An intelligent dosing control method for water treatment based on multi-model fusion

Abstract

The present invention relates to the field of sewage treatment technology, and specifically to an intelligent dosing control method for water treatment based on multi-model fusion. The inlet water quality monitoring data set, the process water quality monitoring data set and the outlet water quality monitoring data set are fused into a fused water quality state data set, and the equipment monitoring data set and the dosing amount monitoring data set are fused into a fused dosing state data set; the fused water quality state data set is input into a deep artificial neural network water quality state sub-model with adjustable parameters, and the fused dosing state data set is input into a deep artificial neural network dosing state sub-model with adjustable parameters; the results of the two models are fused into a deep artificial neural network fusion model with adjustable parameters, and the weight coefficients after percentage error normalization are fused, and the water treatment dosing process is controlled according to the output results of the deep artificial neural network fusion model. The present invention improves response speed and energy conservation and consumption reduction, and realizes efficient, intelligent and sustainable development of the water treatment process.

Description

Technical Field

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

[0002] Directly supplying water from rivers and lakes without treatment poses serious risks, as these water sources may contain pathogenic microorganisms, harmful chemicals, suspended particles and pollutants, which can easily cause the spread of disease and endanger public health. In addition, the volatility and instability of water quality may also make it impossible to guarantee the safety and reliability of drinking water, affecting residents' lives and industrial production. Therefore, a strict water treatment process must be carried out to ensure that the water quality meets the standards and that water use is safe.

[0003] The dosing process in water treatment is crucial. By adding chemical agents such as coagulants, flocculants and disinfectants, suspended solids, organic matter, pathogenic microorganisms and other pollutants in the water can be effectively removed to ensure safe, clean and stable water quality. The dosing process not only improves the efficiency and effectiveness of water treatment, but also ensures that the effluent meets water use standards, protects public health and avoids the risk of waterborne diseases.

[0004] Existing dosing control methods mainly rely on historical experience and manual control, and have significant shortcomings, including slow response to water quality changes, insufficient dosing accuracy, waste of chemicals, high operating costs, and large human operation errors. This method is difficult to dynamically adapt to complex and changing water quality conditions, difficult to achieve optimal control, resulting in unstable water quality and difficulty in fully ensuring water supply safety and economy.

[0005] In summary, the existing technology has a dosing control process that relies on historical experience and manual control, which makes it impossible to ensure that the dosage can accurately meet the needs of the water treatment process. There are technical problems such as the effluent from the water treatment does not meet the drinking water standards and the risk of water quality abnormalities is uncontrollable; therefore, the present invention proposes an intelligent dosing control method for water treatment based on multi-model fusion to improve the above problems. Summary of the Invention

[0006] The purpose of the present invention is to provide an intelligent dosing control method for water treatment based on multi-model fusion, so as to solve the problem that dosing control in the prior art relies solely on historical experience and manual control, resulting in low biochemical reaction efficiency in water treatment and delayed response to abnormal water inflow conditions.

[0007] To achieve the above objectives, the present invention adopts the following technical solution: a water treatment intelligent dosing control method based on multi-model fusion, comprising the following steps:

[0008] S100, sampling water quality indicators based on inlet monitoring to obtain an inlet water quality monitoring data set; sampling water quality indicators based on process monitoring to obtain a process water quality monitoring data set; sampling water quality indicators based on effluent monitoring to obtain an effluent water quality monitoring data set; collecting equipment data from water treatment equipment to obtain an equipment monitoring data set; collecting equipment data from water treatment dosing equipment to obtain a dosing rate monitoring data set;

[0009] S200, fusing the inlet water quality monitoring data set, the process water quality monitoring data set, and the outlet water quality monitoring data set into a fused water quality status data set; fusing the equipment monitoring data set and the dosage monitoring data set into a fused dosage status data set;

[0010] S300: Build a deep artificial neural network adaptive model and perform model training and loss optimization; the deep artificial neural network adaptive model includes a deep artificial neural network water proton sub-model and a deep artificial neural network drug administration sub-model;

[0011] S400, fusing the two model results into a deep artificial neural network fusion model with adjustable parameters, fusing them according to weight coefficients normalized by percentage error, wherein the deep artificial neural network fusion model with adjustable parameters is embedded in a control system terminal of the water treatment dosing equipment;

[0012] S500, controlling the water treatment dosing process according to the output results of the deep artificial neural network fusion model with adjustable parameters.

[0013] Furthermore, in the S100, the inlet water quality indicators include turbidity, chromaticity, temperature, pH value, conductivity, total dissolved solids, total hardness, chemical oxygen demand, biochemical oxygen demand, chloride, sulfate, nitrate, metal ions, total coliform group and total bacteria count at the time of inlet water; the process water quality indicators include turbidity, pH value and chloride of the sedimentation tank mixing area, sedimentation tank flocculation area, sedimentation tank sedimentation area, filter tank, disinfection tank, deep treatment and post-treatment; the effluent water quality indicators include turbidity, chromaticity, pH value, residual chlorine, total dissolved solids, total hardness, fluoride, chloride, nitrate, metal ions, total coliform group, total bacteria count and disinfection by-products at the time of effluent water; the equipment monitoring data includes dosing time, mixing reaction time, flocculation time, sedimentation time, filtration time and disinfection contact time; the dosage monitoring data includes the dosage of chemicals in the mixing area of ​​the sedimentation tank, the flocculation area of ​​the sedimentation tank and the disinfection tank for water treatment equipment.

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

[0015] Furthermore, in the aforementioned S300, the core structure of the deep artificial neural network includes an input layer, a hidden layer, and an output layer; each layer is composed of multiple neurons, which are interconnected through a weight matrix and bias parameters to achieve information transmission and conversion; the input layer of the deep artificial neural network is responsible for receiving raw data and passing it to the hidden layer; the hidden layer contains multiple levels, and the neurons in each hidden layer perform complex feature extraction and transformation on the input data through nonlinear activation functions; the deep artificial neural network uses nonlinear transformations to enable 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 a final prediction result or classification label;

[0016] The deep artificial neural network construction method is as follows: first, prepare and preprocess 30% of the data set and divide it into a training set, a validation set, and a test set; then, randomly initialize the network weights and bias parameters; then, 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 backpropagation 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 1,000 rounds, and output temporary adjustable parameters at the same time.

[0017] Furthermore, the deep artificial neural network adaptive model is as follows:

[0018]

[0019]

[0020] in:

[0021] x i is the input value;

[0022] y i is the output value;

[0023] p i is the true value corresponding to the input value;

[0024] w i is the weight of each layer;

[0025] b i is the bias of each layer;

[0026] g i is the activation function of each layer;

[0027] m is the number of sample points;

[0028] J(θ) is the loss function;

[0029] θ is the loss;

[0030] α is the learning rate.

[0031] Furthermore, in the aforementioned S400, the deep artificial neural network fusion model fusion method is: normalizing the percentage errors of the deep artificial neural network water quality sub-model and the deep artificial neural network drug administration sub-model respectively, and fusing them according to the weight coefficients.

[0032] Furthermore, the step S500 includes the following steps:

[0033] S510, setting the parameters of the deep artificial neural network as temporary adjustable parameters, inputting 100% of the fused data into the deep artificial neural network, and the model outputting a recommended dosage;

[0034] S520, the control module controls the water treatment dosing equipment in the mixing area of ​​the sedimentation tank, the flocculation area of ​​the sedimentation tank, and the disinfection tank during the water treatment dosing process according to the dosing amount recommended by the model;

[0035] S530, monitor the water quality of the water sample collected through the real-time effluent sampling tube, obtain the effluent water quality, and obtain a current effluent water quality monitoring data set;

[0036] S540: Determine whether the water quality index meets the standard. If the water pollution index meets the standard, obtain a sewage discharge instruction.

[0037] Beneficial effects of the present invention:

[0038] It solves the problem in existing technologies that drug dosing control relies solely on historical experience and manual control, resulting in low efficiency of biochemical reactions in water treatment and delayed response to abnormal water inflow conditions; multi-model fusion can accurately adjust the dosage of chemicals based on real-time and historical data, dynamically adapt to changes in water quality, optimize treatment effects, reduce manual intervention and drug waste, improve response speed and energy conservation and consumption reduction, and achieve efficient, intelligent and sustainable development of the water treatment process. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 This is a schematic diagram of the process of intelligent dosing control for water treatment according to the present invention;

[0040] Figure 2 This is a schematic diagram of monitoring sites for intelligent dosing control of water treatment according to the present invention;

[0041] Figure 3 It is a module structure diagram of the intelligent dosing control of water treatment in the present invention. DETAILED DESCRIPTION

[0042] The technical solutions in the embodiments of the present invention will be described clearly and completely below with reference to the accompanying drawings in the embodiments of the present invention.

[0043] Example

[0044] In this embodiment, the intelligent dosing control method for water treatment of the present invention includes the following steps:

[0045] SA100, based on the water quality indicators of inlet monitoring, obtain the inlet water quality monitoring data set; based on the water quality indicators of process monitoring, obtain the process water quality monitoring data set; based on the water quality indicators of outlet monitoring, obtain the outlet water quality monitoring data set;

[0046] SA10011. Obtain the historical influent sampling dataset of the target water treatment plant;

[0047] SA10012. Perform data preprocessing, fill missing values, and delete outliers based on the historical water sampling dataset.

[0048] SA10013. Configure parameters of water quality monitoring instrument according to the fluctuation characteristics of water quality indicators.

[0049] SA10021. Obtain historical sampling data sets for the target water treatment plant's process sedimentation tank mixing zone, sedimentation tank flocculation zone, sedimentation tank settling zone, filter tank, disinfection tank, advanced treatment, and post-treatment.

[0050] SA10022. Sample the data set according to the past process segment, perform data preprocessing, fill in missing values, and delete outliers;

[0051] SA10023. Configure parameters of water quality monitoring instrument according to the fluctuation characteristics of water quality indicators.

[0052] SA10031. Obtain the historical effluent sampling dataset of the target water treatment plant;

[0053] SA10032. Perform data preprocessing, fill missing values, and delete outliers based on the past water discharge sampling data set;

[0054] SA10033. Configure parameters of water quality monitoring instrument according to the fluctuation characteristics of water quality indicators.

[0055] Specifically, in this embodiment, the target water treatment plant inlet water quality indicators include turbidity, chromaticity, temperature, pH value, conductivity, total dissolved solids, total hardness, chemical oxygen demand, biochemical oxygen demand, chloride, sulfate, nitrate, metal ions (such as iron, manganese, copper, lead, etc.), total coliform bacteria and total bacteria count at the time of inlet water; the target water treatment plant process water quality indicators include turbidity, pH value and chloride of the sedimentation tank mixing area, sedimentation tank flocculation area, sedimentation tank sedimentation area, filter tank, disinfection tank, deep treatment and post-treatment; the target water treatment plant effluent water quality indicators include turbidity, chromaticity, pH value, residual chlorine, total dissolved solids, total hardness, fluoride, chloride, nitrate, metal ions, total coliform bacteria, total bacteria count and disinfection by-products at the time of effluent water.

[0056] SB100, by collecting equipment data from water treatment equipment, obtain an equipment status monitoring data set; by collecting equipment data from water treatment dosing equipment, obtain a dosing monitoring data set;

[0057] Specifically, in this embodiment, the sewage undergoes biochemical or physicochemical treatment processes such as inlet regulation, outlet regulation, mixing process, flocculation process, sedimentation process, filtration process, disinfection reaction, deep treatment and post-treatment in the sewage treatment equipment of the target sewage treatment plant. The dosing time, mixing reaction time, flocculation time, sedimentation time, filtration time and disinfection contact time are adjusted and controlled based on the control system to achieve the reasonable arrangement of mixing conditions, flocculation conditions, sedimentation conditions, filtration conditions and disinfection conditions of the water treatment equipment and the normal progress of deep treatment, so that the water supply can effectively undergo treatment processes such as inlet regulation, outlet regulation, mixing process, flocculation process, sedimentation process, filtration process, disinfection reaction, deep treatment and post-treatment.

[0058] Data collection is performed on the control system used to control the water treatment equipment for water treatment, and data collection is performed on the actual operation data of the water treatment equipment to obtain the equipment status monitoring data set. The data types in the equipment status monitoring data set include dosing time, mixing reaction time, flocculation time, sedimentation time, filtration time and disinfection contact time: Data collection is performed on the dosing control system used to control the water treatment equipment for water treatment, and data collection is performed on the dosing data of the water treatment equipment to obtain the dosing monitoring data set. The data types in the dosing monitoring data set include the chemical dosage of the water treatment equipment in the mixing area of ​​the sedimentation tank, the flocculation area of ​​the sedimentation tank, and the disinfection tank.

[0059] S200, merging the multi-source monitoring data sets into a water quality status fusion data set and a drug administration status fusion data set respectively; a data fusion method aligns the data rows and columns of the data sets, and sets the data interval to the minimum sampling time.

[0060] SA200, fuses the inlet water quality monitoring dataset, process water quality monitoring dataset, and outlet water quality monitoring dataset into a water quality state fusion dataset;

[0061] SB200, fuses the equipment status monitoring data set and the dosage monitoring data set into a dosage status fusion data set;

[0062] S300, build a deep artificial neural network adaptive model, perform model training and loss optimization, including the SA300 deep artificial neural network water proton model and the SB300 deep artificial neural network dosing model.

[0063] SA3001 / SB3001, build a deep artificial neural network with adjustable parameters, where the adjustable parameters of the adaptive model include the number of deep artificial neural network data input batches, the size of the deep artificial neural network data input batches, the number of neural network hidden layers, and the dimension of the neural network input data;

[0064] SA3002 / SB3002 uses 30% of the fused data as input variables to train a deep artificial neural network and obtain temporary adjustable parameters. This includes inputting the fused water quality data set into the deep artificial neural network water quality sub-model, adapting the model based on the adjustable parameters, and outputting adaptive parameters; and inputting the fused dosing and medication state data set into the deep artificial neural network medication sub-model, adapting the model based on the adjustable parameters, and outputting adaptive parameters.

[0065] The core structure of a deep artificial neural network consists of an input layer, hidden layers, and an output layer. Each layer consists of multiple neurons, interconnected by weight matrices and bias parameters, enabling information transmission and transformation. The input layer of a deep artificial neural network receives raw data and passes it to the hidden layers. The hidden layers typically consist of multiple layers, with neurons in each layer performing complex feature extraction and transformation on the input data using nonlinear activation functions. These nonlinear transformations enable deep artificial neural networks to capture complex patterns and relationships in the data. The output layer integrates and processes the features extracted by the hidden layers to generate the final prediction or classification label.

[0066] The method for constructing a deep artificial neural network includes the following steps: First, 30% of the dataset is prepared and preprocessed, splitting it into training, validation, and test sets. Next, the network's weights and bias parameters are randomly initialized. Next, the output of each layer is calculated using forward propagation, and the error between the network output and the actual result is calculated using a loss function. Backpropagation is used to calculate the gradient, and an optimization algorithm is used to update the network parameters. This process is repeated until the loss function converges to a percentage error of less than 10% or until a predetermined number of 1,000 training rounds is reached, at which point the adjustable parameters are output.

[0067] The mathematical model of the deep artificial neural network water proton model is:

[0068]

[0069]

[0070] in:

[0071] x i1 It is the input value, including the independent variable for fusing the inlet water quality monitoring dataset, process water quality monitoring dataset, and outlet water quality monitoring dataset into the water quality state fusion dataset;

[0072] y i1 is the output value, which is the predicted value of the water proton model for the dosage of water treatment equipment;

[0073] p i It is the real value corresponding to the input value, which is the real value of the dosage of the water treatment equipment;

[0074] w i is the weight of each layer;

[0075] b i is the bias of each layer;

[0076] g i is the activation function of each layer;

[0077] m is the number of sample points;

[0078] J(θ) is the loss function;

[0079] θ is the loss;

[0080] α is the learning rate.

[0081] The mathematical model of the deep artificial neural network drug delivery sub-model is:

[0082]

[0083] in:

[0084] x i2 It is the input value, including the independent variable for fusing the equipment status monitoring data set and the dosage monitoring data set into the dosage status fusion data set;

[0085] y i2 is the output value, which is the predicted value of the dosing sub-model for water treatment equipment;

[0086] p i It is the real value corresponding to the input value, which is the real value of the dosage of the water treatment equipment;

[0087] wi is the weight of each layer;

[0088] b i is the bias of each layer;

[0089] g i is the activation function of each layer;

[0090] m is the number of sample points;

[0091] J(θ) is the loss function;

[0092] θ is the loss;

[0093] α is the learning rate.

[0094] S400. The deep artificial neural network water quality sub-model and the deep artificial neural network dosing sub-model are integrated into a deep artificial neural network fusion model, wherein the adjustable parameter adaptive model is embedded in the control system terminal of the water treatment equipment; the fusion method of the deep artificial neural network fusion model is to normalize the percentage errors of the deep artificial neural network water quality sub-model and the deep artificial neural network dosing sub-model respectively, and fuse them according to the weight coefficient.

[0095] The mathematical model of the deep artificial neural network fusion model is:

[0096] 1. Calculate RMSE separately

[0097]

[0098] 2. Calculate weights

[0099]

[0100] 3. Calculate normalized weights

[0101]

[0102] 4. Calculate the total value with weight

[0103] y i =w1y i1 +w2y i2

[0104] S500. Input the fusion data into a deep artificial neural network fusion model, and control the water treatment dosing equipment in the mixing area of ​​the sedimentation tank, the flocculation area of ​​the sedimentation tank, and the disinfection tank in the water treatment dosing process according to the deep artificial neural network fusion model and adaptive parameters.

[0105] S510, setting the parameters of the deep artificial neural network as temporary adjustable parameters, inputting 100% of the fused data into the deep artificial neural network, and the model outputting the recommended aeration volume of the blower;

[0106] S520: The control module controls the water treatment dosing equipment in the mixing area of ​​the sedimentation tank, the flocculation area of ​​the sedimentation tank, and the disinfection tank during the water treatment dosing process according to the dosing amount recommended by the model;

[0107] S530, using the data acquisition device to monitor the water quality of the water sample collected through the real-time effluent sampling pipe, obtain the effluent water quality, and obtain a effluent water quality monitoring data set at time T;

[0108] S540: Determine whether the water quality index meets the water quality index allowed by the national standard. If the water quality index meets the water quality index allowed by the national standard, obtain a water supply discharge instruction to allow it to be used as water supply.

[0109] The present invention is not limited to the above-mentioned optimal implementation mode. Anyone can derive other forms of products under the inspiration of the present invention. However, no matter what changes are made in the shape or structure, any technical solution that is the same or similar to that of the present application falls within the scope of protection of the present invention.

Claims

1. A water treatment intelligent dosing control method based on multi-model fusion, characterized in that: The following steps are involved: S100, sampling water quality indicators based on inlet monitoring to obtain an inlet water quality monitoring data set; sampling water quality indicators based on process monitoring to obtain a process water quality monitoring data set; sampling water quality indicators based on effluent monitoring to obtain an effluent water quality monitoring data set; collecting equipment data from water treatment equipment to obtain an equipment monitoring data set; collecting equipment data from water treatment dosing equipment to obtain a dosing rate monitoring data set; S200, fusing the inlet water quality monitoring data set, the process water quality monitoring data set, and the outlet water quality monitoring data set into a fused water quality status data set; fusing the equipment monitoring data set and the dosage monitoring data set into a fused dosage status data set; S300: Build a deep artificial neural network adaptive model and perform model training and loss optimization; the deep artificial neural network adaptive model includes a deep artificial neural network water proton sub-model and a deep artificial neural network drug administration sub-model; The core structure of a deep artificial neural network includes an input layer, a hidden layer, and an output layer. 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 receives raw data and passes it to the hidden layer. The hidden layer consists of 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 transformations, deep artificial neural networks can capture complex patterns and relationships in the data. The output layer integrates the features extracted by the hidden layer to generate the final prediction result or classification label. The deep artificial neural network construction method is as follows: first, prepare and preprocess 30% of the data set and divide it into a training set, a validation set, and a test set; then, randomly initialize the network weights and bias parameters; then, 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 backpropagation 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 1,000 rounds, and output temporary adjustable parameters at the same time; The deep artificial neural network adaptive model is as follows: in: is the input value; is the output value; is the true value corresponding to the input value; is the weight of each layer; is the bias of each layer; is the activation function of each layer; is the number of sample points; is the loss function; It is a loss; is the learning rate; S400, fusing the two model results into a deep artificial neural network fusion model with adjustable parameters, and fusing them according to the weight coefficient after the percentage error is normalized, wherein the deep artificial neural network fusion model with adjustable parameters is embedded in the control system terminal of the water treatment dosing equipment; The fusion method of the deep artificial neural network fusion model is as follows: normalize the percentage errors of the deep artificial neural network water quality sub-model and the deep artificial neural network drug dosing sub-model respectively, and fuse them according to the weight coefficient; S500, based on the deep artificial neural network fusion model with adjustable parameters, the water treatment dosing process is controlled according to the output results.

2. A water treatment intelligent dosing control method based on multi-model fusion according to claim 1, characterized in that: In the S100, the inlet water quality indicators include turbidity, chromaticity, temperature, pH value, conductivity, total dissolved solids, total hardness, chemical oxygen demand, biochemical oxygen demand, chloride, sulfate, nitrate, metal ions, total coliform group and total bacteria count at the time of inlet water; the process water quality indicators include turbidity, pH value and chloride in the mixing area of ​​the sedimentation tank, flocculation area of ​​the sedimentation tank, sedimentation area of ​​the sedimentation tank, filter tank, disinfection tank, deep treatment and post-treatment; the effluent water quality indicators include turbidity, chromaticity, pH value, residual chlorine, total dissolved solids, total hardness, fluoride, chloride, nitrate, metal ions, total coliform group, total bacteria count and disinfection by-products at the time of effluent water; the equipment monitoring data includes dosing time, mixing reaction time, flocculation time, sedimentation time, filtration time and disinfection contact time; the dosage monitoring data includes the dosage of chemicals in the mixing area of ​​the sedimentation tank, the flocculation area of ​​the sedimentation tank and the disinfection tank for water treatment equipment.

3. A water treatment intelligent dosing control method based on multi-model fusion according to claim 1, characterized in that: In the aforementioned S200 , the method for data set fusion is to align the data rows and data columns of the data set, and set the data interval to the minimum sampling time.

4. A water treatment intelligent dosing control method based on multi-model fusion according to claim 1, characterized in that: The step S500 includes the following steps: S510, setting the parameters of the deep artificial neural network as temporary adjustable parameters, inputting 100% of the fused data into the deep artificial neural network, and the model outputting a recommended dosage; S520, the control module controls the water treatment dosing equipment in the mixing area of ​​the sedimentation tank, the flocculation area of ​​the sedimentation tank, and the disinfection tank during the water treatment dosing process according to the dosing amount recommended by the model; S530, monitor the water quality of the water sample collected through the real-time effluent sampling tube, obtain the effluent water quality, and obtain a current effluent water quality monitoring data set; S540: Determine whether the water quality index meets the standard. If the water pollution index meets the standard, obtain a sewage discharge instruction.

Images