Sewage treatment control method, device and system and storage medium thereof

Abstract

The invention provides a sewage treatment control method, device and system and a storage medium thereof, and relates to the technical field of sewage treatment. The control method comprises the following steps: establishing a dynamic simulation model; constructing a full data set; constructing an artificial neural network model based on the full data set and performing training; and acquiring actual inflow water quality data, obtaining actual process control parameters by using the dynamic simulation model and the artificial neural network model, and controlling the operation of sewage treatment. According to the sewage treatment control method, the neural network model is trained by using simulation data, and the control parameters are calculated according to the actual inlet water quality, so that the automation and intelligence of sewage treatment are realized, and the control precision and the system adaptability are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of sewage treatment, in particular to a sewage treatment control method, device and system, and a storage medium thereof. BACKGROUND

[0002] Water pollution problems are becoming increasingly complex. In addition to common pollutants, heavy metals, antibiotics, drug residues and water eutrophication and other new types of pollution are frequently occurring, which significantly exacerbate the threat to the ecosystem and public health. In order to cope with the diversification of water quality, various biochemical treatment processes have been developed in the field of sewage treatment, including CAST, A2O, AO, MBR, SBR, OD and AB, etc. These processes take activated sludge systems as the core, remove pollutants through microbial metabolism, and improve treatment effect by combining operation control strategies.

[0003] Activated sludge models (such as ASM2d) are gradually applied to simulate biochemical processes and predict effluent water quality, and play an important role in assisting process optimization. However, such models are usually limited to biological treatment units and lack comprehensive description of the combined effects of dosing, aeration, reflux and other control links in the system. They rely on a large number of parameters, have complex model structures, and require high input precision, making it difficult to adapt to rapid changes in actual operation.

[0004] In the case of sudden water quality disturbance (such as rainstorm, pollution source leakage), activated sludge models cannot reflect the system changes in real time, and lack the ability to directly convert simulation results into process control instructions. Current sewage plant control still largely relies on manual judgment, and lacks effective data-driven mechanisms to connect prediction models with actual control systems.

[0005] Therefore, although activated sludge models provide a certain basis for water quality prediction, there are still significant deficiencies in dynamic and adaptive control, which limit the further development of sewage treatment processes towards automation and intelligentization. SUMMARY

[0006] Therefore, the present application aims to provide a sewage treatment control method, device, system and storage medium thereof, which utilizes simulation data to train a neural network model and calculates control parameters according to actual influent water quality, achieving automation and intelligentization of sewage treatment and improving control precision and system adaptability.

[0007] In a first aspect, the present application provides a sewage treatment control method, comprising: establishing a dynamic simulation model of the sewage treatment process based on an ASM2d model; Obtain input water quality data of a target water source; obtain output water quality data corresponding to the input water quality data under different operating conditions, as well as simulated process control parameters under the corresponding operating conditions, using the dynamic simulation model; and construct a full data set based on the input water quality data, the output water quality data, and the simulated process control parameters; wherein the full data set excludes the input water quality data, the output water quality data, and the simulated process control parameters that do not meet preset water quality standards; Constructing an artificial neural network model based on the full data set, and training the artificial neural network model to obtain the trained artificial neural network model; wherein the artificial neural network model includes a first model trained based on the input water quality data and a second model trained based on the output water quality data; The actual influent water quality data of the target water source is obtained, the actual process control parameters corresponding to the target water source are obtained using the dynamic simulation model and the artificial neural network model, and the operation of the sewage treatment is controlled based on the actual process control parameters.

[0008] In an optional embodiment, the dynamic simulation model of the sewage treatment process is established based on the ASM2d model, including: Build dynamic simulation models; Acquiring historical operating data of the target water source, and performing steady-state simulation verification on the dynamic simulation model based on the historical operating data; Performing sensitivity analysis on the dynamic simulation model after steady-state simulation verification, and individually optimizing and adjusting the confirmed model parameters based on the sensitivity to obtain the dynamic simulation model with optimized accuracy; Through dynamic simulation, the prediction effect of the dynamic simulation model based on the model parameters after the individual optimization adjustment is verified, and the dynamic simulation model after dynamic simulation is obtained.

[0009] In an optional embodiment, the expression of the sensitivity analysis is: ; Among them, a represents the input model parameter of the dynamic simulation model; a1 and a2 represent the typical value and the changed value of the model parameter respectively; y represents the water quality index that the dynamic simulation model focuses on during the simulation process; y1 and y2 represent the y value of the corresponding simulation output of the water quality index corresponding to the water quality index under the conditions of a1 and a2 respectively; S represents the sensitivity coefficient of the model parameter to the water quality index.

[0010] In an optional embodiment, the individually optimizing and adjusting the confirmed model parameters based on sensitivity includes: Determining whether the absolute value of the sensitivity of the model parameter is greater than a preset sensitivity threshold; If so, the model parameter whose absolute value of sensitivity is greater than the preset sensitivity threshold is used as the parameter to be adjusted; Taking the default value of the dynamic simulation model as the center, setting the variation range and step size of each parameter to be adjusted, and constructing a value sequence of the parameter to be adjusted; Combining all the values ​​of the parameters to be adjusted in an orthogonal manner according to the value sequence to generate a plurality of parameter combinations for simulation analysis; Inputting each set of parameter combinations into the dynamic simulation model to obtain corresponding simulated output water quality indicators; Calculating a corresponding prediction error for each of the simulated output water quality indicators; the prediction error includes a coefficient of determination and a mean square error; Among all the parameter combinations, a set of parameter combinations with the largest determination coefficient and the smallest mean square error is selected as the optimized model parameters, so as to update the dynamic simulation model according to the model parameters.

[0011] In an optional embodiment, the determination coefficient is calculated as follows: ; Among them, R 2 represents the coefficient of determination; Represents the true value of the i-th data point; represents the predicted value of the ith data point output by the dynamic simulation model; represents the sample mean; is the error caused by the prediction, is the error caused by the mean.

[0012] In an optional embodiment, the mean square error is calculated as follows: ; Wherein, MES represents the mean square error; m represents the total number of samples, represents the true value of the i-th data point, represents the predicted value of the ith data point output by the dynamic simulation model; represents the sample mean.

[0013] In an optional embodiment, the training of the artificial neural network model includes: Standardizing the entire dataset, performing principal component analysis on the data in the standardized dataset, and dividing the data into a training set and a validation set; Training is performed using the first model and the second model in the artificial neural network model based on the training set, and the prediction performance of all the artificial neural network models is evaluated according to the absolute percentage error; wherein the training comprises inputting the input water quality data into the first model to obtain the simulated process control parameters corresponding to the input water quality data; and inputting the input water quality data and the output water quality data into the second model to obtain the simulated process control parameters corresponding to the input water quality data and the output water quality data; Determining the type of the preferred artificial neural network model according to the prediction performance, and using a loop algorithm to determine the network structure and hyperparameter configuration of the preferred type of artificial neural network model as the preferred configuration; According to the preferred configuration, the artificial neural network model is retrained until a trained artificial neural network model is obtained.

[0014] In an optional embodiment, the expression for the normalization process is: ; in, represents a standardized score; x represents a specific variable in the entire data set; μ represents the mean value of the input variable in the entire data set; σ represents the standard deviation of the input variable in the entire data set.

[0015] In an optional embodiment, the expression of the principal component analysis is: ; in, represents the principal component analysis matrix; Represents the normalized numerical matrix before PCA; represent and Numeric matrix after principal component analysis.

[0016] In an optional embodiment, the calculation formula of the absolute percentage error is: ; Wherein, MAPE stands for absolute percentage error; n stands for the number of test sets; Represents the analog output value of data point i; Represents the actual output value of data point i.

[0017] In an optional embodiment, the obtaining of actual influent water quality data of the target water source and the use of the dynamic simulation model and the artificial neural network model to obtain actual process control parameters corresponding to the target water source include: Collecting actual influent water quality data of the target water source and obtaining predicted process parameters using the first model; Based on the predicted process parameters, the actual influent water quality data is simulated by the dynamic simulation model to obtain the actual effluent water quality data of the target water source; Determining whether the actual water output data meets the preset water quality standard; If so, output the predicted process parameters as the actual process control parameters; If not, the actual inlet water quality data and the actual outlet water quality data are input into the second model to obtain new process parameters; the new process parameters are used as the predicted process parameters, and the step is returned to simulate the actual inlet water quality data through the dynamic simulation model based on the predicted process parameters until the predicted process parameters that meet the preset water quality standards are obtained, and output as the actual process control parameters.

[0018] In a second aspect, the present invention provides a sewage treatment control device, comprising: Model building module, used to build a dynamic simulation model of the sewage treatment process based on the ASM2d model; The model building module is further configured to obtain input water quality data of a target water source; utilize the dynamic simulation model to obtain output water quality data corresponding to the input water quality data under different operating conditions, as well as simulated process control parameters under the corresponding operating conditions; and construct a full data set based on the input water quality data, the output water quality data, and the simulated process control parameters; wherein the full data set excludes the input water quality data, the output water quality data, and the simulated process control parameters that do not meet preset water quality standards; A model training module is used to construct an artificial neural network model based on the full data set and train the artificial neural network model to obtain the trained artificial neural network model; wherein the artificial neural network model includes a first model trained based on the input water quality data and a second model trained based on the output water quality data; The system control module is used to obtain the actual influent water quality data of the target water source, use the dynamic simulation model and the artificial neural network model to obtain the actual process control parameters corresponding to the target water source, and control the operation of the sewage treatment based on the actual process control parameters.

[0019] In a third aspect, the present invention provides a sewage treatment control system, which includes an information collection device, a central control system, and a PLC execution system; The central control system includes a processor and a memory; the memory stores a computer program, and the processor is used to execute the computer program to implement the sewage treatment control method described in any one of the aforementioned embodiments.

[0020] In a fourth aspect, the present invention provides a computer storage medium storing a computer program, which, when executed on a processor, implements the sewage treatment control method described in any one of the aforementioned embodiments.

[0021] The sewage treatment control method, device, system and computer storage medium provided in the embodiments of the present invention establish a dynamic simulation model of the sewage treatment process based on the ASM2d model, construct a full data set based on the input and output data and process control parameters under different working conditions generated by the model, and then train an artificial neural network model. Based on the model and actual influent water quality data, corresponding process control parameters are obtained to achieve sewage treatment operation control.

[0022] By combining the mechanism simulation capability of the ASM2d model with the nonlinear mapping capability of the artificial neural network, the present invention can accurately calculate the actual process control parameters required for sewage treatment on the basis of constructing an integrated model of dynamic simulation and data training, thereby realizing automated and intelligent control of the sewage treatment process, and improving the system's control accuracy, response speed, and adaptability to complex water quality conditions.

[0023] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, preferred embodiments are given below and described in detail with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the embodiments. It should be understood that the following drawings only illustrate certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other relevant drawings can be obtained based on these drawings without making any creative efforts.

[0025] Figure 1 Schematic diagram of the hardware operating environment involved in the sewage treatment control method embodiment of the present invention; Figure 2 Schematic diagram of the process of sewage treatment control method embodiment 1 of the present invention; Figure 3 This is a schematic diagram of a detailed flow chart of step S100 in Example 2 of the sewage treatment control method of the present invention; Figure 4 This is a schematic diagram of a detailed flow chart of step S130 in Example 2 of the sewage treatment control method of the present invention; Figure 5This is a schematic diagram of a detailed flow chart of step S300 in Example 3 of the sewage treatment control method of the present invention; Figure 6 Schematic diagram of a detailed flow chart of step S400 in embodiment 4 of the sewage treatment control method of the present invention; Figure 7 This is a schematic diagram of module connections of the sewage treatment control device of the present invention. DETAILED DESCRIPTION

[0026] The technical solutions in the embodiments of the present application will be described clearly and completely below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments.

[0027] The components of the embodiments of the present application generally described and illustrated in the drawings herein may be arranged and designed in a variety of different configurations. Therefore, the following detailed description of the embodiments of the present application provided in the drawings is not intended to limit the scope of the claimed application, but rather merely represents selected embodiments of the present application. All other embodiments obtained by those skilled in the art based on the embodiments of the present application without creative effort are within the scope of protection of the present application.

[0028] Hereinafter, the terms "including", "having" and their cognates, which may be used in various embodiments of the present application, are intended only to indicate specific features, numbers, steps, operations, elements, components or combinations of the foregoing items, and should not be understood as first excluding the existence of one or more other features, numbers, steps, operations, elements, components or combinations of the foregoing items or the possibility of adding one or more features, numbers, steps, operations, elements, components or combinations of the foregoing items.

[0029] Furthermore, the terms “first,” “second,” “third,” etc., are merely used for distinguishing descriptions and are not to be understood as indicating or implying relative importance.

[0030] Unless otherwise defined, all terms used herein (including technical and scientific terms) have the same meaning as commonly understood by those skilled in the art to which the various embodiments of the present application belong. The terms (such as those defined in generally used dictionaries) will be interpreted as having the same meaning as in the context of the relevant technical field and will not be interpreted as having an idealized meaning or an overly formal meaning unless clearly defined in the various embodiments of the present application.

[0031] The following describes some embodiments of the present application in detail with reference to the accompanying drawings. In the absence of conflict, the following embodiments and features in the embodiments may be combined with each other.

[0032] like Figure 1, which is a schematic diagram of the structure of the hardware operating environment of the terminal involved in the embodiment of the present invention.

[0033] The sewage treatment control system according to embodiments of the present invention can be a PC or a mobile terminal device such as a smartphone, tablet computer, or laptop computer. The visual navigation sewage treatment control system may include a processor 1001, such as a CPU, a network interface 1004, a user interface 1003, a memory 1005, and a communication bus 1002. Communication bus 1002 is used to enable communication between these components. User interface 1003 may include a display screen, an input unit such as a keyboard, or a remote control. Optionally, user interface 1003 may also include a standard wired interface or a wireless interface. Network interface 1004 may optionally include a standard wired interface or a wireless interface (such as a Wi-Fi interface). Memory 1005 may be a high-speed RAM memory or a stable memory such as a disk drive. Memory 1005 may also be a storage device independent of processor 1001. Optionally, the sewage treatment control system may also include RF (Radio Frequency) circuitry, audio circuitry, a Wi-Fi module, and the like. In addition, the VSLAM visual navigation optimization system can also be configured with other sensors such as gyroscopes, barometers, hygrometers, thermometers, infrared sensors, etc., which will not be described here.

[0034] Those skilled in the art will understand that Figure 1 The sewage treatment control system shown in the figure does not constitute a limitation thereof, and may include more or fewer components than shown in the figure, or combine certain components, or arrange the components differently. Figure 1 As shown, the memory 1005 as a computer-readable storage medium may include an operating system, a data interface control program, a network connection program, and a sewage treatment control program.

[0035] In summary, the method provided by the present invention fully utilizes the ASM2d model's ability to describe biochemical processes and the artificial neural network's ability to learn complex relationships, achieving the beneficial effects of intelligent, precise and automated control of the sewage treatment system.

[0036] Example 1 Reference Figure 2 , this embodiment provides a sewage treatment control method, including: Step S100: establishing a dynamic simulation model of the sewage treatment process based on the ASM2d model.

[0037] It should be noted that ASM2d model is the abbreviation of Activated Sludge Model No. 2d, which is an improved version of the Activated Sludge Model 2d, and is one of the activated sludge model series developed by the International Water Association (IWA). It is specifically used to simulate and optimize wastewater biological treatment processes, especially processes for simultaneous removal of carbon, nitrogen and phosphorus.

[0038] This step establishes a mathematical description model of the wastewater biochemical treatment process by improving the ASM2d model. It is suitable for AAO and other processes, and simulates the removal process of organic matter, nitrogen, phosphorus and other pollutants in wastewater.

[0039] Specifically, based on the ASM2d model, the nitrogen transformation process in the wastewater treatment process is improved to build a dynamic simulation model. The improved model can more accurately simulate the nitrification process and improve the prediction accuracy of the model. Through sensitivity analysis and parameter adjustment, the model parameters are optimized to further improve the accuracy and reliability of the model.

[0040] In this step, a model that can dynamically respond to different water quality conditions is obtained to predict the effluent water quality. The improved model can more accurately simulate the nitrogen transformation mechanism and improve the prediction accuracy. It is suitable for subsequent process parameter calculation.

[0041] Step S200, obtaining input water quality data of a target water source; using the dynamic simulation model to obtain output water quality data corresponding to the input water quality data under different working conditions, and simulation process control parameters under corresponding working conditions, and constructing a full data set based on the input water quality data, the output water quality data and the simulation process control parameters; wherein the full data set excludes the input water quality data, the output water quality data and the simulation process control parameters that do not meet the preset water quality standard.

[0042] The above target water source refers to a specific water source or the source of influent of a wastewater treatment plant that needs to be treated. It is the specific object of the wastewater treatment system or control method. The "target water source" refers to a specific water source that needs to be treated, which can include but is not limited to: influent of a municipal wastewater treatment plant: mixed influent from municipal domestic wastewater, industrial wastewater, etc.; influent of an industrial wastewater treatment plant: wastewater generated in a specific industrial production process; wastewater in a specific area: such as wastewater in an industrial park, residential area or commercial area.

[0043] "Input water quality data" refers to a series of parameters used to describe the characteristics of wastewater, which reflect the pollution level, composition and properties of wastewater. These data are the basis for the operation and control of wastewater treatment systems, and are also important inputs for dynamic simulation models and artificial neural network models.

[0044] The preset water quality standard described above can be Class A, the highest sewage treatment discharge standard specified in China's "Pollutant Discharge Standard for Urban Wastewater Treatment Plants" (GB 18918-2002). This standard primarily applies to the basic requirements for urban sewage treatment plant effluent used as recycled water, or when sewage treatment plant effluent is diverted to rivers and lakes with low dilution capacity for urban landscape water and general reuse. Other water quality standards may also be used. In this embodiment, the preset water quality data used is GB 18918-2002, which is Class A.

[0045] In this step, a dynamic simulation model is used to obtain corresponding output water quality data and simulation process control parameters based on different operating conditions and input water quality data, and a complete data set is constructed. At the same time, data that does not meet the preset water quality standards is excluded.

[0046] Input water quality data is fed into the dynamic simulation model, and the model is run to generate output water quality data and simulated process control parameters. Then, based on the preset water quality standards, data that meets the standards is screened, resulting in a complete data set containing input water quality data, output water quality data, and simulated process control parameters that meet the water quality standards.

[0047] The construction of a full data set ensures the quality of data used to train artificial neural networks and improves the accuracy and reliability of the model.

[0048] It should be noted that in the process of constructing the full data set, after comparing the effluent water quality data with the preset water quality standard (such as Level A standard), the combinations whose effluent water quality does not meet the standard can be deleted, and the remaining data that meet the standard can be combined to form the full data set.

[0049] This step generates a complete dataset containing input and output water quality data under various operating conditions, as well as process control parameters. This provides rich data for training the artificial neural network model, enabling it to learn the input-output relationship under different operating conditions. This improves the generalization and adaptability of the artificial neural network model, enabling it to better cope with the complexities of actual wastewater treatment processes.

[0050] Step S300, constructing an artificial neural network model based on the full data set, and training the artificial neural network model to obtain the trained artificial neural network model; wherein the artificial neural network model includes a first model trained based on the input water quality data and a second model trained based on the output water quality data; Using the full data set constructed in step S200, an artificial neural network model is constructed and trained to enable it to predict corresponding process control parameters based on the input water quality data. This step results in a trained artificial neural network model that can predict corresponding process control parameters based on the input water quality data.

[0051] It should be noted that the above-mentioned “well trained” refers to the completion of the artificial neural network model training, and the completion standard is to achieve a high training accuracy, for example, it may include but is not limited to: the training accuracy is not less than 80%.

[0052] Artificial neural network models have strong nonlinear fitting and adaptive learning capabilities, and can handle complex input-output relationships. During the training process, the input parameters of the first model are influent water quality data, and the output parameter is process parameter A. The input parameters of the second model are influent and effluent water quality data, and the output is process parameter B.

[0053] Through training, the model can learn the potential relationship between input water quality and process control parameters under different working conditions, providing better control parameters for actual operation.

[0054] Step S400, obtaining actual influent water quality data of the target water source, using the dynamic simulation model and the artificial neural network model to obtain actual process control parameters corresponding to the target water source, and controlling the operation of sewage treatment based on the actual process control parameters.

[0055] The actual inlet water quality data may be the water quality data of the target water source at the current stage, or the historical water quality data of the target water source. The data may be the same as or different from the "input water quality data." In this embodiment, the actual inlet water quality data may be the inlet water quality data at a certain stage.

[0056] In this step, the trained artificial neural network model is applied to the actual sewage treatment process. Based on the actual influent water quality data of the target water source, the corresponding process control parameters are predicted and used to control the sewage treatment operation. It should be noted that the obtained process control parameters must meet the preset water quality standards to ensure that the sewage treatment meets the standards.

[0057] For example, actual influent water quality data from a target water source can be obtained in real time. This data is then fed into a dynamic simulation model and an artificial neural network model to generate predicted process control parameters. Based on these predicted process control parameters, a control system such as a PLC (Programmable Logic Controller) can be used to adjust the operating parameters of the sewage treatment equipment, such as aeration rate, chemical dosage, and sludge discharge.

[0058] In this step, process control parameters for sewage treatment are adjusted in real time based on actual influent water quality data to ensure treatment process stability and compliance with effluent quality standards. This achieves intelligent control of the sewage treatment process, enabling automatic adjustment of process control parameters and reducing manual intervention. This improves the system's adaptability and flexibility, enabling rapid response to water quality changes and ensuring compliance with effluent quality standards. This optimizes operating parameters, reduces operating costs, and improves efficiency.

[0059] The sewage treatment control method provided in this embodiment achieves precise simulation, optimized control, and intelligent operation of the sewage treatment process by coupling the ASM2D model with an artificial neural network model. Each step has a clear treatment content, results, and advantages, and the specific implementation method is also relatively clear. This method can effectively improve the stability of sewage treatment and the compliance rate of effluent water quality, while reducing operating costs, and has significant engineering application value.

[0060] Example 2 Reference Figure 3 Based on the above embodiment 1, this embodiment provides a sewage treatment control method, wherein step S100 establishes a dynamic simulation model of the sewage treatment process based on the ASM2d model, including: Step S110: constructing a dynamic simulation model.

[0061] When constructing the dynamic simulation model, the nitrogen conversion process for the ASM2d model was expanded from a one-step reaction to a two-step reaction. Specifically, the nitrite component was added and the autotrophic nitrifying bacteria (AUT) were split into ammonia-oxidizing bacteria (AOB) and nitrite-oxidizing bacteria (NOB). In other words, ammonia nitrogen is first oxidized to nitrite by the ammonia-oxidizing bacteria (AOB), and then nitrite-oxidizing bacteria (NOB) oxidize nitrite to nitrate.

[0062] The dynamic simulation model constructed in step S110 is based on a nitrogen conversion process comprising two steps, namely: (1) Ammonia oxidizing bacteria oxidize ammonia to nitrite; (2) The nitrite is oxidized to nitrate by nitrite-oxidizing bacteria.

[0063] This step yields an improved ASM2d dynamic simulation model that can more accurately simulate the nitrogen transformation process. This improves the model's accuracy in simulating the nitrogen transformation process and enables it to more accurately predict effluent water quality.

[0064] Specifically, a nitrite component can be added to the ASM2d model, and autotrophic nitrifying bacteria (AUT) can be split into ammonia-oxidizing bacteria (AOB) and nitrite-oxidizing bacteria (NOB). A differential equation of reaction rate and a conservation matrix are constructed using mathematical software (such as MATLAB), for example, the reaction matrix constructed in the ASM2D model is solved by MATLAB using the Runge-Kutta-Fehlberg method.

[0065] In step S120, historical operation data of the target water source is obtained, and a steady-state simulation verification is performed on the dynamic simulation model according to the historical operation data.

[0066] The above-mentioned "target water source" refers to a specific water source or an influent source of a wastewater treatment plant that needs to be treated. This name is used to clearly indicate the water source object to which the wastewater treatment system or control method is directed.

[0067] In this step, the historical operation data of the target water source is used to perform a steady-state simulation verification on the dynamic simulation model. The specific data processing needs to collect the historical operation data of the target water source, including influent water quality, effluent water quality, and process control parameters. The steady-state simulation is performed on the dynamic simulation model using these historical operation data to verify the running stability of the model. Thus, the dynamic simulation model after steady-state simulation verification ensures the running stability of the model under steady-state conditions.

[0068] The stability of the model under actual operating conditions is verified in this step, which provides a basis for subsequent sensitivity analysis and optimization adjustment.

[0069] Specifically, the historical operation data of the target water source can be used as input to run the ASM2d dynamic simulation model. The effluent water quality output by the model is compared with the effluent water quality in the actual historical operation data to verify the running stability of the model.

[0070] In step S130, a sensitivity analysis is performed on the dynamic simulation model after steady-state simulation verification, and the identified model parameters are individually optimized and adjusted based on the sensitivity to obtain the dynamic simulation model after optimization of the accuracy.

[0071] In this step, the dynamic simulation model after steady-state simulation verification is subjected to a sensitivity analysis, and the model parameters are individually optimized and adjusted according to the results of the sensitivity analysis. First, the sensitivity of the parameters in the model is analyzed to determine which parameters have a greater impact on the model output. Then, the parameters with high sensitivity are individually adjusted to optimize the model parameters, thereby obtaining the dynamic simulation model after sensitivity analysis and parameter optimization adjustment, which improves the prediction accuracy of the model.

[0072] Through sensitivity analysis and parameter optimization adjustment, the prediction accuracy of the model is improved, and the model can more accurately reflect the actual wastewater treatment process.

[0073] Step S140 , verifying the prediction effect of the dynamic simulation model based on the individually optimized and adjusted model parameters through dynamic simulation, and obtaining the dynamic simulation model after dynamic simulation.

[0074] In this step, dynamic simulations are performed using the optimized and adjusted dynamic simulation model to verify the model's predictions. Dynamic simulations are performed using the optimized and adjusted dynamic simulation model, and then the predicted effluent quality is compared with the effluent quality in the actual operating data to verify the model's predictions.

[0075] The dynamic simulation model has been verified by dynamic simulation to ensure the prediction effect of the model under dynamic conditions. The prediction effect of the model under dynamic conditions has been verified to ensure that the model can accurately predict the effluent water quality under different working conditions.

[0076] Specifically, the system uses historical operating data from the target water source as input and runs a dynamic simulation model with optimized parameters. The model's predicted effluent quality is compared with the actual effluent quality from the operating data to evaluate the model's predictive performance.

[0077] The method provided in this embodiment clarifies the specific steps and methods for constructing a dynamic simulation model. Through these steps, a dynamic simulation model that reflects the actual sewage treatment process can be more accurately constructed, providing a more reliable foundation for subsequent full data set construction and artificial neural network model training.

[0078] Furthermore, in step S130, the expression for sensitivity analysis is: ; Among them, a represents the input model parameter of the dynamic simulation model; a1 and a2 represent the typical value and the changed value of the model parameter respectively; y represents the water quality index that the dynamic simulation model focuses on during the simulation process; y1 and y2 represent the y value of the corresponding simulation output of the water quality index corresponding to the water quality index under the conditions of a1 and a2 respectively; S represents the sensitivity coefficient of the model parameter to the water quality index.

[0079] It’s important to note that sensitivity analysis is a mathematical method used to assess how sensitive a model’s output is to changes in its input parameters. Through sensitivity analysis, we can identify which parameters have the greatest impact on the model’s output, allowing us to focus on these parameters during model optimization.

[0080] Sensitivity analysis can identify parameters with the greatest impact on water quality indicators, allowing you to focus on these parameters during model optimization. Individually adjusting and optimizing highly sensitive parameters can improve model prediction accuracy. Avoiding indiscriminate optimization of all parameters saves computing resources and time.

[0081] Specifically, simulations can be performed using models or mathematical software (MATLAB, R Studio). Each parameter is perturbed individually and the sensitivity coefficient is calculated. Based on the sensitivity coefficient results, parameters with higher sensitivity are optimized. The sensitivity analysis method in this approach systematically evaluates the impact of model parameters on water quality indicators, providing a scientific basis for optimizing dynamic simulation models.

[0082] refer to Figure 4 The step S130 is to individually optimize and adjust the confirmed model parameters based on sensitivity, including: Step S131 , determining whether the absolute value of the sensitivity of the model parameter is greater than a preset sensitivity threshold.

[0083] In this step, we first determine which model parameters have the greatest impact on water quality indicators. We calculate the sensitivity coefficient S for each model parameter and compare the absolute value of the sensitivity coefficient, |S|, with a preset sensitivity threshold. We identify parameters whose absolute sensitivity exceeds the threshold and thus have the greatest impact on the model output.

[0084] This step helps identify the parameters that have a greater impact on the model output and avoids unnecessary adjustments to unimportant parameters.

[0085] Specifically, the sensitivity analysis formula can be used to calculate the sensitivity coefficient S of each parameter. A sensitivity threshold (such as |S|>2) is set to screen out the sensitivity.

[0086] Step S132: If yes, the model parameter whose absolute value of sensitivity is greater than the preset sensitivity threshold is used as a parameter to be adjusted; In this step, the parameters with higher sensitivity are taken as the objects that need to be optimized and adjusted. Among them, the parameters with absolute sensitivity values ​​greater than the preset threshold are marked as parameters to be adjusted, thereby determining the parameter set that needs to be optimized and adjusted.

[0087] This step ensures that optimization adjustments are focused on parameters that have a greater impact on model output, improving optimization efficiency. Specifically, based on the sensitivity analysis results, parameters with absolute sensitivity values ​​greater than a threshold can be selected.

[0088] If not, that is, if the absolute value of the sensitivity of the model parameters is not greater than the preset sensitivity threshold (e.g., |S| ≤ 2), then these parameters have little impact on the model output. In this case, further optimization is usually not required.

[0089] For example, you can: (1) Keep the default values: For parameters whose absolute sensitivity is not greater than the preset threshold, keep the default values ​​in the model. These parameters have little impact on the model output and therefore do not need to be optimized.

[0090] (2) Record and ignore: Mark these parameters as “low sensitivity parameters” and ignore them in the subsequent optimization process. These parameters will not participate in the subsequent optimization adjustment steps.

[0091] Step S133 , taking the default value of the dynamic simulation model as the center, setting the variation range and step size of each parameter to be adjusted, and constructing a value sequence of the parameter to be adjusted.

[0092] In this step, a variation range and step size are set for each parameter to be adjusted to generate a parameter value sequence.

[0093] You can set the parameter range (e.g., 0.5 above and below) with the model default value as the center. Set the step size (e.g., 0.1) to generate a sequence of parameter values, thereby obtaining a sequence of values ​​for each parameter to be adjusted.

[0094] A clear search range and step size are provided for parameter optimization to ensure the systematic and comprehensive optimization process.

[0095] Assuming that the default value of the model is a0, the range of variation is [-0.5, 0.5], that is, it extends up and down by 0.5, and the step size is 0.1, then the value sequence can be: [a0-0.5, a0-0.4,……a0+0.5].

[0096] Step S134 : combining all the values ​​of the parameters to be adjusted in an orthogonal manner according to the value sequence to generate a plurality of parameter combinations for simulation analysis.

[0097] In this step, the values ​​of all parameters to be adjusted are combined to generate multiple parameter combinations for simulation analysis.

[0098] Using the orthogonal design method, the values ​​of all parameters to be adjusted are combined to generate multiple different parameter combinations, thereby obtaining multiple parameter combinations for subsequent simulation analysis.

[0099] The orthogonal design method can efficiently cover the parameter space, reduce the number of simulations, and improve optimization efficiency.

[0100] Generate parameter combinations using an orthogonal design table or algorithm. For example, if there are two parameters to be adjusted, a and b, and their value sequences are [a1, a2, a3] and [b1, b2, b3], respectively, the generated parameter combinations could be (a1, b1), (a1, b2), (a1, b3), (a2, b1), and so on.

[0101] Step S135 , inputting each set of the parameter combinations into the dynamic simulation model to obtain corresponding simulated output water quality indicators.

[0102] In this step, each parameter combination is input into the dynamic simulation model, the simulation is run, and the corresponding output water quality index is obtained. The dynamic simulation model can be run for each parameter combination. The simulated output water quality index corresponding to each parameter combination is recorded to obtain the simulated output water quality index corresponding to each parameter combination.

[0103] Through simulation analysis, the impact of different parameter combinations on model output can be evaluated. Specifically, a dynamic simulation model (such as the ASM2d model) can be used to run each parameter combination and record the output water quality indicators.

[0104] Step S136 , calculating a corresponding prediction error for each of the simulated output water quality indicators; the prediction error includes a coefficient of determination and a mean square error.

[0105] In this step, the prediction error of each simulated output water quality indicator is calculated, including the coefficient of determination (R 2 ) and mean square error (MSE).

[0106] Calculate the prediction error for each parameter combination. Use the coefficient of determination and mean square error as evaluation metrics to determine the prediction error for each parameter combination. Use the prediction error to evaluate the performance of each parameter combination and provide a basis for selecting the optimal parameter combination.

[0107] Furthermore, in step S136, the determination coefficient is calculated as follows: ; Among them, R 2 represents the coefficient of determination; Represents the true value of the i-th data point; represents the predicted value of the ith data point output by the dynamic simulation model; represents the sample mean; is the error caused by the prediction, is the error caused by the mean.

[0108] Furthermore, in step S136, the mean square error is calculated as follows: ; Wherein, MES represents the mean square error; m represents the total number of samples, represents the true value of the i-th data point, represents the predicted value of the ith data point output by the dynamic simulation model; represents the sample mean.

[0109] Step S137 , selecting a set of parameter combinations with the largest determination coefficient and the smallest mean square error among all the parameter combinations as the optimized model parameters, so as to update the dynamic simulation model according to the model parameters.

[0110] In this step, the optimal parameter combination is selected from all possible parameter combinations to update the dynamic simulation model. The prediction errors of all parameter combinations can be compared, and the parameter combination with the largest coefficient of determination and the smallest mean square error is selected to obtain the optimized model parameters.

[0111] By optimizing parameters, the prediction accuracy and reliability of the model can be improved. Specifically, all parameter combinations can be traversed and the R of each combination can be calculated. 2 and MSE. Select R 2 The parameter combination with the largest and smallest MSE.

[0112] This example provides a specific method for optimizing model parameters based on sensitivity analysis. Through these steps, we can systematically identify and optimize parameters that significantly influence model output, improving the model's predictive accuracy and reliability. This method combines sensitivity analysis with orthogonal design to ensure an efficient and comprehensive optimization process.

[0113] Example 3 Reference Figure 5 Based on the above embodiment 1, this embodiment provides a sewage treatment control method, wherein the step S300 of training the artificial neural network model includes: Step S310, performing standardization on the entire dataset, and performing principal component analysis on the data in the standardized entire dataset to divide the data into a training set and a validation set; In this step, the entire dataset is preprocessed before training, including normalization and principal component analysis (PCA), and then divided into training and validation sets. This division allows for efficient evaluation and adjustment of the model during training, thereby improving its performance and reliability.

[0114] Each feature in the full dataset can be normalized to a mean of 0 and a standard deviation of 1. Principal component analysis (PCA) is then performed on the normalized data to reduce the data's dimensionality and remove redundant information. The processed data is then divided into training and test sets, typically in a ratio of 7:3 or 8:2. This results in the normalized full dataset, the full dataset after PCA dimensionality reduction, and the divided training and test sets.

[0115] Normalization can eliminate the influence of different feature dimensions and improve the convergence speed of the model. PCA can reduce data dimensions, remove redundant information, and improve model training efficiency and prediction accuracy.

[0116] Furthermore, the expression for the standardization process is: ; in, represents a standardized score; x represents a specific variable in the entire data set; μ represents the mean value of the input variable in the entire data set; σ represents the standard deviation of the input variable in the entire data set.

[0117] Furthermore, the expression of the principal component analysis is: ; in, represents the principal component analysis matrix; Represents the standardized numerical matrix before principal component analysis; represent and Numeric matrix after principal component analysis.

[0118] Step S320 , performing training using the first model and the second model in the artificial neural network model based on the training set, and evaluating the prediction performance of all the artificial neural network models according to the absolute percentage error.

[0119] The training comprises inputting the input water quality data into the first model to obtain the simulation process control parameters corresponding to the input water quality data; and inputting the input water quality data and the output water quality data into the second model to obtain the simulation process control parameters corresponding to the input water quality data and the output water quality data.

[0120] In this step, different types of artificial neural network models are trained using the training set and their predictive performance is evaluated. You can choose from a variety of artificial neural network models (such as multilayer perceptrons and convolutional neural networks). Each model is trained using the training set. The predictive performance of each model is evaluated using the test set, and the mean absolute percentage error (MAPE) is calculated to provide a predictive performance evaluation of the different types of neural network models.

[0121] This step compares the performance of different models and selects the model that best suits the current data.

[0122] Furthermore, the calculation formula of the absolute percentage error is: ; Wherein, MAPE stands for absolute percentage error; n stands for the number of test sets; Represents the analog output value of data point i; Represents the actual output value of data point i.

[0123] Step S330 , determining the preferred type of artificial neural network model according to the prediction performance, and using a cyclic algorithm to determine the network structure and hyperparameter configuration of the preferred type of artificial neural network model as the preferred configuration.

[0124] In this step, the optimal neural network model type is selected based on the prediction performance, and its network structure and hyperparameters are optimized through a cyclic algorithm.

[0125] You can select the neural network model type with the best prediction performance; use cyclic algorithms (such as grid search and random search) to optimize the model's network structure and hyperparameters; determine the optimal network structure and hyperparameter configuration, and obtain the optimized network structure and hyperparameter configuration.

[0126] By optimizing the network structure and hyperparameters, the prediction accuracy and generalization ability of the model are improved.

[0127] For example, you can use sklearn.model_selection.GridSearchCV or RandomizedSearchCV in Python to perform hyperparameter optimization. For example, for a multilayer perceptron, you can adjust hyperparameters such as the number of hidden layer nodes, learning rate, and activation function.

[0128] Step S340: retrain the artificial neural network model according to the preferred configuration until a trained artificial neural network model is obtained.

[0129] In this step, the neural network model is retrained based on the optimized network structure and hyperparameter configuration until the model converges. The neural network model can be reinitialized using the optimized network structure and hyperparameter configuration; the model is retrained using the training set until the model converges. The final performance of the model is evaluated using the test set, thus completing the training of the artificial neural network model.

[0130] This step ensures that the model achieves optimal performance under the optimized configuration.

[0131] For example, you can use sklearn.neural_network.MLPRegressor or other neural network libraries in Python for model training. Set the termination conditions for training, such as the maximum number of iterations or the convergence threshold.

[0132] This example provides a training process for an artificial neural network model, including data preprocessing, model selection, hyperparameter optimization, and model retraining. Through these steps, the neural network model can be systematically optimized to improve its prediction accuracy and generalization ability. This method combines standardization, PCA dimensionality reduction, model performance evaluation, and hyperparameter optimization to ensure efficient and reliable model training.

[0133] Example 4 Reference Figure 6 Based on the above-mentioned embodiment 1, this embodiment provides a sewage treatment control method. In step S400, actual influent water quality data of the target water source is obtained, and actual process control parameters corresponding to the target water source are obtained using the dynamic simulation model and the artificial neural network model, including: Step S410, collecting actual influent water quality data of the target water source, and obtaining predicted process parameters using the first model; In this step, the actual influent water quality data of the target water source is obtained, which will serve as input for subsequent processing.

[0134] Water quality monitoring equipment (such as online sensors, laboratory analytical instruments) can be used to collect real-time water quality data of the target water source. The data may include but is not limited to COD, NH4-N, NO3 - -N, TP, PO4 3- Thus, the actual influent water quality data is obtained, which reflects the current water quality status of the target water source.

[0135] This step provides real-time water quality information, providing accurate input for subsequent model prediction and control. Specifically, online water quality monitoring systems, such as COD sensors and ammonia nitrogen sensors, can be used to collect data in real time. The collected data is stored in a database or data management system.

[0136] Actual influent water quality data is collected through water quality monitoring equipment and input into the first model to obtain predicted process control parameters. The first model is used to quickly predict process control parameters, providing a basis for subsequent simulation and adjustment.

[0137] Specifically, water quality monitoring equipment, such as COD analyzers and ammonia nitrogen analyzers, can be installed at the water inlet of the sewage treatment plant to collect real-time water quality data and transmit the data to the control system. This data is then input into the first model to obtain predicted process control parameters.

[0138] Assume the first model is a multilayer perceptron (MLP), with the number of neurons in the input layer equal to the dimensionality of the input water quality data, and the number of neurons in the output layer equal to the dimensionality of the process control parameters. The collected water quality data is fed into the model, and the predicted process control parameters are obtained through forward propagation.

[0139] Step S420: Based on the predicted process parameters, the actual inlet water quality data is simulated by the dynamic simulation model to obtain the actual outlet water quality data of the target water source.

[0140] In this step, the dynamic simulation model is used to simulate the wastewater treatment process based on the predicted process control parameters and actual influent water quality data to obtain actual effluent water quality data. The predicted process control parameters and actual influent water quality data can be input into the dynamic simulation model, and the model is run to obtain actual effluent water quality data.

[0141] The effectiveness of the predicted process control parameters is verified through dynamic simulation models to ensure the reliability of effluent water quality.

[0142] The predicted process control parameters and actual influent water quality data can be input into the dynamic simulation model through a computer program, and the model can be run to obtain the actual effluent water quality data.

[0143] Step S430: Determine whether the actual water output data meets the preset water quality standard.

[0144] In this step, the actual effluent water quality data obtained by simulation is checked to see whether it meets the preset water quality standard (such as Class A water quality standard).

[0145] Compare the actual effluent water quality data with the Level A standard to determine whether the effluent water quality meets the standard, thereby ensuring that the effluent water quality meets environmental protection requirements and avoids environmental pollution.

[0146] In this step, the actual effluent water quality data can be compared item by item with the Class A standard through a computer program.

[0147] For example, check whether COD is less than 50mg / L, ammonia nitrogen is less than 5mg / L (or 8mg / L, depending on the specific season), total phosphorus is less than 0.5mg / L, etc. If all indicators meet the standards, the effluent water quality is considered to meet the standards.

[0148] Step S440: If yes, output the predicted process parameters as the actual process control parameters.

[0149] If the actual effluent water quality data meets the preset water quality standards, the predicted process control parameters are used as the actual process control parameters, and the predicted process control parameters are output to the control system, so that the actual process control parameters can be obtained.

[0150] In this step, the predicted process control parameters are directly used to improve control efficiency.

[0151] The predicted process control parameters can be output to the sewage treatment equipment through a control system (such as a PLC). The predicted process control parameters (such as aeration volume, dosage, etc.) are transmitted to the PLC, which adjusts the operation of the equipment based on these parameters.

[0152] Step S450, if not, input the actual inlet water quality data and the actual outlet water quality data into the second model to obtain new process parameters; use the new process parameters as the predicted process parameters, and return to step S420, based on the predicted process parameters, simulate the actual inlet water quality data through the dynamic simulation model until the predicted process parameters that meet the preset water quality standards are obtained, and output and used as the actual process control parameters.

[0153] This step is an iterative process. If the actual effluent water quality data does not meet the preset water quality standards, the actual inlet water quality data and the actual effluent water quality data are input into the second model to obtain new process parameters, and the above steps are repeated until the effluent water quality meets the standards.

[0154] The actual influent water quality data and the actual effluent water quality data are input into the second model, and the model is run to obtain new process parameters. Then, the new process parameters are used as the predicted process parameters, and step S420 is repeated to finally obtain predicted process parameters that meet the preset water quality standards.

[0155] Through cyclic iterative adjustment, the effluent quality is ensured to meet the standards and the accuracy and reliability of control are improved.

[0156] The actual influent and effluent quality data can be input into the second model through a computer program, and the model can be run to obtain new process parameters. The new process parameters can then be used as the predicted process parameters and the simulation and judgment process can be repeated.

[0157] Assume that the second model is also a multi-layer perceptron (MLP), with the number of neurons in the input layer equal to the dimensions of the actual influent water quality data and the actual effluent water quality data, and the number of neurons in the output layer equal to the dimensions of the process control parameters. The actual influent water quality data and the actual effluent water quality data are input into the model, and new process parameters are obtained through forward propagation. Then, using the new process parameters as predicted process parameters, step S420 is repeated until the effluent water quality meets the standards. Furthermore, in sewage treatment control, the actual effluent water quality data is affected by a variety of factors, including fluctuations in influent water quality and changes in equipment operating status. Therefore, directly inputting the output data of the dynamic simulation model into the artificial neural network model may not fully meet the accuracy requirements of the PLC control system. Through iterative optimization, the model output can be gradually adjusted to more closely match actual operational requirements.

[0158] For example, methods for loop iteration optimization may include: Assumptions include: Target water source: a city sewage treatment plant; Preset water quality standard: Level A standard, taking COD (chemical oxygen demand) as an example, the outlet COD limit is 50mg / L.

[0159] Input water quality data: actual influent COD is 300mg / L.

[0160] Process control parameters: aeration rate (unit: m³ / h) and coagulant dosage (unit: kg / h).

[0161] Initial step: Collect actual influent water quality data: Assume that the actual influent COD is 300 mg / L.

[0162] The predicted process parameters are obtained through the first model: Assume that the aeration rate predicted by the first model is 100m³ / h and the coagulant dosage is 5kg / h.

[0163] First iteration: (1) Perform dynamic simulation based on predicted process parameters: Input: actual influent COD = 300 mg / L, predicted aeration rate = 100 m³ / h, coagulant dosage = 5 kg / h; After the dynamic simulation model runs, the actual effluent COD is obtained = 60 mg / L; (2) Determine whether the actual effluent data meets the preset water quality standard: the actual effluent COD = 60 mg / L, which is higher than the Class A standard (50 mg / L) and does not meet the standard. (3) Input the actual influent water quality data and the actual effluent water quality data into the second model: Input: actual influent COD = 300 mg / L, actual effluent COD = 60 mg / L; The second model outputs the new process parameters: aeration rate = 120 m³ / h, coagulant dosage = 6 kg / h.

[0164] Second iteration: (1) Dynamic simulation based on new predicted process parameters: Input: Actual influent COD = 300 mg / L, new predicted aeration rate = 120 m³ / h, coagulant dosage = 6 kg / h; After running the dynamic simulation model, the actual effluent COD = 55 mg / L is obtained; (2) Determine whether the actual effluent data meets the preset water quality standard: Actual effluent COD = 55 mg / L, still higher than the first level A standard (50 mg / L), does not meet the standard. (3) Input actual influent water quality data and actual effluent water quality data into the second model: Input: Actual influent COD = 300 mg / L, actual effluent COD = 55 mg / L; The second model outputs new process parameters: aeration rate = 140 m³ / h, coagulant dosage = 7 kg / h.

[0165] Third iteration: (1) Dynamic simulation based on new predicted process parameters: Input: Actual influent COD = 300 mg / L, new predicted aeration rate = 140 m³ / h, coagulant dosage = 7 kg / h; After running the dynamic simulation model, the actual effluent COD = 48 mg / L is obtained; (2) Determine whether the actual effluent data meets the preset water quality standard: Actual effluent COD = 48 mg / L, lower than the first level A standard (50 mg / L), meets the standard. (3) Output predicted process parameters as actual process control parameters: Output process control parameters: aeration rate = 140 m³ / h, coagulant dosage = 7 kg / h.

[0166] Through the above three iterations, the final process control parameters (aeration rate = 140 m³ / h, coagulant dosage = 7 kg / h) can ensure that the effluent COD meets the first level A standard (50 mg / L). These parameters will be used in the actual wastewater treatment process to ensure that the effluent water quality meets the standard.

[0167] Through the above method, the actual process control parameters in step S420 can be optimized through cyclic iteration to ensure that they meet the requirements of the PLC control system, thereby achieving more accurate and stable wastewater treatment control.

[0168] Further, the operation of the wastewater treatment can be controlled based on the actual process control parameters. According to the obtained actual process control parameters, the operating parameters of the wastewater treatment system are adjusted to achieve the optimal wastewater treatment effect. The obtained actual process control parameters can be transmitted to the control system (such as PLC) of the wastewater treatment system.

[0169] The control system adjusts the operating parameters of the wastewater treatment equipment, such as aeration equipment, dosing equipment, water pumps, etc., according to these parameters, so that the operating parameters of the wastewater treatment system are adjusted to the optimal state, ensuring that the effluent water quality meets the standard.

[0170] The step realizes the automation and intelligent control of the sewage treatment process, improves the treatment efficiency and the stability of the effluent water quality.

[0171] In the embodiment, a dynamic simulation model and an artificial neural network model are used to obtain actual process control parameters according to actual influent water quality data of a target water source, and the operation of the sewage treatment system is controlled accordingly. Through these steps, the automation and intelligent control of the sewage treatment process can be realized, and the treatment efficiency and the stability of the effluent water quality can be improved. This method combines the advantages of dynamic simulation and artificial neural networks, ensuring the prediction accuracy and control effect of the model.

[0172] Reference Figure 7 In the embodiment of the present application, a sewage treatment control device is provided, which comprises: A model construction module 10 is configured to establish a dynamic simulation model of the sewage treatment process based on an ASM2d model. The model construction module 10 is further configured to obtain input water quality data of a target water source, obtain output water quality data corresponding to the input water quality data under different working conditions and simulation process control parameters under corresponding working conditions by using the dynamic simulation model, and construct a full data set based on the input water quality data, the output water quality data and the simulation process control parameters. The input water quality data, the output water quality data and the simulation process control parameters that do not meet the preset water quality standard are excluded from the full data set. A model training module 20 is configured to construct an artificial neural network model based on the full data set, train the artificial neural network model, and obtain a trained artificial neural network model. The artificial neural network model includes a first model trained based on the input water quality data and a second model trained based on the output water quality data. A system control module 30 is configured to obtain actual influent water quality data of the target water source, obtain actual process control parameters corresponding to the target water source by using the dynamic simulation model and the artificial neural network model, and control the operation of the sewage treatment based on the actual process control parameters.

[0173] In the embodiment of the present application, a sewage treatment control system is provided, which comprises an information collection device, a central control system and a PLC (Programmable Logic Controller) execution system. The central control system comprises a processor and a memory. The memory stores a computer program, and the processor is configured to execute the computer program to implement the sewage treatment control method of any one of the preceding embodiments.

[0174] The above-mentioned information collection equipment can collect multiple key parameters of the sewage treatment process in real time, including but not limited to dissolved oxygen (O2), pH value, chemical oxygen demand (COD), ammonia nitrogen (NH4 - ), nitrate (NO4 - ), phosphate (PO4 - ), temperature (T), flow rate (Q), coagulant (PAC) and flocculant (PAM) dosage, chlorine dosage, flow rate, etc. These data will be transmitted to the central control system to provide a basis for subsequent processing and control.

[0175] The central control system, as the core of the entire sewage treatment process, receives various data from information collection equipment. It simulates and calculates the ASM2D model, predicts effluent quality data, and compares it with emission standards. If the predicted effluent quality does not meet emission standards, the central control system, based on pre-set control strategies, issues instructions to the dosing equipment, aeration equipment, and return flow equipment through the PLC system. These instructions adjust the operating parameters of these devices to ensure that the sewage treatment process meets the set standards and that the effluent quality meets emission standards.

[0176] The PLC execution system is closely connected to the dosing equipment, aeration equipment, return flow equipment, and sludge removal equipment. This part receives control commands and data transmitted from the central control system and precisely controls the operation of each device. For example, according to the instructions of the central control system, it adjusts the dosage of the dosing equipment, controls the aeration intensity of the aeration equipment, adjusts the return flow ratio of the return flow equipment, and controls the sludge removal operation of the sludge removal equipment. This ensures that each process link in the sewage treatment process operates stably according to the optimized parameters, thereby achieving efficient and stable sewage treatment results. Through the coordinated operation of information collection equipment, the central control system, and the PLC execution system, the entire sewage treatment process achieves automated and intelligent control, which can effectively respond to water quality fluctuations, optimize operating parameters, improve sewage treatment efficiency, and ensure that the effluent water quality is stable and meets standards.

[0177] In an embodiment of the present application, a computer storage medium is provided, which stores a computer program. When the computer program is executed on a processor, the sewage treatment control method described in any one of the aforementioned embodiments is implemented.

[0178] Finally, it should be noted that the above-described embodiments are only specific implementations of the present invention, which are used to illustrate the technical solutions of the present invention, rather than to limit them. The scope of protection of the present invention is not limited thereto. Although the present invention has been described in detail with reference to the above-described embodiments, those skilled in the art should understand that any person skilled in the art can modify or easily conceive of changes to the technical solutions described in the above-described embodiments within the technical scope disclosed by the present invention, or replace some of the technical features therein with equivalents. However, such modifications, changes, or replacements do not deviate from the spirit and scope of the technical solutions of the embodiments of the present invention. They should all be included in the scope of protection of the present invention. Therefore, the scope of protection of the present invention shall be subject to the scope of protection of the claims.

Claims

1. A sewage treatment control method, characterized in that: include: Establish a dynamic simulation model of the sewage treatment process based on the ASM2d model; Obtain input water quality data of target water sources; Using the dynamic simulation model to obtain output water quality data corresponding to the input water quality data under different operating conditions, as well as simulated process control parameters under the corresponding operating conditions, and constructing a full data set based on the input water quality data, the output water quality data, and the simulated process control parameters; wherein the full data set excludes the input water quality data, the output water quality data, and the simulated process control parameters that do not meet the preset water quality standards; An artificial neural network model is constructed based on the full data set, and the artificial neural network model is trained to obtain the trained artificial neural network model; wherein the artificial neural network model includes a first model trained based on the input water quality data and a second model trained based on the output water quality data; The actual influent water quality data of the target water source is obtained, the actual process control parameters corresponding to the target water source are obtained using the dynamic simulation model and the artificial neural network model, and the operation of the sewage treatment is controlled based on the actual process control parameters.

2. The sewage treatment control method according to claim 1, characterized in that: The dynamic simulation model of the sewage treatment process based on the ASM2d model includes: Build dynamic simulation models; Acquiring historical operating data of the target water source, and performing steady-state simulation verification on the dynamic simulation model based on the historical operating data; Performing sensitivity analysis on the dynamic simulation model after steady-state simulation verification, and optimizing and adjusting the confirmed model parameters based on the sensitivity to obtain the dynamic simulation model with optimized accuracy; Through dynamic simulation, the prediction effect of the dynamic simulation model based on the model parameters after the individual optimization adjustment is verified, and the dynamic simulation model after dynamic simulation is obtained.

3. The sewage treatment control method according to claim 2, characterized in that: The expression of the sensitivity analysis is: ； Wherein, a represents the model parameter input to the dynamic simulation model; a1 and a2 represent the typical value and the changed value of the model parameter, respectively; y represents the water quality index that the dynamic simulation model focuses on during the simulation process; y1 and y2 represent the y value of the corresponding simulation output of the water quality index corresponding to the water quality index under the conditions of a1 and a2, respectively; S represents the sensitivity coefficient of the model parameter to the water quality index; and / or, The individually optimizing and adjusting the confirmed model parameters based on sensitivity includes: Determining whether the absolute value of the sensitivity of the model parameter is greater than a preset sensitivity threshold; If so, the model parameter whose absolute value of sensitivity is greater than the preset sensitivity threshold is used as the parameter to be adjusted; Taking the default value of the dynamic simulation model as the center, setting the variation range and step size of each parameter to be adjusted, and constructing a value sequence of the parameter to be adjusted; Combining all the values ​​of the parameters to be adjusted in an orthogonal manner according to the value sequence to generate a plurality of parameter combinations for simulation analysis; Inputting each set of parameter combinations into the dynamic simulation model to obtain corresponding simulated output water quality indicators; Calculating a corresponding prediction error for each of the simulated output water quality indicators; the prediction error includes a coefficient of determination and a mean square error; Among all the parameter combinations, a set of parameter combinations with the largest determination coefficient and the smallest mean square error is selected as the optimized model parameters, so as to update the dynamic simulation model according to the model parameters.

4. The sewage treatment control method according to claim 3, characterized in that: The calculation method of the determination coefficient is: ； Among them, R 2 represents the coefficient of determination; Represents the true value of the i-th data point; represents the predicted value of the ith data point output by the dynamic simulation model; represents the sample mean; is the error caused by the prediction, is the error in the mean; and / or, The calculation method of the mean square error is: ； Wherein, MES represents the mean square error; m represents the total number of samples, represents the true value of the i-th data point, represents the predicted value of the ith data point output by the dynamic simulation model; represents the sample mean.

5. The sewage treatment control method according to claim 1, wherein: The training of the artificial neural network model comprises: Standardizing the entire dataset, performing principal component analysis on the data in the standardized dataset, and dividing the data into a training set and a validation set; Training is performed using the first model and the second model in the artificial neural network model based on the training set, and the prediction performance of all the artificial neural network models is evaluated according to the absolute percentage error; wherein the training comprises inputting the input water quality data into the first model to obtain the simulated process control parameters corresponding to the input water quality data; and inputting the input water quality data and the output water quality data into the second model to obtain the simulated process control parameters corresponding to the input water quality data and the output water quality data; Determining the type of the preferred artificial neural network model according to the prediction performance, and using a loop algorithm to determine the network structure and hyperparameter configuration of the preferred type of artificial neural network model as the preferred configuration; According to the preferred configuration, the artificial neural network model is retrained until a trained artificial neural network model is obtained.

6. The sewage treatment control method according to claim 5, characterized in that: The expression of the standardization process is: ； in, represents a standardized score; x represents a specific variable in the full data set; μ represents the mean value of the input variable in the full data set; σ represents the standard deviation of the input variable in the full data set; and / or, The expression of the principal component analysis is: ； in, represents the principal component analysis matrix; Represents the standardized numerical matrix before principal component analysis; represent and a numerical matrix after principal component analysis; and / or, The absolute percentage error is calculated as follows: ； Wherein, MAPE stands for absolute percentage error; n stands for the number of test sets; Represents the analog output value of data point i; Represents the actual output value of data point i.

7. The sewage treatment control method according to claim 1, characterized in that: The step of obtaining actual influent water quality data of the target water source and obtaining actual process control parameters corresponding to the target water source using the dynamic simulation model and the artificial neural network model includes: Collecting actual influent water quality data of the target water source and obtaining predicted process parameters using the first model; Based on the predicted process parameters, the actual influent water quality data is simulated by the dynamic simulation model to obtain the actual effluent water quality data of the target water source; Determining whether the actual water output data meets the preset water quality standard; If so, output the predicted process parameters as the actual process control parameters; If not, the actual inlet water quality data and the actual outlet water quality data are input into the second model to obtain new process parameters; the new process parameters are used as the predicted process parameters, and the step is returned to simulate the actual inlet water quality data through the dynamic simulation model based on the predicted process parameters until the predicted process parameters that meet the preset water quality standards are obtained, and output as the actual process control parameters.

8. A sewage treatment control device, characterized in that: include: Model building module, used to build a dynamic simulation model of the sewage treatment process based on the ASM2d model; The model building module is further configured to obtain input water quality data of a target water source; utilize the dynamic simulation model to obtain output water quality data corresponding to the input water quality data under different operating conditions, as well as simulated process control parameters under the corresponding operating conditions; and construct a full data set based on the input water quality data, the output water quality data, and the simulated process control parameters; wherein the full data set excludes the input water quality data, the output water quality data, and the simulated process control parameters that do not meet preset water quality standards; A model training module is used to construct an artificial neural network model based on the full data set and train the artificial neural network model to obtain the trained artificial neural network model; wherein the artificial neural network model includes a first model trained based on the input water quality data and a second model trained based on the output water quality data; The system control module is used to obtain the actual influent water quality data of the target water source, use the dynamic simulation model and the artificial neural network model to obtain the actual process control parameters corresponding to the target water source, and control the operation of the sewage treatment based on the actual process control parameters.

9. A sewage treatment control system, characterized in that: The sewage treatment control system includes information collection equipment, a central control system and a PLC execution system; The central control system includes a processor and a memory; the memory stores a computer program, and the processor is used to execute the computer program to implement the sewage treatment control method according to any one of claims 1 to 7.

10. A computer storage medium, characterized in that It stores a computer program, which, when executed on a processor, implements the sewage treatment control method according to any one of claims 1 to 7.

Images