A micro-nano air floatation phosphorus removal real-time control method and device based on a neural network model, equipment and medium

Abstract

This application discloses a real-time control method, device, equipment, and medium for micro-nano air flotation phosphorus removal based on a neural network model, relating to the field of computer technology. It includes: real-time acquisition of operational data through a preset sensor network, and preprocessing the operational data to obtain target data; generating a recommended PAC dosage value based on influent phosphorus concentration, water turbidity, and historical phosphorus removal rate using a first preset neural network model, and dynamically correcting the recommended PAC dosage value based on potential data using a second preset neural network model to obtain the target PAC dosage; optimizing the dissolved air pressure based on the target PAC dosage, influent phosphorus concentration, water turbidity, historical phosphorus removal rate, and post-coagulation parameters using a third preset neural network model to achieve real-time control of micro-nano air flotation phosphorus removal. This enables real-time, adaptive, and intelligent optimization and control of operating parameters during the micro-nano air flotation phosphorus removal process.

Description

Technical Field

[0001] This invention relates to the field of computer technology, and in particular to a real-time control method, device, equipment and medium for micro-nano air flotation phosphorus removal based on a neural network model. Background Technology

[0002] With the acceleration of industrialization and urbanization, eutrophication of water bodies has become increasingly prominent, and phosphorus emission standards are constantly tightening. As relevant environmental regulations are strictly enforced, phosphorus emission limits for wastewater treatment plant effluents are becoming increasingly stringent. Against this backdrop, the efficiency and intelligence of phosphorus removal processes have become important development directions for the wastewater treatment industry. Micro-nano air flotation technology, as a highly efficient solid-liquid separation method, can achieve efficient capture and removal of suspended particles and flocs by generating microbubbles. When combined with coagulation and chemical dosing processes, this technology can significantly improve phosphorus removal efficiency. However, traditional coagulation-air flotation phosphorus removal control methods still have the following shortcomings:

[0003] 1. Control lag: Relying on offline chemical detection of total phosphorus concentration in effluent, the lag time is as long as several hours, making real-time adjustment difficult.

[0004] II. Fixed process operating parameters: Operating parameters usually rely on manual setting, lack dynamic adjustment capability, and have poor adaptability to fluctuations in influent water quality.

[0005] 3. Dependence on human experience: The operation relies on experienced personnel, and differences exist between different operators, leading to unstable results.

[0006] Fourth, low energy efficiency: Excessive addition of PAC (Poly Aluminum Chloride) or maintenance of excessively high dissolved air pressure not only wastes the reagents but also causes excessive energy consumption, resulting in an increase in the total cost of wastewater treatment.

[0007] Fifth, lack of intelligent decision-making: Most existing systems adjust operation based on a single water quality parameter and lack the ability to integrate and analyze multiple parameters, resulting in a lack of overall consideration in the adjustment of operation.

[0008] As can be seen from the above, how to achieve real-time, adaptive, and intelligent optimization and control of operating parameters in the micro-nano air flotation phosphorus removal process is an urgent problem to be solved. Summary of the Invention

[0009] In view of this, the purpose of this invention is to provide a real-time control method, device, equipment, and medium for micro-nano air flotation phosphorus removal based on a neural network model, which can achieve real-time, adaptive, and intelligent optimization and control of operating parameters during the micro-nano air flotation phosphorus removal process. The specific solution is as follows:

[0010] In a first aspect, this application provides a real-time control method for micro / nano air flotation phosphorus removal based on a neural network model, including:

[0011] Operational data is collected in real time through a preset sensor network, and the operational data is preprocessed to obtain target data; the target data includes influent phosphorus concentration, water turbidity, and potential data;

[0012] A first preset neural network model is used to generate a recommended PAC dosage based on the influent phosphorus concentration, the water turbidity, and the historical phosphorus removal rate. A second preset neural network model is used to dynamically correct the recommended PAC dosage based on the potential data to obtain the target PAC dosage. The first preset neural network model adopts a multi-layer feedforward neural network structure. The second preset neural network model adopts a hybrid architecture combining bidirectional LSTM and fully connected neural networks.

[0013] The dissolved air pressure is optimized using a third preset neural network model based on the target PAC dosage, the influent phosphorus concentration, the water turbidity, the historical phosphorus removal rate, and the parameters after coagulation, so as to achieve real-time control of micro-nano air flotation phosphorus removal; the parameters after coagulation include the water turbidity after coagulation and the potential data after coagulation.

[0014] Optionally, before collecting operational data in real time through a preset sensor network, the method further includes:

[0015] The first preset neural network model, the second preset neural network model, and the third preset neural network model are trained using historical data;

[0016] The sensors in the preset sensor network are initialized, detected, and calibrated; the sensors include a turbidity sensor, a zeta potential meter, and an online total phosphorus analyzer.

[0017] The control actuator is initially configured and calibrated, and initial operating parameters and safe operating boundary conditions are set; the control actuator includes a PAC dosing pump and a dissolved gas pressure control valve.

[0018] Optionally, the step of collecting operational data in real time through a preset sensor network and preprocessing the operational data to obtain target data includes:

[0019] The raw operating data are obtained by real-time collection of water turbidity, potential data and phosphorus concentration in influent and effluent during the coagulation reaction process through sensors in the preset sensor network.

[0020] The original running data is subjected to outlier identification and removal operations, and the original running data after outlier removal is compensated using a preset interpolation algorithm to obtain the compensated running values.

[0021] The compensated running values ​​are sequentially smoothed and normalized, and features are extracted from the processed data to obtain target data containing multi-dimensional feature vectors.

[0022] Optionally, the real-time control method for micro / nano air flotation phosphorus removal based on a neural network model further includes:

[0023] Using historical influent phosphorus concentration, historical average turbidity of water body, historical phosphorus removal rate and corresponding target PAC dosage labels, the first preset neural network model is trained by mean square error loss function and Adam optimizer.

[0024] The first preset neural network model adopts a multi-layer feedforward neural network architecture including an input layer, multiple hidden layers and an output layer; the hidden layers are provided with 64, 32 and 16 neuron nodes respectively, and each hidden layer is configured with a batch normalization layer and a dropout layer respectively.

[0025] Optionally, the real-time control method for micro / nano air flotation phosphorus removal based on a neural network model further includes:

[0026] The second preset neural network model is trained using historical potential data, historical PAC dosage, historical influent phosphorus concentration, historical potential change trends, time interval parameters, and corresponding adjustment coefficient labels; the adjustment coefficient labels are labels generated based on the historical removal rate change direction.

[0027] The second preset neural network model adopts a hybrid architecture combining a bidirectional long short-term memory network and a fully connected neural network; the bidirectional long short-term memory network is used to capture the temporal dependency features of the input parameters; the fully connected neural network is used to output adjustment coefficients.

[0028] Accordingly, the step of dynamically correcting the recommended PAC dosage based on the potential data using a second preset neural network model to obtain the target PAC dosage includes:

[0029] The temporal dependence features of the potential data are obtained by using the bidirectional long short-term memory network in the second preset neural network model, and the target adjustment coefficient for the recommended value of the PAC dosage is output by using the fully connected neural network in the second preset neural network model based on the temporal dependence features of the potential data. The recommended value of the PAC dosage is dynamically corrected based on the target adjustment coefficient to obtain the target PAC dosage.

[0030] Optionally, the real-time control method for micro / nano air flotation phosphorus removal based on a neural network model further includes:

[0031] The third preset neural network model was trained using historical influent phosphorus concentration, historical water turbidity, historical PAC dosage, historical post-coagulation water turbidity, historical post-coagulation potential data, historical phosphorus removal rate, and corresponding target dissolved air pressure labels.

[0032] The third preset neural network model adopts a preset deep neural network architecture, which includes an input layer, multiple hidden layers and an output layer; the hidden layers are provided with 64, 32 and 16 neuron nodes respectively, and each hidden layer is configured with a batch normalization layer and a dropout layer respectively.

[0033] Optionally, the real-time control method for micro / nano air flotation phosphorus removal based on a neural network model further includes:

[0034] An experience database is established, and the data output by the first preset neural network model, the second preset neural network model, and the third preset neural network model are stored in the experience database.

[0035] Based on the data in the experience database and at a preset update interval, the first preset neural network model, the second preset neural network model, and the third preset neural network model are collaboratively optimized and their parameters are iteratively updated.

[0036] Secondly, this application provides a real-time control device for micro-nano air flotation phosphorus removal based on a neural network model, comprising:

[0037] The data acquisition module is used to collect operational data in real time through a preset sensor network and preprocess the operational data to obtain target data; the target data includes influent phosphorus concentration, water turbidity, and potential data;

[0038] The first parameter control module is used to generate a recommended PAC dosage value based on the influent phosphorus concentration, the water turbidity, and historical phosphorus removal rate using a first preset neural network model, and to dynamically correct the recommended PAC dosage value based on the potential data using a second preset neural network model to obtain the target PAC dosage; the first preset neural network model adopts a multi-layer feedforward neural network structure; the second preset neural network model adopts a hybrid architecture combining bidirectional LSTM and fully connected neural networks;

[0039] The second parameter control module is used to optimize the dissolved air pressure based on the target PAC dosage, the influent phosphorus concentration, the water turbidity, the historical phosphorus removal rate, and the parameters after coagulation using a third preset neural network model, so as to achieve real-time control of micro-nano air flotation phosphorus removal; the parameters after coagulation include the water turbidity after coagulation and the potential data after coagulation.

[0040] Thirdly, this application provides an electronic device, comprising:

[0041] Memory, used to store computer programs;

[0042] A processor is used to execute the computer program to implement the aforementioned real-time control method for micro-nano air flotation phosphorus removal based on a neural network model.

[0043] Fourthly, this application provides a computer-readable storage medium for storing a computer program, wherein the computer program, when executed by a processor, implements the aforementioned real-time control method for micro-nano air flotation phosphorus removal based on a neural network model.

[0044] This application provides a real-time control method for micro-nano air flotation phosphorus removal based on a neural network model. The method involves real-time acquisition of operational data through a pre-set sensor network, followed by preprocessing of the data to obtain target data. The target data includes influent phosphorus concentration, water turbidity, and potential data. A first pre-set neural network model generates a recommended PAC dosage based on the influent phosphorus concentration, water turbidity, and historical phosphorus removal rate. A second pre-set neural network model dynamically corrects the recommended PAC dosage based on the potential data to obtain the target PAC dosage. The first pre-set neural network model employs a multi-layer feedforward neural network structure. The second pre-set neural network model uses a hybrid architecture combining bidirectional LSTM and fully connected neural networks. A third pre-set neural network model optimizes the dissolved air pressure based on the target PAC dosage, influent phosphorus concentration, water turbidity, historical phosphorus removal rate, and post-coagulation parameters to achieve real-time control of micro-nano air flotation phosphorus removal. The post-coagulation parameters include post-coagulation water turbidity and post-coagulation potential data.

[0045] As can be seen from the above, this application achieves real-time, adaptive optimization control of PAC dosage and dissolved air pressure in the micro-nano air flotation phosphorus removal process through a multi-neural network model collaborative intelligent decision-making mechanism. This significantly improves the response speed, processing stability, and operational economy of the phosphorus removal process, effectively solving the problems of reliance on manual experience, control lag, parameter rigidity, and low energy efficiency in traditional methods. Thus, it enables real-time, adaptive, and intelligent optimization and control of operating parameters in the micro-nano air flotation phosphorus removal process. Attached Figure Description

[0046] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0047] Figure 1 This is a flowchart of a real-time control method for micro / nano air flotation phosphorus removal based on a neural network model disclosed in this invention.

[0048] Figure 2 This is a schematic diagram of a real-time control device for micro-nano air flotation phosphorus removal based on a neural network model disclosed in this invention.

[0049] Figure 3 This is a structural diagram of an electronic device disclosed in this invention. Detailed Implementation

[0050] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0051] With the acceleration of industrialization and urbanization, eutrophication of water bodies has become increasingly prominent, and phosphorus emission standards are constantly tightening. As relevant environmental regulations are strictly enforced, phosphorus emission limits for wastewater treatment plant effluents are becoming increasingly stringent. Against this backdrop, the efficiency and intelligence of phosphorus removal processes have become important development directions for the wastewater treatment industry. Micro-nano air flotation technology, as a highly efficient solid-liquid separation method, can achieve efficient capture and removal of suspended particles and flocs by generating microbubbles. When combined with coagulation and chemical dosing processes, this technology can significantly improve phosphorus removal efficiency. However, traditional coagulation-air flotation phosphorus removal control methods still suffer from shortcomings such as control lag, fixed process operating parameters, reliance on manual experience, low energy efficiency, and a lack of intelligent decision-making. Therefore, this application provides a real-time control scheme for micro-nano air flotation phosphorus removal based on a neural network model, which can achieve real-time, adaptive, and intelligent optimization and control of operating parameters during the micro-nano air flotation phosphorus removal process.

[0052] See Figure 1 As shown in the figure, this application discloses a real-time control method for micro-nano air flotation phosphorus removal based on a neural network model, including:

[0053] Step S11: Collect operational data in real time through a preset sensor network, and preprocess the operational data to obtain target data.

[0054] In this embodiment, process operation data is collected in real time through a multi-sensor network. The target data includes, but is not limited to, influent phosphorus concentration, water turbidity, and potential data. The sensors in the multi-sensor network include, but are not limited to, turbidity sensors, Zeta potentiometers, online total phosphorus analyzers, temperature sensors, pH sensors, and flow meters. The turbidity sensors are installed in the influent zone and the coagulation reaction zone to continuously monitor changes in water turbidity, with a measurement range covering 0-100 NTU. The Zeta potentiometer uses laser Doppler electrophoresis technology to automatically collect samples during coagulation for potential analysis. The phosphorus concentration in both the influent and the final effluent is also collected in real time.

[0055] It is important to note that before data acquisition, a deep learning model trained on a large amount of historical data must be loaded in advance. This model has learned and mastered the mapping relationship of optimal control strategies under different water quality conditions. Simultaneously, the sensor systems installed at each key node of the process need to be initialized and tested. Zero-point calibration and range verification of the PAC dosing pump must be performed in advance, and position confirmation and response testing of the dissolved gas pressure control valve must be conducted. Specifically, before real-time data acquisition through a preset sensor network, the process may include: training the first, second, and third preset neural network models using historical data; initializing and calibrating the sensors in the preset sensor network; the sensors include turbidity sensors, Zeta potentiometers, and online total phosphorus analyzers; initially configuring and calibrating the control actuators, and setting initial operating parameters and safe operating boundary conditions; the control actuators include the PAC dosing pump and the dissolved gas pressure control valve. In other words, by initializing and testing the sensor systems installed at each key node of the process, it is ensured that each sensor is functioning normally and that the measurement data is accurate and reliable. The initial parameters of the actuators were adjusted in advance. Based on the characteristics of the influent water quality and treatment requirements, initial operating parameters were set, including but not limited to the basic PAC dosage, initial dissolved air pressure, and control cycle. A database of safe operating boundary conditions was also established, setting upper and lower limit protection values ​​for each process parameter to provide a basis for subsequent safety control.

[0056] In this embodiment, raw data is collected through a multi-sensor network and processed to ensure data rationality and accuracy. Specifically, the real-time collection of operational data through a preset sensor network and the preprocessing of the operational data to obtain target data may include: real-time collection of water turbidity, potential data, and influent and effluent phosphorus concentrations during the coagulation reaction process using sensors in the preset sensor network to obtain raw operational data; outlier identification and removal operations are performed on the raw operational data, and a preset interpolation algorithm is used to compensate the raw operational data after outlier removal to obtain compensated operational values; the compensated operational values ​​are then smoothed and normalized sequentially, and feature extraction is performed on the processed data to obtain target data containing multi-dimensional feature vectors. For example, in one specific implementation, abnormal data points that significantly deviate from the normal range can be identified and removed from the data captured by the sensors, and reasonable values ​​can be supplemented using an interpolation algorithm. This ensures the rationality of the process operational data. A sliding window filtering technique is used to smooth the data, effectively suppressing random noise interference while retaining the true process change characteristics. All data undergoes normalization, transforming parameters of different dimensions to the [0,1] interval, providing standardized input features for deep learning models. Furthermore, the module performs feature engineering to extract temporal features of the data, including trends, fluctuation amplitudes, and rates of change, constructing a dataset containing multi-dimensional feature vectors to provide a rich information foundation for intelligent decision-making.

[0057] Step S12: Using a first preset neural network model, a recommended PAC dosage value is generated based on the influent phosphorus concentration, the water turbidity, and the historical phosphorus removal rate. Then, a second preset neural network model is used to dynamically correct the recommended PAC dosage value based on the potential data to obtain the target PAC dosage.

[0058] In this embodiment, the first preset neural network model adopts a multi-layer feedforward neural network architecture, constructing a deep learning neural network prediction model containing an input layer, multiple hidden layers, and an output layer. The hidden layers are configured with 64, 32, and 16 neurons respectively, and a batch normalization layer and a dropout layer are configured after each hidden layer, effectively improving the model's generalization ability and anti-overfitting performance. The system input features have three dimensions, including influent total phosphorus concentration, average influent turbidity, and historical phosphorus removal rate. Through feature engineering, the system extracts the average removal rate corresponding to each influent total phosphorus concentration from historical operating data as an important feature input, fully considering the intrinsic correlation between water quality characteristics and treatment effect. During model training, the system uses a standardized preprocessing module to normalize the input features and target variables, eliminating the influence of dimensional differences on model training. The training algorithm is based on the mean squared error loss function, uses the Adam optimizer for parameter updates, and achieves iterative optimization of model parameters through the backpropagation algorithm. Specifically, using historical influent phosphorus concentration, historical average turbidity, historical phosphorus removal rate, and corresponding target PAC dosage labels, the first preset neural network model is trained using a mean squared error loss function and the Adam optimizer. In the prediction phase, the system receives real-time influent water quality parameters, standardizes them, and inputs them into the fully trained deep neural network, outputting a recommended initial PAC dosage. This system achieves a nonlinear mapping from influent characteristics to the optimal dosage through deep learning technology, overcoming the shortcomings of insufficient accuracy in traditional empirical formulas.

[0059] Furthermore, the second preset neural network model adopts a hybrid architecture combining bidirectional LSTM and fully connected neural networks. The bidirectional LSTM network has 16 hidden units, possessing forward and backward information processing capabilities, and can fully capture the time-dependent characteristics of process parameters such as zeta potential. The fully connected network uses a 32-16-8 three-layer decreasing structure, with a hyperbolic tangent activation function at the end to ensure that the output adjustment coefficient is limited to a reasonable range of [-1, 1]. The system input features are five-dimensional, including real-time zeta potential values, current PAC dosage, influent total phosphorus concentration, zeta potential change trend, and time interval parameters. By setting a time-series sliding window of length 5, the system constructs continuous time-series data, effectively extracting the dynamic characteristics of the process. In terms of data preprocessing, the system is equipped with a historical data buffer management mechanism to maintain the process status data of the most recent 5 sampling periods in real time. By calculating the differential rate of change of zeta potential, the system quantifies the dynamic evolution trend of the colloidal particle surface potential, providing important temporal gradient information for the model. The system extracts continuous time-series data according to experimental groups and automatically generates optimal adjustment coefficient labels based on the direction of change in removal rate between adjacent sampling points. A positive adjustment coefficient is set when the removal rate decreases, and a negative adjustment coefficient is set when the removal rate stabilizes or increases, achieving intelligent mapping from process effect to control strategy. Specifically, the second preset neural network model is trained using historical potential data, historical PAC dosage, historical influent phosphorus concentration, historical potential change trends, time interval parameters, and corresponding adjustment coefficient labels; the adjustment coefficient labels are generated based on the historical direction of removal rate change.

[0060] During the correction decision-making phase, the system combines real-time collected process parameters with historical buffer data into a standard-length time series. After standardization, this series is input into a fully trained hybrid neural network, which outputs a dynamic adjustment coefficient for the PAC dosage. Specifically, the system utilizes a bidirectional long short-term memory network in the second preset neural network model to obtain the time-series dependency features of the potential data. Based on these features, a fully connected neural network in the second preset neural network model outputs a target adjustment coefficient for the recommended PAC dosage. The recommended PAC dosage is then dynamically corrected based on this target adjustment coefficient to obtain the target PAC dosage.

[0061] Step S13: Optimize dissolved air pressure using a third preset neural network model based on the target PAC dosage, the influent phosphorus concentration, the water turbidity, the historical phosphorus removal rate, and parameters after coagulation, so as to achieve real-time control of micro-nano air flotation phosphorus removal.

[0062] In this embodiment, the third preset neural network model adopts an improved deep neural network architecture, constructing an intelligent prediction model containing an input layer, multiple hidden layers, and an output layer. The hidden layer adopts a three-layer structure of 64-32-16, which increases the network capacity compared to the original model. Each hidden layer is followed by a batch normalization layer and a discard layer, significantly improving the model's ability to represent complex process relationships. The system input features are six-dimensional, deeply fusing and analyzing the influent total phosphorus concentration, influent turbidity, PAC dosage, post-coagulation turbidity, post-coagulation Zeta potential, and historical removal rate. This feature combination fully considers the impact of coagulation treatment on subsequent air flotation separation, realizing full-process parameter coupling from influent characteristics to coagulation state to separation effect. During training, the system uses a standardized preprocessing module to normalize the input features and target pressure values, and optimizes the model parameters using the mean squared error loss function and the Adam optimization algorithm. After 1000 training cycles of full iteration, the model deeply learns the complex nonlinear mapping relationship between multi-dimensional process parameters and optimal dissolved air pressure. Specifically, the third preset neural network model is trained using historical influent phosphorus concentration, historical water turbidity, historical PAC dosage, historical post-coagulation turbidity, historical post-coagulation potential data, historical phosphorus removal rate, and corresponding target dissolved air pressure labels. In the prediction and decision-making phase, the system receives real-time process parameters, including influent phosphorus concentration, influent turbidity, PAC dosage, post-coagulation turbidity, post-coagulation Zeta potential, and historical removal rate. After feature standardization, these parameters are input into the fully trained deep neural network, which directly outputs the optimal dissolved air pressure setpoint. This setpoint is learned from historical high-efficiency operating data and can maximize the phosphorus removal rate.

[0063] Furthermore, a unified experience database is established to continuously record data throughout the entire process, from initial prediction and dynamic adjustment to the final effect. By periodically comparing model predictions with actual removal results, online parameter fine-tuning and version iteration are performed on the three deep learning models. Specifically, the real-time control method for micro-nano air flotation phosphorus removal based on neural network models may further include: establishing an experience database and storing the output data of the first, second, and third preset neural network models in the experience database; and performing collaborative optimization and parameter iterative updates on the first, second, and third preset neural network models based on the data in the experience database at preset update intervals. This closed-loop learning mechanism enables the system to continuously adapt to water quality fluctuations and process changes, gradually improving the accuracy and robustness of decisions at each stage, ultimately achieving adaptive optimization control of the entire air flotation phosphorus removal process.

[0064] As can be seen from the above, the embodiments of this application achieve adaptive closed-loop optimization control of PAC dosage and dissolved air pressure by real-time monitoring of water quality indicators and integration with deep learning model calculations. This improves phosphorus removal efficiency, reduces operating costs, and reduces reliance on operator experience. Thus, real-time, adaptive, and intelligent optimization and control of operating parameters can be achieved in the micro-nano flotation phosphorus removal process.

[0065] See Figure 2 As shown in the figure, this application discloses a real-time control device for micro-nano air flotation phosphorus removal based on a neural network model, comprising:

[0066] The data acquisition module 11 is used to collect operational data in real time through a preset sensor network and preprocess the operational data to obtain target data; the target data includes influent phosphorus concentration, water turbidity and potential data.

[0067] The first parameter control module 12 is used to generate a recommended PAC dosage value based on the influent phosphorus concentration, the water turbidity, and the historical phosphorus removal rate using a first preset neural network model, and to dynamically correct the recommended PAC dosage value based on the potential data using a second preset neural network model to obtain the target PAC dosage; the first preset neural network model adopts a multi-layer feedforward neural network structure; the second preset neural network model adopts a hybrid architecture combining bidirectional LSTM and fully connected neural network.

[0068] The second parameter control module 13 is used to optimize the dissolved air pressure based on the target PAC dosage, the influent phosphorus concentration, the water turbidity, the historical phosphorus removal rate, and the parameters after coagulation using a third preset neural network model, so as to achieve real-time control of micro-nano air flotation phosphorus removal; the parameters after coagulation include the water turbidity after coagulation and the potential data after coagulation.

[0069] In some specific embodiments, the data acquisition module 11 may specifically include:

[0070] The data acquisition unit is used to collect data on water turbidity, potential, and phosphorus concentration in the influent and effluent during the coagulation reaction process in real time through sensors in the preset sensor network, so as to obtain raw operating data.

[0071] The first data processing unit is used to perform outlier identification and removal operations on the original running data, and to use a preset interpolation algorithm to perform data compensation on the original running data after outlier removal, so as to obtain the compensated running values.

[0072] The second data processing unit is used to perform smoothing and normalization processing on the compensated running values ​​in sequence, and to extract features from the processed data to obtain target data containing multi-dimensional feature vectors.

[0073] In some specific embodiments, the real-time control device for micro-nano air flotation phosphorus removal based on a neural network model may further include:

[0074] The model training unit is used to train the first preset neural network model, the second preset neural network model, and the third preset neural network model using historical data;

[0075] The first system initialization unit is used to perform initialization detection and calibration of the sensors in the preset sensor network; the sensors include a turbidity sensor, a zeta potential meter, and a total phosphorus online analyzer.

[0076] The second system initialization unit is used to perform initial configuration and calibration of the control actuator, and to set initial operating parameters and safe operating boundary conditions; the control actuator includes a PAC dosing pump and a dissolved gas pressure control valve;

[0077] The first preset neural network model training unit is used to train the first preset neural network model using historical influent phosphorus concentration, historical average turbidity of water, historical phosphorus removal rate and corresponding target PAC dosage labels, through the mean square error loss function and Adam optimizer. The first preset neural network model adopts a multi-layer feedforward neural network architecture including an input layer, multiple hidden layers and an output layer. The hidden layers are provided with 64, 32 and 16 neurons respectively, and each hidden layer is followed by a batch normalization layer and a dropout layer.

[0078] The second preset neural network model training unit is used to train the second preset neural network model using historical potential data, historical PAC dosage, historical influent phosphorus concentration, historical potential change trends, time interval parameters, and corresponding adjustment coefficient labels. The adjustment coefficient labels are generated based on the historical removal rate change direction. The second preset neural network model adopts a hybrid architecture combining a bidirectional long short-term memory network and a fully connected neural network. The bidirectional long short-term memory network is used to capture the temporal dependence features of the input parameters, and the fully connected neural network is used to output the adjustment coefficients.

[0079] The third preset neural network model training unit is used to train the third preset neural network model using historical influent phosphorus concentration, historical water turbidity, historical PAC dosage, historical post-coagulation water turbidity, historical post-coagulation potential data, historical phosphorus removal rate, and corresponding target dissolved air pressure labels. The third preset neural network model adopts a preset deep neural network architecture, including an input layer, multiple hidden layers, and an output layer. The hidden layers have 64, 32, and 16 neurons respectively, and each hidden layer is followed by a batch normalization layer and a discard layer.

[0080] A database establishment unit is used to establish an experience database and store the data output by the first preset neural network model, the second preset neural network model and the third preset neural network model into the experience database.

[0081] The model update unit is used to perform collaborative optimization and parameter iterative update of the first preset neural network model, the second preset neural network model and the third preset neural network model based on the data in the experience database according to a preset update time interval.

[0082] Accordingly, the first parameter control module 12 may specifically include:

[0083] The PAC dosage correction unit is used to obtain the temporal dependence features of the potential data using the bidirectional long short-term memory network in the second preset neural network model, and to output the target adjustment coefficient for the recommended PAC dosage based on the temporal dependence features of the potential data using the fully connected neural network in the second preset neural network model, and to dynamically correct the recommended PAC dosage based on the target adjustment coefficient to obtain the target PAC dosage.

[0084] Furthermore, embodiments of this application also disclose an electronic device, Figure 3 This is a structural diagram of an electronic device 20 according to an exemplary embodiment. The content of the diagram should not be construed as limiting the scope of this application. The electronic device 20 may specifically include: at least one processor 21, at least one memory 22, a power supply 23, a communication interface 24, an input / output interface 25, and a communication bus 26. The memory 22 stores a computer program, which is loaded and executed by the processor 21 to implement the relevant steps in the real-time control method for micro-nano air flotation phosphorus removal based on a neural network model disclosed in any of the foregoing embodiments. Furthermore, the electronic device 20 in this embodiment may specifically be an electronic computer.

[0085] In this embodiment, the power supply 23 is used to provide operating voltage for each hardware device on the electronic device 20; the communication interface 24 can create a data transmission channel between the electronic device 20 and external devices, and the communication protocol it follows can be any communication protocol applicable to the technical solution of this application, and is not specifically limited here; the input / output interface 25 is used to acquire external input data or output data to the outside world, and its specific interface type can be selected according to specific application needs, and is not specifically limited here.

[0086] In addition, the memory 22, as a carrier for resource storage, can be a read-only memory, random access memory, disk or optical disk, etc. The resources stored thereon can include operating system 221, computer program 222, etc., and the storage method can be temporary storage or permanent storage.

[0087] The operating system 221 is used to manage and control the various hardware devices on the electronic device 20 and the computer program 222, which may be Windows Server, Netware, Unix, Linux, etc. In addition to including a computer program capable of performing the real-time control method for micro-nano air flotation phosphorus removal based on a neural network model, which is executed by the electronic device 20 according to any of the foregoing embodiments, the computer program 222 may further include computer programs capable of performing other specific tasks.

[0088] Furthermore, this application also discloses a computer-readable storage medium for storing a computer program; wherein, when the computer program is executed by a processor, it implements the aforementioned real-time control method for micro-nano air flotation phosphorus removal based on a neural network model. Specific steps of this method can be found in the corresponding content disclosed in the foregoing embodiments, and will not be repeated here.

[0089] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatus disclosed in the embodiments, since it corresponds to the method disclosed in the embodiments, the description is relatively simple; relevant parts can be referred to in the method section.

[0090] Those skilled in the art will further recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0091] The steps of the methods or algorithms described in conjunction with the embodiments disclosed herein can be implemented directly by hardware, a software module executed by a processor, or a combination of both. The software module can be located in random access memory (RAM), main memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disk, removable disk, CD-ROM, or any other form of storage medium known in the art.

[0092] Finally, it should be noted that in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0093] The technical solutions provided in this application have been described in detail above. Specific examples have been used to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of this application. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of this application. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A real-time control method for micro / nano air flotation phosphorus removal based on a neural network model, characterized in that, include: Operational data is collected in real time through a preset sensor network, and the operational data is preprocessed to obtain target data; The target data includes influent phosphorus concentration, water turbidity, and potential data; A first preset neural network model is used to generate a recommended PAC dosage based on the influent phosphorus concentration, the water turbidity, and the historical phosphorus removal rate. A second preset neural network model is used to dynamically correct the recommended PAC dosage based on the potential data to obtain the target PAC dosage. The first preset neural network model adopts a multi-layer feedforward neural network structure. The second preset neural network model adopts a hybrid architecture combining bidirectional LSTM and fully connected neural networks. The dissolved air pressure is optimized using a third preset neural network model based on the target PAC dosage, the influent phosphorus concentration, the water turbidity, the historical phosphorus removal rate, and the parameters after coagulation, so as to achieve real-time control of micro-nano air flotation phosphorus removal; the parameters after coagulation include the water turbidity after coagulation and the potential data after coagulation.

2. The real-time control method for micro / nano air flotation phosphorus removal based on a neural network model according to claim 1, characterized in that, Before the real-time acquisition of operational data through a preset sensor network, the following steps are also included: The first preset neural network model, the second preset neural network model, and the third preset neural network model are trained using historical data; The sensors in the preset sensor network are initialized, detected, and calibrated; the sensors include a turbidity sensor, a zeta potential meter, and an online total phosphorus analyzer. The control actuator is initially configured and calibrated, and initial operating parameters and safe operating boundary conditions are set; the control actuator includes a PAC dosing pump and a dissolved gas pressure control valve.

3. The real-time control method for micro / nano air flotation phosphorus removal based on a neural network model according to claim 1, characterized in that, The step of collecting operational data in real time through a preset sensor network and preprocessing the operational data to obtain target data includes: The raw operating data are obtained by real-time collection of water turbidity, potential data and phosphorus concentration in influent and effluent during the coagulation reaction process through sensors in the preset sensor network. The original running data is subjected to outlier identification and removal operations, and the original running data after outlier removal is compensated using a preset interpolation algorithm to obtain the compensated running values. The compensated running values ​​are sequentially smoothed and normalized, and features are extracted from the processed data to obtain target data containing multi-dimensional feature vectors.

4. The real-time control method for micro / nano air flotation phosphorus removal based on a neural network model according to claim 1, characterized in that, Also includes: Using historical influent phosphorus concentration, historical average turbidity of water body, historical phosphorus removal rate and corresponding target PAC dosage labels, the first preset neural network model is trained by mean square error loss function and Adam optimizer. The first preset neural network model adopts a multi-layer feedforward neural network architecture including an input layer, multiple hidden layers and an output layer; the hidden layers are provided with 64, 32 and 16 neuron nodes respectively, and each hidden layer is configured with a batch normalization layer and a dropout layer respectively.

5. The real-time control method for micro / nano air flotation phosphorus removal based on a neural network model according to claim 1, characterized in that, Also includes: The second preset neural network model is trained using historical potential data, historical PAC dosage, historical influent phosphorus concentration, historical potential change trends, time interval parameters, and corresponding adjustment coefficient labels; the adjustment coefficient labels are labels generated based on the historical removal rate change direction. The second preset neural network model adopts a hybrid architecture combining a bidirectional long short-term memory network and a fully connected neural network; the bidirectional long short-term memory network is used to capture the temporal dependency features of the input parameters; the fully connected neural network is used to output adjustment coefficients. Accordingly, the step of dynamically correcting the recommended PAC dosage based on the potential data using a second preset neural network model to obtain the target PAC dosage includes: The temporal dependence features of the potential data are obtained by using the bidirectional long short-term memory network in the second preset neural network model, and the target adjustment coefficient for the recommended value of the PAC dosage is output by using the fully connected neural network in the second preset neural network model based on the temporal dependence features of the potential data. The recommended value of the PAC dosage is dynamically corrected based on the target adjustment coefficient to obtain the target PAC dosage.

6. The real-time control method for micro / nano air flotation phosphorus removal based on a neural network model according to claim 1, characterized in that, Also includes: Using historical influent phosphorus concentration, historical water turbidity, historical PAC dosage, historical post-coagulation water turbidity, historical post-coagulation potential data, historical phosphorus removal rate, and corresponding target dissolved air pressure labels. The third preset neural network model is trained. The third preset neural network model adopts a preset deep neural network architecture, which includes an input layer, multiple hidden layers and an output layer; the hidden layers are provided with 64, 32 and 16 neuron nodes respectively, and each hidden layer is configured with a batch normalization layer and a dropout layer respectively.

7. The real-time control method for micro / nano air flotation phosphorus removal based on a neural network model according to any one of claims 1 to 6, characterized in that, Also includes: An experience database is established, and the data output by the first preset neural network model, the second preset neural network model, and the third preset neural network model are stored in the experience database. Based on the data in the experience database and at a preset update interval, the first preset neural network model, the second preset neural network model, and the third preset neural network model are collaboratively optimized and their parameters are iteratively updated.

8. A real-time control device for micro-nano air flotation phosphorus removal based on a neural network model, characterized in that, include: The data acquisition module is used to collect operational data in real time through a preset sensor network and preprocess the operational data to obtain target data; The target data includes influent phosphorus concentration, water turbidity, and potential data; The first parameter control module is used to generate a recommended PAC dosage value based on the influent phosphorus concentration, the water turbidity, and historical phosphorus removal rate using a first preset neural network model, and to dynamically correct the recommended PAC dosage value based on the potential data using a second preset neural network model to obtain the target PAC dosage; the first preset neural network model adopts a multi-layer feedforward neural network structure; the second preset neural network model adopts a hybrid architecture combining bidirectional LSTM and fully connected neural networks; The second parameter control module is used to optimize the dissolved air pressure based on the target PAC dosage, the influent phosphorus concentration, the water turbidity, the historical phosphorus removal rate, and the parameters after coagulation using a third preset neural network model, so as to achieve real-time control of micro-nano air flotation phosphorus removal; the parameters after coagulation include the water turbidity after coagulation and the potential data after coagulation.

9. An electronic device, characterized in that, include: Memory, used to store computer programs; A processor is configured to execute the computer program to implement the real-time control method for micro / nano air flotation phosphorus removal based on a neural network model as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, Used to store a computer program, wherein the computer program, when executed by a processor, implements the real-time control method for micro-nano air flotation phosphorus removal based on a neural network model as described in any one of claims 1 to 7.

Images