Water index prediction method based on sewage treatment

Abstract

The present invention discloses a water index prediction method based on sewage treatment. The method first stirs and settles the water, then divides the area and depth and establishes a local water supply model and performs a first adjustment on the water inlet index. Then, a logarithmic function model is established to predict the water inlet index at different depths. Finally, five models are established to respectively predict the pH value, COD, ammonia nitrogen content, total phosphorus content, and total nitrogen content, and the predicted values ​​are displayed after corresponding to time. This application first supplies water after standing still and models it in layers, and then combines the vertical distribution characteristics of the fluid after standing still to accurately reflect the dynamic changes of water quality indicators. Based on the law that the concentration of pollutants decays with time during precipitation and diffusion, a logarithmic function is constructed to predict the water inlet index in the depth direction to reduce the number of sampling times. Through grid dynamic correction, the arithmetic mean is used to weight the drainage volume ratio, which reduces the sampling deviation caused by flow disturbance and improves the robustness of the prediction. The above scheme can ensure that the sewage quality is relatively uniform, and then the drainage volume ratio is weighted by the average value of the water inlet index of each area to obtain the optimized water inlet index. The optimized water inlet index is then calculated through the constructed logarithmic function to obtain the water inlet index at different depths, and the water inlet index at different depths is input into the five groups of models set for calculation and prediction to obtain the water outlet index.

Description

Technical Field

[0001] The present invention relates to the field of sewage treatment, and in particular to a water index prediction method based on sewage treatment. Background Art

[0002] Wastewater treatment: The process of purifying wastewater to ensure it meets the required quality for discharge into a water body or reuse. Wastewater treatment is widely used in various fields, including construction, agriculture, transportation, energy, petrochemicals, environmental protection, urban landscapes, healthcare, and catering, and is increasingly becoming part of everyday life.

[0003] Since water quality data cannot be collected frequently during the sewage treatment process, a combination of detection and prediction is generally used to display the inlet and outlet water indicators before and after water treatment instead of frequent measurements. At present, in the water treatment process, it is usually adopted to establish a model and combine it with an algorithm to train the model through a large amount of data, and then optimize the algorithm to obtain a relatively accurate prediction model.

[0004] However, in actual application, due to the difficulties in data collection, the use of large amounts of data for training will significantly increase the initial sampling costs. In the sewage treatment process, direct supply and direct discharge are often adopted. The regularity of various substances in the sewage is difficult to capture during the flow process, and the differences in the uniformity of the sewage are retained, which will further increase the prediction error. Finally, in the sewage treatment process, different depths and the flow of sewage will affect the water inlet indicators. If a single model is used, it will be impossible to accurately predict the water quality. Summary of the Invention

[0005] Based on the problems of low efficiency and high cost in the prior art of directly obtaining data through a large number of sampling tests for training models, as well as the large differences in sewage uniformity caused by direct supply and discharge, which affects the accuracy of the prediction structure, and the inlet flow rate and other indicators, which affect the inlet indicators and lead to the inability of a single model to accurately predict, the present invention provides a water indicator prediction method based on sewage treatment.

[0006] The technical solution adopted by the present invention to solve the above technical problems is:

[0007] Based on the water index prediction method of sewage treatment, sewage treatment includes the following steps in sequence:

[0008] S1, first impurity removal: the inlet water passes through a coarse screen to remove large particles of impurities;

[0009] S2, secondary impurity removal: passing the water through a fine screen to remove small particles of impurities;

[0010] S3, aeration and grit settling: water is passed into an aeration and grit settling tank to remove sand and inorganic impurities in the water;

[0011] S4, enhanced denitrification: decomposition of nitrogen and phosphorus in water by microorganisms;

[0012] S5, chemical purification: reducing phosphorus in water by chemical methods;

[0013] S6, filter tank filtration: filter out tiny particles and chemical sludge in the water;

[0014] S7, disinfection: disinfect water by ultraviolet light and sodium hypochlorite;

[0015] S8, emission;

[0016] Before step S1, the method also includes a preceding step S0: setting at least two identical water supply tanks for alternate water supply, homogenizing and stirring the sewage before injecting it into the water supply tank, and allowing it to stand for a fixed time after injection, dividing the sewage into n depth range segment sequences in the depth direction of the water supply tank, each depth range segment sequence has the same water volume and a depth of D, and the water supply time of each range segment is T, measuring the water inflow index at the top, middle and bottom of each depth range segment sequence, and constructing multiple logarithmic functions corresponding to the depth range segment sequences: , where t is the total water supply time, Y is the predicted water inflow index when the water supply time is t, and the values ​​of a, b and c are obtained by measuring the water inflow index at the top, middle and bottom of each depth range segment sequence respectively. When calculating, the corresponding logarithmic function is selected according to the depth range segment sequence to which the water supply belongs to calculate the water inflow index;

[0017] The water inflow index is measured for each depth range sequence, and then the data is input into a logarithmic function to calculate multiple logarithmic functions. When obtaining the water inflow index for calculation, the corresponding logarithmic function is selected according to the depth range sequence to which the water supply belongs to calculate the water inflow index for prediction; before calculating the water inflow index for prediction, the following steps are performed to reduce the impact of flow on the water inflow index: 1. Each depth range sequence of the water supply tank is divided into several grid areas; 2. The drain outlet is located at a fixed position in the tank; 3. A function of the discharge volume of each grid area and the distance and time from the grid area to the drain outlet is established.

[0018] ; Q(x,y,t) is the discharge volume at position (x, y) and time t, C is the drainage velocity coefficient, g is the acceleration of gravity, h(x,y,t) is the water depth at position (x, y) and time t, Dhj(x,y) is the mapped distance function from position (x, y) to the outlet, Used to express the effect of time on water discharge, ,in( , ) is the coordinate of the drain outlet, is the regularization constant and is calculated by assignment to ensure that the discharge function is close to the data collection position and the outlet position limit The feasibility of calculation, is the time attenuation coefficient, and the coordinates of the center position of each grid area are used to calculate the displacement Calculate the drainage volume for each grid area over multiple time periods The accumulated values ​​are accumulated according to the time periods and the accumulated values ​​are made equal to the total drainage volume of each grid area at the time to optimize the parameters of drainage volume. At the same time, the water inflow index of the top, middle and bottom positions of the center of each grid area is measured and the arithmetic mean of the water inflow index is obtained. The total drainage volume at time t is obtained according to the flow rate and the proportion of drainage volume in each grid area is calculated. The arithmetic mean of the water inflow index of each grid area is multiplied by the proportion of drainage volume of the corresponding grid area and accumulated to obtain the adjustment value of the water inflow index, and then the adjustment value is input into the logarithmic function. The final value used for calculation is obtained by calculating

[0019] Preferably, the effluent indicators are predicted based on the above steps according to the following method, wherein the inlet and outlet indicators include pH value, COD, ammonia nitrogen content, total phosphorus content, and total nitrogen content:

[0020] A. Establish a linear regression model, set the inlet pH as the independent variable and the outlet pH as the dependent variable, collect inlet and outlet pH data and obtain the regression coefficient of the linear regression model, and then predict the pH of the later outlet water based on the linear regression model;

[0021] B. Establish an exponential decay model, collect COD data of the influent and effluent in steps S4-S7 and calculate the removal rate of each step, and then calculate the effluent COD based on the exponential decay model;

[0022] C. Establish a neural network model to simulate the water treatment process, collect ammonia nitrogen content of some influent and effluent water to train the neural network model, and then calculate the ammonia nitrogen content of the effluent based on the neural network model;

[0023] D. Based on the grey model, the total phosphorus content is predicted by cumulative generation sequence and grey differential equation;

[0024] E. Based on the support vector regression model, find the optimal regression hyperplane to predict the total nitrogen content; when calculating the above five water effluent indicators, first substitute the collected data into the above logarithmic function to obtain at least five logarithmic functions;

[0025] After step S8, the preceding step S9 is also included: establishing at least two drainage pools for alternating drainage, dividing the drainage pool into n depth range segment sequences, draining the water at the bottom of each depth range segment sequence respectively, and making the drainage time point t2 correspond to the water supply time point t based on the time t1 required for water treatment, and then converting the measured water inlet index at time t into the corresponding final value through the above-mentioned logarithmic function, and then bringing the final value into the algorithm of step AE to predict the five water outlet indexes respectively, and calibrating the predicted value as the water outlet index at time point t2, where t2=t1+t.

[0026] As an optimization, the calculation formula of the linear regression model is: ,in is the pH value of the effluent, is the influent pH value, b is a constant term, and when the model variable is zero, the constant term is the predicted value. The values ​​of a and b are calculated using the data from steps A and B and the least squares method.

[0027] As an option, calculate the 95% confidence intervals for the coefficient a and the constant term b, and calculate the residual sum of squares. , calculate the error variance estimate , , calculate the standard error SE of a and b, ,SE(b)= , calculate the confidence intervals of a and b respectively , , calculate the range of a single new observation Y1, where t0.025,n-2 represents the critical value of the t distribution with n-2 degrees of freedom; is the sample mean of the influent pH value, and n is the total number of pH values.

[0028] As a preferred method, the calculation formula of the exponential decay model is: ,in is the removal rate in steps S4-S7.

[0029] As a preference, the total phosphorus content is predicted by cumulatively generating the sequence AGF and the ash differential equation, and the prediction is made based on the ash model. ， The gray model formula is: .

[0030] As a preferred method, the specific steps are as follows: collect total phosphorus content data to form an original data sequence Y (0); Calculate the cumulative generation sequence Y (1); Calculate the background value sequence z (1); Construct data matrix B and data vector Y ; Solve the parameters of the grey differential equation by the least squares methoda and b ; Use the obtained parameters to build a prediction model; obtain the predicted value of the original sequence through cumulative generation.

[0031] As a preferred method, the calculation formula of the neural network model is: , the water treatment process is simulated by the influent ammonia nitrogen and effluent ammonia nitrogen, where the activation function For nonlinear mapping, the weight wi and bias b are optimized by inputting the inlet ammonia nitrogen and outlet ammonia nitrogen data. The model is one of Sigmoid, ReLU or Tanh.

[0032] Preferably, the method includes the following steps: setting an input layer, the input layer is the influent ammonia nitrogen concentration; setting a hidden layer, the hidden layer is an abstract representation of the treatment process, including aeration time, agent dosage, microbial activity, and mapping nonlinear relationships through weights and activation functions; setting an output layer, the output layer is the effluent ammonia nitrogen concentration Y.

[0033] As a preference, weights and biases are optimized. The optimization steps are as follows: Define the loss function Loss= , where N is the number of training samples; is the model prediction value; is the measured ammonia nitrogen concentration; the gradient is calculated by the chain rule, and the weight wi and bias b are updated. η , b b , where η is the learning rate, which is used to control the parameter update step size.

[0034] Compared with the existing technology, the advantages of the present invention are as follows: the present application first supplies water after standing still and performs layered modeling, and then combines the vertical distribution characteristics of the fluid after standing still to accurately reflect the dynamic changes of water quality indicators. Based on the law that the concentration of pollutants decays with time during precipitation and diffusion, a logarithmic function is constructed to predict the water inlet index in the depth direction to reduce the number of sampling times. Through grid dynamic correction, the arithmetic mean is used to weight the drainage volume ratio, which reduces the sampling deviation caused by flow disturbance and improves the robustness of the prediction. The above scheme can ensure that the sewage quality is relatively uniform, and then the drainage volume ratio is weighted by the average value of the water inlet index of each area to obtain the optimized water inlet index. The optimized water inlet index is then calculated through the constructed logarithmic function to obtain the water inlet index of different depths, and the water inlet index of different depths is input into the five groups of models set for calculation and prediction to obtain the water outlet index. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] The present invention will be described in further detail below with reference to the accompanying drawings and preferred embodiments. However, those skilled in the art will appreciate that these drawings are drawn only for the purpose of explaining the preferred embodiments and should not be construed as limiting the scope of the present invention. Furthermore, unless otherwise specified, the drawings are merely schematic representations of the composition or structure of the depicted objects and may contain exaggerated representations. Furthermore, the drawings are not necessarily drawn to scale.

[0036] Figure 1 A flowchart of the steps of this application;

[0037] Figure 2-8 This is the sewage treatment monitoring interface display diagram;

[0038] Figure 9 Schematic diagram of the depth range segment sequence and the divided grid area; DETAILED DESCRIPTION

[0039] The preferred embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Those skilled in the art will appreciate that these descriptions are merely illustrative and exemplary and should not be construed as limiting the scope of protection of the present invention. Example

[0040] Water index prediction methods based on sewage treatment, such as Figure 2-8 As shown, sewage treatment includes the following steps in sequence:

[0041] S1, first impurity removal: the inlet water passes through a coarse screen to remove large particles of impurities;

[0042] S2, secondary impurity removal: passing the water through a fine screen to remove small particles of impurities;

[0043] S3, aeration and grit settling: water is passed into an aeration and grit settling tank to remove sand and inorganic impurities in the water;

[0044] S4, enhanced denitrification: decomposition of nitrogen and phosphorus in water by microorganisms;

[0045] S5, chemical purification: reducing phosphorus in water by chemical methods;

[0046] S6, filter tank filtration: filter out tiny particles and chemical sludge in the water;

[0047] S7, disinfection: disinfect water by ultraviolet light and sodium hypochlorite;

[0048] S8, emission;

[0049] Before step S1, the method also includes a preceding step S0: setting at least two identical water supply tanks for alternate water supply, homogenizing and stirring the sewage before injecting it into the water supply tank, and allowing it to stand for a fixed time after injection, dividing the sewage into n depth range segment sequences in the depth direction of the water supply tank, each depth range segment sequence has the same water volume and a depth of D, and the water supply time of each range segment is T, measuring the water inflow index at the top, middle and bottom of each depth range segment sequence, and constructing multiple logarithmic functions corresponding to the depth range segment sequences: , where t is the total water supply time, Y is the predicted water inflow index when the water supply time is t, and the values ​​of a, b and c are obtained by measuring the water inflow index at the top, middle and bottom of each depth range segment sequence respectively. When calculating, the corresponding logarithmic function is selected according to the depth range segment sequence to which the water supply belongs to calculate the water inflow index;

[0050] like Figure 1 As shown, the water inflow index is measured for each depth range segment sequence, and then the data is input into the logarithmic function to calculate multiple logarithmic functions. When obtaining the water inflow index for calculation, the corresponding logarithmic function is selected according to the depth range segment sequence to which the water supply belongs to calculate the water inflow index for prediction; before calculating the water inflow index for prediction, the following steps are performed to reduce the impact of flow on the water inflow index: 1. Each depth range segment sequence of the water supply tank is divided into several grid areas; 2. The drain outlet is located at a fixed position in the tank; 3. A function of the discharge volume of each grid area and the distance and time from the grid area to the drain outlet is established.

[0051] ; Q(x,y,t) is the discharge volume at position (x, y) and time t, C is the drainage velocity coefficient, g is the acceleration of gravity, h(x,y,t) is the water depth at position (x, y) and time t, Dhj(x,y) is the mapped distance function from position (x, y) to the outlet, Used to express the effect of time on water discharge, ,in( , ) is the coordinate of the drain outlet, is the regularization constant and is calculated by assignment to ensure that the discharge function is close to the data collection position and the outlet position limit The feasibility of calculation, is the time attenuation coefficient, and the coordinates of the center position of each grid area are used to calculate the displacement Calculate the drainage volume for each grid area over multiple time periods The accumulated values ​​are accumulated according to the time periods and the accumulated values ​​are made equal to the total drainage volume of each grid area at the time to optimize the parameters of drainage volume. At the same time, the water inflow index of the top, middle and bottom positions of the center of each grid area is measured and the arithmetic mean of the water inflow index is obtained. The total drainage volume at time t is obtained according to the flow rate and the proportion of drainage volume in each grid area is calculated. The arithmetic mean of the water inflow index of each grid area is multiplied by the proportion of drainage volume of the corresponding grid area and accumulated to obtain the adjustment value of the water inflow index, and then the adjustment value is input into the logarithmic function. The final value used for calculation is obtained by calculating in . The depth is segmented by equal thickness, the water supply tank is a rectangular tank, and the grid area is divided into equal areas.

[0052] Preferably, the effluent indicators are predicted based on the above steps according to the following method, wherein the inlet and outlet indicators include pH value, COD, ammonia nitrogen content, total phosphorus content, and total nitrogen content:

[0053] A. Establish a linear regression model, set the inlet pH as the independent variable and the outlet pH as the dependent variable, collect inlet and outlet pH data and obtain the regression coefficient of the linear regression model, and then predict the pH of the later outlet water based on the linear regression model;

[0054] B. Establish an exponential decay model, collect COD data of the influent and effluent in steps S4-S7 and calculate the removal rate of each step, and then calculate the effluent COD based on the exponential decay model;

[0055] C. Establish a neural network model to simulate the water treatment process, collect ammonia nitrogen content of some influent and effluent water to train the neural network model, and then calculate the ammonia nitrogen content of the effluent based on the neural network model;

[0056] D. Based on the grey model, the total phosphorus content is predicted by cumulative generation sequence and grey differential equation;

[0057] E. Based on the support vector regression model, find the optimal regression hyperplane to predict the total nitrogen content; when calculating the above five water effluent indicators, first substitute the collected data into the above logarithmic function to obtain at least five logarithmic functions;

[0058] After step S8, the preceding step S9 is also included: establishing at least two drainage pools for alternating drainage, dividing the drainage pool into n depth range segment sequences, draining the water at the bottom of each depth range segment sequence respectively, and making the drainage time point t2 correspond to the water supply time point t based on the time t1 required for water treatment, and then converting the measured water inlet index at time t into the corresponding final value through the above-mentioned logarithmic function, and then bringing the final value into the algorithm of step AE to predict the five water outlet indexes respectively, and calibrating the predicted value as the water outlet index at time point t2, where t2=t1+t.

[0059] As an optimization, the calculation formula of the linear regression model is: ,in is the pH value of the effluent, is the influent pH value, b is a constant term, and when the model variable is zero, the constant term is the predicted value. The values ​​of a and b are calculated using the data from steps A and B and the least squares method.

[0060] As an option, calculate the 95% confidence intervals for the coefficient a and the constant term b, and calculate the residual sum of squares. , calculate the error variance estimate , , calculate the standard error SE of a and b, ,SE(b)= , calculate the confidence intervals of a and b respectively , , calculate the range of a single new observation Y1, where t0.025,n-2 represents the critical value of the t distribution with n-2 degrees of freedom; is the sample mean of the influent pH value, and n is the total number of pH values.

[0061] As a preferred method, the calculation formula of the exponential decay model is: ,in is the removal rate in steps S4-S7.

[0062] As a preference, the total phosphorus content is predicted by cumulatively generating the sequence AGF and the ash differential equation, and the prediction is made based on the ash model. ， The gray model formula is: .

[0063] As a preferred method, the specific steps are as follows: collect total phosphorus content data to form an original data sequence Y (0); Calculate the cumulative generation sequence Y (1); Calculate the background value sequence z (1); Construct data matrix B and data vector Y ; Solve the parameters of the grey differential equation by the least squares method a and b ; Use the obtained parameters to build a prediction model; and obtain the predicted value of the original sequence through cumulative generation. Where Y(0)={Y(0)(1),}

[0064] As a preferred method, the calculation formula of the neural network model is: , the water treatment process is simulated by the influent ammonia nitrogen and effluent ammonia nitrogen, where the activation function For nonlinear mapping, the weight wi and bias b are optimized by inputting the inlet ammonia nitrogen and outlet ammonia nitrogen data. The model is one of Sigmoid, ReLU or Tanh.

[0065] Preferably, the method includes the following steps: setting an input layer, the input layer is the influent ammonia nitrogen concentration; setting a hidden layer, the hidden layer is an abstract representation of the treatment process, including aeration time, agent dosage, microbial activity, and mapping nonlinear relationships through weights and activation functions; setting an output layer, the output layer is the effluent ammonia nitrogen concentration Y.

[0066] As a preference, weights and biases are optimized. The optimization steps are as follows: Define the loss function Loss= , where N is the number of training samples; is the model prediction value; is the measured ammonia nitrogen concentration; the gradient is calculated by the chain rule, and the weight wi and bias b are updated. η , b b , where η is the learning rate, which is used to control the parameter update step size.

[0067] like Figure 9 As shown, a depth range segment sequence is established in the depth direction, and the grid area is divided in the horizontal direction. A drain outlet is set in each depth range segment sequence, and the water supply from the upper and lower layers is controlled by an electric control valve and a liquid level meter.

[0068] It should be noted that when collecting data for the water supply pool, 1. It is necessary to collect the water inlet index of the center position of the grid in different areas and the actual water inlet index of the outlet of the water supply pool to calculate the logarithmic function; 2. Calculate the drainage volume of the grid area and calculate the total drainage volume to adjust the parameters of the drainage volume corresponding formula; 3. Collect the drainage index of the drainage pool. This drainage index refers to the index discharged from the sewage treatment into the drainage pool or discharged from the drainage pool. Then, bring the actual water inlet index of the outlet of the water supply pool and the drainage index of the drainage pool into the corresponding model for calculation; 4. When it comes to the removal rate, it is necessary to gradually measure the sewage index to determine the removal rate in the sewage treatment process and then bring the removal rate into the corresponding model for calculation; Finally, when making a prediction, it is only necessary to collect the water inlet index of the center position of the grid in different areas, and the remaining parameters are calculated using the formulas of each model to complete the prediction. The water inlet index used for calculation is obtained by performing two calculations, and the water inlet index is input into the five model formulas to calculate the water outlet index.

[0069] The above describes the water index prediction method based on sewage treatment provided by the present invention. Specific examples are used herein to illustrate the principles and implementation methods of the present invention. The above examples are merely intended to facilitate understanding of the present invention and its core concepts. It should be noted that those skilled in the art may make various improvements and modifications to the present invention without departing from the principles of the present invention, and such improvements and modifications fall within the scope of protection of the claims of the present invention.

Claims

1. Based on the water index prediction method of sewage treatment, sewage treatment includes the following steps in sequence: S1, first impurity removal: the inlet water passes through a coarse screen to remove large particles of impurities; S2, secondary impurity removal: passing water through a fine screen to remove small particles of impurities; S3, aeration and grit settling: water is passed into an aeration and grit settling tank to remove sand and inorganic impurities in the water; S4, enhanced denitrification: decomposition of nitrogen and phosphorus in water by microorganisms; S5, chemical purification: reducing phosphorus in water by chemical methods; S6, filter tank filtration: filter out tiny particles and chemical sludge in the water; S7, disinfection: disinfect water by ultraviolet light and sodium hypochlorite; S8, emission; Its characteristics are: Before step S1, the method also includes a preceding step S0: setting at least two identical water supply tanks for alternate water supply, homogenizing and stirring the sewage before injecting it into the water supply tank, and allowing it to stand for a fixed time after injection, dividing the sewage into n depth range segment sequences in the depth direction of the water supply tank, each depth range segment sequence has the same water volume and a depth of D, and the water supply time of each range segment is T, measuring the water inflow index at the top, middle and bottom of each depth range segment sequence, and constructing multiple logarithmic functions corresponding to the depth range segment sequences: , where t is the total water supply time, Y is the water inflow index predicted when the water supply time is t, and when calculating, the corresponding logarithmic function is selected according to the sequence of the depth range segment to which the water supply belongs to calculate the water inflow index; The water inflow index is measured for each depth range sequence, and then the data is input into a logarithmic function to calculate multiple logarithmic functions. When obtaining the water inflow index for calculation, the corresponding logarithmic function is selected according to the depth range sequence to which the water supply belongs to calculate the water inflow index for prediction; before calculating the water inflow index for prediction, the following steps are performed to reduce the impact of flow on the water inflow index:

1. Each depth range sequence of the water supply tank is divided into several grid areas; 2. The drain outlet is located at a fixed position in the tank; 3. A function of the discharge volume of each grid area and the distance and time from the grid area to the drain outlet is established. ; Q(x,y,t) is the discharge volume at position (x, y) and time t, C is the drainage velocity coefficient, g is the acceleration of gravity, h(x,y,t) is the water depth at position (x, y) and time t, Dhj(x,y) is the mapped distance function from position (x, y) to the outlet, Used to express the effect of time on water discharge, ,in( , ) is the coordinate of the drain outlet, is the regularization constant and is calculated by assignment to ensure that the discharge function is close to the data collection position and the drain outlet position limit. The feasibility of calculation, is the time attenuation coefficient, and the coordinates of the center position of each grid area are used to calculate the displacement Calculate the drainage volume for each grid area over multiple time periods The accumulated values ​​are accumulated according to the time periods and the accumulated values ​​are made equal to the total drainage volume of each grid area at the time to optimize the parameters of drainage volume. At the same time, the water inflow index of the top, middle and bottom positions of the center of each grid area is measured and the arithmetic mean of the water inflow index is obtained. The total drainage volume at time t is obtained according to the flow rate and the proportion of drainage volume in each grid area is calculated. The arithmetic mean of the water inflow index of each grid area is multiplied by the proportion of drainage volume of the corresponding grid area and accumulated to obtain the adjustment value of the water inflow index, and then the adjustment value is input into the logarithmic function. The final value used for calculation is obtained by calculating the values ​​of a, b and c. The values ​​of a, b and c are obtained by the following method: measure the water inlet index of the drain outlet and set it = Y, set the adjustment value = , and calculate the values ​​of a, b, and c through multiple sets of data.

2. The water index prediction method based on sewage treatment according to claim 1 is characterized in that: Based on the above steps, the effluent indicators are predicted according to the following method, where the inlet and outlet indicators include pH value, COD, ammonia nitrogen content, total phosphorus content, and total nitrogen content: A. Establish a linear regression model, set the inlet pH as the independent variable and the outlet pH as the dependent variable, collect inlet and outlet pH data and obtain the regression coefficient of the linear regression model, and then predict the pH of the later outlet water based on the linear regression model; B. Establish an exponential decay model, collect COD data of the influent and effluent in steps S4-S7 and calculate the removal rate of each step, and then calculate the effluent COD based on the exponential decay model; C. Establish a neural network model to simulate the water treatment process, collect ammonia nitrogen content of some influent and effluent water to train the neural network model, and then calculate the ammonia nitrogen content of the effluent based on the neural network model; D. Based on the grey model, the total phosphorus content is predicted by cumulative generation sequence and grey differential equation; E. Based on the support vector regression model, find the optimal regression hyperplane to predict the total nitrogen content; when calculating the above five water effluent indicators, first substitute the collected data into the above logarithmic function to obtain at least five logarithmic functions; After step S8, the preceding step S9 is also included: establishing at least two drainage pools for alternating drainage, dividing the drainage pool into n depth range segment sequences, draining the water at the bottom of each depth range segment sequence respectively, and making the drainage time point t2 correspond to the water supply time point t based on the time t1 required for water treatment, and then converting the measured water inlet index at time t into the corresponding final value through the above-mentioned logarithmic function, and then bringing the final value into the algorithm of step AE to predict the five water outlet indexes respectively, and calibrating the predicted value as the water outlet index at time point t2, where t2=t1+t.

3. The water index prediction method based on sewage treatment according to claim 2 is characterized in that: The calculation formula of the linear regression model is: ,in is the pH value of the effluent, is the influent pH value, b is a constant term, and when the model variable is zero, the constant term is the predicted value. The values ​​of a and b are calculated using the data from steps A and B and the least squares method.

4. The water index prediction method based on sewage treatment according to claim 3 is characterized in that: Calculate the 95% confidence interval of coefficient a and constant term b, calculate the residual sum of squares, , calculate the error variance estimate, , calculate the standard error SE of a and b, ,SE(b)= , calculate the confidence intervals of a and b respectively , , calculate the range of a single new observation Y1, where t0.025,n-2 represents the critical value of the t distribution with n-2 degrees of freedom; is the sample mean of the influent pH value, and n is the total number of pH values.

5. The water index prediction method based on sewage treatment according to claim 2 is characterized in that: The calculation formula of the exponential decay model is: ,in is the removal rate in steps S4-S7.

6. The water index prediction method based on sewage treatment according to claim 2, characterized in that: The total phosphorus content is predicted by accumulating the generated sequence AGF and the gray differential equation. The prediction is based on the gray model. The gray model formula is: .

7. The water index prediction method based on sewage treatment according to claim 6, characterized in that: The specific steps are as follows: Collect total phosphorus content data to form the original data sequence Calculate and accumulate generated sequences ; Calculate background value sequence ; Construct data matrix B and data vector Y; solve parameters a and b in the grey differential equation by least squares method; use the obtained parameters to build a prediction model; obtain the predicted value of the original sequence by cumulative subtraction.

8. The water index prediction method based on sewage treatment according to claim 2 is characterized in that: The calculation formula of the neural network model is: , the water treatment process is simulated by the influent ammonia nitrogen and effluent ammonia nitrogen, where the activation function For nonlinear mapping, the weight wi and bias b are optimized by inputting the inlet ammonia nitrogen and outlet ammonia nitrogen data. The model is one of Sigmoid, ReLU or Tanh.

9. The water index prediction method based on sewage treatment according to claim 8, characterized in that: The following steps are involved: Set up an input layer, which is the influent ammonia nitrogen concentration; set up a hidden layer, which is an abstract representation of the treatment process, including aeration time, agent dosage, and microbial activity, and maps nonlinear relationships through weights and activation functions; set up an output layer, which is the effluent ammonia nitrogen concentration Y.

10. The water index prediction method based on sewage treatment according to claim 8, characterized in that: Optimize the weights and biases. The optimization steps are as follows: Define the loss function Loss= , where N is the number of training samples; is the model prediction value; is the measured ammonia nitrogen concentration; the gradient is calculated by the chain rule, and the weight wi and bias b are updated. η , b b , where η is the learning rate, which is used to control the parameter update step size.

Images