A method and system for monitoring industrial wastewater treatment based on LSTM

Abstract

An industrial wastewater treatment monitoring method and system based on LSTM, relating to the technical field of industrial wastewater treatment, setting water quality monitoring points in the current industrial wastewater treatment process, obtaining water quality data of each water quality monitoring point; analyzing historical water quality data of each water quality monitoring point in several monitoring cycles to obtain key water quality monitoring indicators; building a water quality prediction model based on a long short-term memory network LSTM, obtaining the estimated numerical time series of key water quality monitoring indicators of each water quality monitoring point in the current monitoring cycle; real-time monitoring of the prediction results of the water quality prediction model, retraining the water quality prediction model according to the prediction results, and performing overall early warning real-time monitoring at the same time, obtaining control item measures for the next monitoring cycle according to the overall early warning real-time monitoring results, and realizing accurate monitoring and early warning of wastewater treatment effects.

Description

Technical Field

[0001] The present invention relates to the technical field of industrial wastewater treatment, and in particular to an industrial wastewater treatment monitoring method and system based on LSTM. Background Art

[0002] The prior art CN117151557A "A quality monitoring method and system based on industrial wastewater monitoring data" obtains the pollution index and water transparency of each sub-area, and establishes the sewage impact index of the treatment area based on the sewage state index set. If it exceeds the impact threshold, the data in the pollution index set is used as sample data, and the change trend of the sewage impact index in the next monitoring cycle is predicted in combination with the sample data. If the obtained comprehensive prediction value exceeds the impact threshold, the abnormal indicators are screened out according to the abnormality of each pollution indicator, and the corresponding treatment plan is matched from the pre-prepared treatment plan library in combination with the abnormal indicators;

[0003] The prior art CN118817980A "Online rapid detection system and method for industrial wastewater quality" obtains water quality parameters in each water area and performs preprocessing, then extracts features of the water quality parameters in each water area, determines the weight assignment of each water area, and calculates the overall quality coefficient of the water area after weighted summation of the weight assignments in each water area; analyzes the changes in the area of ​​industrial wastewater water areas through a time series model to determine the abnormal change rate of the area of ​​industrial wastewater water areas; after analyzing the overall quality coefficient of the water area and the abnormal change rate of the area of ​​industrial wastewater water areas through a decision tree model, the industrial wastewater water areas are divided into different risk levels, and the water quality detection time of the industrial wastewater water areas is dynamically adjusted according to the risk level.

[0004] Industrial wastewater treatment is an important part of environmental protection. Traditional monitoring methods mainly rely on manual sampling and laboratory analysis, which have problems such as poor real-time performance, high cost, and low efficiency. In recent years, with the development of artificial intelligence technology, especially the application of deep learning technology, new solutions have been provided for industrial wastewater treatment monitoring. As a special recursive neural network, the long short-term memory network (LSTM) can effectively process time series data and is suitable for real-time monitoring and prediction of water quality parameters in the process of industrial wastewater treatment. Summary of the invention

[0005] The purpose of the present invention is to provide an industrial wastewater treatment monitoring method based on LSTM, which realizes accurate monitoring and early warning of wastewater treatment effects by real-time collection and analysis of water quality parameters in the wastewater treatment process, and comprises the following steps:

[0006] Step s1: setting water quality monitoring points in the current industrial wastewater treatment process and obtaining water quality data of each water quality monitoring point;

[0007] Step s2: Analyze the historical water quality data of each water quality monitoring point within several monitoring cycles to obtain key water quality monitoring indicators;

[0008] Step s3: construct a water quality prediction model based on the long short-term memory network LSTM to obtain the estimated numerical time series of key water quality monitoring indicators at each water quality monitoring point in the current monitoring period;

[0009] Step s4: Monitor the prediction results of the water quality prediction model in real time, retrain the water quality prediction model according to the prediction results, and perform overall early warning real-time monitoring at the same time, and obtain the control item measures for the next monitoring cycle according to the overall early warning real-time monitoring results.

[0010] Furthermore, in the current industrial wastewater treatment process, water quality monitoring points are set up, and the process of obtaining water quality data of each water quality monitoring point includes:

[0011] Obtain the process unit characteristics of the current industrial wastewater treatment equipment, and divide the industrial wastewater treatment process into several process subsequences according to the process unit characteristics;

[0012] Set water quality monitoring points in each process subsequence, and use data retrieval to obtain several water quality monitoring indicators at each water quality monitoring point according to the functional characteristics in the process unit characteristics of the corresponding process subsequence;

[0013] The water quality monitoring points obtain water quality data in real time according to the water quality monitoring indicators, mark the monitoring time, and set the monitoring cycle.

[0014] Furthermore, the process of analyzing the historical water quality data of each water quality monitoring point within several monitoring cycles and obtaining key water quality monitoring indicators includes:

[0015] Acquire historical water quality data of each water quality monitoring point within a number of monitoring cycles, and extract numerical time series corresponding to a number of water quality monitoring indicators in the historical water quality data;

[0016] According to the numerical time series corresponding to the several water quality monitoring indicators, the average numerical amplitude of the several water quality monitoring indicators and the cumulative time when the numerical values ​​of the several water quality monitoring indicators are not within the preset standard threshold range are obtained, and the stability coefficient of the several water quality monitoring indicators is obtained according to the average numerical amplitude and the cumulative time;

[0017] The water quality monitoring indicators whose stability coefficients are less than the preset stability coefficient threshold are marked as risk monitoring indicators, the cumulative number of times the water quality monitoring indicators at each water quality monitoring point are marked as risk monitoring indicators is obtained, and the water quality monitoring indicators whose corresponding cumulative number of times is greater than the preset cumulative number threshold are marked as key water quality monitoring indicators.

[0018] Furthermore, a water quality prediction model is constructed based on the long short-term memory network LSTM to obtain the estimated numerical time series of key water quality monitoring indicators at each water quality monitoring point in the current monitoring period, including:

[0019] A water quality prediction model is constructed based on the long short-term memory network LSTM, a training rotation chain list and a chain list length s of the training rotation chain list are constructed, the number of intervals between each historical monitoring cycle and the current monitoring cycle is obtained, each historical monitoring cycle is sorted in reverse order according to the number of intervals, and the numerical time series of key water quality monitoring indicators in s historical monitoring cycles are inserted into the training rotation chain list according to the sorting results;

[0020] Construct a training set and a test set according to the numerical time series of key water quality monitoring indicators in s historical monitoring cycles in the training rotation chain table, input the training set into the water quality prediction model for training until the loss function training is stable, and save the model parameters, test the water quality prediction model through the test set until it meets the preset requirements, and output the water quality prediction model;

[0021] The water quality prediction model is used to obtain the estimated numerical time series of key water quality monitoring indicators at each water quality monitoring point in the current monitoring period.

[0022] Furthermore, the prediction results of the water quality prediction model are monitored in real time, the water quality prediction model is retrained according to the prediction results, and the process of overall early warning and real-time monitoring includes:

[0023] A detection time point is set at the end timestamp of the monitoring cycle, and the estimated alignment coefficients of several key water quality monitoring indicators are obtained at the detection time point. The estimated alignment coefficients of each key water quality monitoring indicator are compared with a preset estimated alignment coefficient threshold. If there is a key water quality monitoring indicator whose estimated alignment coefficient is less than the estimated alignment coefficient threshold, the numerical time series sequence of the key water quality monitoring indicator of the current monitoring cycle is inserted into the training rotation chain table, and the water quality estimation model is retrained based on the training rotation chain table, and overall early warning real-time monitoring is performed;

[0024] If there is no key water quality monitoring indicator whose estimated alignment coefficient is less than the estimated alignment coefficient threshold, the estimated numerical time series of several quality monitoring indicators of each water quality monitoring point in the next monitoring period is obtained according to the water quality estimation model.

[0025] Furthermore, the process of overall early warning real-time monitoring includes:

[0026] Extracting time characteristics and spatial characteristics of the numerical time series of key water quality monitoring indicators of each water quality monitoring node in the current monitoring period, respectively, to obtain the time and space characteristic sequence of the key water quality monitoring indicators;

[0027] Building a water quality impact prediction model based on an RBF neural network, acquiring training data based on a training rotation chain table, and using the training data to train the water quality impact prediction model;

[0028] The numerical time series and spatiotemporal feature series of key water quality monitoring indicators are input into the trained water quality impact prediction model, and the water quality impact prediction information is obtained according to the output layer of the water quality impact prediction model.

[0029] Furthermore, in the specific implementation process, the time characteristics and spatial characteristics of the numerical time series of the key water quality monitoring indicators of each water quality monitoring node in the current monitoring period are extracted respectively, and the process of obtaining the spatiotemporal characteristic sequence of the key water quality monitoring indicators includes:

[0030] Obtain the assembly relationship and assembly sequence between various industrial wastewater treatment equipment in the industrial wastewater treatment process, and construct a processing directed graph according to several process subsequences and the assembly relationship and assembly sequence between various industrial wastewater treatment equipment;

[0031] Several process subsequences are used as nodes of the processing directed graph, and the assembly relationship and assembly order between various industrial wastewater equipment are used as the connection relationship between nodes. The processing directed graph is learned through the graph attention network, and the numerical time series of key water quality monitoring indicators of each water quality monitoring node is imported into the corresponding node;

[0032] The attention weights of the adjacent nodes on the target node are obtained through the attention mechanism, and the attention weights are distributed to each adjacent node to obtain the feature vectors of the target node and the adjacent nodes at each time point. For each time point, the feature vectors of the adjacent nodes are weighted summed using the attention weights to generate an aggregated feature representation of the target node. The aggregated feature representations at each time point are arranged in chronological order to form a spatiotemporal feature sequence.

[0033] Furthermore, the process of obtaining the control item measures for the next monitoring cycle based on the overall early warning real-time monitoring results includes:

[0034] Construct a control item database, extract the numerical time series sequence of key water quality monitoring indicators of each water quality monitoring node in the water quality impact prediction information, perform data retrieval on the numerical time series sequence of key water quality monitoring indicators of each water quality monitoring node in the control item database, and obtain the control item measures for the next monitoring cycle.

[0035] Compared with the prior art, the present invention has the following beneficial effects:

[0036] 1. Strong real-time performance: By real-time collection and analysis of water quality parameters in the wastewater treatment process, abnormal water quality can be discovered and handled in a timely manner.

[0037] 2. High accuracy: The estimation results based on the LSTM model can accurately reflect the trend of water quality changes and improve monitoring accuracy.

[0038] 3. High degree of automation: Through automated data processing and model training, human intervention is reduced and work efficiency is improved.

[0039] 4. Timely early warning: Through real-time monitoring of overall early warning, early warning can be issued in time before abnormal water quality occurs, and measures can be taken to avoid environmental risks. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 This is a schematic diagram of an industrial wastewater treatment monitoring method and system based on LSTM in an embodiment of the present application. DETAILED DESCRIPTION

[0041] The following is a clear and complete description of the technical solutions in the embodiments of the present application in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present application.

[0042] like Figure 1 As shown, an industrial wastewater treatment monitoring method based on LSTM includes the following steps:

[0043] Step s1: setting water quality monitoring points in the current industrial wastewater treatment process and obtaining water quality data of each water quality monitoring point;

[0044] Step s2: Analyze the historical water quality data of each water quality monitoring point within several monitoring cycles to obtain key water quality monitoring indicators;

[0045] Step s3: construct a water quality prediction model based on the long short-term memory network LSTM to obtain the estimated numerical time series of key water quality monitoring indicators at each water quality monitoring point in the current monitoring period;

[0046] Step s4: Monitor the prediction results of the water quality prediction model in real time, retrain the water quality prediction model according to the prediction results, and perform overall early warning real-time monitoring at the same time, and obtain the control item measures for the next monitoring cycle according to the overall early warning real-time monitoring results.

[0047] It should be further explained that, in the specific implementation process, water quality monitoring points are set up in the current industrial wastewater treatment process, and the process of obtaining water quality data at each water quality monitoring point includes:

[0048] Obtaining process unit characteristics of current industrial wastewater treatment equipment, the process unit characteristics include grids and screens, regulating tanks, sedimentation tanks, coagulation and flocculation, biological treatment, oxidation and reduction, membrane separation, disinfection, etc., and dividing the industrial wastewater treatment process into several process subsequences according to the process unit characteristics;

[0049] Set water quality monitoring points in each process subsequence, and use data retrieval to obtain several water quality monitoring indicators of each water quality monitoring point according to the functional characteristics in the process unit characteristics of the corresponding process subsequence. The water quality monitoring indicators include pH value, chemical oxygen demand (COD), biochemical oxygen demand (BOD), suspended solids (SS), ammonia nitrogen, etc.;

[0050] The water quality monitoring points obtain water quality data in real time according to the water quality monitoring indicators, mark the monitoring time, and set the monitoring cycle.

[0051] It should be further explained that, in the specific implementation process, the historical water quality data of each water quality monitoring point within several monitoring cycles are analyzed to obtain the key water quality monitoring indicators, including:

[0052] Obtain historical water quality data of each water quality monitoring point within several monitoring cycles, extract numerical time series corresponding to several water quality monitoring indicators in the historical water quality data, and preset standard threshold intervals corresponding to several water quality monitoring indicators of each process subsequence;

[0053] According to the numerical time series corresponding to the water quality monitoring indicators, the average numerical amplitude of the water quality monitoring indicators and the cumulative time when the values ​​of the water quality monitoring indicators are not within the standard threshold range are obtained, and the stability coefficients of the water quality monitoring indicators are obtained according to the average numerical amplitude and the cumulative time;

[0054] A stability coefficient threshold is preset, and water quality monitoring indicators with stability coefficients less than the stability coefficient threshold are marked as risk monitoring indicators. Statistical analysis is performed on the historical water quality data of each water quality monitoring point within several monitoring cycles to obtain the cumulative number of times the water quality monitoring indicators of each water quality monitoring point are marked as risk monitoring indicators, and water quality monitoring indicators with corresponding cumulative numbers greater than the preset cumulative number threshold are marked as key water quality monitoring indicators.

[0055] The calculation formula for obtaining the floating characteristic coefficients of several water quality monitoring indicators based on the average numerical amplitude and cumulative time is:

[0056]

[0057] in, Represents the floating characteristic coefficient of water quality monitoring index i, represents the average numerical range of water quality monitoring index i, represents the cumulative time of water quality monitoring index i, represents the conversion factor, , Represents the weight coefficient.

[0058] It should be further explained that, in the specific implementation process, the process of building a water quality prediction model based on the long short-term memory network LSTM and obtaining the estimated numerical time series of key water quality monitoring indicators at each water quality monitoring point in the current monitoring period includes:

[0059] A water quality prediction model is constructed based on the long short-term memory network LSTM, a training rotation chain list and a chain list length s of the training rotation chain list are constructed, the number of intervals between each historical monitoring cycle and the current monitoring cycle is obtained, and each historical monitoring cycle is sorted in reverse order according to the number of intervals. The fewer the number of intervals, the higher the ranking. According to the sorting results, the numerical time series of key water quality monitoring indicators in s historical monitoring cycles are inserted into the training rotation chain list;

[0060] A training set and a test set are constructed according to the numerical time series of key water quality monitoring indicators in s historical monitoring cycles in the training rotation chain table, and the hyperparameter settings of the water quality prediction model are obtained by random search. The random search finds the best hyperparameter settings by randomly sampling hyperparameter combinations, which is more efficient than grid search, especially when the hyperparameter space is large. The hyperparameter definition range of the present invention is a learning rate between [0.001, 0.1], a hidden layer size between [32, 128], a batch size between [16, 64], and a number of training rounds between [10, 50]. Random sampling is used to generate a specified number of hyperparameter combinations. For each randomly generated hyperparameter combination, a water quality prediction model is trained and its performance is evaluated. Cross-validation is used to evaluate the generalization ability of the model, and the hyperparameter combination with the best performance is selected. The training set is then input into the water quality prediction model for training until the loss function training is stable, and the model parameters are saved. The water quality prediction model is tested by the test set until it meets the preset requirements, and the water quality prediction model is output;

[0061] The water quality prediction model is used to obtain the estimated numerical time series of key water quality monitoring indicators at each water quality monitoring point in the current monitoring period.

[0062] It should be further explained that, in the specific implementation process, the prediction results of the water quality prediction model are monitored in real time, the water quality prediction model is retrained according to the prediction results, and the overall early warning real-time monitoring process includes:

[0063] A detection time point is set at the end timestamp of the monitoring cycle, and at the detection time point, the numerical time series sequences of several key water quality monitoring indicators at each water quality monitoring point are compared with the corresponding estimated numerical time series sequences to obtain estimated alignment coefficients of several key water quality monitoring indicators, and the estimated alignment coefficients of each key water quality monitoring indicator are compared with a preset estimated alignment coefficient threshold. If there is a key water quality monitoring indicator whose estimated alignment coefficient is less than the estimated alignment coefficient threshold, the numerical time series sequences of the key water quality monitoring indicators of the current monitoring cycle are inserted into the training rotation chain list, and the water quality estimation model is retrained based on the training rotation chain list, and overall early warning real-time monitoring is performed;

[0064] Among them, the calculation formula for obtaining the estimated alignment coefficient of several key water quality monitoring indicators is:

[0065]

[0066] in, represents the estimated alignment coefficient of the key water quality monitoring indicator k, represents the value of the key water quality monitoring indicator k at the tth moment, It represents the estimated value of the key water quality monitoring indicator k at the tth moment, and n represents the total number of moments.

[0067] If there is no key water quality monitoring indicator whose estimated alignment coefficient is less than the estimated alignment coefficient threshold, the estimated numerical time series of several quality monitoring indicators of each water quality monitoring point in the next monitoring period is obtained according to the water quality estimation model.

[0068] It should be further explained that, in the specific implementation process, a control item database is constructed, and the historical water quality data of each process subsequence in several historical monitoring cycles and the corresponding control item measures are matched and stored in the control item database. If there is no key water quality monitoring indicator with an estimated alignment coefficient less than the estimated alignment coefficient threshold, after obtaining the estimated numerical time series sequence of several quality monitoring indicators of each water quality monitoring point in the next monitoring period according to the water quality prediction model, the estimated numerical time series sequence of several quality monitoring indicators of each water quality monitoring point is used to perform data retrieval in the control item database to obtain the control item measures for the next monitoring period. The control item data specifically include dosage, aeration volume, pH adjustment, sludge return ratio, reactor operating parameters, etc.

[0069] It should be further explained that, in the specific implementation process, the process of overall early warning and real-time monitoring includes:

[0070] Extracting time characteristics and spatial characteristics of the numerical time series of key water quality monitoring indicators of each water quality monitoring node in the current monitoring period, respectively, to obtain the time and space characteristic sequence of the key water quality monitoring indicators;

[0071] Building a water quality impact prediction model based on an RBF neural network, acquiring training data based on a training rotation chain table, and using the training data to train the water quality impact prediction model;

[0072] The numerical time series and spatiotemporal feature series of key water quality monitoring indicators of each water quality monitoring node are input into the trained water quality impact prediction model, and the water quality impact prediction information is obtained according to the output layer of the water quality impact prediction model.

[0073] It should be further explained that, in the specific implementation process, the time characteristics and spatial characteristics of the numerical time series of the key water quality monitoring indicators of each water quality monitoring node in the current monitoring period are extracted respectively, and the process of obtaining the spatiotemporal characteristic sequence of the key water quality monitoring indicators includes:

[0074] Obtain the assembly relationship and assembly sequence between various industrial wastewater treatment equipment in the industrial wastewater treatment process, and construct a processing directed graph according to several process subsequences and the assembly relationship and assembly sequence between various industrial wastewater treatment equipment;

[0075] Several process subsequences are used as nodes of the processing directed graph, and the assembly relationship and assembly order between various industrial wastewater equipment are used as the connection relationship between nodes. The processing directed graph is learned through the graph attention network, and the numerical time series of key water quality monitoring indicators of each water quality monitoring node is imported into the corresponding node;

[0076] The attention weights of the adjacent nodes on the target node are obtained through the attention mechanism, and the attention weights are distributed to each adjacent node to obtain the feature vectors of the target node and the adjacent nodes at each time point. For each time point, the feature vectors of the adjacent nodes are weighted summed using the attention weights to generate an aggregated feature representation of the target node. The aggregated feature representations at each time point are arranged in chronological order to form a spatiotemporal feature sequence.

[0077] It should be further explained that in the specific implementation process, the attention weight of the neighboring node j to the target node i is obtained The calculation formula is:

[0078]

[0079] in, () represents the activation function, which is used to introduce nonlinear characteristics. Represents the feature vector and The concatenation operation of , a represents a learnable vector, represents a learnable weight matrix, represents the feature vector of target node i, Represents the feature vector of the adjacent node j; the feature vectors of the adjacent nodes are weighted summed using the attention weights, and the calculation formula for generating the new feature vector of the target node is:

[0080]

[0081] in, represents an activation function, represents the set of adjacent nodes j.

[0082] It should be further explained that, in the specific implementation process, the process of obtaining the control item measures for the next monitoring cycle based on the overall early warning real-time monitoring results includes:

[0083] Construct a control item database, extract the numerical time series sequence of key water quality monitoring indicators of each water quality monitoring node in the water quality impact prediction information, perform data retrieval on the numerical time series sequence of key water quality monitoring indicators of each water quality monitoring node in the control item database, and obtain the control item measures for the next monitoring cycle.

[0084] The above embodiments are only used to illustrate the technical method of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical method of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical method of the present invention.

Claims

1. A method for monitoring industrial wastewater treatment based on LSTM, characterized in that: The following steps are involved: Step s1: setting water quality monitoring points in the current industrial wastewater treatment process and obtaining water quality data of each water quality monitoring point; Step s2: Analyze the historical water quality data of each water quality monitoring point within several monitoring cycles to obtain key water quality monitoring indicators, including: Acquire historical water quality data of each water quality monitoring point within a number of monitoring cycles, and extract numerical time series corresponding to a number of water quality monitoring indicators in the historical water quality data; According to the numerical time series corresponding to the several water quality monitoring indicators, the average numerical amplitude of the several water quality monitoring indicators and the cumulative time when the numerical values ​​of the several water quality monitoring indicators are not within the preset standard threshold range are obtained, and the stability coefficient of the several water quality monitoring indicators is obtained according to the average numerical amplitude and the cumulative time; Mark the water quality monitoring indicators whose stability coefficients are less than the preset stability coefficient threshold as risk monitoring indicators, obtain the cumulative number of times the water quality monitoring indicators at each water quality monitoring point are marked as risk monitoring indicators, and mark the water quality monitoring indicators whose corresponding cumulative number is greater than the preset cumulative number threshold as key water quality monitoring indicators; Step s3: Build a water quality prediction model based on the long short-term memory network LSTM to obtain the estimated numerical time series of key water quality monitoring indicators at each water quality monitoring point in the current monitoring period, including: A water quality prediction model is constructed based on the long short-term memory network LSTM, a training rotation chain list and a chain list length s of the training rotation chain list are constructed, the number of intervals between each historical monitoring cycle and the current monitoring cycle is obtained, each historical monitoring cycle is sorted in reverse order according to the number of intervals, and the numerical time series of key water quality monitoring indicators in s historical monitoring cycles are inserted into the training rotation chain list according to the sorting results; Construct a training set and a test set according to the numerical time series of key water quality monitoring indicators in s historical monitoring cycles in the training rotation chain table, input the training set into the water quality prediction model for training until the loss function training is stable, and save the model parameters, test the water quality prediction model through the test set until it meets the preset requirements, and output the water quality prediction model; The estimated numerical time series of key water quality monitoring indicators at each water quality monitoring point in the current monitoring period are obtained through the water quality prediction model; Step s4: Monitor the prediction results of the water quality prediction model in real time, retrain the water quality prediction model according to the prediction results, and perform overall early warning real-time monitoring at the same time, and obtain the control item measures for the next monitoring cycle according to the overall early warning real-time monitoring results.

2. The industrial wastewater treatment monitoring method based on LSTM according to claim 1 is characterized in that: In the current industrial wastewater treatment process, water quality monitoring points are set up, and the process of obtaining water quality data at each water quality monitoring point includes: Obtain the process unit characteristics of the current industrial wastewater treatment equipment, and divide the industrial wastewater treatment process into several process subsequences according to the process unit characteristics; Set water quality monitoring points in each process subsequence, and use data retrieval to obtain several water quality monitoring indicators at each water quality monitoring point according to the functional characteristics in the process unit characteristics of the corresponding process subsequence; The water quality monitoring points obtain water quality data in real time according to the water quality monitoring indicators, mark the monitoring time, and set the monitoring cycle.

3. The industrial wastewater treatment monitoring method based on LSTM according to claim 2 is characterized in that: The process of real-time monitoring of the prediction results of the water quality prediction model, retraining the water quality prediction model based on the prediction results, and performing overall early warning real-time monitoring includes: A detection time point is set at the end timestamp of the monitoring cycle, and the estimated alignment coefficients of several key water quality monitoring indicators are obtained at the detection time point. The estimated alignment coefficients of each key water quality monitoring indicator are compared with a preset estimated alignment coefficient threshold. If there is a key water quality monitoring indicator whose estimated alignment coefficient is less than the estimated alignment coefficient threshold, the numerical time series sequence of the key water quality monitoring indicator of the current monitoring cycle is inserted into the training rotation chain table, and the water quality estimation model is retrained based on the training rotation chain table, and overall early warning real-time monitoring is performed; If there is no key water quality monitoring indicator whose estimated alignment coefficient is less than the estimated alignment coefficient threshold, the estimated numerical time series of several quality monitoring indicators of each water quality monitoring point in the next monitoring period is obtained according to the water quality estimation model.

4. The industrial wastewater treatment monitoring method based on LSTM according to claim 3 is characterized in that: The process of conducting overall early warning real-time monitoring includes: Extracting time characteristics and spatial characteristics of the numerical time series of key water quality monitoring indicators of each water quality monitoring node in the current monitoring period, respectively, to obtain the time and space characteristic sequence of the key water quality monitoring indicators; Building a water quality impact prediction model based on an RBF neural network, acquiring training data based on a training rotation chain table, and using the training data to train the water quality impact prediction model; The numerical time series and spatiotemporal feature series of key water quality monitoring indicators are input into the trained water quality impact prediction model, and the water quality impact prediction information is obtained according to the output layer of the water quality impact prediction model.

5. The industrial wastewater treatment monitoring method based on LSTM according to claim 4 is characterized in that: In the specific implementation process, the time characteristics and spatial characteristics of the numerical time series of the key water quality monitoring indicators of each water quality monitoring node in the current monitoring period are extracted respectively, and the process of obtaining the spatiotemporal characteristic sequence of the key water quality monitoring indicators includes: Obtain the assembly relationship and assembly sequence between various industrial wastewater treatment equipment in the industrial wastewater treatment process, and construct a processing directed graph according to several process subsequences and the assembly relationship and assembly sequence between various industrial wastewater treatment equipment; Several process subsequences are used as nodes of the processing directed graph, and the assembly relationship and assembly order between various industrial wastewater equipment are used as the connection relationship between nodes. The processing directed graph is learned through the graph attention network, and the numerical time series of key water quality monitoring indicators of each water quality monitoring node is imported into the corresponding node; The attention weights of the adjacent nodes on the target node are obtained through the attention mechanism, and the attention weights are distributed to each adjacent node to obtain the feature vectors of the target node and the adjacent nodes at each time point. For each time point, the feature vectors of the adjacent nodes are weighted summed using the attention weights to generate an aggregated feature representation of the target node. The aggregated feature representations at each time point are arranged in chronological order to form a spatiotemporal feature sequence.

6. The industrial wastewater treatment monitoring method based on LSTM according to claim 5 is characterized in that: The process of obtaining the control item measures for the next monitoring cycle based on the overall early warning real-time monitoring results includes: Construct a control item database, extract the numerical time series sequence of key water quality monitoring indicators of each water quality monitoring node in the water quality impact prediction information, perform data retrieval on the numerical time series sequence of key water quality monitoring indicators of each water quality monitoring node in the control item database, and obtain the control item measures for the next monitoring cycle.

Images