An intelligent control method, system and device for sewage treatment aeration quantity

Abstract

The application discloses a sewage treatment aeration quantity intelligent control method, system and equipment, and belongs to the sewage treatment aeration control technical field.The control method breaks through the limitation of many aeration influencing factors, complex influencing relationship and difficult mechanism simulation of a sewage treatment plant, improves control precision and anti-load impact capacity, and effectively solves the problem of large control deviation of dissolved oxygen in the actual system; the control method quickly captures the evolution trend of multi-dimensional water quality information, improves the timeliness of decision response, and shortens the adjustment time; the control method adopts a decision mechanism of "simulation-optimization", meets the process requirement, effectively avoids excessive aeration, and significantly reduces the energy consumption of a fan.

Description

Technical Field

[0001] This invention belongs to the field of wastewater treatment aeration control technology, and relates to an intelligent control method, system and equipment for wastewater treatment aeration volume. Background Technology

[0002] Wastewater treatment aeration involves complex biochemical reactions and is influenced by the coupling of multiple factors. Furthermore, there is a significant lag between treatment and effluent detection. Therefore, intelligent aeration control often requires pre-adjustment of the treatment process to predict effluent quality and implement feedforward control. Existing intelligent wastewater aeration control methods often calculate the oxygen demand of the biological tank based on mechanistic formulas. Examples include patent applications CN120004404A (Intelligent Control Method for Water Treatment Aeration), CN120255591A (Intelligent Aeration Volume Control System and Aeration Volume Control Method), and CN120523073A (Wastewater Treatment Aeration Control Method and System Based on Artificial Intelligence and Mechanism Models). Currently, artificial intelligence models are only used to assist in predicting water quality changes. A few existing technologies further utilize artificial intelligence methods to predict theoretical oxygen demand, such as patent application CN121135005A (Precision Wastewater Aeration Method). However, actual wastewater systems are complex and deviate significantly from theoretically simplified systems. Theoretically calculated oxygen demand often fails to meet the production needs of actual wastewater treatment plants, resulting in a large discrepancy between the actual dissolved oxygen (DO) value and the target value after aeration using theoretical oxygen demand calculated by mechanisms or artificial intelligence. Therefore, a new technical solution is urgently needed to effectively address these issues. Summary of the Invention

[0003] This invention aims to provide a method, system, and equipment for intelligent control of aeration volume in wastewater treatment. The method effectively solves the problem that the theoretical oxygen demand calculated by existing technologies causes large deviations in dissolved oxygen control in real-world systems.

[0004] To achieve the above objectives, the technical solution adopted by this invention is: an intelligent control method for aeration volume in wastewater treatment, comprising:

[0005] S1. Establish a historical dataset of aeration operation of the wastewater treatment plant; the historical dataset includes a time series dataset with fixed intervals formed by aligning historical influent data, biological tank status data and aeration volume data of the wastewater treatment plant as input features; the biological tank status data includes dissolved oxygen data.

[0006] S2. The historical dataset is preprocessed. The influent data, biological tank status data and aeration volume at the preprocessed time Ts are taken as input data, and the dissolved oxygen data at the next time Ts+1 is taken as prediction label to obtain the sample dataset, which is divided into training set, validation set and test set.

[0007] S3. Use a neural network algorithm to optimize hyperparameters on the training set and the validation set, calibrate the positive monotonic relationship between aeration data and dissolved oxygen prediction value, select the best prediction effect to evaluate on the test set, and obtain the dissolved oxygen prediction model.

[0008] S4. Obtain the current influent data, current biological treatment tank status data, and current aeration volume data; keep the current influent data and the current biological treatment tank status data unchanged, and increase and / or decrease the current aeration volume data according to a preset step size to obtain an adjustable aeration volume set; based on the adjustable aeration volume set, obtain the next time-based dissolved oxygen prediction value through the dissolved oxygen prediction model, and generate a next time-based dissolved oxygen prediction value set; wherein, the current influent data is the influent data of the wastewater treatment plant at the current time Tt, the current biological treatment tank status data is the biological treatment tank status data of the wastewater treatment plant at the current time Tt, and the current aeration volume data is the aeration volume data of the wastewater treatment plant at the current time Tt;

[0009] S5. Obtain the recommended dissolved oxygen value, wherein the recommended dissolved oxygen value is the next time time predicted value with the smallest absolute value of the difference between the next time time predicted value set and the preset next time time target dissolved oxygen value; obtain the recommended aeration rate as the adjusted aeration rate corresponding to the recommended dissolved oxygen value; control the blower according to the recommended aeration rate to complete the aeration feedforward control of the current time Tt;

[0010] S6. Enter the aeration feedforward control of the next moment, repeat steps S4 to S5, and control the aeration volume of the sewage treatment plant in real time.

[0011] The solution provided by this invention overcomes the limitations of numerous influencing factors, complex relationships, and difficult mechanism simulation in wastewater treatment plants by establishing nonlinear relationships between multidimensional water quality information in a real water plant environment. This improves control accuracy and resistance to load shocks, effectively solving the problem of large dissolved oxygen control deviations in real-world systems. The intelligent aeration volume control method avoids designing complex multi-parameter physicochemical models of fluids, quickly captures the evolution trend of multidimensional water quality information, and correlates it with aeration volume to improve the timeliness of decision response and shorten adjustment time. The intelligent aeration volume control method adopts a "simulation-optimization" decision-making mechanism, meeting process requirements while also considering energy saving and consumption reduction. By simulating DO response curves under different aeration volumes and selecting the closest target value, compared to traditional methods relying on empirical formulas or continuous PID oscillation adjustment, the intelligent aeration volume control method can effectively avoid over-aeration, thereby significantly reducing blower energy consumption.

[0012] According to embodiments of the present invention, the present invention can be further optimized, and the optimized technical solution is as follows:

[0013] In one preferred embodiment, in step S1, the influent data includes influent flow rate, influent chemical oxygen demand, influent ammonia nitrogen, influent total nitrogen, and influent pH; the biochemical tank status data also includes the concentration of suspended solids in the mixed liquor of the biochemical tank, the influent return flow rate of the biochemical tank, and the outfluent return flow rate of the biochemical tank.

[0014] In one preferred embodiment, in step S2, the preprocessing of the historical dataset includes deduplication, deskipping, anomaly removal, missing value imputation, and discrete point interpolation; the division of the training set, validation set, and test set adopts a random partitioning method.

[0015] In one preferred embodiment, between step S2 and step S3, the data further includes: performing data standardization on the sample data; the data standardization is 0-1 normalization, log function standardization, linear proportional standardization, range standardization, or Z-score standardization.

[0016] In one preferred embodiment, step S3, the calibration of the positive monotonic relationship between the aeration volume data and the predicted dissolved oxygen value, specifically includes:

[0017] During training, the gradient of the dissolved oxygen prediction model output with respect to the aeration volume feature dimension is obtained, and the negative gradient of the aeration volume feature dimension is set as a constraint. The formula for calculating the overall loss during training is as follows:

[0018] ;

[0019] in, This represents the batch sample size. Let i be the true value of the i-th sample. Let be the predicted value for the i-th sample. This is the proportionality coefficient. The gradient of predicted dissolved oxygen with respect to the aeration rate dimension for the training set samples. Function implementation When grad(Q) is positive, the physical loss is 0; when grad(Q) is negative, the physical loss is... .

[0020] In one preferred embodiment, in step S3, the neural network algorithm is a backpropagation neural network, a recurrent neural network, a long short-term memory neural network, or a convolutional neural network; the hyperparameter optimization method is a grid search, a random search, a Bayesian optimization, a particle swarm optimization algorithm, or a genetic algorithm; the evaluation metrics include the coefficient of determination, mean absolute error, root mean square error, and mean absolute percentage error, calculated using the following formulas:

[0021] ;

[0022] ;

[0023] ;

[0024] ;

[0025] in, The determination coefficient is... The mean absolute error is... The root mean square error is... The mean absolute percentage error, The number of samples in the test set. Let j be the true value of the j-th sample in the test set. Let be the predicted value of the j-th sample in the test set. The mean of the true values ​​of the test set samples. The mean of the predicted values ​​for the test set samples.

[0026] In one preferred embodiment, in step S4, the range of values ​​for the adjusted aeration volume set is: The aeration rate is adjusted as follows: ;in, The current aeration volume data is as follows. This represents the maximum single adjustment range of the blower's aeration volume. , The preset step size is denoted as .

[0027] In one preferred embodiment, in step S6, the aeration feedforward control proceeds to the next time step, prior to which feedback control for the current time step Tt is also included:

[0028] Obtain the difference between the actual dissolved oxygen data at the next moment and the recommended dissolved oxygen value in step S5, and determine whether the difference is not less than a preset threshold. If it is not less than, execute feedback control to obtain a corrected aeration rate, and control the blower according to the corrected aeration rate until the difference is less than the preset threshold.

[0029] Preferably, the feedback control includes an error threshold compensation method, an error proportional compensation method, or a PID control method; when using the error threshold compensation method, the corrected aeration volume = the recommended aeration volume + the compensated aeration volume; preferably, the compensated aeration volume... for:

[0030] .

[0031] The solution provided by this invention integrates feedforward prediction and feedback correction to construct a closed-loop control system. When the feedforward control fails due to sudden changes in influent water quality or when model deviation causes the error between the actual dissolved oxygen (DO) and the target dissolved oxygen value to exceed the preset range, the deviation is corrected through feedback control. This avoids the open-loop risk that may exist in feedforward control, overcomes the lag of simple feedback control, and achieves stable and accurate aeration control.

[0032] Based on the same concept, the present invention also provides a method for obtaining a dissolved oxygen prediction model, comprising:

[0033] A1. Establish a historical dataset of aeration operation of the wastewater treatment plant; the historical dataset includes a time series dataset with fixed intervals formed by aligning historical influent data, biological tank status data and aeration volume data of the wastewater treatment plant as input features; the biological tank status data includes dissolved oxygen data;

[0034] A2. Preprocess the historical dataset, take the influent data, biological tank status data and aeration volume at the preprocessed time Ts as input data, and the dissolved oxygen data at the next time Ts+1 as prediction label to obtain the sample dataset, and divide it into training set, validation set and test set.

[0035] A3. Using a neural network algorithm, hyperparameters are optimized on the training set and the validation set to calibrate the positive monotonic relationship between aeration data and dissolved oxygen prediction values. The optimal prediction model is selected and evaluated on the test set to obtain the dissolved oxygen prediction model.

[0036] Based on the same concept, the present invention also provides an intelligent control system for wastewater treatment aeration volume, comprising:

[0037] A dataset construction module is used to establish a historical dataset of aeration operation in a wastewater treatment plant. This historical dataset includes time-series datasets with fixed intervals formed after time alignment, using historical influent data, biological tank status data, and aeration volume data as input features. The biological tank status data includes dissolved oxygen data. The module is used to preprocess the historical dataset, taking the preprocessed influent data, biological tank status data, and aeration volume at time Ti as input data, and the corresponding dissolved oxygen data at time Ti+1 as the prediction label, to obtain a sample dataset, which is then divided into a training set, a validation set, and a test set.

[0038] The prediction model building module is used to perform hyperparameter optimization on the training set and the validation set using a neural network algorithm, calibrate the positive monotonic relationship between aeration data and dissolved oxygen prediction values, select the optimal prediction model for evaluation on the test set, and obtain the dissolved oxygen prediction model.

[0039] The feedforward control module is used to acquire current influent data, current biological tank status data, and current aeration rate data; keeping the current influent data and current biological tank status data unchanged, it increments or decrements the current aeration rate data according to a preset step size to obtain an adjustable aeration rate set; based on the adjustable aeration rate set, it obtains the next time-of-flight dissolved oxygen prediction value through the dissolved oxygen prediction model, generating a next time-of-flight dissolved oxygen prediction value set; wherein, the current influent data is the influent data of the wastewater treatment plant at the current time Tt, and the current biological tank status data is the wastewater treatment plant at the current time Tt. The data includes the status data of the biological treatment tank, where the current aeration rate is the aeration rate data at the current time Tt of the wastewater treatment plant; it is used to obtain the recommended dissolved oxygen value, which is the next time time's predicted dissolved oxygen value with the smallest absolute value of the difference between the predicted dissolved oxygen value set and the preset next time time's target dissolved oxygen value; it obtains the adjusted aeration rate corresponding to the recommended dissolved oxygen value; it controls the blower according to the recommended aeration rate to complete the aeration feedforward control at the current time Tt; and it enters the aeration feedforward control at the next time to control the aeration rate of the wastewater treatment plant in real time.

[0040] Based on the same concept, the present invention also provides an electronic device, including a memory, a processor, and a computer program or instructions stored in the memory, wherein the processor executes the computer program or instructions to implement the intelligent control method for aeration volume in wastewater treatment or the method for obtaining a dissolved oxygen prediction model as described above.

[0041] Compared with the prior art, the beneficial effects of the present invention are:

[0042] This invention provides an intelligent control method, system, and equipment for aeration volume in wastewater treatment. The control method overcomes the limitations of numerous influencing factors, complex relationships, and difficult mechanism simulation in wastewater treatment plants by establishing nonlinear relationships between multidimensional water quality information in a real water plant environment. This improves control accuracy and resistance to load shocks, effectively solving the problem of large dissolved oxygen control deviations in real-world systems. The control method rapidly captures the evolution trend of multidimensional water quality information, improving the timeliness of decision response and shortening adjustment time. The control method employs a "simulation-optimization" decision-making mechanism, effectively avoiding over-aeration while meeting process requirements, and significantly reducing blower energy consumption. Attached Figure Description

[0043] Figure 1 This is a flowchart of an embodiment of the intelligent control method for aeration volume in wastewater treatment according to the present invention;

[0044] Figure 2 This is a flowchart of an intelligent control method for aeration volume in wastewater treatment, including feedback control, according to an embodiment of the present invention.

[0045] Figure 3This is an evaluation graph of the model on the test set after training, according to another embodiment of the present invention;

[0046] Figure 4 This is a comparison chart of recommended dissolved oxygen values ​​and actual dissolved oxygen values ​​according to another embodiment of the present invention;

[0047] Figure 5 This is a schematic diagram of an intelligent control system for wastewater treatment aeration volume according to another embodiment of the present invention. Detailed Implementation

[0048] The present invention will now be described in detail with reference to the accompanying drawings and embodiments. It should be noted that, unless otherwise specified, the embodiments and features described herein can be combined with each other.

[0049] Example 1

[0050] This embodiment 1 provides an intelligent control method for aeration volume in wastewater treatment, such as... Figure 1 As shown, the specific steps include:

[0051] S1. Establish a historical dataset of aeration operation of the wastewater treatment plant; the historical dataset includes a time series dataset with fixed intervals formed by aligning historical influent data, biological tank status data and aeration volume data of the wastewater treatment plant as input features; the biological tank status data includes dissolved oxygen data.

[0052] The influent data includes influent flow rate, influent chemical oxygen demand, influent ammonia nitrogen, influent total nitrogen, and influent pH; the biochemical tank status data also includes the concentration of suspended solids in the mixed liquor of the biochemical tank, the influent return flow rate of the biochemical tank, and the outfluent return flow rate of the biochemical tank.

[0053] S2. The historical dataset is preprocessed. The influent data, biological tank status data and aeration volume at the preprocessed time Ts are taken as input data, and the dissolved oxygen data at the next time Ts+1 is taken as prediction label to obtain the sample dataset, which is divided into training set, validation set and test set.

[0054] The preprocessing of the historical dataset includes deduplication, deskipping, anomaly removal, missing value imputation, and discrete point interpolation; the division of the training set, validation set, and test set adopts a random partitioning method. Deduplication is the removal of redundant data, that is, deleting duplicate records with overlapping timestamps or continuously constant values; deskipping is the filtering of abnormal mutations, that is, identifying and correcting jump data where the change amplitude between adjacent sampling points exceeds a set physical threshold.

[0055] The sample data is subjected to data standardization processing; the data standardization processing is 0-1 normalization, log function standardization, linear proportional standardization, range standardization, or Z-score standardization.

[0056] S3. Use a neural network algorithm to optimize hyperparameters on the training set and the validation set, calibrate the positive monotonic relationship between aeration volume data and dissolved oxygen prediction value, select the optimal prediction effect model for evaluation on the test set, and obtain the dissolved oxygen prediction model.

[0057] The calibration of the positive monotonic relationship between the aeration volume data and the predicted dissolved oxygen value specifically includes:

[0058] During training, the gradient of the dissolved oxygen prediction model output with respect to the aeration volume feature dimension is extracted, and the negative gradient of the aeration volume feature dimension is set as a constraint in the loss function; the calculation formula of the loss function is as follows:

[0059] ;

[0060] in, This represents the batch sample size. Let i be the true value of the i-th sample. Let be the predicted value for the i-th sample. This is the proportionality coefficient. The gradient of predicted dissolved oxygen with respect to the aeration rate dimension for the training set samples. Function implementation When grad(Q) is positive, the physical loss is 0; when grad(Q) is negative, the physical loss is... .

[0061] The neural network algorithm is a backpropagation neural network, a recurrent neural network, a long short-term memory neural network, or a convolutional neural network; the hyperparameter optimization method is a grid search, a random search, a Bayesian optimization, a particle swarm optimization algorithm, or a genetic algorithm; the evaluation metrics include the coefficient of determination, mean absolute error, root mean square error, and mean absolute percentage error, calculated using the following formulas:

[0062] ;

[0063] ;

[0064] ;

[0065] ;

[0066] in, The determination coefficient is... The mean absolute error is... The root mean square error is... The mean absolute percentage error, The number of samples in the test set. Let j be the true value of the j-th sample in the test set. Let be the predicted value of the j-th sample in the test set. The mean of the true values ​​of the test set samples. The mean of the predicted values ​​for the test set samples.

[0067] S4. Obtain the current influent data, current biological treatment tank status data, and current aeration volume data; keep the current influent data and the current biological treatment tank status data unchanged, and increase and / or decrease the current aeration volume data according to a preset step size to obtain an adjustable aeration volume set; based on the adjustable aeration volume set, obtain the next time-based dissolved oxygen prediction value through the dissolved oxygen prediction model, and generate a next time-based dissolved oxygen prediction value set; wherein, the current influent data is the influent data of the wastewater treatment plant at the current time Tt, the current biological treatment tank status data is the biological treatment tank status data of the wastewater treatment plant at the current time Tt, and the current aeration volume data is the aeration volume data of the wastewater treatment plant at the current time Tt;

[0068] The range of values ​​for the set of adjustable aeration volumes is: The aeration rate is adjusted as follows: ;in, The current aeration volume data is as follows. This represents the maximum single adjustment range of the blower's aeration volume. , The preset step size is [value]. Preferably, [value]. Set to 1m 3 / min, ±10m 3 / min. In practical applications, the preset step size is used. The smaller the value, the higher the precision of adjustment and the amplitude. The aeration rate needs to be moderate to ensure a smooth transition for the blower without damaging the equipment. In this embodiment of the invention, a series of predicted dissolved oxygen values ​​for the next moment are obtained through a dissolved oxygen prediction model, that is, different adjusted aeration rates are combined with other data features and input into the dissolved oxygen prediction model to obtain a set of predicted dissolved oxygen values ​​for the next moment under the influence of different adjusted aeration rates. The value range of the set of predicted dissolved oxygen values ​​for the next moment is... .

[0069] S5. Obtain the recommended dissolved oxygen value, wherein the recommended dissolved oxygen value is the next time time predicted value with the smallest absolute value of the difference between the next time time predicted value set and the preset next time time target dissolved oxygen value; obtain the recommended aeration rate as the adjusted aeration rate corresponding to the recommended dissolved oxygen value; control the blower according to the recommended aeration rate to complete the aeration feedforward control at the current time Tt;

[0070] Setting the target dissolved oxygen (DO) value for the next moment is based on the control requirements of the wastewater treatment plant, specifying the desired DO level for the next moment. target Find the target dissolved oxygen (DO) value in the set of predicted dissolved oxygen values ​​for the next time step. target The closest next time step dissolution prediction , It is used as the recommended value DO that can be regulated. best That is, the recommended dissolved oxygen value, and expressed in DO. best The corresponding adjustment of the aeration rate is used as the recommended aeration rate Q. best Adjust the aeration of the blower.

[0071] S6. Enter the aeration feedforward control of the next moment, repeat steps S4 to S5, and control the aeration volume of the sewage treatment plant in real time.

[0072] like Figure 2 As shown, the aeration feedforward control proceeds to the next moment, preceded by the feedback control of the current moment:

[0073] Obtain the difference between the actual dissolved oxygen data at the next moment and the recommended dissolved oxygen value in step S5, and determine whether the difference is not less than a preset threshold. If it is not less than, execute feedback control to obtain a corrected aeration rate, and control the blower according to the corrected aeration rate until the difference is less than the preset threshold.

[0074] Preferably, the feedback control includes an error threshold compensation method, an error proportional compensation method, or a PID control method; when the error threshold compensation method is used, the corrected aeration volume = the recommended aeration volume + the compensated aeration volume.

[0075] Example 2

[0076] In this embodiment of the invention, the aeration feedforward control employs a back propagation neural network (BPNN). It uses 10 variables as input features: current influent flow rate, influent chemical oxygen demand (COD), influent ammonia nitrogen, influent total nitrogen, influent pH, dissolved oxygen at the end of the aeration zone in the biological treatment tank, mixed liquor suspended solids (MLSS) concentration in the biological treatment tank, internal return flow rate in the biological treatment tank, external return flow rate in the biological treatment tank, and aeration rate. This information is used to predict the dissolved oxygen in the biological treatment tank at the next moment.

[0077] Historical data of 10 variables from a wastewater treatment plant for the entire year of 2025 were collected, including influent flow rate, influent COD, influent ammonia nitrogen, influent total nitrogen, influent pH, dissolved oxygen in the biological treatment tank, MLSS in the biological treatment tank, internal return flow in the biological treatment tank, external return flow in the biological treatment tank, and aeration rate. These data were then aligned at 10-minute intervals to form a time series dataset.

[0078] After the collected historical data underwent preprocessing such as deduplication, dejumping, outlier removal, and missing value interpolation imputation, the dissolved oxygen (DO) of the biochemical pool for each 10-minute interval was selected as the prediction target, and the training, validation, and test sets were divided in a ratio of 7:1.5:1.5.

[0079] Starting from the initial moment of the training set, the aforementioned 10 variables are selected as input data, and the dissolved oxygen (DO) of the biochemical pool at the next moment is used as the output. A BPNN network is constructed and trained using all training set data. The network shape is 10, 360, 180, 90, 45, 1 in sequence. The loss function is the mean squared error loss, and the calculation formula is as follows:

[0080] ;

[0081] in, Number of samples per batch For the true value of the sample, The predicted value for the sample.

[0082] In addition, the physical loss for model training is constructed based on the aeration volume feature dimension. When the gradient of the predicted dissolved oxygen with respect to the aeration volume dimension is negative, 0.05 times (not limited to 0.05 times in practical applications) of the gradient value of the aeration volume dimension is taken as the physical loss. The calculation formula is as follows:

[0083] ;

[0084] in, Number of samples per batch To predict the gradient of dissolved oxygen with respect to aeration rate for the training samples, after... Function, implementation When grad(Q) is positive, the physical loss is 0; when grad(Q) is negative, the physical loss is... .

[0085] Overall loss During training, the model training effect was verified using a validation set, and the hyperparameters of the BPNN network, including the learning rate, the proportion of neurons ignored by dropout, the batch size, and the number of epochs, were optimized using the particle swarm optimization algorithm. The population size was set to 100, and the termination generation of the particle swarm optimization algorithm was set to 200. The optimized hyperparameters of the BPNN dissolved oxygen prediction model were obtained, as shown in Table 1.

[0086]

[0087] like Figure 3 As shown, the optimized model is selected for evaluation on the test set. The model's coefficient of determination R0 is... 2 The mean absolute percentage error (MAPE) was 0.979588, and the mean absolute percentage error (MAPE) was 8.27%, demonstrating that the model has a strong ability to predict dissolved oxygen (DO) in the future.

[0088] Table 2 shows the 10 model input values ​​for the latest time T (taking 12:00 PM on February 3, 2026 as an example). The aeration rate is 74.8 m³ / s. 3 Based on the aeration rate of / min, the aeration rate feature is adjusted with a step size of 1 and a single maximum amplitude of ±10, resulting in a set of aeration rate feature values ​​[64.8, 65.8, 66.8, ..., 82.8, 83.8, 84.8], a total of 21 aeration rates. These aeration rates are then combined with the other 9 features in Table 2 and input into the BPNN dissolved oxygen prediction model to obtain the corresponding set of predicted dissolved oxygen (DO) values ​​at time T+1 [0.94, 0.97, 0.99, 1.02, 1.03, 1.05, 1.06, 1.07, 1.09, 1.10, 1.12, 1.16, 1.21, 1.24, 1.26, 1.29, 1.34, 1.37, 1.42, 1.48, 1.52].

[0089]

[0090] Based on water management experience, and considering the online monitoring of ammonia nitrogen concentration in the effluent from the biological treatment tank and the dissolved oxygen (DO) control targets under different seasonal conditions, the current month (time T) and the monitored online ammonia nitrogen concentration in the effluent from the biological treatment tank are 1.1 mg / L. Substituting these values ​​into Table 3, the future DO target value is calculated to be 1 mg / L. The nearest achievable DO value (1 mg / L) is then found in the predicted DO value set. best The result was 0.99 mg / L, corresponding to an aeration rate of 66.8 m³ / L. 3 / min, which is used as the recommended aeration value Q. best Adjust the aeration rate of the blower to 66.8 m³.3 Feedforward control is completed in / min.

[0091]

[0092] After completing the feedforward control at time T, monitor the actual dissolved oxygen (DO) value at time T+1 (12:10 PM on February 3, 2026). true Given a value of 1.07 and an acceptable error range of ±0.1, calculate the dissolved oxygen (DO) error ε = DO true - DO best = 0.08. At this point, ε is within the acceptable range, so the feedforward control effect is considered to meet the requirements. Immediately monitor the 10 model input values ​​at time T+1 for the next round of feedforward control.

[0093] The actual dissolved oxygen (DO) value is continuously monitored for subsequent feedforward control. When the dissolved oxygen (DO) error ε exceeds the acceptable range, feedback control using the error threshold compensation method is employed to obtain the corrected aeration rate. The magnitude of the compensated aeration rate ΔQ is as follows:

[0094] ;

[0095] Continuously monitor the magnitude of ε and intermittently provide aeration feedback compensation until ε falls below the acceptable range, then begin the next round of feedforward control. Figure 4 The diagram shows the long-term control effect. Using the intelligent aeration control method provided in Example 1, the actual dissolved oxygen (DO) is in good agreement with the recommended DO, proving that the control method can effectively control the actual DO value of the wastewater treatment plant, provide suitable oxygen conditions for the aeration reaction, ensure the reaction proceeds, and avoid energy waste.

[0096] Example 3

[0097] This embodiment 3 also provides a method for obtaining a dissolved oxygen prediction model, with the following specific steps:

[0098] A1. Establish a historical dataset of aeration operation of the wastewater treatment plant; the historical dataset includes a time series dataset with fixed intervals formed by aligning historical influent data, biological tank status data and aeration volume data of the wastewater treatment plant as input features; the biological tank status data includes dissolved oxygen data;

[0099] A2. Preprocess the historical dataset, take the influent data, biological tank status data and aeration volume at the preprocessed time Ts as input data, and the dissolved oxygen data at the next time Ts+1 as prediction label to obtain the sample dataset, and divide it into training set, validation set and test set.

[0100] A3. Using a neural network algorithm, hyperparameters are optimized on the training set and the validation set to calibrate the positive monotonic relationship between aeration data and dissolved oxygen prediction values. The optimal prediction model is selected and evaluated on the test set to obtain the dissolved oxygen prediction model.

[0101] like Figure 5 As shown, this embodiment 3 also provides an intelligent control system for wastewater treatment aeration volume, including:

[0102] A dataset construction module is used to establish a historical dataset of aeration operation in a wastewater treatment plant. This historical dataset includes time-series datasets with fixed intervals formed after time alignment, using historical influent data, biological tank status data, and aeration volume data as input features. The biological tank status data includes dissolved oxygen data. The module is used to preprocess the historical dataset, taking the preprocessed influent data, biological tank status data, and aeration volume at time Ti as input data, and the corresponding dissolved oxygen data at time Ti+1 as the prediction label, to obtain a sample dataset, which is then divided into a training set, a validation set, and a test set.

[0103] The prediction model building module is used to perform hyperparameter optimization on the training set and the validation set using a neural network algorithm, calibrate the positive monotonic relationship between aeration data and dissolved oxygen prediction values, select the optimal prediction model for evaluation on the test set, and obtain the dissolved oxygen prediction model.

[0104] The feedforward control module is used to acquire current influent data, current biological tank status data, and current aeration rate data; keeping the current influent data and current biological tank status data unchanged, it increments or decrements the current aeration rate data according to a preset step size to obtain an adjustable aeration rate set; based on the adjustable aeration rate set, it obtains the next time-of-flight dissolved oxygen prediction value through the dissolved oxygen prediction model, generating a next time-of-flight dissolved oxygen prediction value set; wherein, the current influent data is the influent data of the wastewater treatment plant at the current time Tt, and the current biological tank status data is the wastewater treatment plant at the current time Tt. The data includes the status data of the biological treatment tank, where the current aeration rate is the aeration rate data at the current time Tt of the wastewater treatment plant; the recommended dissolved oxygen value is obtained, where the recommended dissolved oxygen value is the next time time predicted value whose absolute difference from the preset dissolved oxygen target value is the smallest among the set of next time time predicted values; the adjusted aeration rate corresponding to the recommended dissolved oxygen value is obtained; the blower is controlled according to the recommended aeration rate to complete the aeration feedforward control at the current time Tt; the aeration feedforward control is repeated at the next time time to control the aeration rate of the wastewater treatment plant in real time.

[0105] This embodiment 3 also provides an electronic device, which includes: a memory, a processor, and a computer program or instructions stored in the memory. The processor executes the computer program or instructions to implement the intelligent control method for aeration volume in wastewater treatment or the method for obtaining a dissolved oxygen prediction model in this embodiment of the invention.

[0106] Although not shown, the electronic device includes a processor that can perform various appropriate operations and processes based on programs and / or data stored in read-only memory (ROM) or loaded from a storage portion into random access memory (RAM). The processor can be a multi-core processor or may contain multiple processors. In some embodiments, the processor may include a general-purpose main processor and one or more specialized coprocessors, such as a central processing unit, graphics processing unit (GPU), neural network processor (NPU), digital signal processor (DSP), etc. Various programs and data required for device operation are also stored in RAM. The processor, ROM, and RAM are interconnected via a bus. Input / output (I / O) interfaces are also connected to the bus.

[0107] The processor and memory described above are used together to execute programs / instructions stored in the memory. When the program / instructions are executed by the computer, they can implement the methods, steps, or functions described in the above embodiments.

[0108] The above embodiments should be understood as being used only to illustrate the present invention more clearly, and not to limit the scope of the present invention. After reading the present invention, any modifications of the present embodiments by those skilled in the art will fall within the scope defined by the appended claims.

Claims

1. A method for intelligent control of aeration volume in wastewater treatment, characterized in that, include: S1. Establish a historical dataset of aeration operation of the wastewater treatment plant; the historical dataset includes a time series dataset with fixed intervals formed by aligning the historical influent data, biological tank status data and aeration volume data of the wastewater treatment plant as input features. The biochemical tank status data includes dissolved oxygen data; S2. The historical dataset is preprocessed. The influent data, biological tank status data and aeration volume at the preprocessed time Ts are taken as input data, and the dissolved oxygen data at the next time Ts+1 is taken as prediction label to obtain the sample dataset, which is divided into training set, validation set and test set. S3. Use a neural network algorithm to optimize hyperparameters on the training set and the validation set, calibrate the positive monotonic relationship between aeration volume data and dissolved oxygen prediction value, select the optimal prediction effect model for evaluation on the test set, and obtain the dissolved oxygen prediction model. S4. Obtain current influent data, current biological tank status data, and current aeration volume data; Keeping the current influent data and the current biological treatment tank status data unchanged, the current aeration rate data is increased and / or decreased according to a preset step size to obtain an adjustable aeration rate set. Based on the adjustable aeration rate set, the dissolved oxygen prediction value for the next time moment is obtained through the dissolved oxygen prediction model, generating a set of dissolved oxygen prediction values ​​for the next time moment. Wherein, the current influent data is the influent data of the wastewater treatment plant at the current time Tt, the current biological treatment tank status data is the biological treatment tank status data of the wastewater treatment plant at the current time Tt, and the current aeration rate data is the aeration rate data of the wastewater treatment plant at the current time Tt. S5. Obtain the recommended dissolved oxygen value, wherein the recommended dissolved oxygen value is the next time time predicted value with the smallest absolute value of the difference between the next time time predicted value set and the preset next time time target dissolved oxygen value; obtain the recommended aeration rate as the adjusted aeration rate corresponding to the recommended dissolved oxygen value; control the blower according to the recommended aeration rate to complete the aeration feedforward control of the current time Tt; S6. Enter the aeration feedforward control of the next moment, repeat steps S4 to S5, and control the aeration volume of the sewage treatment plant in real time.

2. The intelligent control method for aeration volume in wastewater treatment according to claim 1, characterized in that, In step S1, the influent data includes influent flow rate, influent chemical oxygen demand, influent ammonia nitrogen, influent total nitrogen, and influent pH; the biochemical tank status data also includes the concentration of suspended solids in the mixed liquor of the biochemical tank, the influent return flow rate of the biochemical tank, and the outfluent return flow rate of the biochemical tank.

3. The intelligent control method for aeration volume in wastewater treatment according to claim 1, characterized in that, In step S2, the preprocessing of the historical dataset includes deduplication, deskipping, anomaly removal, missing value imputation, and discrete point interpolation; the division of the training set, validation set, and test set adopts a random partitioning method.

4. The intelligent control method for aeration volume in wastewater treatment according to claim 1, characterized in that, Between steps S2 and S3, the data further includes: performing data standardization on the sample data; the data standardization is 0-1 normalization, log function standardization, linear proportional standardization, range standardization, or Z-score standardization.

5. The intelligent control method for aeration volume in wastewater treatment according to claim 1, characterized in that, In step S3, the calibration of the positive monotonic relationship between the aeration volume data and the predicted dissolved oxygen value specifically includes: During training, the gradient of the dissolved oxygen prediction model output with respect to the aeration volume feature dimension is obtained, and the negative gradient of the aeration volume feature dimension is set as a constraint. The formula for calculating the overall loss during training is as follows: ； in, This represents the batch sample size. Let i be the true value of the i-th sample. Let be the predicted value for the i-th sample. This is the proportionality coefficient. The gradient of predicted dissolved oxygen with respect to the aeration rate dimension for the training set samples. Function implementation When grad(Q) is positive, the physical loss is 0; when grad(Q) is negative, the physical loss is... .

6. The intelligent control method for aeration volume in wastewater treatment according to claim 1, characterized in that, In step S3, the neural network algorithm is a backpropagation neural network, a recurrent neural network, a long short-term memory neural network, or a convolutional neural network; the hyperparameter optimization method is a grid search, a random search, a Bayesian optimization, a particle swarm optimization algorithm, or a genetic algorithm; the evaluation indicators include the coefficient of determination, mean absolute error, root mean square error, and mean absolute percentage error, calculated using the following formulas: ； ； ； ； in, The determination coefficient is... The mean absolute error is... The root mean square error is... The mean absolute percentage error, The number of samples in the test set. Let j be the true value of the j-th sample in the test set. Let be the predicted value of the j-th sample in the test set. The mean of the true values ​​of the test set samples. The mean of the predicted values ​​for the test set samples.

7. The intelligent control method for aeration volume in wastewater treatment according to claim 1, characterized in that, In step S4, the range of values ​​for the set of adjusted aeration volumes is: The aeration rate is adjusted as follows: ;in, The current aeration volume data is as follows. This represents the maximum single adjustment range of the fan. , The preset step size is denoted as .

8. The intelligent control method for aeration volume in wastewater treatment according to claim 1, characterized in that, In step S6, the aeration feedforward control for the next time step begins, preceded by feedback control for the current time step Tt: Obtain the difference between the actual dissolved oxygen data at the next moment and the recommended dissolved oxygen value in step S5, and determine whether the difference is not less than a preset threshold. If it is not less than, execute feedback control to obtain a corrected aeration rate, and control the blower according to the corrected aeration rate until the difference is less than the preset threshold. Preferably, the feedback control includes an error threshold compensation method, an error proportional compensation method, or a PID control method; When using the error threshold compensation method, the corrected aeration volume = the recommended aeration volume + the compensated aeration volume.

9. A method for obtaining a dissolved oxygen prediction model, characterized in that, include: A1. Establish a historical dataset of aeration operation of a wastewater treatment plant; the historical dataset includes a time series dataset with fixed intervals formed after time alignment, using historical influent data, biological tank status data and aeration volume data of the wastewater treatment plant as input features. The biochemical tank status data includes dissolved oxygen data; A2. Preprocess the historical dataset, take the influent data, biological tank status data and aeration volume at the preprocessed time Ts as input data, and the dissolved oxygen data at the next time Ts+1 as prediction label to obtain the sample dataset, and divide it into training set, validation set and test set. A3. Using a neural network algorithm, hyperparameters are optimized on the training set and the validation set to calibrate the positive monotonic relationship between aeration data and dissolved oxygen prediction values. The optimal prediction model is selected and evaluated on the test set to obtain the dissolved oxygen prediction model.

10. An intelligent control system for aeration volume in wastewater treatment, characterized in that, include: Dataset building module; Used to establish a historical dataset of aeration operation in a wastewater treatment plant; the historical dataset includes a time series dataset with fixed intervals formed after time alignment, using historical influent data, biological tank status data and aeration volume data of the wastewater treatment plant as input features. The biochemical tank status data includes dissolved oxygen data; it is used to preprocess the historical dataset, taking the influent data, biochemical tank status data and aeration rate at time Ti after preprocessing as input data, and the dissolved oxygen data at time Ti+1 as prediction label to obtain a sample dataset, which is divided into training set, validation set and test set. The prediction model building module is used to perform hyperparameter optimization on the training set and the validation set using a neural network algorithm, calibrate the positive monotonic relationship between aeration data and dissolved oxygen prediction values, select the optimal prediction model for evaluation on the test set, and obtain the dissolved oxygen prediction model. Feedforward control module; This system is used to acquire current influent data, current biological tank status data, and current aeration volume data; keeping the current influent data and current biological tank status data unchanged, the current aeration volume data is incremented or decremented according to a preset step size to obtain an adjustable aeration volume set; based on the adjustable aeration volume set, the dissolved oxygen prediction value for the next time moment is obtained through the dissolved oxygen prediction model, generating a set of dissolved oxygen prediction values ​​for the next time moment; wherein, the current influent data is the influent data of the wastewater treatment plant at the current time Tt, the current biological tank status data is the biological tank status data of the wastewater treatment plant at the current time Tt, and the current aeration volume data is the aeration volume data of the wastewater treatment plant at the current time Tt; it is used to acquire a recommended dissolved oxygen value, which is the next time moment's recommended dissolved oxygen prediction value with the smallest absolute value of the difference between the next time moment's predicted dissolved oxygen value set and the preset next time moment's target dissolved oxygen value; it acquires the adjustable aeration volume corresponding to the recommended dissolved oxygen value; and it controls the blower according to the recommended aeration volume to complete the aeration feedforward control for the current time Tt; Enter the next moment of aeration feedforward control to control the aeration volume of the sewage treatment plant in real time.

11. An electronic device comprising a memory, a processor, and a computer program or instructions stored in the memory, characterized in that, The processor executes the computer program or instructions to implement the intelligent control method for aeration volume in wastewater treatment as described in any one of claims 1 to 8 or the method for obtaining the dissolved oxygen prediction model as described in claim 9.

Images