A carbon addition system for sewage treatment

Abstract

The present invention provides a carbon addition system for sewage treatment, comprising at least one dosing bin for storing carbon source stock solution; at least one metering pump connected to the dosing bin for delivering the carbon source stock solution to a sewage pool; a detection module for detecting the current data of sewage in the sewage pool; and a processing module for constructing a target carbon addition model based on the acquired comprehensive data of sewage treatment by an activated sludge process, and inputting the current data as a parameter into the target carbon addition model to generate corresponding target carbon addition data; the processing module is also used to generate corresponding carbon addition instructions based on the target carbon addition data, and the metering pump delivers the carbon source stock solution to the sewage pool according to the carbon addition instructions. The carbon addition system for sewage treatment provided by the present invention can improve the accuracy and reliability of carbon addition decisions.

Description

Technical Field

[0001] The invention relates to the field of sewage treatment, and in particular to a carbon addition system for sewage treatment. Background Art

[0002] The activated sludge process is a technology widely used in the field of sewage treatment. It uses microorganisms in activated sludge to remove pollutants such as organic matter in sewage. This process involves introducing the sewage to be treated into the aeration tank together with the activated sludge returned from the secondary sedimentation tank. Air aeration is used in the aeration tank to achieve full mixing and contact between the sewage and the activated sludge, and provide the necessary dissolved oxygen for the growth and reproduction of microorganisms. In this process, the addition of carbon source stock solution plays a vital role in the biochemical reaction of microorganisms and the decomposition of pollutants. However, in the actual sewage treatment process, the problem of insufficient carbon source is often faced, which requires the addition of carbon source stock solution to ensure the treatment efficiency.

[0003] At present, the addition of carbon source stock solution in the sewage treatment process mainly relies on the experience of industry personnel for adjustment. This method not only lacks accuracy, but also cannot guarantee timely response to changing treatment needs. In addition, excessive or insufficient addition of carbon source stock solution will affect the sewage treatment efficiency and treatment cost, and the reliance on manual adjustment of carbon source stock solution significantly increases the labor cost. These factors jointly restrict the optimization and efficiency improvement of the activated sludge process. Therefore, there is room for improvement. Summary of the invention

[0004] The object of the present invention is to provide a carbon dosing system for sewage treatment, which can improve the accuracy and reliability of carbon dosing decisions.

[0005] In order to solve the above technical problems, the present invention is achieved through the following technical solutions:

[0006] The present invention provides a carbon addition system for sewage treatment, comprising:

[0007] At least one dosing chamber for storing carbon source stock solution;

[0008] At least one metering pump, connected to the dosing chamber, for delivering the carbon source stock solution to the sewage pool;

[0009] A detection module, used for detecting current data of sewage in the sewage pool;

[0010] A processing module, for constructing a target carbon addition model based on the acquired comprehensive data of sewage treatment by the activated sludge process, and inputting the current data as parameters into the target carbon addition model to generate corresponding target carbon addition data;

[0011] The processing module is also used to generate a corresponding carbon addition instruction according to the target carbon addition data, and the metering pump delivers the carbon source stock solution to the sewage pool according to the carbon addition instruction;

[0012] The processing module comprises:

[0013] A model building unit, used to obtain comprehensive data of sewage treatment by activated sludge process from a constructed database, and establish an initial carbon addition model based on the comprehensive data; wherein the initial carbon addition model includes a strategy network model for determining carbon addition action, a state value network model for determining sewage state value, and an action value network model for determining the value of carbon addition action;

[0014] A first input unit, used for inputting the current data as a parameter into the initial carbon addition model to obtain first comprehensive data after carbon addition;

[0015] A first optimization unit is used to start with the first comprehensive data, search for multiple groups of comprehensive data that meet preset conditions from the database, express them as strategy network data, optimize the strategy network model according to the strategy network data, and generate an intermediate carbon addition model;

[0016] A second input unit is used to input the current data as a parameter into the intermediate carbon addition model to obtain second comprehensive data after carbon addition;

[0017] A second optimization unit is used to start with the second comprehensive data, search for multiple groups of comprehensive data that meet preset conditions from the database, expressed as value network data, optimize the state value network model and the action value network model according to the value network data, and generate a target carbon-adding model;

[0018] Among them, the preset condition is expressed as: calculating the difference between the current state data after carbon addition in the first comprehensive data or the second comprehensive data and the sewage state data before carbon addition in a certain group of comprehensive data, and the ratio of this difference to the current state data after carbon addition in the first comprehensive data or the second comprehensive data does not exceed a threshold value; calculating the difference between the sewage state data after carbon addition in the certain group of comprehensive data and the sewage state data before carbon addition in the next group of comprehensive data, and the ratio of this difference to the sewage state data after carbon addition in the certain group of comprehensive data does not exceed a threshold value.

[0019] In one embodiment of the present invention, the detection module includes:

[0020] A COD detection unit, used to detect the chemical oxygen demand data of the sewage;

[0021] A total nitrogen detection unit, used for detecting the total nitrogen data of the sewage;

[0022] A total phosphorus detection unit, used to detect the total phosphorus data of the sewage;

[0023] A turbidity monitoring unit, used to monitor the turbidity data of the sewage;

[0024] The liquid level detection unit is used to detect the liquid level data of the sewage.

[0025] In one embodiment of the present invention, the processing module is also used to calculate sewage volume data based on the liquid level data, and the current data includes the chemical oxygen demand data, the total nitrogen data, the total phosphorus data, the turbidity data, and the sewage volume data.

[0026] In one embodiment of the present invention, the processing module includes:

[0027] A model building unit, used to obtain comprehensive data of sewage treatment by activated sludge process from a constructed database, and establish an initial carbon addition model based on the comprehensive data; wherein the initial carbon addition model includes a strategy network model for determining carbon addition action, a state value network model for determining sewage state value, and an action value network model for determining the value of carbon addition action;

[0028] A first input unit, used for inputting the current data as a parameter into the initial carbon addition model to obtain first comprehensive data after carbon addition;

[0029] a first optimization unit, configured to start with the first comprehensive data, search the database for comprehensive data similar to the first comprehensive data, express the comprehensive data as strategy network data, optimize the strategy network model according to the strategy network data, and generate an intermediate carbon addition model;

[0030] A second input unit is used to input the current data as a parameter into the intermediate carbon addition model to obtain second comprehensive data after carbon addition;

[0031] The second optimization unit is used to start with the second comprehensive data, search for comprehensive data similar to the second comprehensive data from the database, express it as value network data, optimize the state value network model and the action value network model according to the value network data, and generate a target carbon addition model.

[0032] In one embodiment of the present invention, the model building unit includes:

[0033] A data acquisition subunit is used to acquire multiple groups of historical data of sewage treated by an activated sludge process; wherein the historical data include status data of sewage before and after carbon addition, and carbon addition amount data; the status data include chemical oxygen demand data, total nitrogen data, total phosphorus data, turbidity data, and sewage amount data;

[0034] A data processing subunit, used for performing weighted processing on the historical data, obtaining sewage status data, and obtaining carbon addition action data based on the sewage status data;

[0035] The benefit calculation subunit is used to obtain carbon addition benefit data according to the sewage state data and the carbon addition action data;

[0036] The data association subunit is used to associate each group of corresponding historical data, sewage status data, carbon addition action data and carbon addition benefit data to express them as the comprehensive data; build a database based on multiple groups of the comprehensive data, and establish an initial carbon addition model based on the comprehensive data.

[0037] In one embodiment of the present invention, the first input unit includes:

[0038] A first data integration subunit, used to determine current state data according to the current data;

[0039] A first input subunit, used to input the current state data as a parameter into the strategy network model of the initial carbon addition model to obtain the corresponding first current carbon addition action data;

[0040] A first benefit calculation subunit is used to obtain first current carbon addition benefit data according to the current state data and the first current carbon addition action data;

[0041] The first data association subunit is used to associate the current state data before and after carbon addition, the first current carbon addition action data, and the first current carbon addition benefit data to obtain first comprehensive data.

[0042] In one embodiment of the present invention, the first optimization unit includes:

[0043] A first screening subunit, used to obtain multiple groups of comprehensive data meeting preset conditions from the database, represented as strategic network data;

[0044] A second input subunit is used to input the plurality of groups of the strategy network data as parameters into the state value network model to obtain the first state value data before carbon addition;

[0045] A third input subunit is used to input the plurality of groups of the strategy network data as parameters into the action value network model to obtain the first action value data before carbon addition;

[0046] A first expectation calculation subunit, used for obtaining first expected advantage data before carbon addition according to the state value data and the action value data;

[0047] A first target finding subunit, used for determining a strategy target function to be optimized in the strategy network model according to the first expected advantage data;

[0048] The first training optimization subunit is used to optimize the parameters of the policy network model according to the gradient of the policy objective function and the conjugate gradient ascent algorithm, and generate an intermediate carbon addition model.

[0049] In one embodiment of the present invention, the second input unit includes:

[0050] The fourth input subunit is used to input the current state data before carbon addition as a parameter into the strategy network model of the intermediate carbon addition model to obtain the corresponding second current carbon addition action data;

[0051] A second benefit calculation subunit is used to obtain second current carbon addition benefit data according to the current state data and the second current carbon addition action data;

[0052] The second data association subunit is used to associate the current state data before and after carbon addition, the second current carbon addition action data, and the second current carbon addition benefit data to obtain second comprehensive data.

[0053] In one embodiment of the present invention, the second optimization unit includes:

[0054] A second screening subunit is used to obtain multiple groups of comprehensive data that meet preset conditions from the database, expressed as value network data;

[0055] A fifth input subunit, used to input the plurality of sets of the value network data as parameters into the state value network model to obtain the second state value data before carbon addition;

[0056] A sixth input subunit, used to input the plurality of sets of the value network data as parameters into the action value network model, and obtain the second action value data before carbon addition;

[0057] A second expectation calculation subunit, used for obtaining second expected advantage data before carbon addition according to the state value data and the action value data;

[0058] A second target search subunit is used to determine the value target function to be optimized in the state value network model and the action value network model according to the second expected advantage data;

[0059] The second training optimization subunit is used to optimize the parameters of the state value network model and the action value network model according to the gradient of the value objective function and the gradient descent algorithm, and generate a target carbon addition model.

[0060] As described above, the present invention provides a carbon dosing system for sewage treatment, which automatically optimizes the calculation process of the carbon dosing amount through a deep learning algorithm, significantly improves the accuracy and reliability of the carbon dosing decision, reduces human errors, and ensures the stability of the treatment effect. By utilizing the real-time optimization capability of the deep reinforcement learning model, it is possible to timely feedback changes in the treatment process, achieve real-time adjustment of the carbon dosing amount, effectively respond to emergencies that may occur during the treatment process, and ensure water treatment efficiency and effectiveness. Through the automated carbon dosing control strategy, the reliance on professionals is reduced, thereby effectively reducing labor costs, while improving the convenience and economic benefits of operation. By accurately controlling the amount of carbon dosing, not only the efficiency of sewage treatment is improved, but it also helps to reduce potential negative impacts on the environment and promotes sustainable environmental development.

[0061] Of course, any product implementing the present invention does not necessarily need to achieve all of the advantages described above at the same time. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for describing the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0063] Figure 1 A schematic diagram of a carbon addition system for sewage treatment according to an embodiment of the present invention;

[0064] Figure 2 is a schematic diagram of a processing module in one embodiment of the present invention;

[0065] Figure 3 is a schematic diagram of a model building unit in one embodiment of the present invention;

[0066] Figure 4 is a schematic diagram of a first input unit in an embodiment of the present invention;

[0067] Figure 5 is a schematic diagram of a first optimization unit in one embodiment of the present invention;

[0068] Figure 6 is a schematic diagram of a second input unit in one embodiment of the present invention;

[0069] Figure 7 FIG. 4 is a schematic diagram of a second optimization unit in an embodiment of the present invention.

[0070] In the figure: 100, dosing chamber; 200, metering pump; 300, detection module; 400, processing module;

[0071] 310, COD detection unit; 320, total nitrogen detection unit; 330, total phosphorus detection unit; 340, turbidity monitoring unit; 350, liquid level detection unit;

[0072] 410, model building unit; 411, data acquisition subunit; 412, data processing subunit; 413, benefit calculation subunit; 414, data association subunit;

[0073] 420, first input unit; 421, first data integration subunit; 422, first input subunit; 423, first benefit calculation subunit; 424, first data association subunit;

[0074] 430, first optimization unit; 431, first screening subunit; 432, second input subunit; 433, third input subunit; 434, first expectation calculation subunit; 435, first target search subunit; 436, first training optimization subunit;

[0075] 440, second input unit; 441, fourth input subunit; 442, second benefit calculation subunit; 443, second data association subunit;

[0076] 450, second optimization unit; 451, second screening subunit; 452, fifth input subunit; 453, sixth input subunit; 454, second expectation calculation subunit; 455, second target search subunit; 456, second training optimization subunit. DETAILED DESCRIPTION

[0077] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0078] See also Figure 1 The present invention provides a carbon addition system for sewage treatment, which can be applied to the treatment process of activated sludge process, and the addition amount of carbon source stock solution is calculated by establishing a network model to determine the specific content of carbon source stock solution to be added in the sewage treatment process. The carbon addition system may include a dosing bin 100, a metering pump 200, a detection module 300 and a processing module 400.

[0079] In some embodiments, the dosing bin 100 can be used to store carbon source stock solution. The carbon source stock solution can be in liquid form, and the carbon source stock solution can include methanol, glucose, sodium acetate, etc. The dosing bin 100 can be designed as a closed container to prevent the carbon source stock solution from being contaminated or volatilized. The bin body material of the dosing bin 100 can be selected from corrosion-resistant stainless steel or plastic. Depending on the specific needs and processing volume of the sewage treatment plant, the capacity of the dosing bin 100 can vary, ranging from tens of liters to several tons. The number of dosing bins 100 can be at least one. When the number of dosing bins 100 is multiple, the liquid delivery ports of multiple dosing bins 100 can be interconnected.

[0080] In some embodiments, the number of metering pumps 200 can be at least one, and the metering pump 200 can be connected to the dosing bin 100 to accurately and quantitatively deliver the carbon source stock solution to the sewage pool. By adjusting the working parameters of the metering pump, accurate control of the amount of carbon source stock solution added can be achieved. Among them, the metering pump 200 has high accuracy and stability and can adapt to different flow and pressure requirements. The types of metering pumps can include diaphragm pumps, plunger pumps, gear pumps, etc.

[0081] In some embodiments, the detection module 300 can be used to detect the current data of sewage in the sewage pool in real time, and the current data may include chemical oxygen demand data (COD), total nitrogen data (TN), total phosphorus data (TP), turbidity data, liquid level data, etc. Among them, the detection module 300 may include a COD detection unit 310, a total nitrogen detection unit 320, a total phosphorus detection unit 330, a turbidity monitoring unit 340, a liquid level detection unit 350, etc.

[0082] In some embodiments, the COD detection unit 310 can be used to detect the chemical oxygen demand data of the sewage. Chemical oxygen demand data (COD) refers to the total amount of organic matter and some inorganic matter (such as sulfide and iron, etc.) oxidized by chemical methods in sewage. The type of COD detection unit 310 can be unlimited, for example, it can be a COD online automatic detector.

[0083] In some embodiments, the total nitrogen detection unit 320 can be used to detect the total nitrogen data of the sewage. The total nitrogen data (TN) refers to the sum of all forms of nitrogen in the sewage, including ammonium nitrogen, nitrite nitrogen, nitrate nitrogen, and organic nitrogen. The type of the total nitrogen detection unit 320 can be unlimited, for example, it can be an online automatic total nitrogen detector.

[0084] In some embodiments, the total phosphorus detection unit 330 can be used to detect the total phosphorus data of the sewage. The total phosphorus data (TP) refers to all forms of phosphorus in the sewage, such as dissolved phosphorus (DP), particulate phosphorus (PP), etc. The type of the total phosphorus detection unit 330 may not be limited, for example, it may be an online automatic total phosphorus detector.

[0085] In some embodiments, the turbidity monitoring unit 340 can be used to monitor the turbidity data of the sewage. Turbidity data refers to the degree to which suspended particles in the sewage make the water unclear. The type of turbidity monitoring unit 340 may not be limited, for example, it may be an online suspended particle monitor.

[0086] In some embodiments, the liquid level detection unit 350 can be used to detect the liquid level data of the sewage. The liquid level data refers to the water level of the sewage in the sewage pool. The type of the liquid level detection unit 350 can be unlimited, for example, it can be an ultrasonic liquid level meter, a float liquid level meter, a pressure liquid level meter, etc.

[0087] In some embodiments, the processing module 400 can be used to calculate the sewage volume data according to the liquid level data. For example, the sewage volume in the sewage pool can be calculated according to the liquid level data provided by the liquid level detection unit 350 and the geometric dimensions (such as length, width, and height) of the sewage pool. After calculating the sewage volume data, the processing module 400 can obtain the current data of the sewage in the sewage pool. The current data may include chemical oxygen demand data, total nitrogen data, total phosphorus data, turbidity data, sewage volume data, etc.

[0088] See also Figure 2 In some embodiments, the processing module 400 may construct a target carbon addition model based on the acquired comprehensive data of sewage treatment by the activated sludge process. The processing module 400 may include a model construction unit 410, a first input unit 420, a first optimization unit 430, a second input unit 440, and a second optimization unit 450.

[0089] See also Figure 3 In some embodiments, the model building unit 410 can be used to obtain comprehensive data of sewage treatment by activated sludge process from the constructed database, and establish an initial carbon addition model based on the comprehensive data. The model building unit 410 can include a data acquisition subunit 411, a data processing subunit 412, a benefit calculation subunit 413, and a data association subunit 414.

[0090] In some embodiments, the data acquisition subunit 411 can be used to obtain multiple sets of historical data of sewage treated by an activated sludge process. Specifically, before establishing a network model, multiple sets of historical data of sewage treated by an activated sludge process can be obtained. These historical data are very important for optimizing the treatment process, improving treatment efficiency, and reducing costs. The historical data can come from the historical operation data of the sewage treatment plant, etc. Among them, the historical data includes the status data of the sewage before and after carbon addition, and the carbon addition amount data. The status data may include chemical oxygen demand data, total nitrogen data, total phosphorus data, turbidity data, and sewage amount data. The carbon addition amount data refers to the amount of carbon source stock solution added to the sewage during the sewage treatment process.

[0091] In some embodiments, the status data may also include expandable parameters added by parameters measured in actual activated sludge process wastewater treatment. The expandable parameters may include but are not limited to biochemical oxygen demand data (BOD), suspended solids (SS) or particulate matter data, dissolved oxygen data (DO), ammonia nitrogen data, pH value data, temperature data, nitrate nitrogen and nitrite nitrogen data, microbial activity indicators, etc. By expanding the parameters, a more comprehensive perspective can be provided to evaluate and optimize the wastewater treatment process.

[0092] In some embodiments, the data processing subunit 412 can be used to perform weighted processing on the historical data, obtain sewage status data, and obtain carbon addition action data based on the sewage status data. Specifically, after the historical data is obtained, the historical data can be weighted processed to obtain sewage status data. , expressed as .in, Indicates chemical oxygen demand data, Represents total nitrogen data, represents total phosphorus data, Represents turbidity data, Represents sewage volume data.

[0093] In some embodiments, when weighting historical data, the actual measured value of each parameter can be multiplied by a predefined weight to reflect the relative importance of the parameter. The selection of weights should be based on the relative contribution of the parameters to environmental impact, processing costs, operational goals, or regulatory requirements. The size of the weight can be set according to actual needs and is not limited here.

[0094] In some embodiments, when the sewage status data is obtained After that, you can get the carbon adding action data , expressed as .in, Indicates carbon addition data.

[0095] In some embodiments, the revenue calculation subunit 413 can be used to obtain carbon addition revenue data based on the sewage status data and the carbon addition action data. Carbon addition action data After that, you can define the corresponding carbon income data , expressed as .in, Indicates the sewage status data before carbon addition, Indicates the sewage status data after carbon addition, Indicates the change in chemical oxygen demand data before and after carbon addition. Indicates the change in total nitrogen data before and after carbon addition. Indicates the change in total phosphorus data before and after adding carbon. Indicates the change in turbidity data before and after adding carbon. Indicates the change in flow rate data before and after carbon addition.

[0096] In some embodiments, the data association subunit 414 can be used to associate each set of corresponding historical data, sewage status data, carbon addition action data, and carbon addition benefit data, and express them as comprehensive data, and build a database based on multiple sets of comprehensive data. Specifically, when obtaining the sewage status data before carbon addition of a set of historical data , Wastewater status data after carbon addition , carbon addition amount data, carbon addition action data , plus carbon benefit data After that, the individual data in the set of historical data can be associated and expressed as a set of comprehensive data. The comprehensive data can be expressed as a set of determined state-action-benefit-state Different comprehensive data can be obtained for different historical data. At this time, the comprehensive data can be saved in the database to build a database for subsequent processing.

[0097] In some embodiments, after the database is constructed, a suitable neural network model can be selected for training based on the comprehensive data in the database. The network structure of the neural network model can include at least one input layer, three hidden layers, and one output layer, wherein the hidden layer can include at least one convolution layer, one activation function layer, and one pooling layer. In this embodiment, the processing module 400 can construct an initial carbon addition model based on the comprehensive data. The initial carbon addition model can include a strategy network model for determining the carbon addition action, a state value network model for determining the sewage state value, an action value network model for determining the value of the carbon addition action, and the like.

[0098] In some embodiments, the purpose of the policy network model is to determine the best carbon addition action to be taken under a specific wastewater treatment state. This model guides the carbon addition strategy by evaluating the current environmental state and outputting a probability distribution of one or more carbon addition actions. The model attempts to maximize long-term benefits by learning the mapping relationship between environmental states and optimal actions. The policy network model can be expressed as , the input is the sewage status data before carbon addition , output is carbon adding action data ,Right now , are the parameters in the policy network model.

[0099] In some embodiments, the purpose of the state value network model is to evaluate the overall value of a given wastewater treatment state, which reflects the maximum cumulative return that can be obtained in the future starting from the current state. The model analyzes the cumulative returns under different states and the potential value of intervention (such as carbon addition) at a specific point. The state value network model can be expressed as , the input is the sewage status data before carbon addition , the output is carbon-added benefit data , ,in, represents the discounted visit frequency, Represents all sewage status data of sewage in the database.

[0100] In some embodiments, the purpose of the action value network model is to evaluate the expected return of carbon addition actions under a given state. The model predicts the long-term benefits of taking a specific carbon addition action and helps optimize the amount and timing of carbon addition. By comparing the action values ​​of different actions, the carbon addition strategy that optimizes the sewage treatment effect and economic benefits can be selected. The action value network model can be expressed as , the input is the sewage status data before carbon addition Carbon addition action data , output is carbon adding action data , , Represents the parameters in the state value network model and the action value network model.

[0101] See also Figure 4 In some embodiments, the first input unit 420 can be used to input current data as parameters into the initial carbon addition model to obtain the first comprehensive data after carbon addition. The first input unit 420 may include a first data integration subunit 421, a first input subunit 422, a first benefit calculation subunit 423, and a first data association subunit 424.

[0102] In some embodiments, the first data integration subunit 421 can be used to determine the current state data based on the current data. Specifically, when it is necessary to perform carbon treatment on the sewage, the current data in the sewage can be obtained, and the current data can include the chemical oxygen demand data, total nitrogen data, total phosphorus data, turbidity data, flow data and other expandable parameters of the sewage. After obtaining the above parameters, these parameters can be integrated into the current state data to determine the current state data of the sewage. .

[0103] In some embodiments, the first input subunit 422 can be used to input the current state data as a parameter into the strategy network model of the initial carbon addition model to obtain the corresponding first current carbon addition action data. Specifically, after obtaining the current state data, the current state data can be input into the strategy network model of the initial carbon addition model to obtain the corresponding first current carbon addition action data. , expressed as .

[0104] In some embodiments, the first benefit calculation subunit 423 can be used to obtain the first current carbon addition benefit data according to the current state data and the first current carbon addition action data. Specifically, after obtaining the first carbon addition action data, the metering pump 200 can be adjusted according to the carbon addition action data. Afterwards, the sewage can be carbonized according to the first carbon addition action data, and the current state data of the sewage after carbonization can be obtained. At this time, the first current carbon addition benefit data can be obtained based on the current state data before carbon addition and the first current carbon addition action data. , expressed as .

[0105] In some embodiments, the first data association subunit 424 can be used to associate the current state data before and after carbon addition, the first current carbon addition action data, and the first current carbon addition benefit data to obtain the first comprehensive data. , Current status data after carbon addition , First, current carbon addition action data , First, current carbon benefit data After that, the above data can be associated to obtain the corresponding first comprehensive data, which is expressed as .

[0106] See also Figure 5 In some embodiments, the first optimization unit 430 can be used to start with the first comprehensive data, search for comprehensive data similar to the first comprehensive data from the database, represent it as policy network data, optimize the policy network model according to the policy network data, and generate an intermediate carbon addition model. The first optimization unit 430 may include a first screening subunit 431, a second input subunit 432, a third input subunit 433, a first expectation calculation subunit 434, a first target search subunit 435, and a first training optimization subunit 436.

[0107] In some embodiments, the first screening subunit 431 can be used to obtain multiple groups of comprehensive data that meet preset conditions from the database, which are expressed as policy network data. Specifically, after the first comprehensive data is obtained, the comprehensive data that meets the preset conditions can be selected from the database according to the preset conditions, which are expressed as policy network data. The preset conditions can be expressed as: calculating the difference between the current state data after carbon addition in the first comprehensive data and the sewage state data before carbon addition in a certain group of comprehensive data, the ratio of this difference to the current state data after carbon addition in the first comprehensive data does not exceed the threshold value, calculating the difference between the sewage state data after carbon addition in the certain group of comprehensive data and the sewage state data before carbon addition in the next group of comprehensive data, the ratio of this difference to the sewage state data after carbon addition in the certain group of comprehensive data does not exceed the threshold value.

[0108] In some embodiments, when the first comprehensive data is obtained After that, you can search from the database Group comprehensive data to determine strategic network data , expressed as ,in, Represents the parameters in the policy network model.

[0109] In some embodiments, during the search process, it is necessary to start with the first comprehensive data and search for comprehensive data in the database that are similar to the first comprehensive data. A certain set of comprehensive data found needs to satisfy: the sewage state data after carbon addition in the first comprehensive data is similar to the sewage state data before carbon addition in a certain set of comprehensive data found. That is, first calculate the difference in parameters between the sewage state data after carbon addition in the first comprehensive data and the sewage state data before carbon addition in a certain set of comprehensive data found, and then calculate the ratio of the difference to the sewage state data after carbon addition in the first comprehensive data, and determine whether the ratio is less than a threshold value. If the ratio is less than the threshold value, it means that the certain set of comprehensive data meets the preset conditions. The size of the threshold can be unlimited, for example, it can be 0.01.

[0110] In some embodiments, when a certain set of comprehensive data is found in the database to meet the preset conditions, the certain set of comprehensive data can be represented as the first Then, according to the The next set of comprehensive data that meets the preset conditions is found from the first set of comprehensive data in the database. Group and The comprehensive data of the group needs to meet the following requirements: Wastewater status data after carbon addition With Wastewater status data before carbon addition Similar, that is, sewage status data and sewage status data The difference between the parameters and the sewage status data The ratio of does not exceed the threshold. is an integer, and , The number of comprehensive data that meets the preset conditions.

[0111] In some embodiments, since there may be multiple sets of comprehensive data with the same sewage status data before carbon addition in the database, the preset conditions can be further limited at this time, and the comprehensive data with the highest carbon addition benefit data can be selected from these comprehensive data as the strategic network data that meets the preset conditions.

[0112] In some embodiments, the second input subunit 432 can be used to input multiple sets of strategy network data as parameters into the state value network model to obtain the first state value data before carbon addition. After that, it can be input into the state value network model In the example, the first state value data of the corresponding sewage before carbon addition is obtained, which is expressed as .

[0113] In some embodiments, the third input subunit 433 can be used to input multiple sets of strategy network data as parameters into the action value network model to obtain the first action value data before carbon addition. Specifically, after obtaining the first state value data, the strategy network data can be input again. Input to the action-value network model In the example, the first action value data of the corresponding sewage before carbon addition is obtained, which is expressed as .

[0114] In some embodiments, the first expectation calculation subunit 434 can be used to obtain the first expected advantage data before carbon addition according to the state value data and the action value data. Specifically, after obtaining the first state value data and the first action value data, the first expected advantage data of the sewage before carbon addition can be obtained, which is expressed as .

[0115] In some embodiments, the first target search subunit 435 can be used to determine the policy target function to be optimized in the policy network model according to the first expected advantage data. Specifically, after obtaining the first expected advantage data, the parameters in the policy network model can be determined according to the first expected advantage data. And the corresponding policy objective function that needs to be optimized. Specifically, the policy objective function can be expressed as ,in, For parameters The optimization goal is represents the original expected discounted return, Represents the state transition probability of all possible sewage states in the strategy network data. After obtaining the strategy objective function, the gradient of the strategy objective function can be determined , expressed as .

[0116] In some embodiments, the first training optimization subunit 436 can be used to optimize the parameters of the policy network model and generate an intermediate carbon addition model according to the gradient of the policy objective function and the conjugate gradient ascent algorithm. Specifically, after obtaining the policy objective function and its gradient, the parameters can be optimized according to the conjugate gradient ascent algorithm. Optimize. The conjugate gradient ascent algorithm is an optimization algorithm that is mainly used to solve unconstrained optimization problems. The conjugate gradient ascent algorithm searches for the maximum value of the function along the conjugate direction in an iterative manner. In each step, the algorithm selects a new search direction, which is not only orthogonal to the gradient direction of the previous step, but also needs to meet the conjugation condition.

[0117] In some embodiments, the parameters to be optimized are It can be expressed as ,in, Representation parameters The step size of the change, , It is expressed as the smallest non-negative integer that can improve the policy and meet the threshold limit. is the threshold value, is the parameter update direction, , It is represented as a Hessian matrix, Represented as transposed.

[0118] In some embodiments, after completing the above parameters After the optimization of , the optimization of the policy network model can be completed. The optimized policy network model can be expressed as At this point, the policy network model in the initial carbon addition model has been optimized, and the intermediate carbon addition model can be obtained. The state value network model and action value network model in the intermediate carbon addition model have not been optimized.

[0119] See also Figure 6 In some embodiments, the second input unit 440 can be used to input the current data as a parameter into the intermediate carbon addition model to obtain the second comprehensive data after carbon addition. The second input unit 440 may include a fourth input subunit 441, a second benefit calculation subunit 442, and a second data association subunit 443.

[0120] In some embodiments, the fourth input subunit 441 can be used to input the current state data before carbon addition as a parameter into the strategy network model of the intermediate carbon addition model to obtain the corresponding second current carbon addition action data. Specifically, in the aforementioned optimization process, the current state data of the sewage before carbon addition has been obtained, and the strategy network model of the intermediate carbon addition model has been optimized. Therefore, the current state data of the sewage before carbon addition can be input into the strategy network model of the intermediate carbon addition model again to obtain the corresponding second current carbon addition action data. , which can be expressed as At this point, the current state data of the wastewater after carbon addition can be expressed as .

[0121] In some embodiments, the second benefit calculation subunit 442 can be used to obtain the second current carbon addition benefit data according to the current state data and the second current carbon addition action data. Specifically, after obtaining the second current carbon addition action data, the second current carbon addition benefit data can be obtained according to the current state data after carbon addition and the second current carbon addition action data. , expressed as .

[0122] In some embodiments, the second data association subunit 443 can be used to associate the current state data before and after carbon addition, the second current carbon addition action data, and the second current carbon addition benefit data to obtain the second comprehensive data. , Current status data after carbon addition 2. Current carbon addition action data And the second current carbon benefit data After that, the above data can be associated to obtain the corresponding second comprehensive data, which is expressed as .

[0123] See also Figure 7 In some embodiments, the second optimization unit 450 can be used to start with the second comprehensive data, search for comprehensive data similar to the second comprehensive data from the database, express it as value network data, optimize the state value network model and the action value network model according to the value network data, and generate a target carbon model. The second optimization unit 450 may include a second screening subunit 451, a fifth input subunit 452, a sixth input subunit 453, a second expectation calculation subunit 454, a second target search subunit 455, and a second training optimization subunit 456.

[0124] In some embodiments, the second screening subunit 451 can be used to obtain multiple groups of comprehensive data that meet the preset conditions from the database, which are expressed as value network data. Specifically, after obtaining the second comprehensive data, comprehensive data that meet the preset conditions can be selected from the database according to the preset conditions, which are expressed as value network data. The preset conditions can be the same as the aforementioned preset conditions. The obtained value network data , which can be expressed as .

[0125] In some embodiments, the fifth input subunit 452 can be used to input multiple sets of value network data as parameters into the state value network model to obtain the second state value data before carbon addition. After that, it can be input into the state value network model In the example, the second state value data of the corresponding sewage before carbon addition is obtained, which is expressed as .

[0126] In some embodiments, the sixth input subunit 453 can be used to input multiple sets of value network data as parameters into the action value network model to obtain the second action value data before carbon addition. Specifically, after obtaining the second state value data, the value network data can be input again. Input to the action-value network model In the example, the second action value data of the corresponding sewage before carbon addition is obtained, which is expressed as .

[0127] In some embodiments, the second expectation calculation subunit 454 can be used to obtain the second expected advantage data before carbon addition according to the state value data and the action value data. Specifically, after obtaining the second state value data and the second action value data, the second expected advantage data of the sewage before carbon addition can be obtained, which is expressed as .

[0128] In some embodiments, the second target search subunit 455 can be used to determine the value objective function that needs to be optimized in the state value network model and the action value network model according to the second expected advantage data. Specifically, after obtaining the second expected advantage data, the parameters in the state value network model and the action value network model can be determined according to the second expected advantage data. And the corresponding value objective function that needs to be optimized. Specifically, the value objective function can be expressed as ,in, For parameters After obtaining the value objective function, the gradient of the value objective function can be determined. , expressed as .

[0129] In some embodiments, the second training optimization subunit 456 can be used to optimize the parameters of the state value network model and the action value network model according to the gradient of the value objective function and the gradient descent algorithm, and generate a target carbon model. Specifically, after obtaining the value objective function and its gradient, the parameters can be optimized according to the gradient descent algorithm. Optimize. Gradient descent is a widely used optimization algorithm that aims to find the local minimum of a function in an iterative manner. In machine learning and deep learning, it is used to minimize the loss function, that is, to adjust the model parameters to reduce the difference between the model's predicted value and the actual value. Parameters to be optimized It can be expressed as ,in, For parameters The step size of the change, .

[0130] In some embodiments, after completing the above parameters After the optimization, the optimization of the state value network model and the action value network model can be completed. The optimized state value network model can be expressed as , the optimized action value network model can be expressed as At this point, the state value network model and the action value network model in the intermediate carbon addition model have been optimized, and the target carbon addition model can be obtained.

[0131] In some embodiments, after the processing module 400 constructs the target carbon addition model, the current data of the sewage can be input as a parameter into the target carbon addition model to generate the corresponding target carbon addition data. The target carbon addition data refers to the amount of carbon source stock solution added that can achieve the best treatment effect. The processing module 400 can generate specific carbon addition instructions based on the amount of carbon source added generated by the target carbon addition model. These carbon addition instructions may include the frequency, flow rate, time and total amount of carbon source stock solution addition. The metering pump 200 can receive the carbon addition instruction generated by the processing module 400, and accurately transport the carbon source stock solution from the dosing bin 100 to the sewage pool according to the parameters in the carbon addition instruction. The control method of the metering pump 200 may include flow control, time control and frequency control. Flow control refers to adjusting the flow of the metering pump 200 to ensure that the carbon source is added at the required rate. Time control refers to controlling the start and stop time of the metering pump 200 to achieve timed dosing. Frequency control refers to adjusting the pulse frequency of the metering pump 200 to achieve precise dosing.

[0132] It can be seen that in the above scheme, the calculation process of the amount of carbon added is automatically optimized through the deep learning algorithm, which significantly improves the accuracy and reliability of the carbon addition decision, reduces human errors, and ensures the stability of the treatment effect. By utilizing the real-time optimization capability of the deep reinforcement learning model, it is possible to timely feedback changes in the treatment process, realize the real-time adjustment of the amount of carbon added, effectively respond to emergencies that may occur during the treatment process, and ensure water treatment efficiency and effectiveness. Through the automated carbon addition control strategy, the dependence on professionals is reduced, thereby effectively reducing labor costs, while improving the convenience and economic benefits of operation. By accurately controlling the amount of carbon added, not only the efficiency of sewage treatment is improved, but also it helps to reduce potential negative impacts on the environment and promote sustainable environmental development.

[0133] The embodiments of the present invention disclosed above are only used to help illustrate the present invention. The embodiments do not describe all the details in detail, nor do they limit the invention to the specific embodiments described. Obviously, many modifications and changes can be made according to the content of this specification. This specification selects and specifically describes these embodiments in order to better explain the principles and practical applications of the present invention, so that those skilled in the art can understand and use the present invention well. The present invention is limited only by the claims and their full scope and equivalents.

Claims

1. A carbon dosing system for sewage treatment, characterized in that: include: At least one dosing chamber for storing carbon source stock solution; At least one metering pump, connected to the dosing chamber, for delivering the carbon source stock solution to the sewage pool; A detection module, used for detecting current data of sewage in the sewage pool; A processing module, for constructing a target carbon addition model based on the acquired comprehensive data of sewage treatment by the activated sludge process, and inputting the current data as parameters into the target carbon addition model to generate corresponding target carbon addition data; The processing module is also used to generate a corresponding carbon addition instruction according to the target carbon addition data, and the metering pump delivers the carbon source stock solution to the sewage pool according to the carbon addition instruction; The processing module comprises: A model building unit, used to obtain comprehensive data of sewage treatment by activated sludge process from a constructed database, and establish an initial carbon addition model based on the comprehensive data; wherein the initial carbon addition model includes a strategy network model for determining carbon addition action, a state value network model for determining sewage state value, and an action value network model for determining the value of carbon addition action; A first input unit, used for inputting the current data as a parameter into the initial carbon addition model to obtain first comprehensive data after carbon addition; A first optimization unit is used to start with the first comprehensive data, search for multiple groups of comprehensive data that meet preset conditions from the database, express them as strategy network data, optimize the strategy network model according to the strategy network data, and generate an intermediate carbon addition model; A second input unit is used to input the current data as a parameter into the intermediate carbon addition model to obtain second comprehensive data after carbon addition; A second optimization unit is used to start with the second comprehensive data, search for multiple groups of comprehensive data that meet preset conditions from the database, expressed as value network data, optimize the state value network model and the action value network model according to the value network data, and generate a target carbon-adding model; Among them, the preset condition is expressed as: calculating the difference between the current state data after carbon addition in the first comprehensive data or the second comprehensive data and the sewage state data before carbon addition in a certain group of comprehensive data, and the ratio of this difference to the current state data after carbon addition in the first comprehensive data or the second comprehensive data does not exceed a threshold value; calculating the difference between the sewage state data after carbon addition in the certain group of comprehensive data and the sewage state data before carbon addition in the next group of comprehensive data, and the ratio of this difference to the sewage state data after carbon addition in the certain group of comprehensive data does not exceed a threshold value.

2. The carbonization system for sewage treatment according to claim 1, characterized in that: The detection module comprises: A COD detection unit, used to detect the chemical oxygen demand data of the sewage; A total nitrogen detection unit, used for detecting the total nitrogen data of the sewage; A total phosphorus detection unit, used to detect the total phosphorus data of the sewage; A turbidity monitoring unit, used to monitor the turbidity data of the sewage; The liquid level detection unit is used to detect the liquid level data of the sewage.

3. The carbonization system for sewage treatment according to claim 2, characterized in that: The processing module is also used to calculate sewage volume data according to the liquid level data, and the current data includes the chemical oxygen demand data, the total nitrogen data, the total phosphorus data, the turbidity data, and the sewage volume data.

4. The carbonization system for sewage treatment according to claim 1, characterized in that: The model building unit comprises: A data acquisition subunit is used to acquire multiple groups of historical data of sewage treated by an activated sludge process; wherein the historical data include status data of sewage before and after carbon addition, and carbon addition amount data; the status data include chemical oxygen demand data, total nitrogen data, total phosphorus data, turbidity data, and sewage amount data; A data processing subunit, used for performing weighted processing on the historical data, obtaining sewage status data, and obtaining carbon addition action data based on the sewage status data; The benefit calculation subunit is used to obtain carbon addition benefit data according to the sewage state data and the carbon addition action data; The data association subunit is used to associate each group of corresponding historical data, sewage status data, carbon addition action data and carbon addition benefit data to express them as the comprehensive data; build a database based on multiple groups of the comprehensive data, and establish an initial carbon addition model based on the comprehensive data.

5. The carbonization system for sewage treatment according to claim 1, characterized in that: The first input unit comprises: A first data integration subunit, used to determine current state data according to the current data; A first input subunit, used to input the current state data as a parameter into the strategy network model of the initial carbon addition model to obtain the corresponding first current carbon addition action data; A first benefit calculation subunit is used to obtain first current carbon addition benefit data according to the current state data and the first current carbon addition action data; The first data association subunit is used to associate the current state data before and after carbon addition, the first current carbon addition action data, and the first current carbon addition benefit data to obtain first comprehensive data.

6. The carbonization system for sewage treatment according to claim 1, characterized in that: The first optimization unit comprises: A first screening subunit, used to obtain multiple groups of comprehensive data meeting preset conditions from the database, represented as strategic network data; A second input subunit is used to input the plurality of groups of the strategy network data as parameters into the state value network model to obtain the first state value data before carbon addition; A third input subunit is used to input the plurality of groups of the strategy network data as parameters into the action value network model to obtain the first action value data before carbon addition; A first expectation calculation subunit, used for obtaining first expected advantage data before carbon addition according to the state value data and the action value data; A first target finding subunit, used for determining a strategy target function to be optimized in the strategy network model according to the first expected advantage data; The first training optimization subunit is used to optimize the parameters of the policy network model according to the gradient of the policy objective function and the conjugate gradient ascent algorithm, and generate an intermediate carbon addition model.

7. The carbonization system for sewage treatment according to claim 1, characterized in that: The second input unit comprises: The fourth input subunit is used to input the current state data before carbon addition as a parameter into the strategy network model of the intermediate carbon addition model to obtain the corresponding second current carbon addition action data; A second benefit calculation subunit is used to obtain second current carbon addition benefit data according to the current state data and the second current carbon addition action data; The second data association subunit is used to associate the current state data before and after carbon addition, the second current carbon addition action data, and the second current carbon addition benefit data to obtain second comprehensive data.

8. The carbonization system for sewage treatment according to claim 1, characterized in that: The second optimization unit comprises: A second screening subunit is used to obtain multiple groups of comprehensive data that meet preset conditions from the database, expressed as value network data; A fifth input subunit, used to input the plurality of sets of the value network data as parameters into the state value network model to obtain the second state value data before carbon addition; A sixth input subunit, used to input the plurality of sets of the value network data as parameters into the action value network model, and obtain the second action value data before carbon addition; A second expectation calculation subunit, used for obtaining second expected advantage data before carbon addition according to the state value data and the action value data; A second target search subunit is used to determine the value target function to be optimized in the state value network model and the action value network model according to the second expected advantage data; The second training optimization subunit is used to optimize the parameters of the state value network model and the action value network model according to the gradient of the value objective function and the gradient descent algorithm, and generate a target carbon addition model.

Images