Sewage treatment detection method and system with diagnosis function

Abstract

The present application provides a sewage treatment detection method and system with a diagnosis function, which first receives multi-source detection data uploaded by each processing unit in the sewage treatment process, the multi-source detection data including sewage pollutant component data, processing equipment operation parameter data and bacterial population activity data in the reaction tank, then performs cross-unit correlation mapping processing on the multi-source detection data to generate a material conversion correlation network between each processing unit, constructs a dynamic balance model of the sewage treatment system based on the network, inputs the real-time collected multi-source detection data into the model, analyzes abnormal correlation nodes deviating from the balance state and corresponding abnormal influence factors, and finally generates a diagnosis report containing an abnormal treatment path and equipment control instructions according to the analysis result, so as to accurately locate the root cause of the sewage treatment system problem, timely adjust the equipment parameters to restore the system balance, and improve the treatment efficiency and stability.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of sewage treatment, in particular to a sewage treatment detection method and system with a diagnosis function. BACKGROUND

[0002] In the field of sewage treatment, ensuring the stable and efficient operation of the sewage treatment system is the key to protecting environmental quality and sustainable use of water resources. Currently, detection devices are usually set up in each treatment unit during sewage treatment to obtain relevant data information of the sewage. However, the existing sewage treatment detection method has many limitations.

[0003] On the one hand, the detection data of each treatment unit is often collected and analyzed independently, lacking in-depth mining of the data correlation between different treatment units. For example, only the removal effect of pollutants in a single reaction tank is focused on, while the influence of the operating parameters of the previous treatment unit equipment on the microbial activity and pollutant treatment efficiency of the reaction tank is ignored, resulting in the inability to fully understand the complex process of material conversion and energy flow in the entire sewage treatment system.

[0004] On the other hand, when the sewage treatment system is abnormal, the existing detection method is difficult to quickly and accurately locate the root cause of the problem. Due to the lack of modeling and analysis of the overall dynamic balance relationship of the system, the operation and maintenance personnel can only check each treatment unit one by one based on experience, which is not only inefficient, but also easy to miss key problems, and cannot take effective treatment measures in time, thereby affecting the normal operation and treatment effect of the sewage treatment system, and even may cause sewage discharge exceeding the standard, causing serious pollution to the environment. SUMMARY

[0005] In view of the above-mentioned problems, in combination with the first aspect of the present application, the embodiments of the present application provide a sewage treatment detection method with a diagnosis function, which comprises:

[0006] receiving multi-source detection data uploaded by each treatment unit in the sewage treatment process, the multi-source detection data including pollutant composition data of the sewage, operating parameter data of the treatment equipment, and microbial activity data in the reaction tank;

[0007] performing cross-unit correlation mapping processing on the multi-source detection data to generate a material conversion correlation network between each treatment unit, the material conversion correlation network being used to reflect the mutual influence relationship between pollutants, equipment operating parameters and microbial activity in different treatment units;

[0008] constructing a dynamic balance model of the sewage treatment system based on the material conversion correlation network, the dynamic balance model being used to reflect the parameter matching relationship of each treatment unit in the stable operation state of the sewage treatment system;

[0009] Input the real-time collected multi-source detection data into the dynamic balance model, and analyze the abnormal correlation nodes deviating from the balance state and the corresponding abnormal influence factors, the abnormal correlation nodes are the nodes in which the parameters are unbalanced in the processing unit, and the abnormal influence factors are the pollutants or equipment operation parameters causing the parameter imbalance;

[0010] Generate a diagnostic report containing an abnormal processing path and an equipment control instruction according to the abnormal correlation nodes and the abnormal influence factors, and the equipment control instruction is used to adjust the equipment operation parameters of the corresponding processing unit to restore the system balance.

[0011] In another aspect, the embodiment of the present application also provides a sewage treatment detection system with a diagnosis function, which comprises a processor and a machine readable storage medium, the machine readable storage medium is connected with the processor, the machine readable storage medium is used to store programs, instructions or codes, and the processor is used to execute the programs, instructions or codes in the machine readable storage medium to realize the above-mentioned method.

[0012] Based on the above aspects, the embodiment of the present application can receive the multi-source detection data uploaded by each processing unit in the sewage treatment process, perform cross-unit correlation mapping processing on the multi-source detection data, generate a material conversion correlation network between each processing unit, and deeply reveal the complex mutual influence relationship between the pollutants, equipment operation parameters and microbial activity in different processing units. The dynamic balance model of the sewage treatment system constructed based on the material conversion correlation network can accurately reflect the parameter matching relationship of each processing unit in the stable operation state. The dynamic balance model can be input with the real-time collected multi-source detection data, and the abnormal correlation nodes deviating from the balance state and the corresponding abnormal influence factors can be quickly analyzed, so that the accurate positioning of the problem source is realized. The diagnostic report containing the abnormal processing path and the equipment control instruction generated according to the analysis result can guide the operation and maintenance personnel to take effective treatment measures in time, adjust the equipment operation parameters of the corresponding processing unit to restore the system balance, significantly improve the operation stability and processing efficiency of the sewage treatment system, and reduce the risk of sewage exceeding the standard. BRIEF DESCRIPTION OF DRAWINGS

[0013] Figure 1 is the execution flow diagram of the sewage treatment detection method with a diagnosis function provided by the embodiment of the present application.

[0014] Figure 2 is the schematic diagram of the exemplary hardware and software components of the sewage treatment detection system with a diagnosis function provided by the embodiment of the present application. DETAILED DESCRIPTION

[0015] The present application will be specifically described below in conjunction with the drawings of the specification, Figure 1is a flowchart of a sewage treatment detection method with a diagnostic function provided by an embodiment of the present application. The sewage treatment detection method with a diagnostic function will be described in detail below.

[0016] The present embodiment takes a municipal domestic sewage treatment plant as an application scenario. The sewage treatment plant comprises a grid treatment unit, a regulating pool treatment unit, a primary sedimentation tank treatment unit, an A / O biological reaction tank treatment unit, a secondary sedimentation tank treatment unit, a depth filtration treatment unit, and an ultraviolet disinfection treatment unit connected in sequence. Each treatment unit is configured with a sensor group and a data transmission module for real-time collection and uploading of data.

[0017] Step S110: receiving multi-source detection data uploaded by each treatment unit in the sewage treatment process, wherein the multi-source detection data comprises pollutant composition data of sewage, operation parameter data of treatment equipment, and bacterial community activity data in the reaction tank.

[0018] The central control system of the municipal domestic sewage treatment plant is connected with the data transmission modules of each treatment unit through industrial Ethernet. The sensor group of the grid treatment unit collects pollutant composition data such as the suspended solids content of sewage after flowing through the grid, the amount of grid machine intercepts, and operation parameter data such as the control parameters corresponding to the running frequency and the grid spacing of the grid machine, and uploads them to the central control system through the data transmission module.

[0019] The sensor group of the regulating pool treatment unit collects pollutant composition related data such as the pH value, water temperature, and water volume of the sewage in the pool, and operation parameter data such as the running speed of the agitator and the liquid level control parameters of the regulating pool, and uploads them through the data transmission module.

[0020] The data uploaded by the primary sedimentation tank treatment unit includes pollutant composition data such as the turbidity and suspended solids concentration of the pool effluent, and operation parameter data such as the running period of the mud scraper and the flow of the sludge pump.

[0021] The A / O biological reaction tank treatment unit, as the core treatment unit, uploads more abundant data. The pollutant composition data includes the chemical oxygen demand, ammonia nitrogen, total nitrogen, and total phosphorus concentrations in the aerobic and anoxic zones; the operation parameter data includes the wind pressure and aeration quantity of the aerobic zone aeration blower, the power of the anoxic zone agitator, and the flow of the reflux pump; and the bacterial community activity data is collected by the biological sensor arranged in the tank, including the types of nitrifying bacteria, aerobic heterotrophic bacteria in the aerobic zone, and denitrifying bacteria in the anoxic zone, as well as the activity related data such as the cell concentration and respiration rate of each bacterial community.

[0022] The pollutant composition data uploaded by the secondary sedimentation tank treatment unit includes the suspended solids concentration and transparency of the effluent, and the operation parameter data includes the running speed of the mud scraper and the sludge reflux ratio.

[0023] The turbidity, COD concentration and other pollutant composition data of the filtered water on the deep filtration processing unit, and the backwashing cycle, backwashing water volume and other operation parameter data of the filter tank.

[0024] The pollutant composition data uploaded by the ultraviolet disinfection processing unit is the fecal coliform count of the effluent, and the operation parameter data includes the power of the ultraviolet lamp, the water flow rate and the like.

[0025] The receiving module of the central control system analyzes the data uploaded by each processing unit, and stores them according to the number of the processing unit and the type of the data. The data storage format adopts a structured data table, and each data record contains the data acquisition time, the processing unit identification, the data type identification and the data value set.

[0026] Step S120: performing cross-unit correlation mapping processing on the multi-source detection data to generate a substance conversion correlation network between the processing units, the substance conversion correlation network being used to reflect the mutual influence relationship between the pollutants, the equipment operation parameters and the bacterial activity in different processing units.

[0027] Step S121: extracting a characteristic pollutant index in the pollutant composition data of each processing unit from the multi-source detection data, the characteristic pollutant index being a pollutant type and concentration that changes significantly in the processing process.

[0028] For the grid processing unit, the pollutant composition data of the influent and the effluent are analyzed, and the removal rates of various pollutants are calculated. The removal rate is calculated by the method of (influent concentration-effluent concentration) / influent concentration. Pollutants with removal rates exceeding a set proportion, such as suspended solids, are selected and determined as the characteristic pollutant index of the grid processing unit, and the concentration data of the pollutant at the influent and the effluent are recorded.

[0029] The characteristic pollutant index of the conditioning tank processing unit is the pH value and the water temperature, because these two indexes will change significantly in the conditioning tank to adapt to the subsequent processing process, and the pH value and the water temperature data at different times need to be extracted.

[0030] The primary sedimentation tank processing unit mainly removes settleable suspended solids and part of organic matter. By comparing the influent and effluent data, the suspended solids and the five-day biochemical oxygen demand are taken as the characteristic pollutant index, and the corresponding concentration data are extracted.

[0031] In the A / O biological reaction tank processing unit, the ammonia nitrogen in the aerobic zone is converted into nitrate nitrogen through nitrification, the nitrate nitrogen in the anoxic zone is converted into nitrogen gas through denitrification, and the chemical oxygen demand is degraded in both the aerobic zone and the anoxic zone. Therefore, ammonia nitrogen, total nitrogen and chemical oxygen demand are determined as the characteristic pollutant index of the processing unit, and the concentration data of the influent in the aerobic zone, the effluent in the aerobic zone, the influent in the anoxic zone and the effluent in the anoxic zone are extracted respectively.

[0032] The characteristic pollutant index of the secondary sedimentation tank treatment unit is suspended solids, and the concentration data of the effluent thereof is extracted.

[0033] The depth filtration treatment unit further removes fine suspended solids and part of dissolved organic matter, and the turbidity and chemical oxygen demand are taken as the characteristic pollutant indexes, and the concentration data before and after filtration is extracted.

[0034] The characteristic pollutant index of the ultraviolet disinfection treatment unit is fecal coliform count, and the quantity data before and after disinfection is extracted.

[0035] Step S122: Extract the key operating parameters in the operating parameter data of each treatment unit, which are the equipment operating parameters that directly affect the pollutant treatment effect.

[0036] Step S1221: Obtain all operating parameters and corresponding parameter values in the operating parameter data of each treatment unit.

[0037] The operating parameters of the grid treatment unit include the operating frequency of the grid machine, the control parameters corresponding to the grid spacing, the motor current, etc., and the specific values of these parameters at different times are extracted from the uploaded data.

[0038] The operating parameters of the conditioning tank treatment unit include the operating speed of the agitator, the conditioning tank liquid level control parameters, the inlet valve opening, the outlet valve opening, etc., and the parameter values of each parameter are extracted.

[0039] The operating parameters of the primary sedimentation tank treatment unit include the operating cycle of the mud scraper, the flow of the sludge pump, the sludge discharge frequency of the sludge hopper, etc., and the corresponding parameter values are obtained.

[0040] The operating parameters of the A / O biological reaction tank treatment unit include the air pressure of the aeration fan in the aerobic zone, the aeration amount, the power of the agitator in the anoxic zone, the flow of the reflux pump, etc., and the specific values of these parameters are extracted.

[0041] The operating parameters of the secondary sedimentation tank treatment unit include the operating speed of the mud scraper, the sludge reflux ratio, the residual sludge discharge amount, etc., and the parameter values are extracted.

[0042] The operating parameters of the depth filtration treatment unit include the backwashing cycle of the filter tank, the backwashing water amount, the filtration flow rate, etc., and the corresponding parameter values are obtained.

[0043] The operating parameters of the ultraviolet disinfection treatment unit include the power of the ultraviolet lamp, the water flow rate, the lamp cleanliness parameter, etc., and the parameter values of each parameter are extracted.

[0044] Step S1222: Collect the pollutant treatment effect data corresponding to each operating parameter at different values, which includes the pollutant removal rate and the pollutant concentration after treatment.

[0045] For the grid machine operating frequency of the grid treatment unit, the removal rate of suspended solids and the concentration of suspended solids after treatment are collected at different operating frequencies. At the same time, for the control parameters corresponding to the grid spacing, the pollutant treatment effect data at different parameter values are collected.

[0046] Adjust the agitator operating speed parameter of the conditioning tank treatment unit, and collect the treatment effect data such as the stability of pH value (reflected by the fluctuation range of pH value) and the uniformity of water temperature (reflected by the difference of water temperature at different monitoring points) at different speeds.

[0047] For the sludge scraper operating cycle parameter of the primary sedimentation tank treatment unit, the removal rate of suspended solids and the concentration of suspended solids after treatment are collected at different cycles. For the sludge pump flow parameter, the pollutant treatment effect data at different flow rates are collected.

[0048] For the aeration amount parameter of the aerobic zone of the A / O biological reaction tank treatment unit, the removal rate of ammonia nitrogen and the concentration of ammonia nitrogen after treatment are collected at different aeration amounts. For the agitator power parameter of the anoxic zone, the removal rate of total nitrogen is collected at different powers. For the reflux pump flow parameter, the total nitrogen treatment effect data at different flow rates are collected.

[0049] The operating parameters of other treatment units are collected in a similar manner to obtain the corresponding pollutant treatment effect data.

[0050] Step S1223: Calculate the correlation coefficient between each operating parameter and the pollutant treatment effect data, which is used to represent the degree of linear correlation between the two.

[0051] Taking the grid machine operating frequency of the grid treatment unit and the removal rate of suspended solids as an example, the parameter values at different operating frequencies and the corresponding removal rate of suspended solids are combined to form two data sequences. Using the correlation coefficient calculation method, the correlation coefficient is obtained by dividing the covariance of the two data sequences by the product of the standard deviations of the two data sequences.

[0052] For the agitator operating speed of the conditioning tank treatment unit and the fluctuation range of pH value, the correlation coefficient is also calculated after forming data sequences to reflect the degree of linear correlation between the two.

[0053] According to the same method, the correlation coefficient between each operating parameter and the corresponding pollutant treatment effect data in each treatment unit is calculated.

[0054] Step S1224: Select the operating parameters with an absolute value of the correlation coefficient exceeding a preset correlation threshold as candidate key operating parameters.

[0055] A preset correlation threshold is set, for the grid processing unit, if the absolute value of the correlation coefficient between the grid machine operating frequency and the suspended solids removal rate exceeds the threshold, the grid machine operating frequency is determined as a candidate key operating parameter; if the absolute value of the correlation coefficient between the control parameter corresponding to the grid spacing and the pollutant treatment effect does not exceed the threshold, it is not included in the candidate range.

[0056] The operating parameters of each processing unit are screened one by one to obtain all candidate key operating parameters.

[0057] Step S1225: Sensitivity analysis is performed on the candidate key operating parameters, the change amount of the pollutant treatment effect when the parameter value changes by one unit is calculated, and a sensitivity coefficient is obtained.

[0058] Taking the aeration amount of the aerobic zone of the A / O biological reaction tank processing unit as an example, under the condition that other parameters remain unchanged, the parameter value of the aeration amount is increased by one unit, and the change amount of the ammonia nitrogen removal rate at this time is recorded. The change amount is the sensitivity coefficient of the parameter at the current value.

[0059] The above operation is repeated in different parameter value intervals to obtain the sensitivity coefficients of the candidate key operating parameter in different intervals, and the average value is taken as the final sensitivity coefficient.

[0060] In the same way, the sensitivity coefficients of all candidate key operating parameters are calculated.

[0061] Step S1226: The sensitivity coefficients are sorted in descending order, and the operating parameters with high ranking are selected as key operating parameters. The key operating parameters are the device operating parameters that directly affect the pollutant treatment effect.

[0062] The sensitivity coefficients of the candidate key operating parameters of the A / O biological reaction tank processing unit are sorted, and if the sensitivity coefficient of the aeration amount of the aerobic zone is ranked first, it is determined as the key operating parameter of the processing unit.

[0063] Other processing units also select the candidate key operating parameters with high ranking as the final key operating parameters in this way. For example, the grid processing unit selects the grid machine operating frequency, the adjustment tank processing unit selects the agitator operating speed, and the primary sedimentation tank processing unit selects the mud scraper operating cycle.

[0064] Step S123: Extract the dominant bacteria group index in the bacteria group activity data in the reaction tank. The dominant bacteria group index is the type and activity of the bacteria group that plays a leading role in the pollutant degradation process.

[0065] Step S1231: Obtain all the bacterial species in the bacterial activity data in the reaction tank and the corresponding activity values, including the number of bacterial species and the metabolic rate.

[0066] The aerobic zone and the anoxic zone of the A / O biological reaction tank treatment unit are both provided with biosensors. Through gene sequencing and metabolic monitoring technology, it is determined that there are nitrifying bacteria (including ammonia-oxidizing bacteria and nitrite-oxidizing bacteria), aerobic heterotrophic bacteria and other bacterial species in the aerobic zone; and there are denitrifying bacteria and other bacterial species in the anoxic zone.

[0067] The number of each bacterial species is extracted from the bacterial activity data, such as the number of cells per milliliter of mixed solution; and the metabolic rate, such as the ammonia oxidation rate of ammonia-oxidizing bacteria, the nitrate reduction rate of denitrifying bacteria, and other activity values.

[0068] Step S1232: Analyze the degradation correlation degree of each bacterial species and the characteristic pollutant index, which is calculated by the ratio of the degradation rate of the characteristic pollutant when the bacterial species exists and the degradation rate when the bacterial species does not exist.

[0069] For ammonia-oxidizing bacteria in the aerobic zone, the corresponding characteristic pollutant index is ammonia nitrogen. Under experimental conditions, the degradation rate of ammonia nitrogen in the aerobic zone of the A / O biological reaction tank when ammonia-oxidizing bacteria exist, and the degradation rate of ammonia nitrogen when ammonia-oxidizing bacteria do not exist through sterile treatment are measured. The ratio of the two degradation rates is taken as the degradation correlation degree of ammonia-oxidizing bacteria and ammonia nitrogen.

[0070] For denitrifying bacteria, the corresponding characteristic pollutant index is total nitrogen. Similarly, under experimental conditions, the degradation rates of total nitrogen when denitrifying bacteria exist and do not exist are measured, and the ratio of the two is calculated to obtain the degradation correlation degree of denitrifying bacteria and total nitrogen.

[0071] According to the same method, the degradation correlation degree of aerobic heterotrophic bacteria and the characteristic pollutant index such as chemical oxygen demand is calculated.

[0072] Step S1233: Screen out bacterial species with a degradation correlation degree exceeding a preset correlation threshold value as candidate dominant bacterial species, and calculate the proportion of the activity value of each candidate dominant bacterial species in the total value of all bacterial activities to obtain the activity proportion.

[0073] The preset correlation threshold value is set. If the degradation correlation degree of ammonia-oxidizing bacteria and ammonia nitrogen exceeds the threshold value, ammonia-oxidizing bacteria is included in the candidate dominant bacterial species; if the degradation correlation degree of denitrifying bacteria and total nitrogen exceeds the threshold value, denitrifying bacteria is also included in the candidate dominant bacterial species.

[0074] The total activity value of the candidate dominant flora is calculated, for example, the number and metabolic rate of ammonia-oxidizing bacteria are added to the corresponding activity values of other candidate dominant flora to obtain the total activity value. The proportion of the activity value of each candidate dominant flora in the total activity value, i.e., the activity proportion, is calculated, such as the proportion of the number of ammonia-oxidizing bacteria in the total number of all candidate dominant flora and the proportion of the metabolic rate in the total metabolic rate, and the activity proportion of ammonia-oxidizing bacteria is comprehensively obtained.

[0075] Step S1234: The candidate dominant flora whose activity proportion exceeds the preset activity proportion threshold value is determined as the dominant flora type, and the activity value corresponding to the dominant flora type is extracted as the dominant flora activity data.

[0076] The preset activity proportion threshold value is set, and if the activity proportion of ammonia-oxidizing bacteria exceeds the threshold value, it is determined as the dominant flora type, and the number and metabolic rate and other activity values are extracted as the dominant flora activity data.

[0077] The denitrifying bacteria and the aerobic heterotrophic bacteria with high correlation to chemical oxygen demand degradation and an activity proportion exceeding the threshold value are both determined as the dominant flora type, and the corresponding activity values are extracted.

[0078] Step S1235: The dominant flora type and the corresponding dominant flora activity data are combined to obtain the dominant flora index in the bacterial flora activity data in the reaction tank, and the dominant flora index is the flora type and activity that plays a leading role in the pollutant degradation process.

[0079] The determined dominant flora type, such as ammonia-oxidizing bacteria, nitrite-oxidizing bacteria, denitrifying bacteria, and specific aerobic heterotrophic bacteria, is combined with the corresponding activity value (number, metabolic rate) to form the dominant flora index of the A / O biological reaction tank treatment unit.

[0080] These dominant flora indexes can clearly reflect the situation of the flora that plays a leading role in the pollutant degradation process, for example, ammonia-oxidizing bacteria dominate the preliminary oxidation of ammonia nitrogen, and denitrifying bacteria dominate the reduction of nitrate nitrogen.

[0081] Step S124: A treatment unit correlation matrix is established, the rows of the treatment unit correlation matrix represent the previous treatment unit, the columns represent the subsequent treatment unit, and the matrix elements represent the influence degree of the output material of the previous treatment unit on the input material of the subsequent treatment unit.

[0082] The treatment unit correlation matrix is constructed based on each treatment unit of the urban sewage treatment plant. Rows of the matrix are in turn a grid treatment unit, a regulating pool treatment unit, a primary sedimentation tank treatment unit, an A / O biological reaction tank treatment unit, a secondary sedimentation tank treatment unit, and a depth filtration treatment unit; columns are in turn the regulating pool treatment unit, the primary sedimentation tank treatment unit, the A / O biological reaction tank treatment unit, the secondary sedimentation tank treatment unit, the depth filtration treatment unit, and the ultraviolet disinfection treatment unit.

[0083] The value of the matrix element is determined by analyzing the influence degree of the output material of the previous treatment unit on the input material of the subsequent treatment unit. For example, the output material of the grid treatment unit enters the regulating pool treatment unit, and the influence degree thereof on the input material of the regulating pool treatment unit is embodied by the efficiency of the grid treatment unit in removing suspended solids. The higher the removal efficiency, the greater the influence degree, and the greater the value of the corresponding element in the matrix.

[0084] Similarly, the influence degree of the regulating pool treatment unit on the primary sedimentation tank treatment unit is determined by the stability of the regulating pool treatment unit in regulating water quality and quantity; the influence degree of the primary sedimentation tank treatment unit on the A / O biological reaction tank treatment unit is determined by the effect of the primary sedimentation tank treatment unit in removing suspended solids and organic matter, and so on, to determine the values of the elements in the treatment unit correlation matrix.

[0085] Step S125: inputting the characteristic pollutant index, the key operation parameter, and the dominant bacterial population index as nodes into the treatment unit correlation matrix, and calculating the correlation strength value between the nodes in different treatment units, the correlation strength value being a quantitative value of the mutual influence between two nodes.

[0086] The characteristic pollutant index (such as suspended solids of the grid treatment unit, ammonia nitrogen of the A / O biological reaction tank treatment unit, etc.), the key operation parameter (such as the running frequency of the grid machine, the aeration amount of the aerobic zone, etc.), and the dominant bacterial population index (such as ammonia-oxidizing bacteria, denitrifying bacteria, etc.) of each treatment unit are taken as independent nodes.

[0087] These nodes are input into the treatment unit correlation matrix. For two nodes in different treatment units, such as the suspended solids node of the grid treatment unit and the chemical oxygen demand node of the A / O biological reaction tank treatment unit, the correlation strength value is calculated by analyzing the material conversion relationship and data correlation between the two nodes.

[0088] The calculation of the correlation strength value comprehensively considers the influence degree of the corresponding treatment unit in the treatment unit correlation matrix and the correlation coefficient between the parameters represented by the two nodes, and obtains the correlation strength value through weighted combination, so as to quantify the mutual influence degree between the two nodes.

[0089] Step S126: According to the correlation strength values, connect the nodes according to the size of the correlation strength values to form a preliminary material conversion correlation network, and optimize the nodes of the preliminary material conversion correlation network, merge the nodes with correlation strength values below the preset threshold, and retain the nodes with correlation strength values above the preset threshold and the connection relationship.

[0090] A preliminary material conversion correlation network is formed by connecting the nodes with correlation strength values above the threshold with line segments. For example, the suspended solids node of the grid treatment unit and the suspended solids node of the primary sedimentation tank treatment unit have high correlation strength values, and the two are connected. The aeration quantity node of the aerobic zone of the A / O biological reaction tank treatment unit and the ammonia-oxidizing bacteria node have high correlation strength values, and the two are connected.

[0091] For nodes with correlation strength values below the preset threshold, such as the turbidity node of the depth filtration treatment unit and the fecal coliform bacteria node of the ultraviolet disinfection treatment unit, since the degree of mutual influence is low, these nodes are merged for simplifying the network structure.

[0092] The nodes with correlation strength values above the preset threshold and the connection relationship are retained to form the basic structure of the optimized material conversion correlation network.

[0093] Step S127: Add the identifiers of the corresponding treatment units to the optimized material conversion correlation network to generate a final material conversion correlation network between the treatment units, which is used to reflect the mutual influence relationship between pollutants, equipment operating parameters, and microbial activity in different treatment units.

[0094] In the optimized material conversion correlation network, each node is added with the identifier of the corresponding treatment unit. For example, the suspended solids node is labeled as “grid treatment unit-suspended solids” and “primary sedimentation tank treatment unit-suspended solids”. The aeration quantity node of the aerobic zone is labeled as “A / O biological reaction tank treatment unit-aeration quantity of aerobic zone”. The ammonia-oxidizing bacteria node is labeled as “A / O biological reaction tank treatment unit-ammonia-oxidizing bacteria”.

[0095] By adding the identifiers of the treatment units, the final generated material conversion correlation network can clearly reflect the mutual influence relationship between pollutants, equipment operating parameters, and microbial activity in different treatment units, such as the connection between the “A / O biological reaction tank treatment unit-aeration quantity of aerobic zone” node and the “A / O biological reaction tank treatment unit-ammonia-oxidizing bacteria” node, which reflects the influence of the aeration quantity of the aerobic zone on the activity of ammonia-oxidizing bacteria, and the activity of ammonia-oxidizing bacteria affects the degradation effect of ammonia nitrogen.

[0096] Step S130: based on the substance conversion correlation network, a dynamic balance model of the sewage treatment system is constructed, and the dynamic balance model is used to reflect the parameter matching relationship of each treatment unit of the sewage treatment system in a stable operation state.

[0097] Step S131: a plurality of sets of substance conversion correlation networks and corresponding parameter data sets of each treatment unit of the sewage treatment system in a stable operation period are collected, and the parameter data set includes stable values of pollutant composition data, operation parameter data and microbial activity data.

[0098] A plurality of stable operation periods of a municipal sewage treatment plant are selected, and each period lasts for a certain length of time. In each stable operation period, the state of the substance conversion correlation network is recorded once every fixed time interval, including the connection relationship and correlation strength value of each node.

[0099] At the same time, the parameter data of each treatment unit in the corresponding period is collected, such as the stable value of suspended solids concentration of the grid treatment unit, the stable value of grid machine operation frequency, the stable value of ammonia nitrogen concentration of the A / O biological reaction tank treatment unit, the stable value of aeration quantity in the aerobic zone, the stable value of the number of ammonia oxidizing bacteria, etc., to form a parameter data set.

[0100] The substance conversion correlation network in each stable operation period is associated with the corresponding parameter data set to form a plurality of training samples.

[0101] Step S132: node feature extraction is performed on each of the substance conversion correlation networks to obtain a feature vector of each node, and the feature vector includes the pollutant concentration, operation parameter value and microbial activity value corresponding to the node.

[0102] For each node in each substance conversion correlation network, such as the "grid treatment unit-suspended solids" node, the stable value of suspended solids concentration in the corresponding stable operation period is extracted; for the "A / O biological reaction tank treatment unit-aeration quantity in the aerobic zone" node, the stable value of aeration quantity in the aerobic zone is extracted; for the "A / O biological reaction tank treatment unit-ammonia oxidizing bacteria" node, the stable values of the number and metabolic rate of ammonia oxidizing bacteria are extracted.

[0103] The extracted values are arranged in a predetermined order to form a feature vector of each node. For example, the feature vector of the "A / O biological reaction tank treatment unit-ammonia oxidizing bacteria" node can be composed of the stable value of the number of ammonia oxidizing bacteria and the stable value of the metabolic rate, which is a two-dimensional numerical combination.

[0104] Step S133: the node feature vector and the parameter data set in the same stable operation period are associated and aligned to establish a mapping relationship between the node features and the stable parameters.

[0105] In the same stable operation period, the feature vector of each node is matched with the corresponding parameter value in the parameter data set. For example, the feature vector (ammonia nitrogen concentration stable value) of the "A / O biological reaction tank treatment unit-ammonia nitrogen" node is associated with the ammonia nitrogen stable value of the A / O biological reaction tank treatment unit in the parameter data set; the feature vector of the "A / O biological reaction tank treatment unit-aerated amount of aerobic zone" node is associated with the stable value of the aerated amount of the aerobic zone in the parameter data set.

[0106] Through the above association alignment, the correspondence between each value in the node feature vector and the specific parameter in the parameter data set is clear, thereby establishing a mapping relationship between the node features and the stable parameters, and ensuring the correspondence of the input and output during model training.

[0107] Step S134: training the node feature vector and the mapping relationship after association alignment using a graph neural network algorithm to construct an initial dynamic balance model, the initial dynamic balance model being capable of outputting a corresponding parameter matching relationship according to an input material conversion correlation network.

[0108] Step S1341: inputting the node feature vector after association alignment into the input layer of the graph neural network, and performing dimension conversion processing on the node feature vector to match the dimension of the node feature vector with the dimension of the hidden layer of the graph neural network.

[0109] The input layer of the graph neural network receives the feature vectors of the nodes, which can have different dimensions. For example, the feature vectors of some nodes are two-dimensional, and the feature vectors of some nodes are three-dimensional. The input layer performs dimension conversion on these feature vectors through matrix transformation to convert them into vectors with consistent dimensions with the hidden layer.

[0110] For example, if the dimension of the hidden layer is a specific value, the input layer will expand the two-dimensional feature vectors and compress the three-dimensional feature vectors to ensure that all vectors entering the hidden layer have the same dimension, facilitating subsequent processing.

[0111] Step S1342: in the hidden layer of the graph neural network, the connection relationship between nodes is described by an adjacency matrix, and the node feature vectors are aggregated based on the connection relationship to obtain aggregated feature vectors containing neighbor node information.

[0112] In the hidden layer, an adjacency matrix is constructed according to the structure of the material conversion correlation network, and the elements in the adjacency matrix represent whether there is a connection between the nodes and the connection strength. For each node, the hidden layer collects the feature vectors of all its neighbor nodes.

[0113] For example, the neighbor nodes of the "A / O biological reaction tank treatment unit-ammonia oxidizing bacteria" node can include the "A / O biological reaction tank treatment unit-aerated amount of aerobic zone" node and the "A / O biological reaction tank treatment unit-ammonia nitrogen" node. The hidden layer combines the feature vectors of these neighbor nodes and the feature vector of the node itself according to the correlation strength to obtain an aggregated feature vector, which contains the comprehensive information of the node and its neighbor nodes.

[0114] Step S1343: Perform nonlinear activation processing on the aggregated feature vector to generate an activated feature vector, and input the activated feature vector into an output layer of the graph neural network. The activated feature vector is processed by a fully connected layer to output a parameter matching relationship prediction value of each treatment unit, which contains a pollutant concentration matching range, an operating parameter matching interval, and a microbial community activity matching interval.

[0115] A nonlinear activation function is applied to the aggregated feature vector to make the feature vector have stronger expression ability and generate an activated feature vector. The activated feature vector is input into the output layer. The fully connected layer of the output layer multiplies the activated feature vector by a preset weight matrix and adds a bias term to obtain a parameter matching relationship prediction value after processing.

[0116] For example, for the A / O biological reaction tank treatment unit, the output parameter matching relationship prediction value can include an ammonia nitrogen concentration matching range, an aerated amount of aerobic zone matching interval, an ammonia oxidizing bacteria activity matching interval, and the like. These intervals represent the reasonable matching range between parameters in a stable operating state.

[0117] Step S1344: Calculate a loss value of the parameter matching relationship prediction value and an actual parameter matching relationship, and adjust the weight parameters of the graph neural network using a backpropagation algorithm to minimize the loss value.

[0118] The parameter matching relationship prediction value obtained by the output layer is compared with the actual parameter matching relationship after correlation alignment. The difference between the two is calculated by a loss function to obtain a loss value. The loss function can comprehensively consider the deviation degree of each parameter matching interval.

[0119] Using the backpropagation algorithm, the gradient of the loss value to the weight parameters of each layer is calculated layer by layer from the output layer. The weight parameters are adjusted according to the gradient direction to reduce the loss value. This process is repeated until the loss value reaches a relatively small level.

[0120] Step S1345: Set a training iteration number threshold. When the training iteration number reaches the threshold, stop training and determine the current graph neural network model as an initial dynamic equilibrium model.

[0121] In the model training process, a training iteration threshold is set. Each iteration uses a set of training samples for forward calculation and backward propagation adjustment. When the number of iterations reaches the set threshold, the training is stopped regardless of whether the loss value reaches the minimum, and the graph neural network model at this time is saved as the initial dynamic equilibrium model.

[0122] Step S135: Select a part of the stable running period of the material conversion correlation network and the parameter data set as the verification set, input the initial dynamic equilibrium model, and obtain the parameter matching relationship prediction result.

[0123] From the collected stable running period data, a part of the material conversion correlation network and the corresponding parameter data set not involved in the training are randomly selected as the verification set. The material conversion correlation network in the verification set is input into the initial dynamic equilibrium model, and the model outputs the parameter matching relationship prediction result of each processing unit.

[0124] Step S136: Calculate the deviation value of the parameter matching relationship prediction result and the actual parameter matching relationship in the verification set. If the deviation value exceeds the preset deviation threshold, adjust the inter-layer connection weight of the graph neural network algorithm.

[0125] Compare the parameter matching relationship prediction result of the verification set with the actual parameter matching relationship, and calculate the deviation value between them. The deviation value can be calculated by comparing the coincidence degree of each parameter matching interval, the center value deviation, etc.

[0126] If the deviation value exceeds the preset deviation threshold, it means that the prediction accuracy of the initial dynamic equilibrium model is not enough, and the connection weight between each layer of the graph neural network needs to be adjusted. The adjustment method is similar to the back propagation in the training process, but only uses the verification set data for fine-tuning.

[0127] Step S137: Repeat the model training and parameter adjustment process until the deviation value is less than or equal to the preset deviation threshold, and obtain the final dynamic equilibrium model of the sewage treatment system. The dynamic equilibrium model is used to reflect the parameter matching relationship of each processing unit of the sewage treatment system in the stable running state.

[0128] Apply the adjusted weight parameters to the graph neural network, retrain using the training set, and verify again using the verification set. Repeat this process, continuously adjust the weight parameters, until the deviation value of the verification set is less than or equal to the preset deviation threshold.

[0129] The graph neural network model obtained at this time is the final dynamic equilibrium model, which can accurately reflect the parameter matching relationship between the pollutant concentration, operating parameter, and microbial activity of each processing unit of the sewage treatment system in the stable running state.

[0130] Step S140: input the real-time collected multi-source detection data into the dynamic balance model, and analyze the abnormal correlation nodes deviating from the balance state and the corresponding abnormal influence factors. The abnormal correlation node is a node in the processing unit that appears parameter imbalance, and the abnormal influence factor is a pollutant or equipment operation parameter that causes parameter imbalance.

[0131] Step S141: perform node mapping processing on the real-time collected multi-source detection data, and correspond the data to each node of the material conversion correlation network to generate a real-time material conversion correlation network.

[0132] The central control system receives the multi-source detection data uploaded by each processing unit in real time, such as the suspended solids concentration and the grid machine running frequency uploaded by the grid processing unit in real time; the ammonia nitrogen concentration, the aerobic zone aeration amount, and the ammonia oxidizing bacteria activity data uploaded by the A / O biological reaction tank processing unit in real time.

[0133] According to the definition of each node in the material conversion correlation network, the real-time collected data is mapped to the corresponding node. For example, the real-time ammonia nitrogen concentration data is mapped to the “A / O biological reaction tank processing unit-ammonia nitrogen” node, and the real-time aerobic zone aeration amount data is mapped to the “A / O biological reaction tank processing unit-aerobic zone aeration amount” node to form a real-time material conversion correlation network. The node values in the real-time material conversion correlation network are all real-time data at the current time.

[0134] Step S142: input the real-time material conversion correlation network into the dynamic balance model to obtain the parameter matching relationship standard value of each processing unit.

[0135] The real-time material conversion correlation network is input into the dynamic balance model, and the model outputs the parameter matching relationship standard value that each processing unit should have under the current running state according to the internal parameter matching relationship mapping.

[0136] For example, for the A / O biological reaction tank processing unit, the parameter matching relationship standard value output by the dynamic balance model may include the standard range of the ammonia nitrogen concentration, the standard interval of the aerobic zone aeration amount, and the standard range of the ammonia oxidizing bacteria activity. These standard values are determined based on the stable running state of the system.

[0137] Step S143: extract the real-time parameter values of each node in the real-time material conversion correlation network, and compare them with the corresponding parameter matching relationship standard values to calculate the deviation rate.

[0138] The real-time parameter values of each node in the real-time material conversion correlation network are extracted, such as the real-time concentration value of the “A / O biological reaction tank processing unit-ammonia nitrogen” node and the real-time parameter value of the “A / O biological reaction tank processing unit-aerobic zone aeration amount” node.

[0139] The real-time parameter value of each node is compared with the corresponding parameter matching relationship standard value output by the dynamic balance model, and a deviation rate is calculated. The deviation rate is calculated by (real-time parameter value - standard value center value) / standard value range, wherein the standard value center value is the midpoint of the parameter matching relationship standard value interval, and the standard value range is the difference between the maximum and minimum values of the interval.

[0140] Step S144: Nodes with a deviation rate exceeding a preset deviation rate threshold are screened out and marked as candidate abnormal correlation nodes.

[0141] A preset deviation rate threshold is set, and the deviation rate of each node is judged. If the deviation rate of a certain node exceeds the threshold, such as the real-time concentration value of the "A / O biological reaction tank treatment unit-ammonia nitrogen" node being much higher than the standard value range, the deviation rate of which exceeds the preset deviation rate threshold, the node is marked as a candidate abnormal correlation node.

[0142] After screening, a set of all candidate abnormal correlation nodes is obtained.

[0143] Step S145: The influence range of the candidate abnormal correlation node is analyzed, the influence degree of the abnormal state of the candidate abnormal correlation node on adjacent nodes is determined, and an influence degree score is generated.

[0144] Step S1451: A list of adjacent nodes of the candidate abnormal correlation node in the material conversion correlation network is obtained, wherein the adjacent nodes are nodes having a direct connection relationship with the candidate abnormal correlation node.

[0145] Taking the candidate abnormal correlation node "A / O biological reaction tank treatment unit-ammonia nitrogen" as an example, nodes having a direct connection relationship with it in the material conversion correlation network are found, such as the "A / O biological reaction tank treatment unit-ammonia oxidizing bacteria" node, the "A / O biological reaction tank treatment unit-aerobic zone aeration amount" node, and the "secondary sedimentation tank treatment unit-suspended solids" node, to form a list of adjacent nodes.

[0146] Step S1452: The correlation coefficient of the deviation rate of the candidate abnormal correlation node and the deviation rate of the adjacent node is calculated, and the correlation coefficient is used to represent the synchronization degree of the change of the deviation rates of the two.

[0147] The data sequence of the change of the deviation rate of the candidate abnormal correlation node "A / O biological reaction tank treatment unit-ammonia nitrogen" over time is compared with the data sequence of the change of the deviation rate of each adjacent node over time.

[0148] The correlation degree of the two groups of data sequences is calculated to obtain the correlation coefficient. The correlation coefficient has a value range of -1 to 1, close to 1 indicating a high degree of synchronization of the change of the deviation rates, and close to -1 indicating a low degree of synchronization.

[0149] Step S1453: According to the correlation coefficient and the number of adjacent nodes, an influence diffusion index is calculated, which is positively correlated with the correlation coefficient and the number of adjacent nodes.

[0150] The calculation of the influence diffusion index comprehensively considers the correlation coefficient and the number of adjacent nodes. For each candidate abnormal correlation node, the absolute values of the correlation coefficients of each adjacent node are summed and multiplied by the number of adjacent nodes to obtain the influence diffusion index.

[0151] For example, if a candidate abnormal correlation node has 3 adjacent nodes, the absolute values of the corresponding correlation coefficients are 0.8, 0.6, and 0.7, respectively, the sum is 2.1, and the number of adjacent nodes is 3, the influence diffusion index is 6.3.

[0152] Step S1454: Extract the centrality value of the candidate abnormal correlation node in the material conversion correlation network, which is used to represent the connection importance of the candidate abnormal correlation node in the material conversion correlation network.

[0153] The centrality value is determined by the number of connections and the correlation strength of the candidate abnormal correlation node in the material conversion correlation network. The more the number of connections and the greater the correlation strength, the higher the centrality value, indicating that the node is more important in the network, and its abnormal state has greater potential impact on other nodes.

[0154] For example, the "A / O biological reaction tank treatment unit-ammonia nitrogen" node may have connections with multiple nodes, and the centrality value is relatively high.

[0155] Step S1455: The influence diffusion index and the centrality value are weighted and summed to obtain an influence degree score, and the weight of the weighted sum is preset according to the node type.

[0156] According to the type of the candidate abnormal correlation node, such as a pollutant node, an operating parameter node, and a microbial activity node, the weights of the influence diffusion index and the centrality value are preset. For example, for a pollutant node, the weight of the influence diffusion index may be set to 0.6, and the weight of the centrality value is set to 0.4.

[0157] The influence diffusion index is multiplied by the corresponding weight, and the centrality value is multiplied by the corresponding weight to obtain the influence degree score.

[0158] Step S1456: The influence degree score is sorted in descending order, and a set number of candidate abnormal correlation nodes with high ranking are selected as key analysis objects.

[0159] The influence degree scores of all candidate abnormal correlation nodes are sorted, and the top-ranked nodes are selected as key analysis objects. The abnormal state of these nodes has a greater impact on the system and needs to be prioritized.

[0160] Step S1457: Based on the impact degree score of the key analysis object, determine the impact degree level of the adjacent node.

[0161] According to the impact degree score of the key analysis object, divide the level into three levels: high, medium and low. The score above a certain interval is the high impact degree level, indicating that the abnormal state has a significant impact on the adjacent node; the score within a certain interval is the medium impact degree level; and the score below a certain interval is the low impact degree level.

[0162] Step S146: Determine the abnormal associated node deviating from the balanced state from the candidate abnormal associated nodes according to the impact degree score.

[0163] The candidate abnormal associated node with an impact degree score exceeding a preset impact threshold is determined as an abnormal associated node. For example, if the score corresponding to the preset impact threshold is 5.0, and the impact degree score of the candidate abnormal associated node is 6.3, the candidate abnormal associated node is determined as an abnormal associated node.

[0164] These abnormal associated nodes are nodes in the processing unit that have parameter imbalance and need further analysis of the abnormal causes.

[0165] Step S147: Extract the real-time parameter value and parameter matching relationship standard value corresponding to the abnormal associated node, analyze the parameter type causing the deviation, and determine it as an abnormal impact factor, which is a pollutant or equipment operating parameter causing parameter imbalance.

[0166] For the abnormal associated node "A / O biological reaction tank processing unit-ammonia nitrogen", extract the real-time ammonia nitrogen concentration value and the ammonia nitrogen concentration standard range in the parameter matching relationship standard value, and analyze the deviation reason.

[0167] If it is found that the real-time value of the aeration amount in the aerobic zone is lower than the interval in the parameter matching relationship standard value, and the ammonia oxidizing bacteria activity is also lower than the standard range, combined with the connection relationship of the nodes in the material conversion correlation network, it is judged that the aeration amount in the aerobic zone may be the cause of the abnormal ammonia nitrogen concentration, and the aeration amount in the aerobic zone is determined as the abnormal impact factor.

[0168] If it is found through analysis that the high concentration of a certain pollutant in the upstream inflow causes the parameter abnormality of the downstream processing unit, the pollutant is determined as the abnormal impact factor.

[0169] Step S150: Generate a diagnosis report containing an abnormal treatment path and a device control instruction according to the abnormal associated node and the abnormal impact factor, the device control instruction being used to adjust the equipment operating parameter of the corresponding processing unit to restore system balance.

[0170] Step S151: retrieve historical treatment cases corresponding to the abnormal associated node and abnormal impact factor from the preset fault diagnosis knowledge base, wherein the historical treatment cases include abnormal reason analysis, treatment path, and device adjustment records.

[0171] The fault diagnosis knowledge base stores treatment cases of various abnormal situations occurred in the past in the municipal domestic sewage treatment plant. According to the determined abnormal associated node (such as “A / O biological reaction tank treatment unit-ammonia nitrogen”) and abnormal impact factor (such as aeration amount in the aerobic zone), all relevant historical treatment cases are retrieved in the knowledge base.

[0172] These historical treatment cases record in detail the abnormal reason at that time, such as insufficient aeration amount caused by failure of the aeration fan in the aerobic zone; the treatment path, such as overhauling the aeration fan and adjusting the aeration amount; and the specific records of device adjustment, such as adjusting the aeration amount from a certain value to another value.

[0173] Step S152: perform similarity matching on the historical treatment cases, and screen out the case with the highest similarity to the current abnormal situation as the reference case.

[0174] Step S1521: extract the abnormal associated node features, abnormal impact factor features, and treatment unit state features in the historical treatment cases, and construct a case feature vector.

[0175] For each historical treatment case, the type of abnormal associated node (such as a pollutant node), the treatment unit where it is located, and the deviation rate range are extracted as the abnormal associated node features; the type of abnormal impact factor (such as an operating parameter), the parameter value deviation range, and the like are extracted as the abnormal impact factor features; and the operating state parameter range of other related treatment units at that time, such as the operating frequency range of the grid treatment unit and the pH value range of the adjusting tank treatment unit, are extracted as the treatment unit state features.

[0176] These features are arranged in a preset order to form a case feature vector of each historical treatment case. Each element in the case feature vector corresponds to a specific feature value, which collectively describes the abnormal situation of the historical case.

[0177] Step S1522: extract the abnormal associated node features, abnormal impact factor features, and treatment unit state features in the current abnormal situation, and construct a current feature vector.

[0178] For the current abnormal situation, the type of abnormal associated node, the treatment unit where it is located, and the deviation rate range are also extracted as the features; the type of abnormal impact factor, the parameter value deviation range, and the like are extracted as the features; and the operating state parameter range of other related treatment units at that time are extracted as the treatment unit state features.

[0179] The features are arranged in the same order as the case feature vector to construct a current feature vector, which is consistent with the structure of the case feature vector, so as to compare the similarity.

[0180] Step S1523: Calculate the similarity between the case feature vector and the current feature vector to obtain a similarity score, and sort the historical processing cases according to the similarity score from high to low, and select the historical processing case with the highest score as the preliminary reference case.

[0181] When calculating the similarity between the case feature vector and the current feature vector, the similarity of each corresponding feature element is calculated. For example, for the feature that the abnormal association node is in the processing unit, if the historical case is the same as the current situation, the similarity of this feature is 1; if not, it is 0.

[0182] For numerical features such as deviation rate range and parameter value deviation range, the degree of overlap between the two ranges is calculated. The higher the overlap, the higher the similarity. The similarity scores of all features are weighted and summed to obtain the similarity score of the historical processing case and the current abnormal situation.

[0183] Sort all historical processing cases according to the similarity score from high to low, and select the historical processing case with the highest score as the preliminary reference case.

[0184] Step S1524: Verify the feasibility of the processing path of the preliminary reference case in the current sewage treatment system, check whether the equipment and processing units involved in the processing path are consistent with the current sewage treatment system, if feasible, determine the preliminary reference case as the reference case; if not, select the historical processing case with the next highest score for feasibility verification until a feasible reference case is found.

[0185] Check the equipment involved in the processing path of the preliminary reference case, such as the model and number of aeration blowers, the structure of processing units, etc., and compare them with the equipment and processing units of the current municipal sewage treatment plant.

[0186] If the equipment model, processing unit structure, etc. of the two are consistent, and the operation steps in the processing path can be implemented in the current system, such as adjusting the aeration amount in the current control system, it is considered that the preliminary reference case is feasible, and it is determined as the reference case.

[0187] If the equipment involved in the preliminary reference case does not exist in the current system, or the processing steps cannot be implemented, the case is not feasible, and the historical processing case with the next highest score is selected for the same feasibility verification until a feasible reference case is found.

[0188] Step S153: According to the processing path of the reference case, combined with the material conversion correlation network of the current sewage treatment system, an abnormal processing path from the abnormal correlation node to the normal state is planned, which includes the processing unit sequence that needs to be adjusted and the parameter adjustment direction.

[0189] The processing path of the reference case may be "check the aeration blower of the A / O biological reaction tank processing unit-adjust the aeration amount of the aerobic zone-monitor the ammonia nitrogen concentration change". Combined with the current material conversion correlation network, the connection relationship and influence degree between each node are analyzed, and the processing path of the reference case is adjusted.

[0190] For example, the current material conversion correlation network shows that the adjustment of the aeration amount of the aerobic zone will also affect the activity of ammonia-oxidizing bacteria, and the activity of ammonia-oxidizing bacteria will affect the degradation of ammonia nitrogen. Therefore, when planning the abnormal processing path, in addition to adjusting the aeration amount, a step of monitoring the activity of ammonia-oxidizing bacteria is also needed.

[0191] The final determined abnormal processing path is "check the aeration blower of the A / O biological reaction tank processing unit-gradually increase the aeration amount of the aerobic zone to the parameter matching relationship standard value interval-real-time monitor the activity of ammonia-oxidizing bacteria and the concentration of ammonia nitrogen-when the concentration of ammonia nitrogen decreases to the standard range, stabilize the aeration amount of the aerobic zone", which clearly indicates that the processing unit that needs to be adjusted is the A / O biological reaction tank processing unit, and the parameter adjustment direction is to increase the aeration amount of the aerobic zone.

[0192] Step S154: Based on the location of the abnormal correlation node and the type of the abnormal influence factor, the core content of the diagnosis report is generated, which includes abnormal phenomenon description, abnormal reason inference and abnormal processing path explanation.

[0193] The abnormal correlation node is "A / O biological reaction tank processing unit-ammonia nitrogen", and the abnormal influence factor is the aeration amount of the aerobic zone. In the core content of the diagnosis report, the abnormal phenomenon description is "the real-time value of the ammonia nitrogen concentration of the A / O biological reaction tank processing unit is higher than the parameter matching relationship standard value range".

[0194] The abnormal reason inference combined with the reference case and the current data infers that "the aeration amount of the aerobic zone is lower than the standard interval, which leads to insufficient activity of ammonia-oxidizing bacteria and decreased ammonia nitrogen degradation efficiency".

[0195] The abnormal processing path explanation describes the planned abnormal processing path in detail, including the operation purpose and expected effect of each step.

[0196] Step S155: Add the identification of the abnormal correlation node, the specific parameters of the abnormal influence factor and the step decomposition of the processing path in the diagnosis report to form the target diagnosis report.

[0197] The identification of the abnormal correlation node is marked as "A / O biological reaction tank treatment unit-ammonia nitrogen" in the diagnosis report, and the specific parameters of the abnormal influence factor are "real-time value of aeration amount in the aerobic zone is X, and the parameter matching relationship standard value interval is Y-Z".

[0198] The abnormal treatment path is decomposed into specific steps, such as step one: check the running state of the aeration fan and confirm whether there is a fault; step two: if the fan is normal, gradually increase the aeration amount in the aerobic zone from X to Y through the control system; step three: record the ammonia-oxidizing bacteria activity and ammonia nitrogen concentration data every fixed time interval; step four: when the ammonia nitrogen concentration decreases to the standard range, maintain the current aeration amount stable operation.

[0199] By adding these contents, a complete and clear target diagnosis report is formed.

[0200] Step S156: According to the abnormal treatment path and the parameter adjustment direction in the target diagnosis report, determine the device identification and corresponding parameter adjustment value that need to be adjusted.

[0201] The abnormal treatment path involves adjusting the aeration amount in the aerobic zone of the A / O biological reaction tank treatment unit, and the corresponding device is the aeration fan of the treatment unit, and the device identification is "aeration fan-F01".

[0202] According to the parameter matching relationship standard value interval Y-Z, combined with the current aeration amount X, the parameter adjustment value is determined as adjusting the aeration amount from X to a value within the interval Y-Z, which is determined according to historical data and current system state, such as adjusting to (Y+Z) / 2.

[0203] Step S157: Combine the device identification, parameter adjustment value and adjustment sequence to generate a device control instruction, which is used to adjust the running parameters of the corresponding treatment unit device.

[0204] The content of the device control instruction is "device identification: aeration fan-F01; parameter adjustment value: adjust the aeration amount from X to (Y+Z) / 2; adjustment sequence: first check the device state, and then gradually adjust after no fault, and monitor the changes of related parameters every fixed time interval after each adjustment".

[0205] The device control instruction is sent to the device control module of the A / O biological reaction tank treatment unit through the central control system, and the device control module adjusts the running parameters of the aeration fan according to the device control instruction to restore the balance state of the system.

[0206] Figure 2An exemplary hardware and software components of the sewage treatment detection system 100 with diagnostic function provided by some embodiments of the present application are shown in the schematic diagram. For example, the processor 120 can be used in the sewage treatment detection system 100 with diagnostic function and used to perform the functions in the present application.

[0207] The sewage treatment detection system 100 with diagnostic function can be a general server or a special purpose server, both of which can be used to implement the sewage treatment detection method with diagnostic function of the present application. Although only one server is shown in the present application, for the convenience, the functions described in the present application can be implemented in a distributed manner on multiple similar platforms to balance the processing load.

[0208] For example, the sewage treatment detection system 100 with diagnostic function can include a network port 110 connected to a network, one or more processors 120 for executing program instructions, a communication bus 130, and different forms of storage media 140, such as a disk, a ROM, or a RAM, or any combination thereof. Exemplarily, the sewage treatment detection system 100 with diagnostic function can also include program instructions stored in a ROM, a RAM, or other types of non-transitory storage media, or any combination thereof. The method of the present application can be implemented according to these program instructions. The sewage treatment detection system 100 with diagnostic function also includes an I / O interface 150 between the computer and other input / output devices.

[0209] For the convenience of illustration, only one processor is described in the sewage treatment detection system 100 with diagnostic function. However, it should be noted that the sewage treatment detection system 100 with diagnostic function in the present application can also include multiple processors, so the steps performed by one processor described in the present application can also be jointly performed or separately performed by multiple processors. For example, if the processor of the sewage treatment detection system 100 with diagnostic function performs steps A and B, it should be understood that steps A and B can also be jointly performed by two different processors or separately performed in one processor. For example, a first processor performs step A, a second processor performs step B, or the first processor and the second processor jointly perform steps A and B.

[0210] In addition, the present application also provides a readable storage medium, in which computer executable instructions are pre-set, and when the processor executes the computer executable instructions, the sewage treatment detection method with diagnostic function as described above is implemented.

[0211] It should be noted that, in order to simplify the description of the present application and to help understand one or more embodiments of the present application, in the foregoing description of the embodiments of the present application, various features are sometimes combined into one embodiment, drawing or description thereof.

Claims

1. A sewage treatment detection method with a diagnosis function, characterized by, The method comprises: receiving multi-source detection data uploaded by each treatment unit in a sewage treatment process, the multi-source detection data including pollutant composition data of sewage, operation parameter data of treatment equipment, and bacterial community activity data in a reaction tank; performing cross-unit correlation mapping processing on the multi-source detection data to generate a material conversion correlation network between each treatment unit, the material conversion correlation network being used to reflect mutual influence relationships between pollutants, equipment operation parameters, and bacterial community activity in different treatment units; constructing a dynamic balance model of the sewage treatment system based on the material conversion correlation network, the dynamic balance model being used to reflect parameter matching relationships of each treatment unit in a stable operation state of the sewage treatment system; inputting real-time collected multi-source detection data into the dynamic balance model to analyze abnormal correlation nodes deviating from a balanced state and corresponding abnormal influence factors, the abnormal correlation nodes being nodes in which parameters are unbalanced in the treatment unit, and the abnormal influence factors being pollutants or equipment operation parameters causing parameter imbalance; generating a diagnosis report including an abnormal treatment path and equipment control instructions according to the abnormal correlation nodes and the abnormal influence factors, the equipment control instructions being used to adjust equipment operation parameters of the corresponding treatment unit to restore system balance; the cross-unit correlation mapping processing on the multi-source detection data to generate the material conversion correlation network between each treatment unit comprises: extracting feature pollutant indicators in the pollutant composition data of each treatment unit from the multi-source detection data, the feature pollutant indicators being pollutant types and concentrations that significantly change in the treatment process; extracting key operation parameters in the operation parameter data of each treatment unit, the key operation parameters being equipment operation parameters that directly affect pollutant treatment effects; extracting dominant bacterial community indicators in the bacterial community activity data in the reaction tank, the dominant bacterial community indicators being bacterial community types and activity that play a leading role in the pollutant degradation process; establishing a treatment unit correlation matrix, a row of the treatment unit correlation matrix representing a preceding treatment unit, a column representing a subsequent treatment unit, and a matrix element representing an influence degree of an output material of the preceding treatment unit on an input material of the subsequent treatment unit; inputting the feature pollutant indicators, the key operation parameters, and the dominant bacterial community indicators as nodes into the treatment unit correlation matrix to calculate correlation strength values between the nodes in different treatment units, the correlation strength values being degree quantization values of mutual influence between two nodes; connecting each node according to the correlation strength values to form a preliminary material conversion correlation network, and performing node optimization on the preliminary material conversion correlation network, merging nodes with correlation strength values lower than a preset threshold value, and retaining nodes with correlation strength values higher than the preset threshold value and connection relationships; adding treatment unit identifiers corresponding to each node in the optimized material conversion correlation network to generate a final material conversion correlation network between each treatment unit, the material conversion correlation network being used to reflect mutual influence relationships between pollutants, equipment operation parameters, and bacterial community activity in different treatment units.

2. The sewage treatment detection method with a diagnosis function according to claim 1, characterized in that, The key operating parameters in the operating parameter data of each treatment unit are extracted, including: All operating parameters and corresponding parameter values in the operating parameter data of each treatment unit are obtained; Pollutant treatment effect data corresponding to each operating parameter at different values is collected, which includes pollutant removal rate and pollutant concentration after treatment; The correlation coefficient between each operating parameter and pollutant treatment effect data is calculated, which is used to represent the degree of linear correlation between the two; Operating parameters with absolute values of correlation coefficients exceeding a preset correlation threshold are screened out as candidate key operating parameters; Sensitivity analysis is performed on the candidate key operating parameters, and the change amount of pollutant treatment effect when the parameter value changes by a unit is calculated to obtain a sensitivity coefficient; The sensitivity coefficients are sorted in descending order, and operating parameters with high ranking are selected as key operating parameters, which are device operating parameters that have a direct impact on pollutant treatment effect; In addition, the dominant bacteria index in the bacteria activity data in the reaction tank is extracted, which is the type and activity of the bacteria that play a leading role in the pollutant degradation process, including: All bacteria species and corresponding activity values in the bacteria activity data in the reaction tank are obtained, which include bacteria quantity and metabolic rate; The degradation correlation degree of each bacteria species and the characteristic pollutant index is analyzed, which is calculated by the ratio of the degradation rate of the characteristic pollutant when the bacteria exist to the degradation rate when the bacteria do not exist; Bacteria species with degradation correlation degrees exceeding a preset correlation threshold are screened out as candidate dominant bacteria, and the proportion of the activity value of each candidate dominant bacteria in the total activity value of all bacteria is calculated to obtain an activity proportion; Candidate dominant bacteria with activity proportions exceeding a preset activity proportion threshold are determined as dominant bacteria species, and the corresponding activity values of the dominant bacteria species are extracted as dominant bacteria activity data; The dominant bacteria species and corresponding dominant bacteria activity data are combined to obtain the dominant bacteria index in the bacteria activity data in the reaction tank, which is the type and activity of the bacteria that play a leading role in the pollutant degradation process.

3. The sewage treatment detection method with a diagnosis function according to claim 1, characterized in that, The dynamic balance model of the sewage treatment system is constructed based on the material conversion correlation network, including: Multiple sets of material conversion correlation networks and corresponding parameter data sets of each treatment unit during stable operation period of the sewage treatment system are collected, which include stable values of pollutant composition data, operating parameter data and bacteria activity data; Node feature extraction is performed on each set of material conversion correlation networks to obtain feature vectors of each node, which include pollutant concentration, operating parameter value and bacteria activity value corresponding to the node; The node feature vectors and parameter data sets in the same stable operation period are associated and aligned to establish a mapping relationship between node features and stable parameters; The graph neural network algorithm is used to train the node feature vectors and mapping relationship after association alignment, and an initial dynamic balance model is constructed. The initial dynamic balance model can output corresponding parameter matching relationship according to the input material conversion association network. The material conversion association network and parameter data set of a part of stable operation period are selected as a verification set, which is input into the initial dynamic balance model to obtain a parameter matching relationship prediction result. The deviation value of the parameter matching relationship prediction result and the actual parameter matching relationship in the verification set is calculated. If the deviation value exceeds a preset deviation threshold, the inter-layer connection weight of the graph neural network algorithm is adjusted. The model training and parameter adjustment process is repeated until the deviation value is less than or equal to the preset deviation threshold, and a final dynamic balance model of the sewage treatment system is obtained. The dynamic balance model is used to reflect the parameter matching relationship of each processing unit of the sewage treatment system in the stable operation state.

4. The sewage treatment detection method with a diagnosis function according to claim 3, characterized by, The graph neural network algorithm is used to train the node feature vectors and mapping relationship after association alignment, and an initial dynamic balance model is constructed. The initial dynamic balance model can output corresponding parameter matching relationship according to the input material conversion association network. The node feature vectors after association alignment are input into the input layer of the graph neural network, and the dimension conversion processing is performed on the node feature vectors, so that the dimension of the node feature vectors matches the dimension of the hidden layer of the graph neural network. In the hidden layer of the graph neural network, the connection relationship between nodes is described by an adjacency matrix, and the node feature vectors are aggregated based on the connection relationship to obtain aggregated feature vectors containing neighbor node information. The aggregated feature vectors are subjected to nonlinear activation processing to generate activated feature vectors, and the activated feature vectors are input into the output layer of the graph neural network. The activated feature vectors are processed through a fully connected layer to output parameter matching relationship prediction values of each processing unit. The parameter matching relationship prediction values include pollutant concentration matching range, operation parameter matching interval and microbial community activity matching interval. The loss value of the parameter matching relationship prediction values and the actual parameter matching relationship is calculated, and the weight parameters of the graph neural network are adjusted by using a back propagation algorithm to minimize the loss value. A training iteration number threshold is set. When the training iteration number reaches the threshold, the training is stopped, and the current graph neural network model is determined as the initial dynamic balance model.

5. The sewage treatment detection method with a diagnosis function according to claim 1, characterized in that, The real-time collected multi-source detection data is input into the dynamic balance model to analyze abnormal association nodes deviating from the balance state and corresponding abnormal influence factors, including: The node mapping processing is performed on the real-time collected multi-source detection data, and the data is mapped to each node of the material conversion association network to generate a real-time material conversion association network. The real-time material conversion association network is input into the dynamic balance model to obtain parameter matching relationship standard values of each processing unit. The real-time parameter values of each node in the real-time material conversion association network are extracted, and compared with the corresponding parameter matching relationship standard values to calculate the deviation rate. The nodes with a deviation rate exceeding a preset deviation rate threshold are screened out and marked as candidate abnormal association nodes. Performing influence range analysis on the candidate abnormal correlation node to determine the influence degree of the abnormal state of the candidate abnormal correlation node on adjacent nodes, and generating an influence degree score; Determining the abnormal correlation node deviating from the balanced state from the candidate abnormal correlation nodes according to the influence degree score; Extracting real-time parameter values and parameter matching relationship standard values corresponding to the abnormal correlation node, analyzing the parameter type causing the deviation, and determining the abnormal influence factor, which is the pollutant or equipment operation parameter causing the parameter imbalance.

6. The sewage treatment detection method with a diagnosis function according to claim 5, characterized in that, The influence range analysis on the candidate abnormal correlation node to determine the influence degree of the abnormal state of the candidate abnormal correlation node on adjacent nodes, and generating an influence degree score, comprises: Obtaining a list of adjacent nodes of the candidate abnormal correlation node in the material conversion correlation network, the adjacent nodes being nodes having a direct connection relationship with the candidate abnormal correlation node; Calculating the correlation coefficient of the deviation rate of the candidate abnormal correlation node and the deviation rate of the adjacent nodes, the correlation coefficient being used to represent the synchronization degree of the change of the deviation rates of the two; According to the correlation coefficient and the number of adjacent nodes, calculating an influence diffusion index, the influence diffusion index being positively correlated with the correlation coefficient and the number of adjacent nodes; Extracting the centrality value of the candidate abnormal correlation node in the material conversion correlation network, the centrality value being used to represent the connection importance of the candidate abnormal correlation node in the material conversion correlation network; Performing weighted summation on the influence diffusion index and the centrality value to obtain an influence degree score, the weight of the weighted summation being preset according to the node type; Sorting the influence degree scores in descending order, and selecting a set number of candidate abnormal correlation nodes with high ranking as key analysis objects; Determining the influence degree level of the key analysis objects on adjacent nodes based on the influence degree scores of the key analysis objects.

7. The sewage treatment detection method with a diagnosis function according to claim 1, characterized by, The generation of the diagnostic report containing the abnormal treatment path and the device control instruction according to the abnormal correlation node and the abnormal influence factor, comprises: Retrieving historical treatment cases corresponding to the abnormal correlation node and the abnormal influence factor from a preset fault diagnosis knowledge base, the historical treatment cases containing abnormal reason analysis, treatment path and device adjustment record; Performing similarity matching on the historical treatment cases to filter out the case with the highest similarity to the current abnormal situation as a reference case; According to the treatment path of the reference case, combining the material conversion correlation network of the current sewage treatment system, and planning an abnormal treatment path from the abnormal correlation node to the normal state, the abnormal treatment path containing the processing unit sequence and parameter adjustment direction that need to be adjusted; Generating the core content of the diagnostic report based on the position of the abnormal correlation node and the type of the abnormal influence factor, the core content containing abnormal phenomenon description, abnormal reason inference and abnormal treatment path explanation; Adding the identification of the abnormal correlation node, the specific parameters of the abnormal influence factor and the step decomposition of the treatment path in the diagnostic report to form a target diagnostic report; According to the abnormality processing path and the parameter adjustment direction in the target diagnosis report, a device identifier and a corresponding parameter adjustment value that need to be regulated are determined; The device identifier, the parameter adjustment value and the adjustment sequence are combined to generate a device regulation instruction, and the device regulation instruction is used to adjust the device operation parameter of the corresponding processing unit.

8. The sewage treatment detection method with a diagnosis function according to claim 7, characterized in that, The similarity matching of the historical processing cases is performed to screen out a case with the highest similarity to the current abnormality as a reference case, and the similarity matching of the historical processing cases includes: Extracting abnormal associated node features, abnormal influence factor features and processing unit state features in the historical processing cases to construct a case feature vector; Extracting abnormal associated node features, abnormal influence factor features and processing unit state features in the current abnormality to construct a current feature vector; Calculating the similarity of the case feature vector and the current feature vector to obtain a similarity score, and sorting the historical processing cases in a descending order of the similarity score, and selecting a historical processing case at a top position as a preliminary reference case; Verifying the feasibility of the processing path of the preliminary reference case in the current sewage treatment system, checking whether the devices and processing units involved in the processing path are consistent with the current sewage treatment system, and if feasible, determining the preliminary reference case as the reference case; if not feasible, selecting a historical processing case at a next position to perform the feasibility verification until a feasible reference case is found.

9. A sewage treatment detection system with a diagnosis function, characterized by, The sewage treatment detection method with a diagnosis function includes a processor and a memory, the memory is connected with the processor, the memory is used to store programs, instructions or codes, and the processor is used to execute the programs, instructions or codes in the memory to realize the sewage treatment detection method with a diagnosis function in any one of claims 1-8.

Images