Weighted fusion model-based intelligent aeration control method for sewage treatment

The intelligent aeration control method based on a weighted fusion model, which combines deep artificial neural networks and long short-term memory neural networks, solves the problems of low efficiency and insufficient accuracy in manual aeration control, and achieves efficient, stable and precise aeration control for wastewater treatment.

CN118954809AActive Publication Date: 2024-11-15HARBIN INSTITUTE OF TECHNOLOGY (SHENZHEN) (INSTITUTE OF SCIENCE AND TECHNOLOGY INNOVATION HARBIN INSTITUTE OF TECHNOLOGY SHENZHEN)

Patent Information

Application Number
CN202410936945.8
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-12
Publication Date
2024-11-15
Estimated Expiration
2044-07-12

AI Technical Summary

Technical Problem

In existing technologies, the aeration control of the wastewater biochemical treatment process relies on manual operation, which leads to low efficiency, slow response speed, insufficient precision, difficulty in achieving precise adjustment, increased labor costs and operational errors, and affects the wastewater treatment effect.

Method used

An intelligent aeration control method based on a weighted fusion model is adopted. By combining a deep artificial neural network and a long short-term memory neural network fusion model with water quality and equipment operating data, the aeration volume is monitored and optimized in real time to achieve efficient and precise sewage treatment.

Benefits of technology

It achieves efficient, stable and precise control of aeration during the sewage treatment process, reduces labor costs, minimizes operational errors, and ensures that sewage is discharged in compliance with standards.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118954809A_ABST
    Figure CN118954809A_ABST
Patent Text Reader

Abstract

The invention relates to the technical field of sewage treatment, in particular to an intelligent aeration control method for sewage treatment based on a weighted fusion model. Acquiring an inlet water quality monitoring data set, a process water quality monitoring data set, an outlet water quality monitoring data set, an aeration rate monitoring data set and an equipment working condition monitoring data set; the inlet water quality monitoring data set, the process water quality monitoring data set, the outlet water quality monitoring data set and the aeration rate monitoring data set are fused into a water quality aeration fusion data set, and the equipment working condition monitoring data set and the aeration rate monitoring data set are fused into a working condition aeration fusion data set. And inputting the water quality aeration fusion data set into the parameter-adjustable deep artificial neural network sub-model, and inputting the working condition aeration fusion data set into the parameter-adjustable long-short term memory neural network sub-model. And fusing the results of the two models into a parameter-adjustable artificial neural network and long and short-term memory network fusion model, and fusing according to a weight coefficient after percentage error normalization.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of sewage treatment, and in particular to an intelligent aeration control method for sewage treatment based on a weighted fusion model. Background Art

[0002] Direct discharge of industrial and domestic sewage without treatment can cause serious environmental pollution, water ecological damage and public health risks. Untreated sewage contains a large number of harmful substances, and direct discharge can pollute rivers, lakes and groundwater, causing water quality deterioration. Pollutants can also damage the water ecosystem and endanger the survival of aquatic organisms. In addition, polluted water bodies may spread diseases and threaten the health and quality of life of surrounding residents. This practice not only causes long-term damage to the environment, but also brings huge social and economic costs.

[0003] Biochemical treatment and aeration play a key role in sewage treatment. Biochemical treatment uses microorganisms to degrade organic pollutants in sewage and convert them into harmless substances, thereby purifying water quality. Aeration introduces oxygen into sewage to enhance the activity and reproduction rate of microorganisms and improve their efficiency in decomposing organic pollutants. Aeration not only promotes the biochemical treatment process, but also effectively prevents the formation of anaerobic environment in sewage and reduces the generation of odor and harmful gases.

[0004] At present, the control of the aeration stage of the sewage biochemical treatment process in China is still mainly concentrated on manual control. The disadvantages of manual control of aeration mainly include low efficiency, slow response speed and insufficient precision. Manual control requires operators to constantly monitor and adjust, which is labor-intensive and difficult to respond to water quality changes in real time, resulting in unstable aeration effect. In addition, manual adjustment is difficult to be precise, which may cause excessive or insufficient aeration, affecting the sewage treatment effect. High labor costs and the risk of operational errors and mistakes also increase management complexity and operating costs.

[0005] In summary, the existing technology relies on manual control during the aeration stage, which makes it impossible to ensure that the aeration volume can accurately meet the biochemical process. There are technical problems such as excessive sewage discharge causing environmental pollution and adaptive adjustment under abnormal environmental conditions. Therefore, it is necessary to use intelligent models to improve the accuracy of control. Summary of the invention

[0006] The purpose of the present invention is to provide an intelligent aeration control method for sewage treatment based on a weighted fusion model to solve the problem that aeration control in the prior art only relies on manual control and ignores the guiding role of the intelligent model on the change of aeration state.

[0007] To achieve the above object, the present invention adopts the following technical solution: a sewage treatment intelligent aeration control method based on a weighted fusion model, comprising the following steps:

[0008] S100, sampling the water quality indicators of inlet monitoring to obtain an inlet water quality monitoring data set, sampling the water quality indicators of process monitoring to obtain a process water quality monitoring data set, sampling the water quality indicators of effluent monitoring to obtain an effluent water quality monitoring data set; collecting equipment data from the aeration equipment of the aerobic tank in the biochemical section of the sewage treatment to obtain an aeration volume monitoring data set; collecting equipment data from the sewage treatment equipment to obtain an equipment operating condition monitoring data set;

[0009] S200, merging the inlet water quality monitoring data set, the process water quality monitoring data set, the outlet water quality monitoring data set and the aeration volume monitoring data set into a water quality aeration fusion data set; merging the equipment operating condition monitoring data set and the aeration volume monitoring data set into an operating condition aeration fusion data set;

[0010] S300, inputting the water quality aeration fusion data set into the deep artificial neural network sub-model with adjustable parameters; inputting the working condition aeration fusion data set into the long short-term memory neural network sub-model with adjustable parameters;

[0011] S400, fusing the results of the two models into an artificial neural network and a long short-term memory network fusion model with adjustable parameters, and fusing them according to the weight coefficient after the percentage error is normalized, wherein the adjustable parameter adaptive model is embedded in the control system terminal of the sewage treatment aeration equipment;

[0012] S500, based on the artificial neural network and long short-term memory network fusion model, the sewage treatment aeration process is controlled according to the output results.

[0013] Furthermore, in the S100, the inlet water quality monitoring data set, the process water quality monitoring data set and the effluent water quality monitoring data set are data sets of the historical period of the target sewage treatment plant; the data type in the aeration volume monitoring data set includes the aeration volume of the sewage treatment equipment in the aerobic tank of the process section; the data type in the equipment condition monitoring data set includes sewage anaerobic hydraulic retention time data, anoxic hydraulic retention time data, and aerobic hydraulic retention time data.

[0014] Furthermore, in the above S200, the method for merging the data sets is to align the data rows and data columns of the data sets, and the data interval is set to the minimum sampling time.

[0015] Furthermore, in the S300, the method for constructing a deep artificial neural network is as follows: first, prepare and preprocess 20% of the data set and divide it into a training set, a validation set and a test set; then, randomly initialize the weights and bias parameters of the network; then, calculate the output of each layer through forward propagation, and then calculate the error between the network output and the actual result according to the loss function; calculate the gradient through the back propagation algorithm, and use the optimization algorithm to update the network parameters; repeat the above process until the loss function percentage error is less than 10% and converges or reaches a predetermined number of training rounds of 500, and output the adjustable parameters at the same time.

[0016] Furthermore, in the S300, the method for constructing the long short-term memory neural network is as follows: first, prepare and preprocess 20% of the data set and divide it into a training set, a validation set and a test set; then, randomly initialize the weights and bias parameters of the network; then, calculate the output of each layer of cell units through forward propagation, and then calculate the error between the network output and the actual result according to the loss function; calculate the gradient through the back propagation algorithm, and use the optimization algorithm to update the network parameters; repeat the above process until the loss function percentage error is less than 10% convergence or reaches a predetermined number of training rounds of 500, and output the adjustable parameters at the same time.

[0017] Furthermore, the S500 includes the following steps:

[0018] S510, setting the parameters of the neural network and long short-term memory network fusion model as temporary adjustable parameters, inputting 100% of the fusion data into the neural network and long short-term memory network fusion model, and the model outputting the recommended aeration amount of the aeration device;

[0019] S520, the control module controls the sewage treatment aeration equipment of the aerobic tank in the biochemical section of the sewage treatment aeration process according to the aeration amount recommended by the model;

[0020] S530, monitoring the water quality of the water samples collected through the real-time effluent sampling tube, obtaining the effluent water quality, and obtaining a effluent water quality monitoring data set at time T;

[0021] S540: Determine whether the water quality index meets the standard. If the water quality index meets the standard, obtain a sewage discharge instruction to allow discharge.

[0022] Beneficial effects of the present invention:

[0023] 1. The method of the present invention solves the problem that manual control in the prior art is usually inefficient, slow in response, and has high labor costs, and requires operators to constantly monitor and adjust, making it difficult to achieve real-time response and precise adjustment, resulting in unstable aeration effects;

[0024] 2. The method of the present invention solves the problem of high labor costs and the risk of operational errors and mistakes, and difficulty in achieving 24-hour uninterrupted monitoring, and realizes efficient, intelligent, precise and stable control of the sewage treatment aeration process. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Figure 1 This is a schematic diagram of the process of the present invention;

[0026] Figure 2 This is a schematic diagram of monitoring sites for the method of the present invention;

[0027] Figure 3 This is a structural diagram of the control module of the method of the present invention. DETAILED DESCRIPTION

[0028] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the accompanying drawings in the embodiments of the present invention.

[0029] Embodiments of the present invention:

[0030] An intelligent aeration control method for sewage treatment based on weighted fusion model, such as Figure 1 As shown, the following steps are included:

[0031] S100, monitor water quality, equipment operating conditions, aeration volume and other information separately

[0032] SA100, monitor water quality indicators of inlet, process and outlet water;

[0033] SA1001. Monitor the water quality of water samples collected through the inlet sampling pipe and obtain an inlet water quality monitoring data set;

[0034] SA10011. Obtain the historical influent sampling data set of the target sewage treatment plant;

[0035] SA10012. Perform data preprocessing, fill missing values, and delete outliers according to the historical water sampling data set;

[0036] SA10013. Configure parameters of water quality monitoring instrument according to the fluctuation characteristics of water quality indicators.

[0037] SA1002, monitor the water quality of water samples collected through the sampling tube of the biochemical section and obtain the process water quality monitoring data set;

[0038] SA10021. Obtain the water quality sampling data set of the process biochemical section of the target sewage treatment plant during the historical period;

[0039] SA10022. Sampling data sets according to past process segments, preprocessing data, filling missing values, and removing outliers;

[0040] SA10023. Configure parameters of water quality monitoring instrument according to the fluctuation characteristics of water quality indicators.

[0041] SA1003, monitor the water quality of water samples collected through the effluent sampling tube to obtain an effluent water quality monitoring data set;

[0042] SA10031. Obtain the effluent sampling dataset of the target sewage treatment plant in the historical period;

[0043] SA10032. Perform data preprocessing, fill missing values, and delete outliers according to the past water discharge section sampling data set;

[0044] SA10033. Configure parameters of water quality monitoring instrument according to the fluctuation characteristics of water quality indicators.

[0045] Among them, data collection requires corresponding data collection devices, and the water quality data collection device includes at least three water quality monitors at each site to take the average value to reduce errors;

[0046] Specifically, in the present embodiments SA1001, SA1002 and SA1003, the target sewage treatment plant is a professional sewage treatment organization that adopts biochemical treatment to treat sewage. The sewage undergoes sewage biochemical treatment processes such as sewage inlet regulation, effluent regulation, anaerobic reaction, anoxic reaction, and aerobic reaction in the sewage treatment equipment of the target sewage treatment plant.

[0047] SB100, monitor the aeration index of the aerobic pool in the biochemical section;

[0048] The aeration volume data of the sewage treatment equipment is collected to obtain an aeration volume monitoring data set, wherein the data types in the aeration volume monitoring data set include the aeration volume of the sewage treatment equipment in the process section aerobic tank.

[0049] SC100, monitor the equipment working condition

[0050] By collecting equipment data from the water treatment equipment, the equipment condition monitoring data set is obtained; the sewage undergoes sewage inlet regulation, effluent regulation, anaerobic reaction, anoxic reaction, aerobic reaction and other sewage biochemical treatment processes in the sewage treatment equipment of the target sewage treatment plant, and the sewage anaerobic hydraulic retention time, anoxic hydraulic retention time, aerobic hydraulic retention time and deep treatment agent addition time are adjusted and controlled based on the control system to achieve reasonable space for building anaerobic conditions, anoxic conditions and aerobic conditions of the sewage treatment equipment and the normal progress of deep treatment, so that the sewage can effectively undergo sewage inlet regulation, effluent regulation, anaerobic reaction, anoxic reaction, aerobic reaction and other sewage biochemical treatment processes.

[0051] Data collection is performed on the control system used to control the sewage treatment equipment for sewage treatment, and data collection is performed on the actual operation data of the sewage treatment equipment to obtain the equipment condition monitoring data set. The data types in the equipment condition monitoring data set include sewage anaerobic hydraulic retention time data, anoxic hydraulic retention time data, and aerobic hydraulic retention time data.

[0052] S200, merge the data sets used for building multiple models into water quality aeration fusion data set and working condition aeration fusion data set

[0053] The data fusion method is: align the data rows and columns of the data set, and set the data interval to the minimum sampling time.

[0054] SA200, merge the inlet water quality monitoring data set, process water quality monitoring data set, effluent water quality monitoring data set and aeration volume monitoring data set into a water quality aeration fusion data set;

[0055] SB200, merge the equipment condition monitoring data set and the aeration volume monitoring data set into a condition aeration fusion data set.

[0056] S300, build adaptive model, perform model training and loss optimization

[0057] These include the SA300 deep artificial neural network sub-model and the SB300 long short-term memory neural network sub-model.

[0058] SA3001. Build a deep artificial neural network sub-model with adjustable parameters, where the adjustable parameter adaptive model includes the number of deep artificial neural network data input batches, the size of deep artificial neural network data input batches, the number of neural network hidden layers, and the dimension of neural network input data;

[0059] SA3002, using 20% ​​of the water quality aeration fusion data set as input variables, training the deep artificial neural network and obtaining temporary adjustable parameters. This includes inputting the water quality aeration fusion data set into the deep artificial neural network sub-model, and outputting adaptive parameters based on the adjustable parameter adaptive model.

[0060] The core structure of a deep artificial neural network includes an input layer, a hidden layer, an output layer, and an activation function. The main structure of the deep artificial neural network sub-model includes 1 input layer, 6 hidden layers, 1 output layer, and 6 activation functions. Each layer consists of multiple neurons, which are interconnected through weight matrices and bias parameters to achieve information transmission and conversion. The input layer of a deep artificial neural network is responsible for receiving raw data and passing it to the hidden layer. The hidden layer usually contains multiple layers, and the neurons in each hidden layer perform complex feature extraction and transformation on the input data through nonlinear activation functions. Through nonlinear transformation, the deep artificial neural network enables the network to capture complex patterns and relationships in the data. The output layer comprehensively processes the features extracted by the hidden layer to generate the final prediction results or classification labels. The layer structure of the deep artificial neural network sub-model is distributed as follows: first, there is 1 input layer, then 6 sets of identical hidden layers and activation functions, and finally 1 output layer.

[0061] The construction method of a deep artificial neural network includes the following steps: First, prepare and preprocess 20% of the data set and divide it into a training set, a validation set, and a test set. Then, randomly initialize the weights and bias parameters of the network. Next, calculate the output of each layer through forward propagation, and then calculate the error between the network output and the actual result based on the loss function. Calculate the gradient through the back-propagation algorithm, and use the optimization algorithm to update the network parameters. Repeat the above process until the loss function percentage error is less than 10% convergence or reaches the predetermined number of training rounds of 500 rounds, and output the adjustable parameters at the same time.

[0062] The deep artificial neural network sub-model is as follows:

[0063]

[0064]

[0065]

[0066] in:

[0067] x i is the input value, including the inlet water quality monitoring data set, the process water quality monitoring data set, and the outlet water quality monitoring data set;

[0068] y i is the output value, including the predicted value of the aeration rate monitoring data set;

[0069] p i is the true value corresponding to the output value, including the true value corresponding to the predicted value of the aeration monitoring data set

[0070] w i is the weight of each layer;

[0071] b i is the bias of each layer;

[0072] g i is the activation function of each layer;

[0073] m is the number of sample points;

[0074] J(θ) is the loss function;

[0075] θ is the loss;

[0076] α is the learning rate.

[0077] SB3001. Build a long short-term memory neural network sub-model with adjustable parameters, where the adjustable parameter adaptive model includes the number of long short-term memory neural network data input batches, the size of the long short-term memory neural network data input batches, the number of long short-term memory neural network hidden layers, the dimension of the long short-term memory neural network input data, and the time step of the long short-term memory neural network.

[0078] SB3002, using 20% ​​of the water quality condition aeration fusion data set as input variables, training the long short-term memory neural network and obtaining temporary adjustable parameters. This includes inputting the condition aeration fusion data set into the long short-term memory neural network sub-model, outputting adaptive parameters according to the adjustable parameter adaptive model.

[0079] The core structure of the long short-term memory neural network includes a memory cell unit composed of a forget gate, an input gate, and an output gate. Information is transmitted and updated in an orderly manner through the memory cell unit. First, the forget gate determines how much of the previous memory to forget based on the current input and the previous hidden state. Then, the input gate generates new candidate memories based on the current input and the previous hidden state, and determines the update ratio of these new information. Next, the memory cell state is updated with the forgotten old memory and the newly added memory. Finally, the output gate generates a new hidden state based on the current memory cell state and input and outputs it, thereby achieving effective transmission and retention of information between time steps. The main structure of the deep long short-term memory neural network includes 128 memory cell units.

[0080] The construction method of the long short-term memory neural network includes the following steps: First, prepare and preprocess 20% of the data set and divide it into a training set, a validation set, and a test set. Then, randomly initialize the weights and bias parameters of the network. Next, calculate the output of each layer of cells through forward propagation, and then calculate the error between the network output and the actual result according to the loss function. Calculate the gradient through the back propagation algorithm, and use the optimization algorithm to update the network parameters. Repeat the above process until the loss function percentage error is less than 10% convergence or reaches the predetermined number of training rounds of 500 rounds, and output the adjustable parameters at the same time.

[0081] The LSTM neural network sub-model is as follows:

[0082]

[0083] F t =σ(w F x t +w F H t-1 +b F )

[0084] I t =σ(w I x t +w I H t-1 +b I )

[0085]

[0086] C t =F t *C t-1 +I t *tanh(w C x t +w C H t-1 +b C )

[0087] O t =σ(w O x t +w O H t-1 +b O )

[0088] H t =O t *tanh(C t-1 )

[0089] in:

[0090] σ(x)=sigmoid(x) is the sigmoid activation function;

[0091] x t is the input value, including the equipment condition monitoring data set;

[0092] F t It is the Gate of Oblivion;

[0093] w F is the weight matrix of the forget gate, applied to the current input x t ;

[0094] w F H t-1is the weight matrix of the forget gate, applied to the hidden state H at the previous moment t-1 ;

[0095] b F is the bias term of the forget gate;

[0096] I t is the input gate;

[0097] w I is the weight matrix of the input gate, applied to the current input x t ;

[0098] w I H t-1 is the weight matrix of the input gate, applied to the hidden state H at the previous moment t-1 ;

[0099] b I is the bias term of the input gate;

[0100] tanh(x) is the hyperbolic tangent activation function;

[0101] C t , C t-1 It is the memory unit state update;

[0102] w C is the weight matrix of the memory cell state, applied to the current input x t and the hidden state H at the previous moment t-1 ;

[0103] b C is the bias term of the memory cell state;

[0104] O t is the output gate, including the aeration monitoring dataset;

[0105] w O is the weight matrix of the output gate, applied to the current input x t ;

[0106] w O H t-1 is the weight matrix of the output gate, applied to the hidden state H at the previous moment t-1 ;

[0107] b O is the bias term of the output gate;

[0108] H t is a hidden state.

[0109] The results of the two models are fused into an artificial neural network and long short-term memory network fusion model with adjustable parameters, and fused according to the weight coefficient after the percentage error is normalized. Among them, the adjustable parameter adaptive model is embedded in the control system terminal of the sewage treatment aeration equipment. According to the artificial neural network and long short-term memory network fusion model, the sewage treatment aeration process is controlled according to the output results.

[0110] S400, the deep artificial neural network sub-model and the long short-term memory network sub-model are integrated into an artificial neural network and long short-term memory network fusion model, and the weight coefficients after the percentage error is normalized are integrated. The adjustable parameter adaptive model is embedded in the control system terminal of the sewage treatment aeration equipment; the sewage treatment intelligent aeration control system module based on the fusion of neural network and long short-term memory network is as follows Figure 3 shown.

[0111] The mathematical model of the artificial neural network and long short-term memory network fusion model is:

[0112] 1. Calculate MAPE separately

[0113]

[0114]

[0115] 2. Calculate weights

[0116]

[0117]

[0118] 3. Calculate normalized weights

[0119]

[0120]

[0121] 4. Calculate the weighted fusion value

[0122] y i =w 1 y i1 +w 2 y i2

[0123] in:

[0124] x i1 is the input value of the artificial neural network sub-model;

[0125] y i1 is the output value of the artificial neural network sub-model;

[0126] x i1is the input value of the long short-term memory network sub-model;

[0127] y i1 is the output value of the long short-term memory network sub-model;

[0128] p i is the true value corresponding to the input value;

[0129] MAPE(p i ,y i1 ) is the mean absolute percentage error of the ANN submodel;

[0130] MAPE(p i ,y i2 ) is the mean absolute percentage error of the LSTM sub-model;

[0131] is the weight of the artificial neural network sub-model;

[0132] is the weight of the LSTM sub-model;

[0133] w 1 is the normalized weight of the ANN sub-model;

[0134] w 2 is the normalized weight of the LSTM sub-model;

[0135] y i Weighted fusion value.

[0136] S500, inputting the fusion data into the neural network and the long short-term memory network fusion model, and controlling the sewage treatment aeration equipment of the aerobic tank in the biochemical section of the sewage treatment aeration process according to the neural network and the long short-term memory network fusion model according to the adaptive parameters.

[0137] S510, setting the parameters of the neural network and long short-term memory network fusion model as temporary adjustable parameters, inputting 100% of the fusion data into the neural network and long short-term memory network fusion model, and the model outputting the recommended aeration amount of the aeration device;

[0138] S520, the control module controls the sewage treatment aeration equipment of the aerobic tank in the biochemical section of the sewage treatment aeration process according to the aeration amount recommended by the model;

[0139] S530, monitoring the water quality of the water samples collected through the real-time effluent sampling tube according to the data acquisition device, obtaining the effluent water quality, and obtaining a effluent water quality monitoring data set at time T;

[0140] S540, determining whether the water quality index meets the sewage water quality index permitted by the national standard; if the water quality index meets the national standard water quality index, obtaining a sewage discharge instruction to allow discharge.

[0141] The present invention is not limited to the above-mentioned optimal implementation mode. Anyone can derive other various forms of products under the inspiration of the present invention. However, no matter what changes are made in the shape or structure, all technical solutions that are the same or similar to those of the present application fall within the protection scope of the present invention.

Claims

1. A sewage treatment intelligent aeration control method based on a weighted fusion model, characterized in that: The following steps are involved: S100, sampling the water quality indicators of inlet monitoring to obtain an inlet water quality monitoring data set, sampling the water quality indicators of process monitoring to obtain a process water quality monitoring data set, sampling the water quality indicators of effluent monitoring to obtain an effluent water quality monitoring data set; collecting equipment data from the aeration equipment of the aerobic tank in the biochemical section of the sewage treatment to obtain an aeration volume monitoring data set; collecting equipment data from the sewage treatment equipment to obtain an equipment operating condition monitoring data set; S200, merging the inlet water quality monitoring data set, the process water quality monitoring data set, the outlet water quality monitoring data set and the aeration volume monitoring data set into a water quality aeration fusion data set; The equipment condition monitoring data set and the aeration volume monitoring data set are merged into a condition aeration fusion data set; S300, inputting the water quality aeration fusion data set into the deep artificial neural network sub-model with adjustable parameters; inputting the operating condition aeration fusion data set into the long short-term memory neural network sub-model with adjustable parameters; S400, fusing the results of the two models into an artificial neural network and a long short-term memory network fusion model with adjustable parameters, and fusing them according to the weight coefficient after the percentage error is normalized, wherein the adjustable parameter adaptive model is embedded in the control system terminal of the sewage treatment aeration equipment; S500, based on the artificial neural network and long short-term memory network fusion model, the sewage treatment aeration process is controlled according to the output results.

2. According to claim 1, a sewage treatment intelligent aeration control method based on a weighted fusion model is characterized by: In the S100, the inlet water quality monitoring data set, the process water quality monitoring data set and the effluent water quality monitoring data set are data sets of the historical period of the target sewage treatment plant; the data type in the aeration volume monitoring data set includes the aeration volume of the sewage treatment equipment in the aerobic tank of the process section; the data type in the equipment condition monitoring data set includes sewage anaerobic hydraulic retention time data, anoxic hydraulic retention time data, and aerobic hydraulic retention time data.

3. According to claim 1, a sewage treatment intelligent aeration control method based on a weighted fusion model is characterized by: In the above-mentioned S200, the method for merging the data sets is to align the data rows and data columns of the data sets, and set the data interval to the minimum sampling time.

4. A sewage treatment intelligent aeration control method based on a weighted fusion model according to claim 1, characterized in that: In the S300, the method for constructing the deep artificial neural network is as follows: first, prepare and preprocess 20% of the data set and divide it into a training set, a validation set and a test set; then, randomly initialize the weights and bias parameters of the network; then, calculate the output of each layer by forward propagation, and then calculate the error between the network output and the actual result according to the loss function; calculate the gradient by the back propagation algorithm, and use the optimization algorithm to update the network parameters; The above process is repeated until the loss function percentage error is less than 10% and converges or the predetermined number of training rounds, 500, is reached, and the adjustable parameters are output at the same time.

5. A sewage treatment intelligent aeration control method based on a weighted fusion model according to claim 1, characterized in that: In the S300, the method for constructing the long short-term memory neural network is as follows: first, prepare and preprocess 20% of the data set and divide it into a training set, a validation set and a test set; then, randomly initialize the weights and bias parameters of the network; then, calculate the output of each layer of cell units through forward propagation, and then calculate the error between the network output and the actual result according to the loss function; calculate the gradient through the back propagation algorithm, and use the optimization algorithm to update the network parameters; repeat the above process until the loss function percentage error is less than 10% and converges or reaches a predetermined number of training rounds of 500, and output the adjustable parameters at the same time.

6. A sewage treatment intelligent aeration control method based on a weighted fusion model according to claim 1, characterized in that: The S500 includes the following steps: S510, setting the parameters of the neural network and long short-term memory network fusion model as temporary adjustable parameters, inputting 100% of the fusion data into the neural network and long short-term memory network fusion model, and the model outputting the recommended aeration amount of the aeration device; S520, the control module controls the sewage treatment aeration equipment of the aerobic tank in the biochemical section of the sewage treatment aeration process according to the aeration amount recommended by the model; S530, monitoring the water quality of the water samples collected through the real-time effluent sampling tube, obtaining the effluent water quality, and obtaining a effluent water quality monitoring data set at time T; S540: Determine whether the water quality index meets the standard. If the water quality index meets the standard, obtain a sewage discharge instruction to allow discharge.

Citation Information

Patent Citations

  • Sewage treatment aeration intelligent control method and system based on deep learning

    CN112782975A

  • Water quality prediction method, parameter recommendation and optimization method and model training method

    CN115796050A

  • Method and device for predicting hydrogen sulfide in urban sewage pipeline based on WATS-LSTM coupling model

    CN116306299A

  • Intelligent aeration method and system for sewage treatment based on artificial intelligence

    CN116956156A

  • Cultivation tail water prediction processing method based on LSTM neural network model

    CN117875506A

Cited By

  • Method for realizing high-efficiency low-consumption micro-aerobic hydrolytic acidification of petrochemical wastewater by regulating and controlling aeration rate in combination with space-time diagram neural network and reinforced learning

    CN119430460A

  • Method for regulating aeration volume by combining spatio-temporal graph neural network and reinforcement learning to achieve efficient and low-consumption micro-aerobic hydrolysis acidification of petrochemical wastewater

    CN119430460B

  • Sewage treatment aeration optimization control system and method

    CN119689964A