example, in an embodiment of the present invention, the anomaly score of a data point is determined according to the importance of the data point in the INFLO model to achieve monitoring of waste gas from lithium carbonate production, including: setting an anomaly threshold; if the anomaly score of a data point is greater than the anomaly threshold, the data point is an abnormal data point; if the anomaly score of the data point is not greater than the anomaly threshold, the data point is a normal data point.

[0073] Among them, the abnormal threshold can be set to 1; the abnormal threshold can be specifically set according to actual needs, and the embodiment of the present invention will not be described in detail here.

[0074] It is understandable that when the abnormal score of a data point is greater than the abnormal threshold, it means that the exhaust gas content at the data point exceeds the safety threshold. The exhaust gas content exceeding the safety threshold may be caused by a variety of factors, which poses a great threat to air quality and human health. In order to effectively reduce the impact of excessive exhaust gas and reduce safety hazards, timely warnings can be issued after abnormal data is monitored.

[0075] For example, in an embodiment of the present invention, the monitoring of waste gas from the production of lithium carbonate is achieved, and then the method further includes: marking abnormal data points and issuing abnormal warnings.

[0076] In this way, the embodiment of the present invention can issue an early warning through technology to remind the staff to take timely measures and improve the exhaust gas emissions by adjusting the production process, repairing equipment, etc.

[0077] It can be seen that in an embodiment of the present invention, when monitoring waste gas from lithium carbonate production, the k nearest neighbors and reverse k nearest neighbors of each data point in the waste gas data set can be obtained, and the different data points in the k nearest neighbors and reverse k nearest neighbors are recorded as difference points of the data points, and the density characteristics of the data points are determined, and the density characteristics are used to describe the distribution of the difference points in the data set; the influence of the difference points on the data points is calculated; the ratio of the variance of the data point in the k nearest neighbors of the difference point to the Frobenius norm of the k nearest neighbor correlation matrix is ​​normalized to obtain a first ratio, and the absolute value of the difference between the first ratio and the influence degree is used as the actual influence degree of the difference point on the data point; the cumulative and normalized value of the actual influence degree of all the difference points of the data point on the data point is recorded as the importance of the data point; in the INFLO model, the abnormal score of the data point is determined according to the importance of the data point to achieve monitoring of waste gas from lithium carbonate production.

[0078] In this way, the embodiment of the present invention monitors abnormal and excessive waste gas data in a huge amount of waste gas data through an anomaly detection algorithm. In this process, the embodiment of the present invention takes into account that not all data points require the participation of reverse k-nearest neighbors when determining the anomaly score. Therefore, the influence of the abnormal points in the k-nearest neighbors and reverse k-nearest neighbors of the data point on the data point is obtained, thereby determining the influence space of the data point; on this basis, the embodiment of the present invention also takes into account that noise data will affect the accuracy of the influence degree. Therefore, the influence degree is corrected by further obtaining the deviation degree and correlation between the abnormal point and the data point in its k-nearest neighbors, thereby effectively improving the accuracy of determining the abnormal score of the data point, thereby effectively improving the accuracy of obtaining the monitoring results of lithium carbonate production waste gas.

[0079] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for monitoring waste gas from lithium carbonate production, characterized in that: include: The k nearest neighbors and reverse k nearest neighbors of each data point in the exhaust gas data set are obtained, the different data points in the k nearest neighbors and reverse k nearest neighbors are recorded as the difference points of the data points, and the density characteristics of the data points are determined, and the density characteristics are used to describe the distribution of the difference points in the data set; the influence of the difference points on the data points is calculated: ; In the formula, Indicates the influence of the j-th difference point on the i-th data point, represents the density feature of the j-th difference point, represents the density feature of the i-th data point, represents the distance between the jth difference point and the i-th data point, represents the exponential function with base e, represents the linear normalization function; Normalizing the ratio of the variance of the data point in the k nearest neighbors of the difference point to the Frobenius norm of the k nearest neighbor correlation matrix to obtain a first ratio, and taking the absolute value of the difference between the first ratio and the influence degree as the actual influence degree of the difference point on the data point; The accumulated and normalized value of the actual influence of all the difference points of the data point on the data point is recorded as the importance of the data point; In the INFLO model, the abnormal score of the data point is determined according to the importance of the data point to achieve the monitoring of lithium carbonate production waste gas.

2. A method for monitoring waste gas from lithium carbonate production according to claim 1, characterized in that: The step of obtaining the k nearest neighbors and reverse k nearest neighbors of each data point in the exhaust gas data set includes: Deploy gas sensors to collect waste gas data generated in lithium carbonate production, and pre-process the waste gas data to obtain data points in the waste gas data set; Preset the number k of neighboring data points of each data point, and use the k data points with the closest Euclidean distance to the data point as the k nearest neighbors of the data point; The data points that have the data point as k nearest neighbors are obtained, and all the data points that have the data point as k nearest neighbors constitute the reverse k nearest neighbors of the data point.

3. A method for monitoring waste gas from lithium carbonate production according to claim 1, characterized in that: The step of determining the density characteristics of the data points comprises: The ratio of the cumulative sum of distances between the k nearest neighbors of the data point and the data point to the number of data points in the k nearest neighbors is obtained, and recorded as the density feature of the data point.

4. A method for monitoring waste gas from lithium carbonate production according to claim 1, characterized in that: The method for obtaining the k nearest neighbor correlation matrix includes: Obtaining the correlation between the difference point and each data point in the k nearest neighbors of the difference point by using the DTW algorithm; The matrix composed of the correlations between the difference point and each data point is recorded as the k-nearest neighbor correlation matrix of the difference point.

5. A method for monitoring waste gas from lithium carbonate production according to claim 1, characterized in that: Determining the abnormality score of the data point according to the importance of the data point in the INFLO model includes: Preset threshold value; If the importance of the data point is greater than a preset threshold, the k nearest neighbors of the data point are used as the influence space of the data point; If the importance of the data point is not greater than a preset threshold, the k nearest neighbors and reverse k nearest neighbors of the data point are used as the influence space of the data point; The influence space of the data point is used in the INFLO model to obtain the anomaly score of the data point.

6. A method for monitoring waste gas from lithium carbonate production according to claim 5, characterized in that: The using the influence space of the data point in the INFLO model to obtain the abnormality score of the data point includes: The reciprocal of the distance between the data point and the nearest data point in its influence space is recorded as the local density of the data point; The ratio of the mean of the local densities of all data points in the influence space of the data point to the local density of the data point is recorded as the abnormal score of the data point.

7. A method for monitoring waste gas from lithium carbonate production according to claim 1, characterized in that: Determining the abnormal score of the data point according to the importance of the data point in the INFLO model to achieve monitoring of lithium carbonate production waste gas includes: Set anomaly thresholds; If the anomaly score of the data point is greater than the anomaly threshold, the data point is an abnormal data point; If the anomaly score of the data point is not greater than the anomaly threshold, the data point is a normal data point.

8. A method for monitoring waste gas from lithium carbonate production according to claim 7, characterized in that: The method of realizing the monitoring of waste gas from lithium carbonate production further includes: After marking the abnormal data points, an abnormal warning is issued.

Images