Intelligent sewage treatment monitoring method and system

Abstract

The invention discloses an intelligent sewage treatment monitoring method and system. The method comprises the steps of data acquisition, data preprocessing, establishment of a sewage treatment monitoring model and sewage treatment monitoring. The invention belongs to the field of sewage monitoring, and particularly relates to an intelligent sewage treatment monitoring method and system.According to the scheme, the periodic rule of sewage treatment monitoring data is extracted on the basis of periodic position coding, and the dynamic dependency relationship in the sewage treatment process is comprehensively captured on the basis of dynamic association; the overall influence of a context event chain is comprehensively considered through a global aggregation mechanism, and the accuracy of sewage treatment monitoring is improved; a label matrix is constructed based on construction of a feature distance matrix and sewage treatment target output, fine sewage treatment parameter changes can be captured, and the influence of sewage treatment input parameter changes on a sewage treatment monitoring result is more sensitively reflected; a loss function for sewage treatment is constructed based on matrix combination, a monitoring target is better adapted and output, and accurate prediction of abnormal sewage treatment conditions is realized.

Description

Technical Field

[0001] The present invention relates to the technical field of sewage monitoring, and in particular to an intelligent sewage treatment monitoring method and system. Background Art

[0002] Sewage treatment monitoring methods refer to technical solutions that use technical means to collect, analyze and evaluate key parameters, operating conditions, treatment effects and abnormal conditions in the sewage treatment process in real time or periodically. The core goal of these methods is to ensure the stable and efficient operation of the sewage treatment system through data monitoring and processing, and to achieve accurate prediction and control of water quality changes. However, general sewage treatment monitoring methods are unable to fully express the characteristics of sewage treatment events, have poor adaptability to sudden sewage treatment events, noise data and abnormal conditions, resulting in inaccurate monitoring results; general sewage treatment monitoring methods have problems with improper handling of differences between operating conditions, resulting in low accuracy of monitoring results. Summary of the invention

[0003] In view of the above situation, in order to overcome the defects of the prior art, the present invention provides an intelligent sewage treatment monitoring method and system. In view of the fact that general sewage treatment monitoring methods are unable to fully express the characteristics of sewage treatment events, have poor adaptability to sudden events, noise data and abnormal conditions in sewage treatment, and lead to inaccurate monitoring results, this scheme extracts the periodic law of sewage treatment monitoring data based on periodic position coding, and comprehensively captures the dynamic dependencies in the sewage treatment process based on dynamic association; through a global aggregation mechanism, the overall impact of the context event chain is comprehensively considered to improve the accuracy of sewage treatment monitoring; in view of the fact that general sewage treatment monitoring methods have improper handling of differences between operating states, resulting in low accuracy of monitoring results, this scheme is based on constructing a feature distance matrix and constructing a label matrix based on sewage treatment target output, which can capture subtle changes in sewage treatment parameters and more keenly reflect the impact of changes in sewage treatment input parameters on sewage treatment monitoring results; based on matrix combination, a loss function for sewage treatment is constructed to better adapt to the output monitoring target and achieve accurate prediction of abnormal sewage treatment conditions.

[0004] The technical solution adopted by the present invention is as follows: The present invention provides an intelligent sewage treatment monitoring method, which comprises the following steps:

[0005] Step S1: data collection;

[0006] Step S2: data preprocessing;

[0007] Step S3: Establishing a sewage treatment monitoring model;

[0008] Step S4: sewage treatment monitoring.

[0009] Furthermore, in step S1, the data collection is to collect historical sewage treatment monitoring data; the historical sewage treatment monitoring data includes sewage treatment input parameter data, sewage treatment process operation status data, timestamp, sewage treatment output result data and operation monitoring results; the operation monitoring results are used as data labels.

[0010] Furthermore, in step S2, the data preprocessing is to clean the collected data, convert the data, standardize the data and define key events; the data cleaning is to process missing values ​​and duplicate values; the data conversion is to convert the data into vector form; the standardization is based on the maximum and minimum normalization method to standardize the data; the key event definition specifically defines input parameter events, process control events, abnormal events and output result events; each event Expressed as ; and define the context event chain for , define the candidate event set for ; where v i is the event code; s i is the sensor data; o i is an operational behavior; i It is the result of operation monitoring; It is all the historical events in context; are candidate events, i.e., all possible operational monitoring results.

[0011] Furthermore, in step S3, the establishment of the sewage treatment monitoring model specifically includes the following steps:

[0012] Step S31: construct a parameter encoding layer; use Transformer to capture the fine-grained dependencies between event parameters, and introduce time and position encoding; specifically, embed each parameter into a d-dimensional vector, combine to obtain the initial input matrix X0, and capture the dependencies by stacking Transformers, expressed as: ; Event embedding is represented as: ; Add position encoding to express it as: ; ;in, and They are the Transformer layer outputs of the nth layer and the n-1th layer respectively; It is processed by the Transformer layer of the nth layer; is the final embedded representation of the event; is the tanh function; It is vector concatenation; It is the embedding of various parameters of the event; is the trainable weight matrix of the parameter encoding layer; is the initial input matrix after position encoding; is the periodic position encoding matrix; t is the timestamp, k is the feature dimension index; It is an element-by-element interaction;

[0013] Step S32: construct an association structure layer; specifically: construct a graph G=(V,E), where the node V represents an event and the edge E represents the time sequence between events; update the node embedding in the graph through a multi-hop gated graph neural network, and the node embedding update is expressed as: ; Finally embed V L It is expressed as: ; in, and The events Embedding representation of level l+1 and level l skip; is the embedding of neighbor node events; h is the neighbor node index; N(·) is the node neighbor set; is the attention weight between events; W is the weight matrix of the association structure layer, and b is the bias vector of the association structure layer; is the sigmoid activation function; It is the embedding of each event in the following event chain; is the embedding of the candidate event; yes Activation function;

[0014] Step S33: construct a state alignment layer; combine local and global matching mechanisms to calculate the correlation between the context event chain and the candidate event; local matching uses cosine similarity, expressed as: ; Event matching vector It is expressed as: ; Get the aggregated local matching vector , expressed as: ; ;in, It is an event Embedding is the L2 norm; is the element-wise product; softmax(·) is the softmax function; are all context events The set of matching vectors of ; is a linear transformation function; T is a matrix transpose;

[0015] Step S34: Constructing a state assessment layer; based on the global matching vector Predicted relevance score, expressed as: ; and get the regularized normalized correlation score , expressed as: ; Select the event with the highest score as the prediction result , expressed as: ; where s(·) is the relevance score; and is the trainable weight matrix of the state evaluation layer; and is the bias vector of the state assessment layer; is the total number of candidate events, is the j1th candidate event; is the regularization coefficient;

[0016] Step S35: loss function design; define the feature distance matrix as the pairwise distance matrix between input parameter features in the sewage treatment process, expressed as: ; Construct a label matrix based on the sewage treatment target output and add a regularization disturbance term, expressed as: ; Get the loss function of the sewage treatment system , expressed as: ; Calculate the gradient for back propagation, expressed as: ;in, is the feature distance matrix element; and are the eigenvalues ​​of the i1th and jth samples on the kth input dimension respectively; d is the number of dimensions; is the label matrix element; is the normalization coefficient, used to adjust the scale of label distance; and are the label values ​​of the i1th sample and the jth sample respectively; is the intensity coefficient; c is the disturbance amplitude; is the total number of samples; n is the total input dimension;

[0017] Step S36: model determination; a prediction threshold is set in advance, and the pre-treated historical sewage treatment monitoring data is randomly divided into a training set and a test set; when the model converges on the training set loss, the sewage treatment monitoring model training is completed; if the prediction accuracy of the trained sewage treatment monitoring model for the test set is higher than the prediction threshold, the sewage treatment monitoring model is established, otherwise the data set is re-divided, and the initial weight and bias training model are adjusted.

[0018] Furthermore, in step S4, the sewage treatment monitoring collects sewage treatment input parameter data, sewage treatment process operation status data and sewage treatment output result data in real time, and inputs them into the established sewage treatment monitoring model after pre-processing, and outputs the operation monitoring results output by the model as monitoring results of the real-time collected data.

[0019] The present invention provides an intelligent sewage treatment monitoring system, comprising a data acquisition module, a data preprocessing module, a sewage treatment monitoring model building module and a sewage treatment monitoring module;

[0020] The data acquisition module collects historical sewage treatment monitoring data and sends the data to the data preprocessing module;

[0021] The data preprocessing module performs data cleaning, data conversion, standardization and key event definition on the collected data; and sends the data to the sewage treatment monitoring model establishment module;

[0022] The sewage treatment monitoring model establishment module extracts the periodic law of sewage treatment monitoring data based on periodic position coding; comprehensively considers the overall impact of the context event chain through a global aggregation mechanism; constructs a label matrix based on the construction of a feature distance matrix and based on the sewage treatment target output; constructs a loss function for sewage treatment based on matrix combination; finally realizes the establishment of a sewage treatment monitoring model; and sends the data to the sewage treatment monitoring module;

[0023] The sewage treatment monitoring module implements sewage treatment monitoring on sewage collected in real time based on the established sewage treatment monitoring model.

[0024] The beneficial effects achieved by the present invention using the above scheme are as follows:

[0025] (1) In view of the fact that general sewage treatment monitoring methods are unable to fully express the characteristics of sewage treatment events and have poor adaptability to sudden sewage treatment events, noise data and abnormal operating conditions, resulting in inaccurate monitoring results, this scheme extracts the periodic laws of sewage treatment monitoring data based on periodic position coding and comprehensively captures the dynamic dependencies in the sewage treatment process based on dynamic association. The overall impact of the context event chain is comprehensively considered through a global aggregation mechanism to improve the accuracy of sewage treatment monitoring.

[0026] (2) In view of the problem that general sewage treatment monitoring methods have improper handling of differences between operating states, resulting in low accuracy of monitoring results, this scheme is based on constructing a feature distance matrix and a label matrix based on the sewage treatment target output. It can capture subtle changes in sewage treatment parameters and more sensitively reflect the impact of changes in sewage treatment input parameters on sewage treatment monitoring results. Based on the matrix combination, a loss function for sewage treatment is constructed to better adapt to the output monitoring target and achieve accurate prediction of abnormal sewage treatment conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 A schematic diagram of a flow chart of an intelligent sewage treatment monitoring method provided by the present invention;

[0028] Figure 2 A schematic diagram of an intelligent sewage treatment monitoring system provided by the present invention;

[0029] Figure 3 It is a schematic diagram of the process of step S3.

[0030] The accompanying drawings are used to provide further understanding of the present invention and constitute a part of the specification. They are used to explain the present invention together with the embodiments of the present invention and do not constitute a limitation of the present invention. DETAILED DESCRIPTION

[0031] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments; based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0032] In the description of the present invention, it is necessary to understand that terms such as “upper”, “lower”, “front”, “back”, “left”, “right”, “top”, “bottom”, “inside” and “outside” indicating directions or positional relationships are based on the directions or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the referred system or element must have a specific direction, be constructed and operated in a specific direction, and therefore cannot be understood as a limitation on the present invention.

[0033] Example 1, see Figure 1 The present invention provides an intelligent sewage treatment monitoring method, which comprises the following steps:

[0034] Step S1: data collection, collecting historical sewage treatment monitoring data;

[0035] Step S2: Data preprocessing, data cleaning, data conversion, standardization and key event definition of the collected data;

[0036] Step S3: Establish a sewage treatment monitoring model, extract the periodicity of sewage treatment monitoring data based on periodic position coding; comprehensively consider the overall impact of the context event chain through a global aggregation mechanism; construct a label matrix based on the construction of a feature distance matrix and based on the sewage treatment target output; construct a loss function for sewage treatment based on the matrix combination; and finally realize the establishment of a sewage treatment monitoring model;

[0037] Step S4: sewage treatment monitoring, based on the established sewage treatment monitoring model, sewage treatment data collected in real time is monitored.

[0038] Example 2, see Figure 1This embodiment is based on the above embodiment. In step S1, the historical sewage treatment monitoring data includes sewage treatment input parameter data, sewage treatment process operation status data, timestamp, sewage treatment output result data and operation monitoring results; the operation monitoring results are used as data labels; sewage treatment input parameter data include chemical substance concentration, pH value and inlet flow rate; sewage treatment process operation status data include aeration intensity, sludge concentration, dosage and water pump status; sewage treatment output result data include effluent water quality parameters, treatment efficiency and operation energy consumption; operation monitoring results include normal operation, mild abnormal operation and severe abnormal operation.

[0039] Example 3, see Figure 1 This embodiment is based on the above embodiment. In step S2, data cleaning is to process missing values ​​and duplicate values; the data conversion is to convert the data into vector form; the standardization process is to standardize the data based on the maximum and minimum normalization method; the key event definition is specifically to define input parameter events, process control events, abnormal events and output result events; each event Expressed as ; and define the context event chain for , define the candidate event set for ; where v i is the event code; s i is the sensor data; o i is an operational behavior; i It is the result of operation monitoring; It is all the historical events in context; are candidate events, i.e., all possible operational monitoring results.

[0040] Example 4, see Figure 1 and Figure 3 Based on the above embodiment, in step S3, establishing a sewage treatment monitoring model specifically includes the following steps:

[0041] Step S31: construct a parameter encoding layer; use Transformer to capture the fine-grained dependencies between event parameters, and introduce time and position encoding; specifically, embed each parameter into a d-dimensional vector, combine to obtain the initial input matrix X0, and capture the dependencies by stacking Transformers, expressed as: ; Event embedding is represented as: ; Add position encoding to express it as: ; ;in, and They are the Transformer layer outputs of the nth layer and the n-1th layer respectively; It is processed by the Transformer layer of the nth layer; is the final embedded representation of the event; is the tanh function; It is vector concatenation; It is the embedding of various parameters of the event; is the trainable weight matrix of the parameter encoding layer; is the initial input matrix after position encoding; is the periodic position encoding matrix; t is the timestamp, k is the feature dimension index; It is an element-by-element interaction; the parameter encoding layer is used to accurately capture the complex relationship between event parameters and events, and provide high-quality input to support efficient monitoring and abnormal prediction of the sewage treatment process;

[0042] Step S32: construct an association structure layer; specifically: construct a graph G=(V,E), where the node V represents an event and the edge E represents the time sequence between events; update the node embedding in the graph through a multi-hop gated graph neural network, and the node embedding update is expressed as: ; Finally embed V L It is expressed as: ; in, and The events Embedding representation of level l+1 and level l skip; is the embedding of neighbor node events; h is the neighbor node index; N(·) is the node neighbor set; is the attention weight between events; W is the weight matrix of the association structure layer, and b is the bias vector of the association structure layer; is the sigmoid activation function; It is the embedding of each event in the following event chain; is the embedding of the candidate event; yes Activation function; the association structure layer is used to capture the dynamic dependencies between sewage treatment events, enrich the context information of events, and enhance the adaptability to noise;

[0043] Step S33: construct a state alignment layer; combine local and global matching mechanisms to calculate the correlation between the context event chain and the candidate event; local matching uses cosine similarity, expressed as: ; Event matching vector It is expressed as: ; Get the aggregated local matching vector , expressed as: ; ;in, It is an event Embedding is the L2 norm; is the element-wise product; softmax(·) is the softmax function; are all context events The set of matching vectors of ; is a linear transformation function; T is a matrix transpose; the state alignment layer is used to measure the similarity between a single context event and a candidate event;

[0044] Step S34: Constructing a state assessment layer; based on the global matching vector Predicted relevance score, expressed as: ; and get the regularized normalized correlation score , expressed as: ; Select the event with the highest score as the prediction result , expressed as: ; where s(·) is the relevance score; and is the trainable weight matrix of the state evaluation layer; and is the bias vector of the state assessment layer; is the total number of candidate events, is the j1th candidate event; is the regularization coefficient; the state evaluation layer combines the context event chain with the correlation of the candidate events and outputs the final predicted operation monitoring result;

[0045] Step S35: loss function design;

[0046] Step S36: model determination; a prediction threshold is set in advance, and the pre-treated historical sewage treatment monitoring data is randomly divided into a training set and a test set; when the model converges on the training set loss, the sewage treatment monitoring model training is completed; if the prediction accuracy of the trained sewage treatment monitoring model for the test set is higher than the prediction threshold, the sewage treatment monitoring model is established, otherwise the data set is re-divided, and the initial weight and bias training model are adjusted.

[0047] By performing the above operations, the general sewage treatment monitoring methods are unable to fully express the characteristics of sewage treatment events, and have poor adaptability to sudden sewage treatment events, noise data and abnormal conditions, resulting in inaccurate monitoring results. This scheme extracts the periodic laws of sewage treatment monitoring data based on periodic position coding, and comprehensively captures the dynamic dependencies in the sewage treatment process based on dynamic association. The overall impact of the context event chain is comprehensively considered through the global aggregation mechanism to improve the accuracy of sewage treatment monitoring.

[0048] Example 5, see Figure 3 This embodiment is based on the above embodiment. In step S3, the loss function is specifically designed as follows:

[0049] The characteristic distance matrix is ​​defined as the pairwise distance matrix between the input parameter features in the sewage treatment process, expressed as: ; Construct a label matrix based on the sewage treatment target output and add a regularization disturbance term, expressed as: ; Get the loss function of the sewage treatment system , expressed as: ; Calculate the gradient for back propagation, expressed as: ;in, is the feature distance matrix element; and are the eigenvalues ​​of the i1th and jth samples on the kth input dimension respectively; d is the number of dimensions; is the label matrix element; is the normalization coefficient used to adjust the scale of label distance; and are the label values ​​of the i1th sample and the jth sample respectively; is the intensity coefficient; c is the disturbance amplitude; is the total number of samples; n is the total input dimension.

[0050] By performing the above operations, in order to address the problem that general sewage treatment monitoring methods do not properly handle the differences between operating states, resulting in low accuracy of monitoring results, this solution is based on constructing a feature distance matrix and a label matrix based on the sewage treatment target output. It can capture subtle changes in sewage treatment parameters and more sensitively reflect the impact of changes in sewage treatment input parameters on sewage treatment monitoring results. It also constructs a loss function for sewage treatment based on the matrix combination to better adapt to the output monitoring target and achieve accurate prediction of abnormal sewage treatment conditions.

[0051] Example 6, see Figure 1 This embodiment is based on the above embodiment. In step S4, sewage treatment monitoring collects sewage treatment input parameter data, sewage treatment process operation status data and sewage treatment output result data in real time, and inputs them into the established sewage treatment monitoring model after pre-processing, and outputs the operation monitoring results output by the model as the monitoring results of the real-time collected data; when the operation monitoring results are seriously abnormal, early warning processing is carried out on relevant personnel.

[0052] Embodiment 7, see Figure 2 , this embodiment is based on the above embodiment, and the present invention provides an intelligent sewage treatment monitoring system, including a data acquisition module, a data preprocessing module, a sewage treatment monitoring model establishment module and a sewage treatment monitoring module;

[0053] The data acquisition module collects historical sewage treatment monitoring data and sends the data to the data preprocessing module;

[0054] The data preprocessing module performs data cleaning, data conversion, standardization and key event definition on the collected data; and sends the data to the sewage treatment monitoring model establishment module;

[0055] The sewage treatment monitoring model establishment module extracts the periodic law of sewage treatment monitoring data based on periodic position coding; comprehensively considers the overall impact of the context event chain through a global aggregation mechanism; constructs a label matrix based on the construction of a feature distance matrix and based on the sewage treatment target output; constructs a loss function for sewage treatment based on matrix combination; finally realizes the establishment of a sewage treatment monitoring model; and sends the data to the sewage treatment monitoring module;

[0056] The sewage treatment monitoring module implements sewage treatment monitoring on sewage collected in real time based on the established sewage treatment monitoring model.

[0057] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.

[0058] While the embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that many changes, modifications, substitutions and variations can be made to the embodiments without departing from the principles and spirit of the invention.

[0059] The present invention and its embodiments are described above, and such description is not restrictive. The drawings show only one embodiment of the present invention, and the actual structure is not limited thereto. In short, if ordinary technicians in the field are inspired by it, without departing from the purpose of the invention, they can design a structure and embodiment similar to the technical solution without creativity, which should belong to the protection scope of the present invention.

Claims

1. An intelligent sewage treatment monitoring method, characterized in that: The method comprises the following steps: Step S1: data collection, collecting historical sewage treatment monitoring data; Step S2: data preprocessing; Step S3: Establishing a sewage treatment monitoring model; Step S4: sewage treatment monitoring, based on the established sewage treatment monitoring model, sewage treatment monitoring is performed on the sewage collected in real time; Step S3 includes step S31: constructing a parameter encoding layer; using Transformer to capture fine-grained dependencies between event parameters, and introducing time and position encoding; specifically, each parameter is embedded into a d-dimensional vector, combined to obtain the initial input matrix X0, and the dependency is captured by stacking Transformer, which is expressed as: ; Event embedding is represented as: ; Add position encoding to express it as: ; ;in, and They are the Transformer layer outputs of the nth layer and the n-1th layer respectively; It is processed by the Transformer layer of the nth layer; is the final embedded representation of the event; is the tanh function; It is vector concatenation; It is the embedding of various parameters of the event; is the trainable weight matrix of the parameter encoding layer; is the initial input matrix after position encoding; is the periodic position encoding matrix; t is the timestamp, k is the feature dimension index; It is an element-by-element interaction.

2. An intelligent sewage treatment monitoring method according to claim 1, characterized in that: In step S3, the establishment of the sewage treatment monitoring model further includes the following steps: Step S32: construct an association structure layer; specifically: construct a graph G=(V,E), where the node V represents an event and the edge E represents the time sequence between events; update the node embedding in the graph through a multi-hop gated graph neural network, and the node embedding update is expressed as: ; Finally embed V L It is expressed as: ; in, and The events Embedding representation of level l+1 and level l skip; is the embedding of neighbor node events; h is the neighbor node index; N(·) is the node neighbor set; is the attention weight between events; W is the weight matrix of the association structure layer, and b is the bias vector of the association structure layer; is the sigmoid activation function; It is the embedding of each event in the following event chain; is the embedding of the candidate event; yes Activation function; Step S33: construct a state alignment layer; combine local and global matching mechanisms to calculate the correlation between the context event chain and the candidate event; local matching uses cosine similarity, expressed as: ; Event matching vector It is expressed as: ; Get the aggregated local matching vector , expressed as: ; ;in, It is an event Embedding is the L2 norm; is the element-wise product; softmax(·) is the softmax function; are all context events The set of matching vectors of ; is a linear transformation function; T is a matrix transpose; Step S34: Constructing a state assessment layer; based on the global matching vector Predicted relevance score, expressed as: ; and get the regularized normalized correlation score , expressed as: ; Select the event with the highest score as the prediction result , expressed as: ; where s(·) is the relevance score; and is the trainable weight matrix of the state evaluation layer; and is the bias vector of the state assessment layer; is the total number of candidate events, is the j1th candidate event; is the regularization coefficient; Step S35: loss function design; Step S36: model determination; a prediction threshold is set in advance, and the pre-treated historical sewage treatment monitoring data is randomly divided into a training set and a test set; when the model converges on the training set loss, the sewage treatment monitoring model training is completed; if the prediction accuracy of the trained sewage treatment monitoring model for the test set is higher than the prediction threshold, the sewage treatment monitoring model is established, otherwise the data set is re-divided, and the initial weight and bias training model are adjusted.

3. An intelligent sewage treatment monitoring method according to claim 2, characterized in that: In step S35, the loss function is specifically designed as follows: The characteristic distance matrix is ​​defined as the pairwise distance matrix between the input parameter features in the sewage treatment process, expressed as: ; Construct a label matrix based on the sewage treatment target output and add a regularization disturbance term, expressed as: ; Get the loss function of the sewage treatment system , expressed as: ; Calculate the gradient for backpropagation, expressed as: ;in, is the feature distance matrix element; and are the eigenvalues ​​of the i1th and jth samples on the kth input dimension respectively; d is the number of dimensions; is the label matrix element; is the normalization coefficient used to adjust the scale of label distance; and are the label values ​​of the i1th sample and the jth sample respectively; is the strength coefficient; c is the disturbance amplitude; is the total number of samples; n is the total input dimension.

4. The intelligent sewage treatment monitoring method according to claim 1 is characterized in that: In step S1, the historical sewage treatment monitoring data includes sewage treatment input parameter data, sewage treatment process operation status data, timestamp, sewage treatment output result data and operation monitoring results; the operation monitoring results are used as data labels.

5. The intelligent sewage treatment monitoring method according to claim 1 is characterized in that: In step S4, the sewage treatment monitoring collects sewage treatment input parameter data, sewage treatment process operation status data and sewage treatment output result data in real time, inputs them into the established sewage treatment monitoring model after pre-processing, and outputs the operation monitoring results output by the model as the monitoring results of the real-time collected data.

6. An intelligent sewage treatment monitoring method according to claim 1, characterized in that: In step S2, the data preprocessing is to clean the collected data, convert the data, standardize the data and define key events; the data cleaning is to process missing values ​​and duplicate values; the data conversion is to convert the data into vector form; the standardization is to standardize the data based on the maximum and minimum normalization method; the key event definition is specifically to define input parameter events, process control events, abnormal events and output result events; each event Expressed as ; and define the context event chain for , define the candidate event set for ; Among them, v i is the event code; s i is the sensor data; o i is an operational behavior; i It is the result of operation monitoring; It is all the historical events in context; are candidate events, i.e., all possible operational monitoring results.

7. An intelligent sewage treatment monitoring system, used to implement an intelligent sewage treatment monitoring method as claimed in any one of claims 1 to 6, characterized in that: It includes data acquisition module, data preprocessing module, sewage treatment monitoring model building module and sewage treatment monitoring module; The data acquisition module collects historical sewage treatment monitoring data and sends the data to the data preprocessing module; The data preprocessing module performs data cleaning, data conversion, standardization and key event definition on the collected data; and sends the data to the sewage treatment monitoring model establishment module; The sewage treatment monitoring model building module extracts the periodicity of sewage treatment monitoring data based on periodic position coding; The overall impact of the contextual event chain is considered through a global aggregation mechanism; a label matrix is ​​constructed based on the feature distance matrix and the sewage treatment target output; a loss function for sewage treatment is constructed based on the matrix combination; and finally a sewage treatment monitoring model is established; and the data is sent to the sewage treatment monitoring module; The sewage treatment monitoring module implements sewage treatment monitoring on sewage collected in real time based on the established sewage treatment monitoring model.

Images