A model-based self-learning control method applied to water treatment

Abstract

The application provides a self-learning control method applied to water treatment and based on model ratio, which comprises the following steps: a model construction step of constructing a three-dimensional simulation environment, a strategy construction step of constructing an execution strategy library, a parameter configuration step of inputting variable parameter information to obtain a simulation mechanics model, a mapping construction step of constructing a feature mapping relationship of a measured distribution sub-data and an average distribution feature sub-data of a mapping position, a detection mapping step of determining corresponding average distribution feature sub-data according to the feature mapping relationship, and a dynamic execution step of controlling corresponding execution equipment to work according to the execution strategy. The self-learning control method applied to water treatment and based on model ratio can accurately obtain a dosing position and accurately dose a medicament for water treatment.

Description

Technical Field

[0001] This invention relates to the field of water treatment technology, and more specifically to a self-learning control method based on model ratio applied to water treatment. Background Technology

[0002] Water treatment refers to the physical and chemical measures taken to bring water quality up to certain standards. Current water treatment methods include activated carbon adsorption, permeation filtration, water softening, and biochemical methods. This solution focuses on the biochemical method. Existing water treatment technologies involve collecting the water to be treated in a reaction tank, preparing a chemical mixture based on the volume of the water to be treated, and then injecting the prepared chemical mixture into the reaction tank through an actuation device. The mixing component and the foaming component at the bottom work together to ensure that the microorganisms in the water to be treated in the reaction tank can fully contact the chemical mixture under agitation and aerobic conditions, thereby completing the biological water treatment.

[0003] However, the accuracy of content detection in current biological water treatment processes is relatively low. Conventional techniques involve vertically inserting detection sensors into the water to be treated from a few points. Due to differences in the stirring frequency of the stirring components, the foaming amount of the foaming components, and other uncontrollable variables in the treatment tank, the content of the reagents in the water to be treated is uneven, or the content of the reagents in some places is too high or too low. This makes it difficult for the treated water quality to meet the requirements, which increases the treatment cost and reduces the treatment efficiency. Summary of the Invention

[0004] In view of the shortcomings of the existing technology, the purpose of this invention is to provide a self-learning control method based on model ratio for water treatment. This self-learning control method for water treatment can accurately obtain the dosing position and accurately add the agent for water treatment.

[0005] To achieve the above objectives, the present invention provides the following technical solution:

[0006] A self-learning control method based on model ratio applied to water treatment includes the following steps:

[0007] The model building steps include constructing a three-dimensional simulation environment, which is generated based on real-world water treatment information.

[0008] The strategy construction steps include building an execution strategy library, which stores several execution strategies, and the execution strategies are indexed by the evenly distributed feature sub-data.

[0009] The parameter configuration steps involve inputting variable parameter information into a 3D simulation environment to obtain a simulation mechanical model. The simulation mechanical model reflects the simulation results at different simulation times. The simulation results include target feature distribution data, which reflects the distribution of target features.

[0010] The mapping construction step involves determining the sensor's position in the real scene to generate the mapping position, and constructing the feature mapping relationship between the measured distribution sub-data and the average distribution feature sub-data of the mapping position based on the acquired target feature distribution data.

[0011] The detection mapping step involves acquiring the measured distribution sub-data from the sensor and determining the corresponding average distribution feature sub-data based on the feature mapping relationship.

[0012] The dynamic execution steps utilize the execution strategy corresponding to the obtained average distribution feature sub-data index, and control the corresponding execution device to operate according to the execution strategy.

[0013] Furthermore, the real-time water treatment information includes the dimensions of the reaction tank, the location of the stirring component, the location of the foaming component, and the location of the chemical dosing.

[0014] Furthermore, the execution strategy includes the amount of drug input, the drug input components, the stirring speed of the stirring component, and the foaming speed of the bubble component.

[0015] Furthermore, the average distribution feature sub-data is the average concentration value of the target analyte, calculated using a definite integral algorithm. The algorithm is as follows:

[0016] Furthermore, it also includes a simulation correction step, which involves obtaining the measured distribution sub-data of the sensor, obtaining correction data by comparing the measured distribution sub-data with the average distribution feature sub-data in the target feature distribution data, and correcting the simulation mechanical model based on the correction data.

[0017] Furthermore, the mapping construction step also includes a sensor mapping suggestion sub-step, which calculates the sample test function for each coordinate in the simulation mechanics model, calculates the slope value of the sample test function, and determines the sample test function with the highest slope value as the target function.

[0018] Furthermore, the target feature distribution data is in the form of a concentration distribution map.

[0019] Furthermore, the measured distribution sub-data is the actual concentration value of the target object detected by the sensor at the mapped position, the average distribution feature sub-data is the average concentration value of the target object in the simulation mechanical model, and the mapping relationship is the ratio of the actual concentration value of the target object to the average concentration value of the target object.

[0020] The beneficial effects of this invention are as follows: A simulation model is obtained by performing three-dimensional simulation modeling on the actual scene. A simulation mechanical model can be obtained based on the simulation model and various variable parameter information. The concentration of each point and the overall thermogram can be obtained through the simulation mechanical model. By comparing the theoretical reagent concentration value mapped in the simulation mechanical model with the actual reagent concentration detected by the sensor, the simulation mechanical model can be corrected based on the comparison value. The optimal theoretical dosing point can be obtained based on the thermogram distribution in the simulation mechanical model. The optimal actual dosing point can be obtained based on the mapping relationship, so as to achieve precise dosing of reagents for water treatment. Attached Figure Description

[0021] Figure 1 This is a flowchart of the invention;

[0022] Figure 2 This is the first heat map of the present invention;

[0023] Figure 3 This is the second heat map of the present invention;

[0024] Figure 4 This is the third heat map of the present invention. Detailed Implementation

[0025] The present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Identical components are denoted by the same reference numerals. It should be noted that the terms "front," "rear," "left," "right," "upper," and "lower" used in the following description refer to directions in the accompanying drawings, and the terms "bottom surface," "top surface," "inner," and "outer" refer to directions toward or away from the geometric center of a specific component, respectively.

[0026] Because the detection accuracy of reagents in current biological water treatment processes is relatively low, conventional techniques involve vertically inserting sensors into the water at a few points for detection. However, due to variations in the stirring frequency of the mixing components, the foaming capacity of the foaming components, and other uncontrollable variables within the treatment tank, the reagent content in the water becomes uneven, leading to either excessively high or low concentrations at certain points. This results in water quality that fails to meet requirements, increasing treatment costs and reducing efficiency. Therefore, this invention designs a model-based self-learning control method for water treatment. Figure 1 As shown, it includes the following steps:

[0027] The model construction steps involve building a 3D simulation environment, which is generated based on real-world water treatment information. This information includes the dimensions of the reaction tank, the locations of the agitator components, the foaming components, and the dosing points. First, the depth, length, and width of the reaction tank are obtained, and a model of the reaction tank is created. Then, the agitator components are attached to the reaction tank model according to their positions and structures. Similarly, the foaming components are attached to the reaction tank model according to their positions and structures. Finally, the dosing points are marked on the 3D model of the reaction tank.

[0028] The strategy construction steps include building an execution strategy library, which stores several execution strategies. These strategies are indexed by average distribution feature data, which represents the average concentration of the target analyte (i.e., the average concentration of the entire reaction tank). The strategies are calculated using a definite integral algorithm. Different execution strategies contain different average concentration values. The execution strategy includes drug input amount, drug input components, stirring speed of stirring components and foaming speed of bubble components. Based on the average concentration of the reaction tank, the corresponding drug input content, drug input components, stirring speed of stirring components and foaming speed and other variable factors can be indexed in the execution strategy.

[0029] The parameter configuration steps involve inputting variable parameter information into the 3D simulation environment to obtain a simulation mechanical model. These variable parameters include factors such as the inlet, dosage, stirring speed, foaming speed, and outlet. By combining this variable parameter information with the 3D model of the reaction tank, a simulation model can be obtained through simulation, such as... Figure 2-4 As shown, the simulation mechanical model reflects the simulation results at different simulation times. The simulation results include a heat map decomposed from the simulation mechanical model at a certain time. The heat map includes the drug concentration values ​​at each point. The simulation results also include target feature distribution data, which reflects the distribution of target features. The distribution of target features is fed back by the heat map. The target feature distribution data is in the format of a concentration distribution map.

[0030] The mapping construction steps involve determining the sensor's position in the real-world scene to generate the mapping location. Based on the acquired target feature distribution data, a feature mapping relationship is constructed between the measured distribution sub-data and the average distribution feature sub-data of the mapping location. The measured distribution sub-data and the average distribution feature sub-data have already been converted in the target feature distribution data. In the actual scene, there is a mapping relationship between the data collected by the sensor at the mapping location and the data in the heat map. The measured distribution sub-data is the actual concentration value of the target object detected by the sensor at the mapping location, and the average distribution feature sub-data is the average concentration value of the target object in the simulation mechanical model. The mapping relationship is the ratio of the actual concentration value of the target object to the average concentration value of the target object.

[0031] The detection mapping step involves acquiring the measured distribution sub-data from the sensor and determining the corresponding average distribution feature sub-data based on the feature mapping relationship.

[0032] The dynamic execution steps utilize the execution strategy corresponding to the obtained average distribution feature sub-data index, and control the corresponding execution device to operate according to the execution strategy.

[0033] A simulation model is obtained by performing 3D simulation modeling on the actual scene. Based on the simulation model and various variable parameter information, a simulation mechanical model can be obtained. Then, the concentration of each point and the overall thermogram can be obtained through the simulation mechanical model. By comparing the actual drug concentration detected by the sensor with the theoretical drug concentration value mapped in the simulation mechanical model, the simulation mechanical model can be corrected based on the comparison value. The optimal theoretical dosing point can be obtained based on the thermogram distribution in the simulation mechanical model. The optimal actual dosing point can be obtained based on the mapping relationship, so as to achieve precise drug dosing for water treatment.

[0034] By establishing the diffusion coefficient of a drug, the effective diffusion law and reliable minimum reaction time can be derived through equation solving and simulation. The initial measurement value from the sensor can be used to obtain the dosage limit (concentration x volume) at that moment, and the next dosage limit value can be obtained at least after one minimum reaction time interval, thus forming a dosing control mechanism.

[0035] Since there will be some deviation between the simulated mechanical model and the actual detected concentration value, the simulation correction step is also included. This step involves obtaining the measured distribution sub-data of the sensor, comparing the measured distribution sub-data with the average distribution feature sub-data in the target feature distribution data to obtain correction data, and then correcting the simulated mechanical model based on the correction data to ensure that the subsequent execution strategy can be more accurate.

[0036] The mapping construction step also includes a sensor mapping suggestion sub-step, which calculates the sample test function for each coordinate in the simulation mechanical model, calculates the slope value of the sample test function, and determines the sample test function with the highest slope value as the target function.

[0037] First, a 3D model of the reaction tank is created. This model is then divided into several regions. Variables such as fluid properties, drug properties, flow conditions, and initial boundary conditions are substituted into the 3D model. A 3D simulation model is obtained through network calculus and CFD simulation. This model can be converted into a heat map, which reflects the relationship between the stirring speed of the agitator and the bubble generation rate at different times and the drug concentration at various points in the reaction tank. Then, the drug concentration is obtained from several difficult-to-measure locations in the reaction tank using sensors. The drug concentration values ​​at the mapped points of the actual sensor locations in the heat map are then compared to update and correct the 3D simulation model, resulting in an accurate model. Finally, based on the drug concentration data collected by the sensors, corresponding dosing control, stirring control, and bubble generation control are matched within the 3D simulation model to precisely control the drug input and improve water treatment efficiency.

[0038] The above are merely preferred embodiments of the present invention. The scope of protection of the present invention is not limited to the above embodiments. All technical solutions falling within the scope of the present invention's concept are within the scope of protection of the present invention. It should be noted that for those skilled in the art, any improvements and modifications made without departing from the principle of the present invention should also be considered within the scope of protection of the present invention.

Claims

1. A self-learning control method based on model ratio applied to water treatment, characterized in that: Includes the following steps: The model building steps include constructing a three-dimensional simulation environment, which is generated based on real-world water treatment information. The strategy construction steps include building an execution strategy library, which stores several execution strategies, and the execution strategies are indexed by the evenly distributed feature sub-data. The parameter configuration steps involve inputting variable parameter information into a 3D simulation environment to obtain a simulation mechanical model. The simulation mechanical model reflects the simulation results at different simulation times. The simulation results include target feature distribution data, which reflects the distribution of target features. The mapping construction step involves determining the sensor's position in the real scene to generate the mapping position, and constructing the feature mapping relationship between the measured distribution sub-data and the average distribution feature sub-data of the mapping position based on the acquired target feature distribution data. The detection mapping step involves acquiring the measured distribution sub-data from the sensor and determining the corresponding average distribution feature sub-data based on the feature mapping relationship. The dynamic execution steps are based on the execution strategy corresponding to the obtained average distribution feature sub-data index, and the corresponding execution device is controlled to work according to the execution strategy. The real-time water treatment information includes the dimensions of the reaction tank, the location of the stirring components, the location of the foaming components, and the location of the chemical dosing. The execution strategy includes the amount of drug input, the drug input components, the stirring speed of the stirring component, and the foaming speed of the bubble component. The average distribution feature subdata is the average concentration value of the target analyte, calculated using a definite integral algorithm. The algorithm is as follows: It also includes a simulation correction step, which involves obtaining the measured distribution sub-data of the sensor, obtaining correction data by comparing the measured distribution sub-data with the average distribution feature sub-data in the target feature distribution data, and correcting the simulation mechanical model based on the correction data.

2. The self-learning control method based on model ratio applied to water treatment according to claim 1, characterized in that: The mapping construction step also includes a sensor mapping suggestion sub-step, which calculates the sample test function for each coordinate in the simulation mechanical model, calculates the slope value of the sample test function, and determines the sample test function with the highest slope value as the target function.

3. The self-learning control method based on model ratio applied to water treatment according to claim 2, characterized in that: The target feature distribution data is presented in the form of a concentration distribution map.

4. The self-learning control method based on model ratio applied to water treatment according to claim 3, characterized in that: The measured distribution sub-data is the actual concentration value of the target object detected by the sensor at the mapped position, the average distribution feature sub-data is the average concentration value of the target object in the simulation mechanical model, and the mapping relationship is the ratio of the actual concentration value of the target object to the average concentration value of the target object.

Images