Monitoring and alarm system for environmental protection equipment and method thereof

Abstract

The application relates to the technical field of intelligent monitoring, and particularly discloses an environmental protection equipment monitoring alarm system and method thereof, which adopts a deep learning technology to pretreat and clean sewage discharge data, converts the data into a unified format, carries out embedding coding and context semantic information extraction, and judges whether the running state of the environmental protection equipment is normal based on the same. In this way, the running abnormal problems of the environmental protection equipment can be found and solved in a timely manner, so that the compliance of sewage treatment and the effectiveness of environmental protection can be ensured.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent monitoring, and more particularly, to a monitoring and alarm system for environmental protection equipment and a method thereof. BACKGROUND

[0002] With the rapid development of economy, people's environmental protection consciousness is becoming stronger and stronger, and real-time monitoring of environmental quality is very important. Industrial enterprises are an important driving force for economic development and social progress in China, and are also one of the important sources of environmental pollution.

[0003] In the production process of industrial enterprises, a large amount of broad-spectrum pollutants and specific pollutants will be generated, among which wastewater discharge is one of the main pollution sources. The discharge of huge industrial wastewater exacerbates water pollution and water resource shortage. Moreover, various harmful substances such as heavy metals and organic matter in wastewater can accumulate in water, which can directly threaten health if people or animals directly ingest these pollutants.

[0004] At present, most factories use environmental protection equipment to treat wastewater before discharging, but in actual application, environmental protection equipment will inevitably have abnormalities, which may cause the treated wastewater to still be non-compliant.

[0005] Therefore, a monitoring and alarm system for environmental protection equipment and a method thereof are expected, which can timely discover and solve the operation abnormality problem of the environmental protection equipment. SUMMARY

[0006] In order to solve the above technical problems, the present application is proposed. The embodiments of the present application provide a monitoring and alarm system for environmental protection equipment and a method thereof, which uses deep learning technology to preprocess and clean wastewater discharge data, converts the data into a unified format, and then performs embedding coding and context semantic information extraction, and judges whether the running state of the environmental protection equipment is normal based on this. In this way, the operation abnormality problem of the environmental protection equipment can be timely discovered and solved, so as to ensure the compliance of wastewater treatment and the effectiveness of environmental protection.

[0007] Correspondingly, according to one aspect of the present application, a monitoring and alarm system for environmental protection equipment is provided, which comprises:

[0008] a data acquisition module configured to acquire wastewater discharge data after treatment by the environmental protection equipment, wherein the wastewater discharge data comprises discharge flow, pH value, chemical oxygen demand, ammonia nitrogen, and heavy metal content;

[0009] a data cleaning module configured to clean the wastewater discharge data to obtain a plurality of data items, wherein the plurality of data items contain data attributes and data values;

[0010] The attribute embedding coding module is configured to respectively perform word embedding coding on the data attributes of each data item in the plurality of data items by using a word embedding layer to obtain a plurality of data item attribute word embedding vectors.

[0011] The data adding module is configured to add the data values of each data item in the plurality of data items to the end of the attribute word embedding vector of the data item to obtain a plurality of data item embedding vectors.

[0012] The sewage discharge data semantic understanding module is configured to input the plurality of data item embedding vectors into a transformer-based context encoder to obtain a sewage discharge data semantic feature vector.

[0013] The Gaussian enhancement module is configured to perform feature enhancement on the sewage discharge data semantic feature vector based on a Gaussian density map to obtain a classification feature matrix.

[0014] The optimization module is configured to perform feature distribution optimization on the classification feature matrix to obtain an optimized classification feature matrix.

[0015] The monitoring result generation module is configured to input the optimized classification feature matrix into a classifier to obtain a classification result, which is used to indicate whether the operation state of the sewage environmental protection equipment is normal.

[0016] In the above-mentioned monitoring and alarm system for environmental protection equipment, the attribute embedding coding module is configured to use a learnable embedding matrix of the word embedding layer to respectively perform word embedding coding on the data attributes of each data item in the plurality of data items to obtain the plurality of data item attribute word embedding vectors.

[0017] In the above-mentioned monitoring and alarm system for environmental protection equipment, the sewage discharge data semantic understanding module includes: a one-dimensional arrangement unit configured to arrange the plurality of data item embedding vectors in one dimension to obtain a data global embedding vector; a self-attention unit configured to calculate the product between the data global embedding vector and the transpose vector of each data item embedding vector in the plurality of data item embedding vectors to obtain a plurality of self-attention correlation matrices; a normalized self-attention unit configured to respectively perform normalization processing on each self-attention correlation matrix in the plurality of self-attention correlation matrices to obtain a plurality of normalized self-attention correlation matrices; an attention calculation unit configured to input each normalized self-attention correlation matrix in the plurality of normalized self-attention correlation matrices into a classification function to obtain a plurality of probability values; an attention application unit configured to respectively use each probability value in the plurality of probability values as a weight to weight each data item embedding vector in the plurality of data item embedding vectors to obtain a plurality of data item feature vectors; and a concatenation unit configured to concatenate the plurality of data item feature vectors to obtain the sewage discharge data semantic feature vector.

[0018] In the monitoring and alarm system of the environmental protection equipment, the Gaussian enhancement module comprises a Gaussian density map construction unit configured to construct a Gaussian density map of the sewage discharge data semantic feature vector according to a Gaussian formula, wherein the Gaussian formula is: wherein μ is the sewage discharge data semantic feature vector, σ is a variance between feature values of two corresponding positions in the sewage discharge data semantic feature vector, x represents a variable of the Gaussian density map; and a Gaussian discretization unit configured to perform Gaussian discretization processing on Gaussian distribution of each position in the Gaussian density map to obtain the classification feature matrix.

[0019] In the monitoring and alarm system of the environmental protection equipment, the optimization module comprises a global pooling unit configured to perform global mean pooling on the classification feature matrix to obtain a global semantic feature vector; a covariance matrix calculation unit configured to calculate a covariance matrix of the global semantic feature vector; an eigenvalue decomposition unit configured to perform eigenvalue decomposition on the covariance matrix to obtain a plurality of eigenvalues and a plurality of eigenvector corresponding to the plurality of eigenvalues; a class homography subspace construction unit configured to extract K eigenvectors corresponding to K largest eigenvalues from the plurality of eigenvectors, and the K eigenvectors constitute a k-dimensional class homography subspace; an unfolding unit configured to unfold each row vector of the classification feature matrix to obtain a plurality of classification feature local unfolding feature vectors; a mapping unit configured to map the plurality of classification feature local unfolding feature vectors to the class homography subspace respectively to obtain a plurality of mapped classification feature local unfolding feature vectors; a probabilistic unit configured to activate the plurality of mapped classification feature local unfolding feature vectors through a Sigmoid activation function to obtain a plurality of mapped probabilistic classification feature local unfolding feature vectors; and a dimension reconstruction unit configured to perform dimension reconstruction on the plurality of mapped probabilistic classification feature local unfolding feature vectors to obtain an optimized classification feature matrix.

[0020] In the monitoring and alarm system of the environmental protection equipment, the monitoring result generation module comprises a matrix unfolding unit configured to unfold the optimized classification feature matrix into a classification feature vector according to a row vector or a column vector; a fully connected coding unit configured to perform fully connected coding on the classification feature vector using a fully connected layer of the classifier to obtain a fully connected coding feature vector; a probability obtaining unit configured to obtain a first probability belonging to a normal sewage environmental protection equipment operation state and a second probability belonging to an abnormal sewage environmental protection equipment operation state by using a Softmax classification function of the classifier on the fully connected coding feature vector; and a classification result determination unit configured to determine the classification result based on a comparison between the first probability and the second probability.

[0021] According to another aspect of the present application, a monitoring and alarming method of an environmental protection device is provided, which comprises:

[0022] obtaining sewage discharge data of the environmental protection device after treatment, wherein the sewage discharge data comprises discharge flow, pH value, chemical oxygen demand, ammonia nitrogen and heavy metal content;

[0023] performing data cleaning on the sewage discharge data to obtain a plurality of data items, wherein the plurality of data items comprise data attributes and data values;

[0024] respectively passing the data attributes of each data item in the plurality of data items through a word embedding layer to obtain a plurality of data item attribute word embedding vectors;

[0025] respectively adding the data values of each data item in the plurality of data items to the end of the data item attribute word embedding vector of the each data item to obtain a plurality of data item embedding vectors;

[0026] inputting the plurality of data item embedding vectors into a context encoder based on a converter to obtain a sewage discharge data semantic feature vector;

[0027] performing feature enhancement on the sewage discharge data semantic feature vector based on a Gaussian density map to obtain a classification feature matrix;

[0028] performing feature distribution optimization on the classification feature matrix to obtain an optimized classification feature matrix;

[0029] passing the optimized classification feature matrix through a classifier to obtain a classification result, wherein the classification result is used to represent whether the operation state of the sewage environmental protection device is normal.

[0030] In the above-mentioned monitoring and alarming method of the environmental protection device, respectively passing the data attributes of each data item in the plurality of data items through a word embedding layer to obtain a plurality of data item attribute word embedding vectors comprises: using a learnable embedding matrix of the word embedding layer to respectively perform word embedding coding on the data attributes of each data item in the plurality of data items to obtain the plurality of data item attribute word embedding vectors.

[0031] In the monitoring and alarming method of the environmental protection equipment, the plurality of data items are embedded into a vector input of a transformer-based context encoder to obtain a sewage discharge data semantic feature vector, including: one-dimensionally arranging the plurality of data item embedding vectors to obtain a data global embedding vector; calculating a product between the data global embedding vector and a transpose vector of each data item embedding vector in the plurality of data item embedding vectors to obtain a plurality of self-attention correlation matrices; respectively performing standardization processing on each self-attention correlation matrix in the plurality of self-attention correlation matrices to obtain a plurality of standardized self-attention correlation matrices; passing each standardized self-attention correlation matrix in the plurality of standardized self-attention correlation matrices through a classification function to obtain a plurality of probability values; respectively weighting each data item embedding vector in the plurality of data item embedding vectors with each probability value in the plurality of probability values to obtain a plurality of data item feature vectors; and concatenating the plurality of data item feature vectors to obtain the sewage discharge data semantic feature vector.

[0032] In the monitoring and alarming method of the environmental protection equipment, the sewage discharge data semantic feature vector is subjected to feature enhancement based on a Gaussian density map to obtain a classification feature matrix, including: constructing a Gaussian density map of the sewage discharge data semantic feature vector by using a Gaussian formula as follows: wherein μ is the sewage discharge data semantic feature vector, and σ is a variance between feature values of corresponding two positions in the sewage discharge data semantic feature vector. wherein μ is the sewage discharge data semantic feature vector, and σ is a variance between feature values of corresponding two positions in the sewage discharge data semantic feature vector. wherein μ is the sewage discharge data semantic feature vector, and σ is a variance between feature values of corresponding two positions in the sewage discharge data semantic feature vector.

[0033] Compared with the prior art, the monitoring and alarming system of the environmental protection equipment and the method thereof provided by the present application adopt deep learning technology, pre-process and clean sewage discharge data, convert the data into a unified format, then perform embedding coding and context semantic information extraction, and judge whether the running state of the environmental protection equipment is normal based on the same. In this way, the running abnormal problems of the environmental protection equipment can be found and solved in time, so as to ensure the compliance of sewage treatment and the effectiveness of environmental protection. BRIEF DESCRIPTION OF DRAWINGS

[0034] The above and other objects, features and advantages of the present application will become more apparent from the following detailed description of embodiments of the present application, when taken in conjunction with the accompanying drawings. The drawings provided in the present application are used to provide further understanding of the embodiments of the present application, and constitute a part of the specification, and are used to explain the present application together with the embodiments of the present application, and do not constitute a limitation on the present application. In the drawings, the same reference numerals generally represent the same components or steps.

[0035] Figure 1A block diagram of the monitoring and alarm system of the environmental protection equipment according to the embodiment of the present application.

[0036] Figure 2 A schematic diagram of the architecture of the monitoring and alarm system of the environmental protection equipment according to the embodiment of the present application.

[0037] Figure 3 A block diagram of the sewage discharge data semantic understanding module in the monitoring and alarm system of the environmental protection equipment according to the embodiment of the present application.

[0038] Figure 4 A block diagram of the Gaussian enhancement module in the monitoring and alarm system of the environmental protection equipment according to the embodiment of the present application.

[0039] Figure 5 A block diagram of the monitoring result generation module in the monitoring and alarm system of the environmental protection equipment according to the embodiment of the present application.

[0040] Figure 6 A flowchart of the monitoring and alarm method of the environmental protection equipment according to the embodiment of the present application. DETAILED DESCRIPTION

[0041] In the following, the example embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the example embodiments described herein.

[0042] SUMMARY

[0043] As described above, in the production process of industrial enterprises, a large amount of sewage will be generated. The huge industrial wastewater discharge aggravates the problem of water pollution and water resource shortage. At present, most factories use environmental protection equipment to treat the sewage before discharging, but in actual application, the environmental protection equipment will inevitably have abnormalities, which may cause the treated sewage to still be non-compliant. Therefore, an environmental protection equipment monitoring and alarm system and method are expected, which can timely find and solve the operation abnormality problem of the environmental protection equipment.

[0044] At present, deep learning and neural networks have been widely used in computer vision, natural language processing, speech signal processing and other fields. In addition, deep learning and neural networks have also shown a level close to or even surpassing human beings in image classification, object detection, semantic segmentation, text translation and other fields.

[0045] In recent years, the development of deep learning and neural networks has provided new solutions for the operation state monitoring of environmental protection equipment.

[0046] Accordingly, considering that in the actual operation state monitoring of the environmental protection equipment, the key lies in sufficient semantic understanding of the sewage discharge data after the environmental protection equipment treatment. However, due to the sewage discharge data including many data items and data sources, such as discharge flow, PH value, chemical oxygen demand, ammonia nitrogen, and heavy metal content. These data format standards are not unified, which brings adverse effects to the semantic information understanding of the sewage discharge data. Based on this, in the technical solution of the present application, the deep learning technology is used to preprocess and clean the sewage discharge data, convert the data into a unified format, and then perform embedding coding and context semantic information extraction, and judge whether the operation state of the environmental protection equipment is normal based on this. In this way, the operation abnormal problems of the environmental protection equipment can be found and solved in time, so as to ensure the compliance of sewage treatment and the effectiveness of environmental protection.

[0047] Specifically, in the technical solution of the present application, first, the sewage discharge data after the environmental protection equipment treatment is obtained, wherein the sewage discharge data includes discharge flow, PH value, chemical oxygen demand, ammonia nitrogen, and heavy metal content. Then, considering that sewage discharge data from different sources often has different formats and structures, which is not conducive to subsequent data processing; and the sewage discharge data may be affected by sensor errors, measurement deviations, incomplete data acquisition or noise, etc. to reduce data quality. Therefore, the sewage discharge data is cleaned to exclude outliers and outliers, thereby improving the quality and accuracy of the data. And through data cleaning, the data is standardized and consistent, which is converted into a more operable form, thereby obtaining a plurality of data items, wherein the plurality of data items contain data attributes and data values.

[0048] Then, the data attributes in the plurality of data items are respectively converted through a word embedding layer to convert the data attributes into a numerical representation that can be understood and processed by a machine learning model, to obtain a plurality of data item attribute word embedding vectors. It can be understood that the data attributes are key features describing the sewage discharge data, such as discharge flow, PH value, chemical oxygen demand, etc. By converting the data attributes into word embedding vectors, they can be represented as a continuous numerical vector, capturing the semantic association and similarity between attributes, providing more rich information for subsequent feature extraction and modeling. And after converting the data attributes into vector representation through the word embedding layer, the data attributes can be normalized to have similar scales and ranges, to avoid the influence of numerical differences between different data attributes on model training and prediction, and to improve the stability and accuracy of the model.

[0049] Then, the data value of each data item in the plurality of data items is added to the end of the attribute word embedding vector of the respective data item, respectively, to combine the data attribute and the corresponding numerical value together to form a complete data item representation, thereby obtaining a plurality of data item embedding vectors. It should be understood that the data item is composed of a data attribute and a corresponding numerical value, which together describe the characteristics of the sewage discharge data. By adding the data value to the end of the data item attribute word embedding vector, the inconsistency or loss of information caused by separate processing of the data attribute and the numerical value is avoided, thereby providing a more accurate and reliable data representation.

[0050] Further, the plurality of data item embedding vectors are input into a transformer-based context encoder to extract semantic feature information of the sewage discharge data, thereby obtaining a sewage discharge data semantic feature vector. It should be understood that the sewage discharge data is composed of a plurality of data items, and there is a context relationship between the data items. The transformer-based context encoder has strong modeling capability and can effectively capture the context information between the data items. By inputting the plurality of data item embedding vectors into the context encoder, multi-level feature extraction can be performed on the plurality of data item embedding vectors, and the self-attention mechanism thereof can be used to learn the relevance between the data items to capture local features and global relationships in the data items.

[0051] Then, the sewage discharge data semantic feature vector is enhanced using a Gaussian density map to obtain a classification feature matrix. Accordingly, the semantic feature vector of the sewage discharge data usually has a certain spatial distribution, in which similar data points are clustered together. By using a Gaussian density map, this distribution is converted into a probabilistic representation, in which high-density areas correspond to clustered data points. The Gaussian density map focuses attention on high-density areas by assigning probability weights at each data point, so that the influence of data points in low-density areas on classification is reduced, while the influence of data points in high-density areas on classification is enhanced. That is, different weights are applied to features of different importance, increasing the attention to important features and reducing the influence of irrelevant features and noise, thereby improving the accuracy and robustness of classification.

[0052] In particular, considering that there can be noise and outliers in the data, these outliers can interfere with the distribution of feature points. Outliers can be caused by measurement or recording errors, data collection problems or other abnormal situations. These outliers can be located at any position in the feature space, resulting in uneven distribution of feature points. At the same time, in the classification problem, if the number of samples of different categories is very different, the distribution of feature points can be uneven. For example, the number of samples of a certain category is very small, while the number of samples of another category is very large, so the distribution of feature points in the feature space can be biased towards the category with more samples. And when data cleaning and feature extraction are performed, the distribution of feature points can also be uneven. Some important features can be ignored, while some irrelevant or redundant features can be retained, resulting in uneven distribution of feature points in the feature space.

[0053] Therefore, the feature distribution optimization is performed on the classification feature matrix to obtain an optimized classification feature matrix, including: performing global mean pooling on the classification feature matrix to obtain a global semantic feature vector; calculating a covariance matrix of the global semantic feature vector; performing eigenvalue decomposition on the covariance matrix to obtain a plurality of eigenvalues and a plurality of eigenvector corresponding to the plurality of eigenvalues; extracting K eigenvectors corresponding to the K largest eigenvalues from the plurality of eigenvectors, the K eigenvectors constituting a k-dimensional category homography subspace; expanding each row vector of the classification feature matrix to obtain a plurality of classification feature local expansion feature vectors; mapping the plurality of classification feature local expansion feature vectors to the category homography subspace respectively to obtain a plurality of mapped classification feature local expansion feature vectors; activating the plurality of mapped classification feature local expansion feature vectors through a Sigmoid activation function to obtain a plurality of mapped probabilistic classification feature local expansion feature vectors; and reconstructing the dimensions of the plurality of mapped probabilistic classification feature local expansion feature vectors to obtain the optimized classification feature matrix.

[0054] In particular, in the technical solution of the present application, the global semantic feature distribution of the classification feature matrix is first calculated, and then the category homography mapping subspace of the classification feature matrix is constructed based on the global semantic feature distribution. Then, each local feature of the classification feature matrix is mapped to the category homography mapping subspace for feature dimension sparsification, that is, the category homography mapping subspace is used as a pivot mapping subspace to perform cluster-based feature distribution modulation on each local feature of the classification feature matrix. In this way, because the distribution of feature points in the category homography mapping subspace is more uniform and compact, the interference of outliers can be reduced, and the feature robustness of the classification feature matrix can be enhanced.

[0055] Finally, the optimized classification feature matrix is ​​passed through a classifier to obtain classification results indicating whether the wastewater environmental protection equipment is operating normally. Accordingly, if the classification result indicates that the wastewater environmental protection equipment is operating normally, it means that the equipment is operating normally and wastewater is being discharged in compliance with regulations; if the classification result indicates that the wastewater environmental protection equipment is operating abnormally, it means that there is an abnormality in the equipment and further inspection, repair, or other necessary measures are required.

[0056] After introducing the basic principles of this application, various non-limiting embodiments of this application will be described in detail below with reference to the accompanying drawings.

[0057] Exemplary System

[0058] Figure 1 This is a block diagram of a monitoring and alarm system for environmental protection equipment according to an embodiment of this application. Figure 1 As shown, the environmental protection equipment monitoring and alarm system 100 according to an embodiment of this application includes: a data acquisition module 110, used to acquire wastewater discharge data after treatment by the environmental protection equipment, wherein the wastewater discharge data includes discharge flow rate, pH value, chemical oxygen demand, ammonia nitrogen, and heavy metal content; a data cleaning module 120, used to clean the wastewater discharge data to obtain multiple data items, wherein the multiple data items include data attributes and data values; an attribute embedding and encoding module 130, used to pass the data attributes of each data item in the multiple data items through a word embedding layer to obtain multiple data item attribute word embedding vectors; and a data adding module 140, used to add the data values ​​of each data item in the multiple data items to... The embedding vectors of the attribute words of each data item are added to the end to obtain multiple data item embedding vectors; the sewage discharge data semantic understanding module 150 is used to input the multiple data item embedding vectors into a context encoder based on a converter to obtain a sewage discharge data semantic feature vector; the Gaussian enhancement module 160 is used to perform feature enhancement on the sewage discharge data semantic feature vector based on a Gaussian density map to obtain a classification feature matrix; the optimization module 170 is used to optimize the feature distribution of the classification feature matrix to obtain an optimized classification feature matrix; the monitoring result generation module 180 is used to pass the optimized classification feature matrix through a classifier to obtain a classification result, which is used to indicate whether the sewage environmental protection equipment is operating normally.

[0059] Figure 2 This is a schematic diagram of the architecture of a monitoring and alarm system for environmental protection equipment according to an embodiment of this application. Figure 2As shown, first, sewage discharge data after treatment by an environmental protection device is acquired, wherein the sewage discharge data includes discharge flow, PH value, chemical oxygen demand, ammonia nitrogen, and heavy metal content. Next, the sewage discharge data is data cleaned to obtain a plurality of data items, wherein the plurality of data items include data attributes and data values. Then, the data attributes of each data item in the plurality of data items are respectively passed through a word embedding layer to obtain a plurality of data item attribute word embedding vectors. Secondly, the data values of each data item in the plurality of data items are respectively added to the end of the attribute word embedding vector of the each data item to obtain a plurality of data item embedding vectors. Then, the plurality of data item embedding vectors are input into a transformer-based context encoder to obtain a sewage discharge data semantic feature vector. Next, the sewage discharge data semantic feature vector is feature enhanced based on a Gaussian density map to obtain a classification feature matrix. Then, the classification feature matrix is feature distribution optimized to obtain an optimized classification feature matrix. Finally, the optimized classification feature matrix is passed through a classifier to obtain a classification result, which is used to represent whether the sewage environmental protection device is in a normal operating state.

[0060] In the above-mentioned environmental protection device monitoring alarm system 100, the data acquisition module 110 is configured to acquire sewage discharge data after treatment by an environmental protection device, wherein the sewage discharge data includes discharge flow, PH value, chemical oxygen demand, ammonia nitrogen, and heavy metal content. As described in the above background, a large amount of sewage is generated in the production process of an industrial enterprise. The discharge of a large amount of industrial wastewater exacerbates water pollution and water resource shortage. At present, most factories use environmental protection devices to treat sewage before discharging. However, in actual application, the environmental protection device is likely to be abnormal, which may cause the treated sewage to still be non-compliant. Therefore, an environmental protection device monitoring alarm system is expected to timely discover and solve the operating abnormality problem of the environmental protection device.

[0061] Accordingly, considering that in the actual operation state monitoring of the environmental protection equipment, the key lies in sufficient semantic understanding of the sewage discharge data after the treatment of the environmental protection equipment. However, due to the sewage discharge data including many data items and data sources, such as discharge flow, PH value, chemical oxygen demand, ammonia nitrogen, and heavy metal content. These data format standards are not unified, which adversely affects the semantic information understanding of the sewage discharge data. Based on this, in the technical scheme of the present application, the deep learning technology is used to preprocess and clean the sewage discharge data, convert the data into a unified format, and then perform embedding coding and context semantic information extraction, and judge whether the operation state of the environmental protection equipment is normal based on this. In this way, the operation abnormal problems of the environmental protection equipment can be found and solved in time, so as to ensure the compliance of sewage treatment and the effectiveness of environmental protection. Specifically, in the technical scheme of the present application, first, the sewage discharge data after the treatment of the environmental protection equipment is obtained, wherein the sewage discharge data includes discharge flow, PH value, chemical oxygen demand, ammonia nitrogen, and heavy metal content.

[0062] In the above-mentioned monitoring and alarm system 100 of the environmental protection equipment, the data cleaning module 120 is used for data cleaning of the sewage discharge data to obtain a plurality of data items, wherein the plurality of data items contain data attributes and data values. Considering that sewage discharge data from different sources often has different formats and structures, which is not conducive to subsequent data processing; and the sewage discharge data may be affected by sensor errors, measurement deviations, incomplete data acquisition or noise, etc. to reduce data quality. Therefore, the sewage discharge data is cleaned to exclude outliers and outliers, thereby improving the quality and accuracy of the data. And through data cleaning, the data is standardized and consistent, and is converted into a more operable form, thereby obtaining a plurality of data items, wherein the plurality of data items contain data attributes and data values.

[0063] In the above-mentioned monitoring and alarm system 100 of the environmental protection equipment, the attribute embedding coding module 130 is used for embedding the data attributes of each data item in the plurality of data items through a word embedding layer to obtain a plurality of data item attribute word embedding vectors. It can be understood that the data attribute is a key feature for describing the sewage discharge data, such as discharge flow, PH value, chemical oxygen demand, etc. By converting the data attribute into a word embedding vector, it can be characterized as a continuous numerical vector, capturing the semantic association and similarity between attributes, providing more rich information for subsequent feature extraction and modeling. And after converting the data attribute into a vector representation through the word embedding layer, the data attribute can be normalized to have similar scales and ranges to avoid the influence of numerical differences between different data attributes on model training and prediction, and improve the stability and accuracy of the model.

[0064] Correspondingly, in one specific example, the attribute embedding encoding module 130 is configured to: use the learnable embedding matrix of the word embedding layer to respectively perform word embedding encoding on the data attributes of each data item in the plurality of data items to obtain a plurality of data item attribute word embedding vectors.

[0065] In the above-mentioned environmental protection equipment monitoring and alarm system 100, the data adding module 140 is configured to add the data value of each data item in the plurality of data items to the end of the attribute word embedding vector of the data item to obtain a plurality of data item embedding vectors. It should be understood that the data item is composed of data attributes and corresponding values, which together describe the characteristics of the sewage discharge data. By adding the data value to the end of the data item attribute word embedding vector, the data attribute and the corresponding value are combined together to form a complete data item representation, avoiding the inconsistency or loss of information caused by separate processing of data attributes and values, thereby providing more accurate and reliable data representation.

[0066] In the above-mentioned environmental protection equipment monitoring and alarm system 100, the sewage discharge data semantic understanding module 150 is configured to input the plurality of data item embedding vectors into the transformer-based context encoder to obtain a sewage discharge data semantic feature vector. It should be understood that the sewage discharge data is composed of a plurality of data items, and there is a context relationship between the data items. The transformer-based context encoder has strong modeling capability and can effectively capture the context information between the data items. By inputting the plurality of data item embedding vectors into the context encoder, multi-level feature extraction can be performed on the plurality of data item embedding vectors, and the self-attention mechanism thereof can be used to learn the relevance between the data items to capture local features and global relationships in the data items.

[0067] Figure 3 The block diagram of the sewage discharge data semantic understanding module in the environmental protection equipment monitoring and alarm system according to the embodiment of the present application. As shown in Figure 3As shown, the sewage discharge data semantic understanding module 150 comprises: a one-dimensional arrangement unit 151, configured to arrange the plurality of data item embedding vectors in one dimension to obtain a data global embedding vector; a self-attention unit 152, configured to calculate the product between the data global embedding vector and the transpose vector of each data item embedding vector in the plurality of data item embedding vectors to obtain a plurality of self-attention correlation matrices; a normalized self-attention unit 153, configured to respectively perform normalization processing on each self-attention correlation matrix in the plurality of self-attention correlation matrices to obtain a plurality of normalized self-attention correlation matrices; an attention calculation unit 154, configured to pass each normalized self-attention correlation matrix in the plurality of normalized self-attention correlation matrices through a classification function to obtain a plurality of probability values; an attention application unit 155, configured to respectively weight each data item embedding vector in the plurality of data item embedding vectors with each probability value in the plurality of probability values to obtain a plurality of data item feature vectors; and a concatenation unit 156, configured to concatenate the plurality of data item feature vectors to obtain the sewage discharge data semantic feature vector.

[0068] In the above-mentioned environmental protection equipment monitoring alarm system 100, the Gaussian enhancement module 160 is configured to perform feature enhancement on the sewage discharge data semantic feature vector based on a Gaussian density map to obtain a classification feature matrix. Accordingly, the semantic feature vector of the sewage discharge data usually has a certain spatial distribution, in which similar data points are clustered together. By using a Gaussian density map, this distribution is converted into a probability representation, in which high-density areas correspond to clustered data points. The Gaussian density map assigns a probability weight at each data point, focusing attention on high-density areas, so that the influence of data points in low-density areas on classification is reduced, while the influence of data points in high-density areas on classification is enhanced. That is, different weights are applied to features of different importance, improving the attention to important features and reducing the influence of irrelevant features and noise, thereby improving the accuracy and robustness of classification.

[0069] Figure 4 A block diagram of the Gaussian enhancement module in the environmental protection equipment monitoring alarm system according to the embodiments of the present application. As shown in Figure 4 The Gaussian enhancement module 160 comprises: a Gaussian density map construction unit 161, configured to construct a Gaussian density map of the sewage discharge data semantic feature vector using the following Gaussian formula: wherein μ is the sewage discharge data semantic feature vector, σ is the variance between the feature values of the corresponding two positions in the sewage discharge data semantic feature vector, represents a Gaussian density probability function, x represents a variable of the Gaussian density map; a Gaussian discretization unit 162 is configured to perform Gaussian discretization processing on Gaussian distributions of various positions in the Gaussian density map to obtain the classification feature matrix.

[0070] In the monitoring and alarm system 100 of the environmental protection equipment, the optimization module 170 is configured to perform feature distribution optimization on the classification feature matrix to obtain an optimized classification feature matrix. In particular, considering that there may be noise and abnormal points in the data, these abnormal points may interfere with the distribution of feature points. Abnormal points may be caused by measurement or recording errors, data acquisition problems or other abnormal situations. These abnormal points may be located at any position in the feature space, resulting in uneven distribution of feature points. At the same time, in the classification problem, if the number of samples of different categories is very different, the distribution of feature points may be uneven. For example, the number of samples of a certain category is very small, while the number of samples of another category is very large, so the distribution of feature points in the feature space may be biased towards the category with more samples. Moreover, during data cleaning and feature extraction, the distribution of feature points may also be uneven. Some important features may be ignored, while some irrelevant or redundant features may be retained, resulting in uneven distribution of feature points in the feature space. Therefore, the classification feature matrix is subjected to feature distribution optimization to obtain an optimized classification feature matrix.

[0071] Correspondingly, in one specific example, the optimization module 170 includes: a global pooling unit configured to perform global mean pooling on the classification feature matrix to obtain a global semantic feature vector; a covariance matrix calculation unit configured to calculate a covariance matrix of the global semantic feature vector; an eigenvalue decomposition unit configured to perform eigenvalue decomposition on the covariance matrix to obtain a plurality of eigenvalues and a plurality of eigenvector corresponding to the plurality of eigenvalues; a class homography subspace construction unit configured to extract K eigenvectors corresponding to K largest eigenvalues from the plurality of eigenvectors, the K eigenvectors constituting a k-dimensional class homography subspace; an unfolding unit configured to unfold each row vector of the classification feature matrix to obtain a plurality of classification feature local unfolding feature vectors; a mapping unit configured to map the plurality of classification feature local unfolding feature vectors to the class homography subspace respectively to obtain a plurality of mapped classification feature local unfolding feature vectors; a probabilistic unit configured to activate the plurality of mapped classification feature local unfolding feature vectors through a Sigmoid activation function to obtain a plurality of mapped probabilistic classification feature local unfolding feature vectors; and a dimension reconstruction unit configured to perform dimension reconstruction on the plurality of mapped probabilistic classification feature local unfolding feature vectors to obtain an optimized classification feature matrix.

[0072] In particular, in the technical solution of the present application, firstly, the global semantic feature distribution of the classification feature matrix is calculated, and then the class homography mapping subspace of the classification feature matrix is constructed based on the global semantic feature distribution. Next, each local feature of the classification feature matrix is mapped to the class homography mapping subspace for feature dimension sparsification, that is, each local feature of the classification feature matrix is clustered and modulated in feature distribution with the class homography mapping subspace as the pivot mapping subspace. In this way, since the distribution of feature points is more uniform and compact in the class homography mapping subspace, the interference of abnormal points can be reduced, and the feature robustness of the classification feature matrix can be enhanced.

[0073] In the monitoring and alarm system 100 of the environmental protection equipment described above, the monitoring result generation module 180 is configured to pass the optimized classification feature matrix through a classifier to obtain a classification result, and the classification result is used to indicate whether the operation state of the sewage environmental protection equipment is normal. Accordingly, if the classification result indicates that the operation state of the sewage environmental protection equipment is normal, it means that the equipment is running normally and the sewage is discharged in compliance with the regulations. If the classification result indicates that the operation state of the sewage environmental protection equipment is abnormal, it means that the equipment has an abnormal situation and needs to be further checked, repaired or other necessary measures are taken.

[0074] Figure 5 A block diagram of the monitoring result generation module in the monitoring and alarm system of the environmental protection equipment according to the embodiments of the present application is shown in FIG. 8. As shown in FIG. 8, the monitoring result generation module 180 includes a matrix expansion unit 181, a fully connected coding unit 182, a probability obtaining unit 183, and a classification result determination unit 184. Figure 5 The matrix expansion unit 181 is configured to expand the optimized classification feature matrix into a classification feature vector according to a row vector or a column vector. The fully connected coding unit 182 is configured to perform fully connected coding on the classification feature vector using the fully connected layer of the classifier to obtain a fully connected coding feature vector. The probability obtaining unit 183 is configured to pass the fully connected coding feature vector through the Softmax classification function of the classifier to obtain a first probability belonging to the normal operation state of the sewage environmental protection equipment and a second probability belonging to the abnormal operation state of the sewage environmental protection equipment. The classification result determination unit 184 is configured to determine the classification result based on the comparison between the first probability and the second probability.

[0075] In summary, the monitoring and alarm system of the environmental protection equipment according to the embodiments of the present application is illustrated, which uses deep learning technology to pre-process and clean the sewage discharge data, converts the data into a unified format, and then performs embedding coding and context semantic information extraction, and judges whether the operation state of the environmental protection equipment is normal based on this. In this way, the running abnormal problems of the environmental protection equipment can be found and solved in time, so as to ensure the compliance of sewage treatment and the effectiveness of environmental protection.

[0076] Exemplary method

[0077] Figure 6 This is a flowchart of a monitoring and alarm method for environmental protection equipment according to an embodiment of this application. Figure 6 As shown, the monitoring and alarm method for environmental protection equipment according to an embodiment of this application includes the following steps: S110, acquiring wastewater discharge data after treatment by the environmental protection equipment, wherein the wastewater discharge data includes discharge flow rate, pH value, chemical oxygen demand, ammonia nitrogen, and heavy metal content; S120, performing data cleaning on the wastewater discharge data to obtain multiple data items, wherein the multiple data items include data attributes and data values; S130, passing the data attributes of each data item in the multiple data items through a word embedding layer to obtain multiple data item attribute word embedding vectors; S140, passing the data values ​​of each data item in the multiple data items through a word embedding layer to obtain multiple data item attribute word embedding vectors. Adding to the end of the attribute word embedding vector of each data item to obtain multiple data item embedding vectors; S150, inputting the multiple data item embedding vectors into a context encoder based on a converter to obtain a semantic feature vector of wastewater discharge data; S160, performing feature enhancement on the semantic feature vector of wastewater discharge data based on a Gaussian density map to obtain a classification feature matrix; S170, optimizing the feature distribution of the classification feature matrix to obtain an optimized classification feature matrix; S180, passing the optimized classification feature matrix through a classifier to obtain a classification result, the classification result being used to indicate whether the wastewater environmental protection equipment is operating normally.

[0078] In a specific example, in the above-mentioned monitoring and alarm method for environmental protection equipment, step S130, which involves passing the data attributes of each data item in the plurality of data items through a word embedding layer to obtain multiple data item attribute word embedding vectors, includes: using the learnable embedding matrix of the word embedding layer to perform word embedding encoding on the data attributes of each data item in the plurality of data items to obtain multiple data item attribute word embedding vectors.

[0079] In one specific example, in the monitoring and alarming method of the environmental protection equipment, the step S150 of embedding the plurality of data items into a vector to input a transformer-based context encoder to obtain a sewage discharge data semantic feature vector includes: arranging the plurality of data items into a one-dimensional vector to obtain a data global embedding vector; calculating the product between the data global embedding vector and the transpose vector of each data item embedding vector in the plurality of data item embedding vectors to obtain a plurality of self-attention correlation matrices; respectively performing standardization processing on each self-attention correlation matrix in the plurality of self-attention correlation matrices to obtain a plurality of standardized self-attention correlation matrices; passing each standardized self-attention correlation matrix in the plurality of standardized self-attention correlation matrices through a classification function to obtain a plurality of probability values; respectively weighting each data item embedding vector in the plurality of data item embedding vectors with each probability value in the plurality of probability values to obtain a plurality of data item feature vectors; and concatenating the plurality of data item feature vectors to obtain the sewage discharge data semantic feature vector.

[0080] In one specific example, in the monitoring and alarming method of the environmental protection equipment, the step S160 of performing feature enhancement on the sewage discharge data semantic feature vector based on a Gaussian density map to obtain a classification feature matrix includes: constructing a Gaussian density map of the sewage discharge data semantic feature vector by using the following Gaussian formula; wherein the Gaussian formula is: wherein μ is the sewage discharge data semantic feature vector, and σ is the variance between the feature values of the corresponding two positions in the sewage discharge data semantic feature vector, wherein represents a Gaussian density probability function, and x represents the variable of the Gaussian density map; and performing Gaussian discretization processing on the Gaussian distribution of each position in the Gaussian density map to obtain the classification feature matrix.

[0081] In a specific example, in the above-mentioned monitoring and alarm method for environmental protection equipment, step S170, optimizing the feature distribution of the classification feature matrix to obtain an optimized classification feature matrix, includes: performing global mean pooling on the classification feature matrix to obtain a global semantic feature vector; calculating the covariance matrix of the global semantic feature vector; performing eigenvalue decomposition on the covariance matrix to obtain multiple eigenvalues ​​and multiple eigenvalue vectors corresponding to the multiple eigenvalues; extracting K eigenvalue vectors corresponding to the K largest eigenvalues ​​from the multiple eigenvalue vectors, the K eigenvalue vectors forming a k-dimensional homography subspace; expanding each row vector of the classification feature matrix to obtain multiple locally expanded classification feature vectors; mapping the multiple locally expanded classification feature vectors to the homography subspace to obtain multiple mapped locally expanded classification feature vectors; activating the multiple mapped locally expanded classification feature vectors using the Sigmoid activation function to obtain multiple mapped probabilistic locally expanded classification feature vectors; and reconstructing the dimensions of the multiple mapped probabilistic locally expanded classification feature vectors to obtain the optimized classification feature matrix.

[0082] In a specific example, in the above-mentioned monitoring and alarm method for environmental protection equipment, step S180, which involves passing the optimized classification feature matrix through a classifier to obtain a classification result, wherein the classification result is used to indicate whether the wastewater environmental protection equipment is operating normally, includes: expanding the optimized classification feature matrix into a classification feature vector according to row vectors or column vectors; using the fully connected layer of the classifier to perform fully connected encoding on the classification feature vector to obtain a fully connected encoded feature vector; passing the fully connected encoded feature vector through the Softmax classification function of the classifier to obtain a first probability of classifying the wastewater environmental protection equipment as operating normally and a second probability of classifying it as operating abnormally; and determining the classification result based on the comparison between the first probability and the second probability.

[0083] Here, those skilled in the art will understand that the specific operations of each step in the above-mentioned monitoring and alarm method for environmental protection equipment have been referenced above. Figures 1 to 5 The description of the monitoring and alarm system for environmental protection equipment is detailed here, and therefore, its repeated description will be omitted.

Claims

1. A monitoring alarm system for an environmental protection device, characterized by, The method comprises the following steps: A data acquisition module is used to acquire sewage discharge data after treatment by an environmental protection device, wherein the sewage discharge data comprises discharge flow, pH value, chemical oxygen demand, ammonia nitrogen and heavy metal content; A data cleaning module is used to clean the sewage discharge data to obtain a plurality of data items, wherein the plurality of data items comprise data attributes and data values; An attribute embedding encoding module is used to respectively embed the data attributes of each data item in the plurality of data items through a word embedding layer to obtain a plurality of data item attribute word embedding vectors; A data adding module is used to add the data values of each data item in the plurality of data items to the end of the attribute word embedding vector of the data item to obtain a plurality of data item embedding vectors; A sewage discharge data semantic understanding module is used to input the plurality of data item embedding vectors into a context encoder based on a converter to obtain a sewage discharge data semantic feature vector; A Gaussian enhancement module is used to enhance the sewage discharge data semantic feature vector based on a Gaussian density map to obtain a classification feature matrix; An optimization module is used to optimize the feature distribution of the classification feature matrix to obtain an optimized classification feature matrix; A monitoring result generation module is used to input the optimized classification feature matrix into a classifier to obtain a classification result, wherein the classification result is used to represent whether the operation state of the sewage environmental protection device is normal. The optimization module comprises: A global pooling unit is used to perform global mean pooling on the classification feature matrix to obtain a global semantic feature vector; A covariance matrix calculation unit is used to calculate the covariance matrix of the global semantic feature vector; An eigenvalue decomposition unit is used to perform eigenvalue decomposition on the covariance matrix to obtain a plurality of eigenvalues and a plurality of eigenvalue vectors corresponding to the plurality of eigenvalues; A homographic subspace construction unit is used to extract K eigenvalue vectors corresponding to K largest eigenvalues from the plurality of eigenvalue vectors, wherein the K eigenvalue vectors constitute a K-dimensional homographic subspace; An unfolding unit is used to unfold each row vector of the classification feature matrix to obtain a plurality of classification feature local unfolding feature vectors; A mapping unit is used to map the plurality of classification feature local unfolding feature vectors to the homographic subspace to obtain a plurality of mapped classification feature local unfolding feature vectors; A probabilistic unit is used to activate the plurality of mapped classification feature local unfolding feature vectors through a Sigmoid activation function to obtain a plurality of mapped probabilistic classification feature local unfolding feature vectors; A dimension reconstruction unit is used to reconstruct the dimensions of the plurality of mapped probabilistic classification feature local unfolding feature vectors to obtain an optimized classification feature matrix.

2. The monitoring alarm system for environmental protection equipment according to claim 1, wherein The attribute embedding encoding module is used to respectively embed the data attributes of each data item in the plurality of data items through a learnable embedding matrix of the word embedding layer to obtain the plurality of data item attribute word embedding vectors.

3. The monitoring alarm system for environmental protection equipment according to claim 2, wherein The sewage discharge data semantic understanding module comprises: A one-dimensional arrangement unit is used to arrange the plurality of data item embedding vectors in one dimension to obtain a data global embedding vector. a self-attention unit configured to calculate a product between the data global embedding vector and a transpose vector of each of the plurality of data item embedding vectors to obtain a plurality of self-attention correlation matrices; a normalization self-attention unit configured to respectively normalize each of the plurality of self-attention correlation matrices to obtain a plurality of normalized self-attention correlation matrices; an attention calculation unit configured to pass each of the plurality of normalized self-attention correlation matrices through a classification function to obtain a plurality of probability values; an attention application unit configured to respectively weight each of the plurality of data item embedding vectors with each of the plurality of probability values as a weight to obtain a plurality of data item feature vectors; a concatenation unit configured to concatenate the plurality of data item feature vectors to obtain the sewage discharge data semantic feature vector.

4. The monitoring alarm system for environmental protection equipment according to claim 3, wherein The Gaussian enhancement module comprises: a Gaussian density map construction unit configured to construct a Gaussian density map of the sewage discharge data semantic feature vector using a Gaussian formula as follows: ， wherein, is the sewage discharge data semantic feature vector, and is the variance between the feature values of the corresponding two positions in the sewage discharge data semantic feature vector, denotes a Gaussian density probability function, denotes a variable of the Gaussian density map; a Gaussian discretization unit configured to perform Gaussian discretization processing on Gaussian distribution of each position in the Gaussian density map to obtain the classification feature matrix.

5. The monitoring alarm system for environmental protection equipment according to claim 4, wherein The monitoring result generation module comprises: a matrix expansion unit configured to expand the optimized classification feature matrix into a classification feature vector according to a row vector or a column vector; a fully connected coding unit configured to perform fully connected coding on the classification feature vector using a fully connected layer of the classifier to obtain a fully connected coding feature vector; a probability obtaining unit configured to pass the fully connected coding feature vector through a Softmax classification function of the classifier to obtain a first probability belonging to a normal sewage environmental protection equipment operating state and a second probability belonging to an abnormal sewage environmental protection equipment operating state; a classification result determination unit configured to determine the classification result based on a comparison between the first probability and the second probability.

6. An environmental equipment monitoring alarm method, characterized by, comprises: obtaining sewage discharge data processed by an environmental protection equipment, wherein the sewage discharge data comprises discharge flow, pH value, chemical oxygen demand, ammonia nitrogen, and heavy metal content; performing data cleaning on the sewage discharge data to obtain a plurality of data items, wherein the plurality of data items comprise data attributes and data values; respectively passing data attributes of each of the plurality of data items through a word embedding layer to obtain a plurality of data item attribute word embedding vectors; respectively adding data values of each of the plurality of data items at the end of the data item attribute word embedding vector to obtain a plurality of data item embedding vectors; inputting the plurality of data item embedding vectors into a context encoder based on a converter to obtain a sewage discharge data semantic feature vector; performing feature enhancement on the sewage discharge data semantic feature vector based on a Gaussian density map to obtain a classification feature matrix; performing feature distribution optimization on the classification feature matrix to obtain an optimized classification feature matrix; passing the optimized classification feature matrix through a classifier to obtain a classification result, wherein the classification result is used to represent whether a sewage environmental protection equipment operating state is normal or not. The feature distribution optimization is performed on the classification feature matrix to obtain an optimized classification feature matrix, including: global mean pooling is performed on the classification feature matrix to obtain a global semantic feature vector; a covariance matrix of the global semantic feature vector is calculated; eigenvalue decomposition is performed on the covariance matrix to obtain a plurality of eigenvalues and a plurality of eigenvectors corresponding to the plurality of eigenvalues; K eigenvectors corresponding to K largest eigenvalues are extracted from the plurality of eigenvectors, and the K eigenvectors constitute a k-dimensional homographic subspace; each row vector of the classification feature matrix is unfolded to obtain a plurality of classification feature local unfolded feature vectors; the plurality of classification feature local unfolded feature vectors are respectively mapped to the homographic subspace to obtain a plurality of mapped classification feature local unfolded feature vectors; the plurality of mapped classification feature local unfolded feature vectors are activated by a Sigmoid activation function to obtain a plurality of mapped probabilistic classification feature local unfolded feature vectors; dimension reconstruction is performed on the plurality of mapped probabilistic classification feature local unfolded feature vectors to obtain an optimized classification feature matrix.

7. The method of claim 6, wherein the alarm is activated when the concentration of the pollutant exceeds the predetermined threshold. The data attributes of each data item in the plurality of data items are respectively input into a word embedding layer to obtain a plurality of data item attribute word embedding vectors, including: using a learnable embedding matrix of the word embedding layer to respectively perform word embedding coding on the data attributes of each data item in the plurality of data items to obtain the plurality of data item attribute word embedding vectors.

8. The method of claim 7, wherein the alarm is activated when the concentration of the pollutant exceeds the predetermined threshold. The plurality of data item embedding vectors are input into a transformer-based context encoder to obtain a sewage discharge data semantic feature vector, including: one-dimensional arrangement is performed on the plurality of data item embedding vectors to obtain a data global embedding vector; the product between the data global embedding vector and the transpose vector of each data item embedding vector in the plurality of data item embedding vectors is calculated to obtain a plurality of self-attention association matrices; standardization processing is respectively performed on each self-attention association matrix in the plurality of self-attention association matrices to obtain a plurality of standardized self-attention association matrices; each standardized self-attention association matrix in the plurality of standardized self-attention association matrices is input into a classification function to obtain a plurality of probability values; each probability value in the plurality of probability values is used as a weight to weight each data item embedding vector in the plurality of data item embedding vectors to obtain a plurality of data item feature vectors; the plurality of data item feature vectors are concatenated to obtain the sewage discharge data semantic feature vector.

9. The method of claim 8, wherein the alarm is activated when the concentration of the pollutant exceeds a predetermined threshold. Feature enhancement is performed on the sewage discharge data semantic feature vector based on a Gaussian density map to obtain a classification feature matrix, including: a Gaussian density map of the sewage discharge data semantic feature vector is constructed by using the following Gaussian formula; wherein the Gaussian formula is: ， wherein, is the sewage discharge data semantic feature vector, and is the variance between the feature values of the corresponding two positions in the sewage discharge data semantic feature vector, denotes a Gaussian density probability function, denotes a variable of the Gaussian density map; Gaussian discretization processing is performed on the Gaussian distribution of each position in the Gaussian density map to obtain the classification feature matrix.

Images