Nitrification liquid backflow control system of integrated sewage treatment equipment

Abstract

The invention relates to the technical field of sewage treatment, and discloses an integrated sewage treatment nitrification liquid backflow control system. The system comprises a human-computer interaction interface, a water quality detection sensor acquisition module, a total nitrogen prediction module, a reflux pump control module and a data storage module which are connected through 485 communication interfaces. According to the method, a BP + LSTM mixed neural network is adopted, water quality detection parameters including the pH value, the chemical oxygen demand (COD), the biological oxygen demand (BOD), the sludge concentration (SS) and the dissolved oxygen (DO) are combined, the total nitrogen solubility in an anoxic pond and an aerobic pond in the integrated sewage treatment equipment is predicted, the nitrification liquid backflow proportion is calculated, and therefore operation of a backflow pump is accurately controlled. The system can adjust the backflow ratio of the nitrification liquid in real time, solves the problem that the existing equipment cannot adjust the backflow ratio in real time according to the total nitrogen content, and avoids the phenomenon of poor denitrification effect caused by low backflow ratio or energy waste caused by high backflow ratio.

Description

Technical Field

[0001] The present invention relates to the technical field of sewage treatment, and specifically to a nitrification liquid reflux control system for an integrated sewage treatment device. Background Art

[0002] With the continuous development of society and the improvement of living standards, the discharge of rural domestic sewage has been increasing year by year. As a relatively stable fresh water resource, if rural sewage can be effectively treated by a perfect integrated sewage treatment device, not only can the recycling of fresh water be realized, the problem of water resource shortage be alleviated, but also the demand for natural water bodies can be reduced, and the risk of water environmental pollution can be effectively reduced.

[0003] At present, the electrical control systems of traditional integrated sewage treatment devices mostly rely on pure hardware circuits or simple PLC controls. However, when nitrite nitrogen generated by the nitrification reaction in the aerobic tank is refluxed to the anoxic tank (denitrification tank) for denitrification reaction, the reflux ratio of the nitrification liquid is usually set depending on the experience of the operator. An excessively low internal reflux ratio will lead to a decrease in denitrification efficiency and cause the total nitrogen (TN) concentration in the effluent to exceed the standard; while an excessively high internal reflux ratio, although helpful for nitrogen removal, will carry too much dissolved oxygen (DO), destroy the anoxic environment, consume additional carbon sources, and significantly increase the energy consumption cost. When the internal reflux ratio exceeds 600%, the further increase in the reflux ratio has limited effect on improving the denitrification efficiency, with low cost performance and resulting in waste of resources.

[0004] The precise control of the reflux liquid ratio can be achieved by real-time monitoring of the total nitrogen concentration. However, the current measurement of total nitrogen mainly relies on the chemical reagent method. This method not only has a long measurement period, consumes a large amount of manpower and material resources, but also lacks timeliness, and cannot adjust the internal reflux ratio of the nitrification liquid in a timely manner, resulting in poor sewage treatment effect and increased energy consumption. Therefore, in view of these limitations of the prior art, there is an urgent need to develop a system that can predict the total nitrogen concentration in real time and automatically adjust the reflux ratio to improve the sewage treatment efficiency and economic benefits. Summary of the Invention

[0005] In view of the above problems, the present invention proposes a nitrification liquid reflux control system for an integrated sewage treatment device. The present invention designs a complete total nitrogen prediction system and nitrification liquid reflux control system, which can predict the total nitrogen concentration in sewage in real time, and adjust the internal reflux ratio of the nitrification liquid by controlling the reflux pump in the aerobic tank to optimize the sewage treatment effect, while reducing energy consumption and achieving efficient and energy-saving sewage treatment.

[0006] To achieve the above object, the present invention adopts the following technical solutions: An integrated sewage treatment equipment nitrification liquid reflux control system proposed by the present invention includes: a human-computer interaction interface, a water quality detection sensor acquisition module, a total nitrogen prediction module, a reflux pump control module, and a data storage module. Among them, the human-computer interaction interface is respectively connected to the water quality detection sensor acquisition module, the total nitrogen prediction module, and the reflux pump control module; the water quality detection sensor acquisition module is connected to the total nitrogen prediction module; the total nitrogen prediction module is connected to the reflux pump control module; the data storage module is connected to the water quality detection sensor acquisition module, the total nitrogen prediction module, and the reflux pump control module, and all modules achieve data interaction through a 485 communication interface.

[0007] Preferably, the total nitrogen prediction module: adopts a hybrid model combining LSTM and BP neural network, and predicts the total nitrogen concentration in the anoxic tank and aerobic tank of the sewage treatment equipment by inputting water quality parameters (such as pH value, COD, BOD, SS, DO, etc.). This model has strong timeliness and accuracy, and can reflect the water quality changes in real time and dynamically adjust the reflux ratio.

[0008] Preferably, the training process of the total nitrogen prediction module prediction model includes the following steps: (1) Receive the water quality parameter data provided by the water quality detection sensor acquisition module, and perform normalization processing to eliminate the influence of different dimensions on the prediction result. Among them, the normalization processing method is as follows: ; is the original data, are the maximum and minimum values of the original data set respectively; (2) Initialize the LSTM neural network and the BP neural network. According to the set input sequence and output sequence , determine the number of input nodes n of the neural network, the number of output nodes m , and the number of hidden layer nodes , initialize the weights of the network, the connection weights between the hidden layer and the output layer, and initially set the learning rate η and neuron excitation. Initialize the weights of the network, the connection weights between the hidden layer and the output layer, and initially set the learning rate η and neuron excitation; (3) Forward propagation of the LSTM network, process data through time step sequence information, and the output is the predicted value of the total nitrogen concentration. The BP neural network is responsible for capturing the non-linear relationship of the input data and calculating the predicted value through the connection between the hidden layer and the output layer; (4) Combine the prediction results of the LSTM and BP networks, and use the weighted average ensemble learning method to obtain the final predicted value, and calculate the predicted output ; (5) Calculate the output error of the model, and obtain the prediction error of the neural network according to the output and the expected output sequence Y: : ; (6) Update the weights. Based on the calculated model output error update the connection weights of the LSTM and BP neural networks and , and use the ensemble learning method to optimize the predictions of both: ; ; where and represent the activation values in the LSTM and BP neural networks, and represent the weight matrices of the LSTM and BP neural networks; (7) Update the thresholds. Update the thresholds of the LSTM and BP neural network nodes according to the prediction error value and ; ; (8) According to the difference between the predicted output and the expected value, determine whether the preset convergence condition is satisfied. If the error does not reach the set threshold, return to step (3) to continue training and adjustment until the stop condition is met.

[0009] Preferably, according to the total nitrogen prediction value, the reflux ratio of the nitrification liquid is adjusted in real time. Through the feedback control mechanism, the internal reflux ratio is automatically adjusted to avoid the negative impact of too low or too high reflux ratio on the denitrification efficiency. Especially when the reflux ratio exceeds 600%, the control system can automatically optimize the reflux ratio to avoid waste of resources and ensure the efficient operation of the system. The formula for the denitrification efficiency is: ; where R is the external reflux ratio, r is the internal reflux ratio. Among them, the external reflux ratio is controlled relatively small (30 - 50%), and the denitrification efficiency reflux formula can be simplified to: ; Rearrange the above formula and convert it to: ; The formula for the nitrogen removal efficiency is: ; Preferably, the water quality detection sensor acquisition module adopts a fully closed-loop processing solution and is respectively arranged in the anoxic tank and the aerobic tank. The water quality detection sensor acquisition module includes a liquid level gauge, a pH sensor, a suspended solid concentration sensor, a dissolved oxygen sensor, a chemical oxygen demand sensor, and a biochemical oxygen demand sensor.

[0010] Preferably, the human-machine interface has real-time monitoring, prediction, and control functions, can display the operating status, monitoring data, prediction values, alarm status, and operating efficiency information of the current sewage treatment equipment. The interface supports remote access and control functions, can display data in the form of charts, curves, etc., and has the training and optimization functions of the prediction model. The model is optimized according to user feedback to improve the prediction accuracy and precision.

[0011] The present invention also proposes a control method for the nitrification liquid reflux control system of the integrated sewage treatment equipment, including the following steps: (1) Collect the pH value, COD, BOD, SS, and DO parameters in the sewage treatment system and record them through the water quality detection sensor acquisition module; (2) Preprocess the collected parameter data and standardize the data using the normalization method to eliminate the influence of dimensions in subsequent processing and ensure the consistency and accuracy of the model input; (3) Determine the topological structure of the neural network, combine the LSTM and BP neural networks, set the structures of the input layer, hidden layer, output layer, and long short-term memory unit (LSTM layer), and capture the dynamic changes in the sewage treatment process by using time series features; (4) According to the collected data, train the hybrid neural network using the training set to improve the accuracy of the model in processing time series data, and at the same time improve the overall prediction accuracy by means of the feedback mechanism of the BP neural network; (5) Use the trained model to predict the total nitrogen, input the newly collected water quality parameters into the model for prediction, and perform anti-normalization processing on the prediction results to obtain the actual total nitrogen concentration prediction value; (6) Transmit the predicted total nitrogen concentration value to the reflux pump control module to control the output flow of the reflux pump and adjust the nitrification liquid reflux ratio to optimize the nitrogen removal efficiency in the sewage treatment process; at the same time, the prediction results will be stored in the data storage module to provide data basis for subsequent analysis, model optimization, and decision support.

[0012] Compared with the prior art, the present invention has the following beneficial effects: The nitrification liquid reflux control system of the integrated sewage treatment equipment proposed by the present invention solves the problems of lag in real-time measurement of total nitrogen (TN) concentration and adjustment of reflux ratio in the prior art by accurately predicting the total nitrogen concentration in sewage and introducing a hybrid prediction model of LSTM and BP neural network. This system can automatically adjust the reflux ratio, avoiding resource waste and inefficient operation caused by artificially setting the reflux ratio in traditional methods, thus significantly improving the treatment efficiency, nitrogen removal effect and economic benefits of sewage treatment equipment. Compared with the traditional chemical reagent method, the present invention has higher real-time performance and accuracy, can reduce manual intervention, and lower operation costs and energy consumption. In addition, the automated and intelligent control of the system makes the sewage treatment process more precise and flexible, meeting the requirements of the development of modern sewage treatment equipment towards high efficiency, intelligence and energy conservation. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] Figure 1 It is a schematic diagram of the overall system of the present invention; Figure 2 It is a schematic diagram of the overall integrated sewage treatment equipment; Figure 3 It is a detailed schematic diagram of the system of the present invention; Figure 4 It is a flowchart of the algorithm. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0014] The present invention will be described below with reference to the accompanying drawings and embodiments.

[0015] The nitrification reflux liquid control system of the integrated sewage treatment equipment of the integrated sewage treatment equipment is implemented in the anoxic tank and aerobic tank of the equipment. Based on the AAO process, the entire process flow is as Figure 1 and Figure 2 shown. The influent water to be treated is the rural domestic sewage pretreated by grille filtration and regulating tank. The influent water of the anoxic tank comes from the sewage treated by the anaerobic tank, while the influent water of the aerobic tank is the sewage treated by the anoxic tank.

[0016] The domestic sewage treated by grille, regulating tank and anaerobic treatment flows into the anoxic tank. Under anoxic environment (dissolved oxygen < 0.5), denitrifying bacteria reduce nitrite and nitrate to nitrogen gas, which escapes from the water body to complete the nitrogen removal process.

[0017] In the aerobic tank, the sewage from the anoxic tank undergoes nitrification reaction under aerobic conditions, and ammonia nitrogen is oxidized by nitrifying bacteria to nitrite and nitrate. The nitrification reaction includes two steps: ammonia nitrogen is converted to nitrite by nitrite bacteria, and then nitrite is converted to nitrate by nitrate bacteria. Both nitrite bacteria and nitrate bacteria are chemolithoautotrophic bacteria, using CO2, CO3 2- , HCO3 - as carbon sources, and through NH3, NH4 +, NO2 - Obtain energy through the redox reaction.

[0018] Combining nitrification and denitrification reactions achieves the nitrogen removal goal.

[0019] Such as Figure 1 Figure

[0020] The human - machine interface S101 is connected to other modules through a 485 communication interface to achieve real - time monitoring, prediction, and control. The interface displays the device operation status, monitoring data, prediction values, and alarm information, and supports remote monitoring and management, enhancing the intelligent and automated operation capabilities of the device. The interface also has the function of training and optimizing the prediction model to improve the prediction accuracy.

[0021] The water quality detection sensor acquisition module S102 is responsible for monitoring water quality parameters, including level gauges, pH values, suspended solid concentration SS, chemical oxygen demand COD, and biological oxygen demand BOD. The anoxic tank Z201 and the aerobic tank Z202 are respectively equipped with corresponding sensors. At the same time, a level gauge is also set in the anoxic tank, which can prevent sewage overflow caused by excessive reflux. The specific information and connection methods of the sensors are as Figure 3 shown.

[0022] The data collected by the water quality detection sensor acquisition module S102 is pre - processed and then imported into the total nitrogen prediction module S103, and the original data is stored for maintenance reference.

[0023] To establish a total nitrogen prediction model for sewage treatment, the present invention adopts a hybrid neural network model that combines the LSTM long - short - term memory neural network and the BP neural network, and constructs the model based on 300 sets of historical sewage water quality parameter data collected. After the model is connected to the system, in addition to recording historical sewage water quality parameters, new collected data will also be input into the prediction model for update and iteration, thereby continuously improving the prediction accuracy of the model. The update period of the model is set to every 2 hours.

[0024] Such as Figure 4 shown, the input of the total nitrogen prediction model S103 includes water quality parameters after normalization, such as pH value, COD, BOD, SS, and DO, and the output is the predicted total nitrogen concentration. Its processing mainly includes the following processes: (1) Receive the water quality parameter data provided by the water quality detection sensor acquisition module S102, and perform normalization processing to eliminate the influence of different dimensions on the prediction results. The normalization processing method is as follows: ; is the original data, are the maximum and minimum values of the original data set respectively; (2) Initialize the LSTM neural network and the BP neural network. According to the set input sequence and output sequence , determine the number of input nodes n of the neural network, the number of output nodes m , and the number of hidden layer nodes . Initialize the weights of the network, the connection weights between the hidden layer and the output layer, and initially set the learning rate η and neuron excitation; (3) Forward propagation of the LSTM network, process data through the time step sequence information, and the output is the predicted value of the total nitrogen concentration. The BP neural network is responsible for capturing the non-linear relationship of the input data and calculating the predicted value through the connection between the hidden layer and the output layer; (4) Combine the prediction results of the LSTM and BP networks, and use the weighted average ensemble learning method to obtain the final predicted value, and calculate the predicted output ] of the hybrid model; (5) Calculate the output error of the model. According to the output and the expected output sequence Y, obtain the prediction error of the neural network: ; (6) Weight update, update the connection weights of the LSTM and BP neural networks based on the calculated model output error and , and use the ensemble learning method to optimize the predictions of both: ; ; Among them, and represent the activation values in the LSTM and BP neural networks, and represent the weight matrices of the LSTM and BP neural networks; (7) Update the threshold, update the thresholds of the LSTM and BP neural network nodes according to the prediction error value and ; (8) Based on the difference between the predicted output and the expected value, determine whether the preset convergence condition is satisfied. If the error does not reach the set threshold, return to step (3) to continue training and adjustment until the stop condition is met.

[0025] The predicted total nitrogen concentrations in the anoxic tank and the aerobic tank are used to calculate the nitrified liquid reflux ratio. The data is input into the reflux pump control module S104, and the corresponding nitrified liquid reflux ratio is calculated to control the reflux of the nitrified liquid to the anoxic tank.

[0026] The reflux pump control module S104 controls the reflux of the nitrified liquid. The nitrified liquid reflux ratio is calculated in real time and the nitrification and denitrification efficiencies are optimized according to the settings of the external reflux ratio and the internal reflux ratio to improve the nitrogen removal effect. The formula for the denitrification efficiency is: ; where R is the external reflux ratio, r is the internal reflux ratio. Among them, the external reflux ratio is controlled relatively small (30 - 50%), and the denitrification efficiency reflux formula can be simplified to: ; After arranging the above formula, it is converted to: ; The formula for the nitrogen removal efficiency is: ; Then, the flow rate of the reflux is precisely controlled through the flow meter Z204 to ensure that the nitrified liquid reflux ratio is accurately controlled.

[0027] The present invention also includes a data storage module S105 for storing the water quality parameters detected by the water quality sensor, the total nitrogen prediction information, and the status information of the pump. This module is provided with a data export interface to facilitate the subsequent maintenance personnel to extract the original data and provide the reference support required for maintenance.

[0028] Traditional sewage treatment processes rely on open-loop or semi-open-loop processes and require on-site professional personnel for debugging. The treatment effect is easily affected by factors such as the environment and equipment aging. The present invention adopts a full-closed-loop detection and treatment solution, dynamically updates the total nitrogen prediction model based on the water quality parameters at each stage, and improves the prediction accuracy. By accurately predicting the total nitrogen concentrations in the anoxic tank and the aerobic tank, optimizing the nitrified liquid reflux ratio, the sewage treatment effect is significantly improved, energy is saved, and the deficiencies in the traditional process are overcome.

[0029] The above embodiments are only used to illustrate the specific implementation manners of the present invention. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present invention, several deformations and changes can still be made, and these deformations and changes should all fall within the protection scope of the present invention.

Claims

1. An integrated sewage treatment equipment nitrification liquid reflux control system, characterized in that: Including: Human-computer interaction interface (S101); Water quality detection sensor acquisition module (S102); Total nitrogen prediction module (S103); Return pump control module (S104); Data storage module (S105); The human-computer interaction interface (S101) is respectively connected to the water quality detection sensor acquisition module (S102), the total nitrogen prediction module (S103) and the return pump control module (S104); the water quality detection sensor acquisition module (S102) is connected to the total nitrogen prediction module (S103); the total nitrogen prediction module (S103) is connected to the return pump control module (S104); the data storage module (S105) is connected to the water quality detection sensor acquisition module (S102), the total nitrogen prediction module (S103) and the return pump control module (S104), and all modules achieve data interaction through the 485 communication interface.

2. The nitrification liquid reflux control system of an integrated sewage treatment device according to claim 1, wherein: The total nitrogen prediction module (S103) adopts a hybrid neural network model, combining the LSTM long short-term memory neural network and the traditional BP neural network, and is used to predict the total nitrogen concentration. The input is the pH value, COD, BOD, SS, and DO water quality parameters after normalization processing, and the output is the predicted total nitrogen content. This hybrid model effectively combines the time series processing ability of LSTM and the non-linear feature learning ability of the BP neural network through the ensemble learning method, enhancing the accuracy and stability of the prediction.

3. An integrated sewage treatment equipment nitrification liquid reflux control system according to claim 2, characterized in that: The total nitrogen prediction module (S103) adopts a hybrid neural network model, combining the LSTM model and the BP neural network. The training process of this model includes the following steps: (1) Receive the water quality parameter data provided by the water quality detection sensor acquisition module (S102), and perform normalization processing to eliminate the influence of different dimensions on the prediction result. Among them, the normalization processing method is as follows: ； is the original data, are the maximum and minimum values of the original data set, respectively; (2) Initialize the LSTM neural network and the BP neural network. According to the set input sequence and output sequence , determine the number of input nodes of the neural network n , the number of output nodes m , and the number of hidden layer nodes . Initialize the weights of the network, the connection weights between the hidden layer and the output layer, and initially set the learning rate η and neuron excitation; (3) Forward propagation of the LSTM network, process data through time step sequence information, and the output is the predicted value of the total nitrogen concentration. The BP neural network is responsible for capturing the non-linear relationship of the input data and calculating the predicted value through the connection between the hidden layer and the output layer; (4) Combining the prediction results of the LSTM and BP networks, an ensemble learning method of weighted average is adopted to obtain the final predicted value, and the predicted output of the hybrid model is calculated. ; (5) Calculate the output error of the model, and obtain the prediction error of the neural network based on the output and the expected output sequence Y : ； (6) Weight update, based on the calculated model output error Update the connection weights of the LSTM and BP neural networks And , and use the ensemble learning method to optimize the predictions of both: ； ； Among them, and represent the activation values in the LSTM and BP neural networks, and represent the weight matrices of the LSTM and BP neural networks; (7) Update the threshold value to update the threshold values of the LSTM and BP neural network nodes according to the prediction error value ; and ; (8) According to the difference between the predicted output and the expected value, judge whether the preset convergence condition is satisfied. If the error does not reach the set threshold, return to step (3) to continue training and adjustment until the stop condition is met.

4. The nitrification liquid reflux control system of an integrated sewage treatment device according to claim 2, characterized in that: The total nitrogen prediction module (S103) is trained based on the historical operation data of the integrated sewage treatment equipment, and a hybrid model of LSTM and BP neural network. The input of this model is the historical pH value, COD, BOD, SS, and DO water quality parameters, and the output is the corresponding historical total nitrogen concentration. The hybrid model can simultaneously consider the time series characteristics and non-linear relationships of the data, effectively improving the prediction accuracy.

5. The nitrification liquid reflux control system of an integrated sewage treatment device according to claim 1, characterized in that: The return pump control module (S104) controls the return of nitrification liquid. The return ratio of nitrification liquid is calculated in real time, and according to the settings of the external return ratio and the internal return ratio, the nitrification and denitrification efficiencies are optimized to improve the nitrogen removal effect. The formula for the denitrification efficiency is: ； Among them R is the external reflux ratio r is the internal reflux ratio. Among them, the external reflux ratio is controlled relatively small at 30 - 50%, and the denitrification efficiency reflux formula can be simplified as: ； After organizing the above formula, it is converted to: ； The formula for the nitrogen removal efficiency is: ； is the total nitrogen concentration in the anoxic tank, is the total nitrogen concentration in the aerobic tank, is the total nitrogen removal efficiency of the equipment.

6. The nitrification liquid reflux control system of an integrated sewage treatment device according to claim 1, characterized in that: The water quality detection sensor acquisition module (S102) adopts a full-closed-loop processing solution and is respectively arranged in the anoxic tank and the aerobic tank. The water quality detection sensor acquisition module (S102) includes a liquid level gauge, a pH sensor, a suspended solid concentration sensor, a dissolved oxygen sensor, a chemical oxygen demand sensor, and a biochemical oxygen demand sensor.

7. An integrated sewage treatment equipment nitrification liquid reflux control system according to claim 1, characterized in that: The human-computer interaction interface (S101) has real-time monitoring, prediction, and control functions. It can display the operating status, monitoring data, prediction values, alarm status, and operating efficiency information of the current sewage treatment equipment. The interface supports remote access and control functions, can display data in the form of charts, curves, etc., and has the functions of training and optimizing the prediction model. The model is optimized according to user feedback to improve the prediction accuracy and precision.

8. The nitrification liquid reflux control system of the integrated sewage treatment equipment according to claim 1, characterized in that: It includes the following steps: (1) Collect the pH value, COD, BOD, SS, and DO parameters in the sewage treatment system and record them through the water quality detection sensor acquisition module (S102); (2) Preprocess the collected parameter data and standardize the data using the normalization method to eliminate the influence of dimensions in subsequent processing and ensure the consistency and accuracy of the model input; (3) Determine the topological structure of the neural network, combine the LSTM and BP neural networks, set the structures of the input layer, hidden layer, output layer, and long short-term memory unit, and capture the dynamic changes in the sewage treatment process by using the time series characteristics; (4) According to the collected data, train the hybrid neural network using the training set to improve the accuracy of the model in processing time series data, and at the same time improve the overall prediction accuracy by means of the feedback mechanism of the BP neural network; (5) Use the trained model to predict the total nitrogen. Input the newly collected water quality parameters into the model for prediction, and perform anti-normalization processing on the prediction result to obtain the actual total nitrogen concentration prediction value; (6) Transmit the predicted total nitrogen concentration value to the reflux pump control module (S104) to control the output flow of the reflux pump and adjust the nitrification liquid reflux ratio to optimize the nitrogen removal efficiency in the sewage treatment process; at the same time, the prediction result will be stored in the data storage module (S105) to provide a data basis for subsequent analysis, model optimization, and decision support.

Images