Machine learning-based dynamic optimization setting method and system for process parameters of biological reaction tank of sewage treatment plant

Abstract

The invention discloses a dynamic optimization setting method and system for process parameters of a biological reaction tank of a sewage treatment plant based on machine learning. The method comprises the following steps: S1, data acquisition; s2, data integration and preprocessing; s3, establishing an effluent quality and concentration simulation prediction matrix; s4, carrying out model fusion; and S5, carrying out dynamic optimization setting on the technological parameters of the biological reaction tank based on the effluent quality simulation prediction model matrix established in the steps S1 to S4. The method comprises the following steps: firstly, establishing an outlet water quality high-precision simulation prediction model matrix, performing setting optimization on process parameters of a green pool and a reverse pool by adopting a particle swarm algorithm according to the model matrix, and finally obtaining process parameter set values of a future operation period; the advanced data analysis technology is fully utilized, the operation efficiency and effect of the sewage treatment plant are improved through the steps of collecting, processing, predicting, optimizing and the like, and energy conservation and consumption reduction are achieved on the premise that it is guaranteed that the effluent of the sewage treatment plant stably reaches the standard.

Description

Technical Field

[0001] The present invention belongs to the technical field of sewage treatment, and particularly relates to a method and system for dynamically optimizing and setting the process parameters of the biological reaction tank in a sewage treatment plant based on machine learning. Background Art

[0002] With the acceleration of the urbanization process and the expansion of industrial production, sewage treatment plants play a crucial role in maintaining environmental quality and ensuring public health. Among them, the biochemical reaction tank, as the core process means affecting nitrogen removal and carbon reduction in sewage treatment plants, its operation effectiveness is not only closely related to the effluent quality, but also largely determines the operation energy consumption of the sewage treatment plant. In traditional sewage treatment plants, the process control method of the biological reaction tank usually relies on the fixed process parameter values set by the operators for control. Once these parameters are set, they will remain constant for a long time. However, in the actual sewage treatment process, there are many uncertain factors, such as fluctuations in influent water quality and seasonal changes, etc. These factors make the fixed treatment parameters difficult to adapt to real-time changes, thereby affecting the effluent quality and increasing the operation cost at the same time. Summary of the Invention

[0003] In view of this, the purpose of the present invention is to provide a method and system for dynamically optimizing and setting the process parameters of the biological reaction tank in a sewage treatment plant based on machine learning to solve the deficiencies in the prior art.

[0004] To achieve the above purpose, the present invention is realized through the following technical solutions:

[0005] Provide a method for dynamically optimizing and setting the process parameters of the biological reaction tank in a sewage treatment plant based on machine learning, which includes the following steps:

[0006] S1. Data collection;

[0007] S2. Data integration and preprocessing;

[0008] S3. Establish a simulation prediction matrix of the effluent water quality concentration;

[0009] S4. Model fusion;

[0010] S5. Dynamically optimize and set the process parameters of the biological reaction tank dissolution based on the simulation prediction model matrix of the effluent water quality after model fusion established in steps S1 to S4.

[0011] As the method for dynamically optimizing and setting the process parameters of the biological reaction tank in a sewage treatment plant based on machine learning, step S2 includes:

[0012] S21. Data integration

[0013] Define the hydraulic retention time calculated from the influent of the sewage treatment plant to the process parameter measuring instrument of the anoxic-oxic tank as the first hydraulic retention time HRT1, and calculate the data translation step S1 based on HRT1 and the maximum acquisition interval T;

[0014] According to the calculated S1 value, move the process parameter measurement values in the original acquisition sequence forward by the corresponding number of sampling periods, and recombine them with the remaining collected data to form a new data set;

[0015] For the missing process parameter values caused by the forward movement operation, use the process parameter control reference value set by the sewage treatment plant to supplement them;

[0016] S22. Data preprocessing

[0017] It includes filling missing values, identifying and correcting abnormal readings. The filling of missing values is achieved by interpolation method. At the same time, box plot analysis and 3σ principle statistical method are used to detect potential abnormal points, and the same technical means as filling missing values are used to adjust or replace these abnormal values. Finally, data normalization means are used to normalize the entire data set.

[0018] As described in the dynamic optimization setting method of the anoxic-oxic tank process parameters of the sewage treatment plant based on machine learning, wherein step S3 includes:

[0019] S31. Selection of effluent water quality concentration simulation prediction model

[0020] In the model selection method, the ensemble learning algorithm suitable for fitting non-linear feature relationships is organically combined with the recurrent neural network algorithm good at capturing time series characteristics;

[0021] Where the model matrix is defined as F T =[F1,F2…F n , F T Is a column vector with dimension n, where each element is an ensemble learning class or a recurrent neural network class algorithm;

[0022] S32. Custom loss function

[0023] Define the hydraulic retention time calculated from the process parameter measuring instrument of the anoxic-oxic tank to the measuring position of the effluent instrument of the sewage treatment plant as the second hydraulic retention time HRT2, and the simulation prediction step of each element in the model matrix is:

[0024] S2 = HRT2 / T

[0025] According to the simulation prediction step, design the loss function:

[0026]

[0027] Among them, S2 is the simulation prediction step size, t is the time point of simulation prediction, α and β are constants between 0 and 1, and γ, σ, and δ are positive real numbers; αt - T ensures that the error of recent prediction has a greater weight; βyt - yt + 1 - βyt - yt2 is a mixed term of absolute error and squared error; is an additional penalty for large errors;

[0028] S33. Establish a multi-dimensional simulation prediction result matrix for the effluent water quality concentration

[0029] Establish a multi-dimensional simulation prediction result matrix based on the model matrix elements and the defined loss function. Through different matrix elements, the characteristics of each dimension of historical data are mined and learned to generate corresponding simulation prediction results, and the results are stored in the simulation prediction result matrix:

[0030]

[0031] Among them, F ij represents the simulation prediction result of the jth step of the ith element of the algorithm matrix.

[0032] As described in the method for dynamically optimizing and setting the process parameters of the biological reaction tank in a sewage treatment plant based on machine learning, among them, the ensemble learning algorithm suitable for fitting non-linear characteristic relationships in step S31 includes but is not limited to XGBoost, CatBoost, and LightGBM.

[0033] As described in the method for dynamically optimizing and setting the process parameters of the biological reaction tank in a sewage treatment plant based on machine learning, among them, the recurrent neural network algorithm that is good at capturing time series characteristics in step S31 includes but is not limited to LSTM, GRU, BiLSTM, and BiGRU.

[0034] As described in the method for dynamically optimizing and setting the process parameters of the biological reaction tank in a sewage treatment plant based on machine learning, among them, step S4 includes:

[0035] S41. Define the model fusion coefficient

[0036] The row vector K of the model fusion coefficient = [k1, k2..k n , and the column vector B of constant coefficients = [b1, b2...bS2], where the elements in K are model fusion coefficients and are obtained by fitting using the multiple linear regression algorithm;

[0037] S42. Calculate the final result of the effluent water quality concentration simulation prediction

[0038] R f = K × R + B

[0039] Among them, Rf = r1, r1…r S2is a column vector of simulation prediction results, representing the final simulation prediction results of the effluent water quality concentration. ri represents the simulation prediction results of the effluent water quality concentration at the i-th moment.

[0040] As described in the method for dynamically optimizing and setting the process parameters of the biological reaction tank in a sewage treatment plant based on machine learning, wherein step S5 includes:

[0041] S51. Data collection and preprocessing

[0042] Obtain sampling data within a hydraulic retention time (HRT) cycle at the current moment. This dataset should comprehensively cover influent characteristics, process control variables, and effluent status. Perform integration and necessary preprocessing operations on the collected data according to established methods.

[0043] S52. Simulation prediction of effluent water quality concentration

[0044] Import the data preprocessed in S51 into the effluent water quality simulation prediction model matrix after model fusion constructed in steps S3 - S4. Use this model matrix to predict the effluent water quality concentration within the next S2 time steps, generating a sequence of predicted values.

[0045] S53. Error analysis and optimization adjustment

[0046] Compare the difference between the predicted results of the effluent water quality concentration within the next S2 time steps output by the model matrix and the target control value set within the sewage treatment plant. Calculate the error between the two. Taking the minimization of this error as the objective function, use the particle swarm optimization (PSO) algorithm to find the optimal process parameter setting strategy through iterative search. During this process, continuously adjust and optimize the process parameter setting values until the error between the predicted effluent water quality concentration and the target value is less than a predetermined threshold.

[0047] S54. Implement control decision

[0048] Once the process parameter setting scheme that enables the predicted value of the effluent water quality concentration to reach the optimal matching degree is determined through the particle swarm optimization algorithm, use the optimized process parameter setting values as the actual process parameter control parameters for the biological reaction tank in the next M operation cycles to achieve dynamic and precise control of the process parameters in the biological reaction tank, where M = (HRT1 - HRT2) / T.

[0049] On the other hand, provide a system for dynamically optimizing and setting the process parameters of the biological reaction tank in a sewage treatment plant based on machine learning, which is implemented based on the method for dynamically optimizing and setting the process parameters of the biological reaction tank in a sewage treatment plant as described in any one of the above. It includes a data collection module, a data integration and preprocessing module, an effluent water quality concentration simulation prediction matrix module, a model fusion module, and a dynamic optimization setting module for the process parameters of the biological reaction tank, which are connected in sequence.

[0050] The beneficial effects of the technical solution of the present invention are as follows:

[0051] Firstly, the system establishes a high-precision simulation prediction model matrix for the effluent water quality. According to the model matrix, the particle swarm optimization algorithm is used to optimize the setting of the process parameter concentrations in the biological reaction tank, and finally the set values of the process parameters for a future operation cycle are obtained. The present invention makes full use of advanced data analysis technologies, and through steps such as collection, processing, prediction, and optimization, is committed to improving the operation efficiency and effect of the sewage treatment plant, and realizing energy conservation and consumption reduction on the premise of ensuring that the effluent of the sewage plant meets the standards stably. Description of the Drawings

[0052] To further illustrate the above-mentioned objects, structural features, and effects of the present invention, the present invention will be described in detail below with reference to the drawings.

[0053] Figure 1 It is a schematic flow chart of the method for the preferred embodiment of the present invention. Detailed Embodiments

[0054] The terms "invention" and "the present invention" used in this specification are intended in a broad sense to refer to all the subject matters of this specification and any subsequent patent claims. Statements containing these terms should not be construed as limiting the subject matter described herein or the meaning or scope of any subsequent patent claims. In addition, this specification does not attempt to describe or limit the subject matter covered by any specific component, paragraph, statement, or claim of this application. The subject matter should be understood with reference to the entire specification, all the drawings, and any subsequent claims. The present invention may have other embodiments and may be practiced or implemented in other ways. Moreover, it should be understood that the wording and terms used herein are for illustrative purposes and should not be considered limiting.

[0055] The details of the present invention will now be discussed with reference to the drawings, which illustrate the present invention by way of example only. In the drawings, like features or components may be labeled with the same reference numerals.

[0056] The use of the terms "comprising", "having", and "including" and their variants herein means including the items listed hereinafter, their equivalents, and additional items. Although directions such as above, below, upward, downward, backward, bottom, top, front, and back may be referred to in the description of the drawings for convenience with reference to the drawings, these directions are not intended to be literally accepted or limit the present invention in any form. In addition, terms such as "first", "second", "third", etc. are used herein for illustrative purposes and are not intended to indicate or imply importance or significance.

[0057] This embodiment studies a sewage treatment plant with a daily treatment capacity of 100,000 tons. The sewage treatment plant is equipped with an advanced automatic control system to achieve precise process control. Preferably, the automatic control system includes, but is not limited to, a precise aeration system and an internal reflux control system. Different process parameters have a significant impact on the effluent quality. The process parameters include, but are not limited to, the dissolved oxygen concentration in the aerobic zone, the reflux ratio, and the reflux flow rate. The effluent quality includes, but is not limited to, the effluent ammonia nitrogen concentration and the effluent total nitrogen concentration. Further, the dissolved oxygen concentration mainly affects the removal effect of ammonia nitrogen, and the internal reflux flow rate and the internal reflux ratio mainly affect the removal rate of total nitrogen. In this embodiment, by predicting different effluent quality indicators in real time and dynamically adjusting relevant process parameters according to the prediction results, the effluent quality is ensured to meet the standards stably, and on this basis, the energy consumption and operation efficiency are maximally optimized.

[0058] See Figure 1 As shown, the method for dynamically optimizing and setting the process parameters of the biological reaction tank in a sewage treatment plant based on machine learning according to the present invention includes the following steps:

[0059] S1; Data collection.

[0060] S2: Data integration and preprocessing.

[0061] S3: Establishment of an effluent quality concentration simulation prediction matrix.

[0062] S4: Model fusion.

[0063] S5: Dynamically optimize and set the process parameters of the biological reaction tank based on the effluent quality simulation prediction model matrix after model fusion established in steps S1 - S4.

[0064] In this embodiment, in order to construct a comprehensive and accurate dynamic optimization setting model for the process parameters of the biological reaction tank in a sewage treatment plant based on machine learning, data collection is the primary link. The specific implementation process of S1 is as follows:

[0065] When collecting the operation data of the sewage treatment plant, the set time collection range should be greater than 6 months. In addition, to ensure the timeliness and accuracy of the data set, it is further stipulated that the maximum allowable sampling interval between each monitoring point should not exceed 1 hour (this interval is defined as T). This means that within the entire predetermined data collection period, all relevant operation parameters need to be regularly recorded at a fixed frequency of T time units.

[0066] The collected data cover all key nodes of the sewage treatment process, including influent data, process control data, and effluent data. The influent data include flow rate, temperature, and key water quality indicators such as ammonia nitrogen, total nitrogen, and COD. The process control data include key process parameters such as dissolved oxygen (DO) concentration in each section of the biological reaction tank, internal return flow rate, external return flow rate, blower flow rate, and valve opening. The effluent data include effluent flow rate, temperature, and water quality indicators such as ammonia nitrogen, total nitrogen, and COD.

[0067] In this embodiment, in order for the model to better learn the biochemical reaction rules of the biological reaction tank in the sewage treatment plant, the specific implementation process of S2 is as follows:

[0068] S21. Data integration. During the sewage treatment process, the hydraulic retention time calculated from the sewage influent to the process parameter measuring instrument in the biological reaction tank is defined as the first hydraulic retention time HRT1, and the data translation step S1 is calculated based on HRT1 and the maximum collection interval T. Subsequently, the measured values of the process parameters in the original collection sequence are moved forward by the corresponding number of sampling periods according to the calculated S1 value, and recombined with the remaining collected data to form a new data set; for the missing process parameter values generated by the forward movement operation, the process parameter control reference values set by the sewage treatment plant are used for supplementation.

[0069] S22. Data preprocessing. Preprocessing steps are implemented on the recombined data. The preprocessing process mainly includes filling missing values, identifying and correcting abnormal readings. Among them, the filling of missing values is achieved through interpolation. At the same time, statistical methods such as box plot analysis and 3σ principle are used to detect potential abnormal points, and the same technical means as filling missing values are used to adjust or replace these abnormal values to ensure the consistency and reliability of the final data set. Finally, data normalization means are used to normalize the entire data set.

[0070] In this embodiment, in order to combine the advantages of different models and improve the simulation prediction ability and generalization performance of the model, multiple machine learning models are used for modeling. The specific implementation process of establishing the S3 effluent water quality concentration simulation prediction matrix is as follows:

[0071] S31. Selection of the effluent water quality concentration simulation prediction model

[0072] In the described model selection method, we organically combine the ensemble learning algorithms applicable to fitting non-linear feature relationships (including but not limited to XGBoost, CatBoost, LightGBM, etc.) with the recurrent neural network algorithms that are good at capturing time series characteristics (including but not limited to LSTM, GRU, BiLSTM, BiGRU, etc.). Through the ensemble learning algorithm, the complex interactions between multi-dimensional features can be effectively processed, while the recurrent neural network can be used to well capture the dynamic change patterns in the time series. This modeling strategy not only enhances the model's ability to understand the influence of different factors, but also improves its robustness in the face of data with outliers or missing values, thus significantly improving the accuracy and reliability of the predicted effluent water quality concentration, where the model matrix is defined as F T =[F1,F2…F n , F T is a column vector of dimension n, where each element is an ensemble learning class or a recurrent neural network class algorithm.

[0073] S32. Custom loss function

[0074] Define the hydraulic retention time calculated from the raw and return pond process parameter measuring instrument to the measuring position of the sewage treatment plant effluent instrument as the second hydraulic retention time HRT2, and the simulation prediction step size of each element in the model matrix is:

[0075] S2 = HRT2 / T

[0076] According to the simulation prediction step size, design the loss function:

[0077]

[0078] S2 is the simulation prediction step size, t is the time point of simulation prediction, α, β are constants between 0 and 1, and γ, σ, δ are positive real numbers.

[0079] αt - T ensures that the error of recent predictions has a greater weight.

[0080] βyt - yt+1 - βyt - yt2 is a mixed term of absolute error and squared error.

[0081] is an additional penalty for large errors.

[0082] By adjusting these parameters, the model can pay more attention to recent predictions and give greater penalties to those abnormally large errors.

[0083] S33. Establish a multi-dimensional simulation prediction result matrix of effluent water quality concentration

[0084] Establish a multi-dimensional simulation prediction result matrix based on the model matrix elements and a customized loss function. Through different matrix elements, mine and learn the characteristics of each dimension of historical data, generate corresponding simulation prediction results, and store the results in the simulation prediction result matrix.

[0085]

[0086] Among them, F ij represents the simulation prediction result of the j-th step of the i-th element of the algorithm matrix.

[0087] In order to fuse the simulation prediction results of each element in the multi-dimensional simulation prediction model matrix of the effluent water quality, and improve the overall simulation prediction accuracy and generalization ability, the specific implementation process of S4 model fusion is as follows:

[0088] S41. Define the model fusion coefficient

[0089] The model fusion coefficient row vector K = [k1, k2..k n and the constant coefficient column vector B = [b1, b2...bS2], where the elements in K are model fusion coefficients, which are obtained by fitting using the multiple linear regression algorithm.

[0090] S42. Calculate the final simulation prediction result of the effluent water quality concentration

[0091] R f = K × R + B

[0092] Among them, Rf = r1, r1…rS2 is a column vector of simulation prediction results, representing the final simulation prediction result of the effluent water quality concentration, and ri represents the simulation prediction result at the i-th moment for the effluent water quality concentration.

[0093] Based on the effluent water quality simulation prediction model matrix established in steps S1 to S4, perform dynamic optimization setting of the process parameters of the anoxic-oxic reactor. The specific implementation process of S5 is as follows:

[0094] S51. Data collection and preprocessing: First, obtain the sampling data within an HRT (Hydraulic Retention Time) cycle at the current moment. This dataset should comprehensively cover the influent characteristics (including but not limited to flow rate, temperature, and key water quality parameters such as ammonia nitrogen, total nitrogen, COD, etc.), process control variables (key process parameters such as dissolved oxygen (DO) concentration in each section of the anoxic-oxic reactor, internal reflux flow rate, external reflux flow rate, blower flow rate, valve opening, etc.), and effluent status (such as effluent flow rate, temperature, and corresponding water quality indicators). Subsequently, perform integration and necessary preprocessing operations on the collected data according to established methods to ensure the data quality input into the model.

[0095] S52. Effluent quality concentration simulation prediction: Import the data preprocessed in S51 into the effluent quality simulation prediction model matrix after the models constructed in steps S3 - S4 are fused, and use this model matrix to predict the effluent quality concentration within the next S2 time steps, generating a sequence of predicted values.

[0096] S53. Error analysis and optimization adjustment: Compare the difference between the predicted results of the effluent quality concentration within the next S2 time steps output by the model matrix and the target control value set within the sewage treatment plant, and calculate the error between the two. Taking the minimization of this error as the objective function, adopt the particle swarm optimization algorithm (PSO), and find the optimal process parameter setting strategy through iterative search. During this process, continuously adjust and optimize the process parameter setting values until the error between the predicted effluent quality concentration and the target value is less than the predetermined threshold.

[0097] S54. Implement control decision: Once the process parameter setting scheme that enables the predicted value of the effluent quality concentration to reach the optimal matching degree is determined through the particle swarm optimization algorithm, use the optimized process parameter setting values as the actual process parameter control parameters of the biological reaction tank in the next M operation cycles, realizing the dynamic and precise control of the process parameters in the biological reaction tank, where M=(HRT1 - HRT2) / T.

[0098] The above are only the preferred embodiments of the present invention, and do not limit the implementation manners and protection scope of the present invention accordingly. For those skilled in the art, it should be able to realize that any equivalent substitution and obvious changes made by using the description and illustrations of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for dynamically optimizing process parameters of a sewage treatment plant's retort tank based on machine learning, characterized in that: The following steps are involved: S1, data collection; S2, data integration and preprocessing; S3, establish a simulation prediction matrix for effluent water quality concentration; S4, model fusion; S5. Based on the effluent water quality simulation prediction model matrix after the model fusion established in steps S1 to S4, dynamic optimization and setting of the process parameters of the bioremediation tank are performed.

2. The method for dynamically optimizing the process parameters of the sewage treatment plant's reaction tank based on machine learning as claimed in claim 1, characterized in that: Step S2 includes: S21. Data Integration Define the hydraulic retention time calculated by the process parameter measuring instrument from the sewage plant inlet to the raw water tank as the first hydraulic retention time HRT1, and calculate the data translation step length S1 based on HRT1 and the maximum acquisition interval T; According to the calculated S1 value, the process parameter measurement values ​​in the original acquisition sequence are moved forward by a corresponding number of sampling cycles, and recombined with the remaining collected data to form a new data set; For the missing process parameter values ​​caused by the forward operation, the process parameter control benchmark values ​​set by the sewage treatment plant are used to supplement them; S22. Data preprocessing It includes filling missing values ​​and identifying and correcting abnormal readings. The filling of missing values ​​is achieved through interpolation. At the same time, box plot analysis and 3σ principle statistical methods are used to detect potential abnormal points. These abnormal values ​​are adjusted or replaced using the same technical means as those for filling missing values. Finally, data normalization is used to normalize the entire data set.

3. The method for dynamically optimizing and setting process parameters of a sewage treatment plant's bioremediation tank based on machine learning as claimed in claim 1, characterized in that: Step S3 includes: S31. Selection of simulation prediction model for effluent water quality concentration In the model selection method, the ensemble learning algorithm suitable for fitting nonlinear feature relationships is organically combined with the recurrent neural network algorithm that is good at capturing time series characteristics; The model matrix is ​​defined as F T =[F1,F2…F n ],F T is a column vector of dimension n, where each element is an ensemble learning or recurrent neural network algorithm; S32. Custom loss function The hydraulic retention time calculated from the process parameter measuring instrument of the reactor to the measuring position of the sewage treatment plant effluent meter is defined as the second hydraulic retention time HRT2. The simulation prediction step length of each element in the model matrix is: S2=HRT2 / T According to the simulation prediction step length, the loss function is designed: Among them, S2 is the simulation prediction step, t is the time point of simulation prediction, α and β are constants between 0 and 1, γ, σ, and δ are positive real numbers; αt-T ensures that the error of recent prediction has a greater weight; βyt-yt+1-βyt-yt2 is a mixed term of absolute error and square error; It is an additional penalty for large errors; S33. Establish a multi-dimensional simulation prediction result matrix for effluent water quality concentration A multi-dimensional simulation prediction result matrix is ​​established based on the model matrix elements and the custom loss function. Through different matrix elements, the characteristics of each dimension of historical data are mined and learned to generate corresponding simulation prediction results, and the results are stored in the simulation prediction result matrix: Where Fij represents the j-th simulation prediction result of the ith element of the algorithm matrix.

4. The method for dynamically optimizing the process parameters of the sewage treatment plant's bioreaction tank based on machine learning as claimed in claim 3, characterized in that: The ensemble learning algorithm suitable for fitting nonlinear feature relationships in step S31 includes but is not limited to XGBoost, CatBoost, and LightGBM.

5. The method for dynamically optimizing the process parameters of the sewage treatment plant's reaction tank based on machine learning as claimed in claim 3, characterized in that: The recurrent neural network algorithms in step S31 that are good at capturing time series characteristics include but are not limited to LSTM, GRU, BiLSTM, and BiGRU.

6. The method for dynamically optimizing and setting process parameters of a sewage treatment plant's bioreaction tank based on machine learning as claimed in claim 1, characterized in that: Step S4 includes: S41. Define model fusion coefficient Model fusion coefficient row vector K = [k1, k2..k n ] and constant coefficient column vector B = [b1, b2...bS2], where the elements in K are model fusion coefficients, which are fitted using a multivariate linear regression algorithm; S42, calculate the final result of water quality concentration simulation prediction R f =K×R+B Where R f =r1,r1…r S2 is a simulation prediction result column vector, which represents the final simulation prediction result of the effluent water quality concentration. i It represents the simulation prediction result of the effluent water quality concentration at the i-th moment.

7. The method for dynamically optimizing and setting process parameters of a sewage treatment plant's bioreaction tank based on machine learning as claimed in claim 1, characterized in that: Step S5 includes: S51. Data collection and preprocessing At the current moment, obtain sampling data within a hydraulic retention time HRT cycle. The data set should fully cover the influent characteristics, process control variables, and effluent status. Perform integration and necessary preprocessing operations on the above collected data according to the established method; S52, effluent water quality concentration simulation prediction The data pre-processed in step S51 is imported into the effluent water quality simulation prediction model matrix after the model fusion constructed in steps S3 to S4, and the effluent water quality concentration within the next S2 time steps is predicted using the model matrix to generate a prediction value sequence; S53, Error analysis and optimization adjustment Compare the difference between the predicted effluent water quality concentration in the future S2 time steps output by the model matrix and the target control value set inside the sewage treatment plant, calculate the error between the two, and take minimizing this error as the objective function. Use the particle swarm optimization algorithm PSO to find the best process parameter setting strategy through iterative search. In this process, continuously adjust and optimize the process parameter setting value until the error between the predicted effluent water quality concentration and the target value is less than the predetermined threshold. S54. Implement control decisions Once the particle swarm optimization algorithm is used to determine the process parameter setting scheme that makes the predicted value of the effluent water quality concentration reach the optimal matching degree, the optimized process parameter setting value is used as the actual process parameter control parameter of the bioreaction tank in the next M operation cycles to achieve dynamic and precise regulation of the process parameters in the bioreaction tank, where M = (HRT1-HRT2) / T.

8. A system for dynamically optimizing and setting process parameters of a sewage treatment plant's bioremediation tank based on machine learning, characterized in that: The method for dynamically optimizing and setting process parameters of a biological reaction tank in a sewage treatment plant based on machine learning as described in any one of claims 1 to 7 is implemented, comprising a data acquisition module, a data integration and preprocessing module, an effluent water quality concentration simulation prediction matrix module, a model fusion module and a dynamic optimization and setting module for process parameters of a biological reaction tank connected in sequence.

Images