Ozone Dosing Control Method Based on Real-time Ultraviolet Spectroscopy Monitoring

Abstract

This invention relates to the field of wastewater treatment technology, providing a method for ozone dosing control based on real-time ultraviolet spectroscopy monitoring. The method includes a step S of determining whether ozone needs to be added and a step P of quantitatively controlling ozone dosing. In step S, water samples are continuously acquired, and absorbance data is obtained using an ultraviolet-visible full-wavelength scanner to establish a dataset. LDA linear discriminant analysis is then performed to determine the ozone dosing discrimination formula. Step P involves continuously monitoring water samples, screening data, determining ultraviolet absorption spectral values, calculating the total ozone dosage, and constructing a dataset E for multiple linear regression analysis to establish the quantitative control formula for ozone dosing. This invention achieves precise control of ozone dosing in secondary treated effluent through ultraviolet spectral data and chemometric methods, optimizing ozone use, saving energy, improving wastewater treatment efficiency, and reducing operating costs.

Description

Technical Field

[0001] This invention relates to the field of wastewater treatment technology, specifically to an ozone dosing control method based on real-time ultraviolet spectroscopy monitoring. Background Technology

[0002] Industrial wastewater contains complex, recalcitrant organic compounds that are difficult to remove effectively using traditional biological treatment methods. These organic compounds significantly contribute to the chemical oxygen demand (COD) in the effluent from biological treatment processes; COD is a key indicator for measuring the degree of organic pollution in water bodies.

[0003] To address this issue, ozone oxidation technology, as an advanced oxidation process, is widely used. Ozone oxidation is not only simple to operate but also does not produce secondary pollution, thus playing a crucial role in the advanced treatment processes of wastewater treatment plants. Ozone oxidizes recalcitrant organic matter in two ways: firstly, by directly reacting with the organic matter; and secondly, by indirectly oxidizing the organic matter through hydroxyl radicals generated during its decomposition in water.

[0004] Although ozone oxidation technology is effective, it faces energy consumption and efficiency issues in practical operation. This is particularly relevant when pollutant concentrations are 5-10 g DOC / m³. 3 At this time, an effective ozone dosage of 0.8 grams is required per gram of COD, and the energy consumption is approximately 0.06-0.13 kWh / m³. 3 Ozone accounts for 20-40% of the total energy consumption in wastewater treatment. Existing ozone dosing systems often over-dosing, which not only results in poor performance but also significantly increases the power consumption of the ozone generator, becoming a major energy cost in wastewater treatment processes.

[0005] Most organic pollutants exhibit strong absorption characteristics in the ultraviolet (UV) region, and UV absorbance can serve as an indicator for detecting the concentration of organic pollutants in water bodies. By discovering and establishing the relationship between UV spectral data and ozone dosing control, the ozone dosing control process can be optimized. This helps reduce unnecessary ozone consumption, optimize ozone use, thereby saving energy and improving the overall efficiency of wastewater treatment. It ensures that wastewater treatment meets environmental standards while reducing operating costs, achieving both economic and environmental benefits for the wastewater treatment process. Summary of the Invention

[0006] To address the problems existing in the background technology, this invention provides an ozone dosing control method based on real-time ultraviolet spectroscopy monitoring, which includes a determination step S of whether ozone needs to be added and a quantitative control step P of ozone dosing, as detailed below:

[0007] The specific process of step S includes:

[0008] S1. Continuously acquire water samples from the ozone oxidation deep treatment system;

[0009] S2. Select several samples and use a UV-Vis full-wavelength scanner to obtain the absorbance of organic matter in the ozone oxidation deep treatment system in the UV range.

[0010] S3. Plot the ultraviolet absorption spectrum with wavelength as the abscissa and absorbance as the ordinate, and determine the absorption peaks.

[0011] S4. Extract the absorption peak wavelengths and corresponding absorbance values ​​to create dataset A. Preprocess the extracted data to ensure that the number of variables is much smaller than the number of data groups (i.e., the number of columns is much smaller than the number of rows) and that each data group has an absorption peak. Create dataset B. Each row of dataset B represents a different observation value, and each column represents a spectral variable, i.e., the absorbance value at a specific wavelength.

[0012] S5. Perform COD determination on all samples;

[0013] S6. Construct a new dataset and perform LDA linear discriminant analysis to determine the ozone dosing discrimination formula;

[0014] The specific process of step P includes:

[0015] P1. Continuously acquire water samples from the ozone oxidation deep treatment system and determine the COD value;

[0016] P2. Select several samples, set filtering conditions, filter the COD data, and create dataset B';

[0017] P3. Determine the ultraviolet absorption spectrum value corresponding to the screened COD data;

[0018] P4. Determine the total ozone dosage based on the operating conditions of the ozone oxidation deep treatment system.

[0019] P5. Add a "Total Ozone Dosage" variable to dataset B' to construct dataset E;

[0020] P6. Perform multiple linear regression analysis on dataset E to determine the quantitative control formula for ozone dosing.

[0021] In the preferred embodiment, in step S4, the preprocessing procedure is to delete wavelengths whose absorption peaks occur 12 times or less in all observations.

[0022] In the preferred embodiment, in step S3, the absorption peak is determined from the spectrum. The absorption peak is defined as the place with the highest absorbance on the curve. The peak finding is set to the five maximum values ​​around the peak, and the peak finding baseline is set to y = 0.

[0023] In the preferred embodiment, step S6 specifically includes:

[0024] S61: Determine the training set data: Add a column "Ozone Dosing Switch" variable to dataset B to obtain dataset C; the "Ozone Dosing Switch" variable uses numbers to represent whether ozone needs to be added. According to the switch control principle, "1" represents yes and "0" represents no. "1" and "0" are distinguished according to the COD value. For samples with COD value > 27 mg / L, the "Ozone Dosing Switch" is assigned a value of "1", which means that ozone needs to be added; for samples with COD value ≤ 27 mg / L, the "Ozone Dosing Switch" is assigned a value of "0", which means that ozone does not need to be added; select several groups from dataset C to build a model. The ratio of "1" to "0" in the several groups of data is (8~10):(4~6).

[0025] S62: Performing LDA is linear discriminant analysis: Let the entire sample dataset be Z = {(x1,y1),(x2,y2),…,(x...}. N ,y N )}, where x i Let y be a d-dimensional vector, representing the number of samples and the features of the samples. i Let N be the class label for the corresponding sample, and N be the number of samples. Assuming there are K classes in total, the dataset for each class can be represented as follows:

[0026] This dataset x i y is a 10-dimensional vector. i Here, the value is either 0 or 1, resulting in two types of samples.

[0027] The goal of LDA is to map the original high-dimensional samples to have the minimum intra-class distance d in the low-dimensional feature space. w and the maximum inter-class distance d b d w and d b It is expressed as follows:

[0028]

[0029] In the formula, Let be the basis vector matrix in the low-dimensional feature space, m be the dimension of the low-dimensional space, and m < d; combining the above two objectives, the following optimization objective can be derived as the loss function of LDA:

[0030]

[0031] Define the within-class scatter matrix S respectively w and the inter-class scatter matrix S b As shown in the following formula:

[0032]

[0033] The loss function can then be transformed into:

[0034]

[0035] Since the solution of the loss function depends only on the direction of W and not on its magnitude, let W... T When W=1, the loss function becomes a constrained form, i.e.:

[0036]

[0037] The solution can be obtained using the Lagrange multiplier method; the solution to W is a matrix. If x is the matrix formed by the eigenvectors corresponding to the first m largest eigenvalues, then x i The projection, i.e., the dimensionality-reduced sample features, can be represented as x. i =W T x i The specific process is as follows:

[0038] S621: Calculate the mean vector of all samples. and the mean vector of each class of samples

[0039] S622: Calculate the intra-class scatter matrix S w and the inter-class scatter matrix S b ;

[0040] S623: Find the matrix eigenvalues ​​and eigenvectors;

[0041] S624: Take the eigenvectors corresponding to the first m largest eigenvalues ​​to obtain the projection matrix ∈W;

[0042] S625: Calculate the dimensionality-reduced sample features x i .

[0043] In the preferred embodiment, the ozone addition discrimination formula determined in step S6 is:

[0044] y = b + k1A 228 +k2A 229.5 +k3A 237 -k4A 253.5 -k5A 254 +k6A 255.5 -k7A 266 -k8A 272.5 -k9A 362.5 -k 10 A 371

[0045] In the formula A xxx , representing the absorbance value at a wavelength of XXX nm, k iRepresent the fitting parameters (i = 1, 2, …, 10); y has no unit. When y ≥ 0.5, it is considered that y = 1, that is, ozone needs to be added; when Y < 0.5, it is considered that y = 0, that is, ozone does not need to be added.

[0046] The finally established formula is as follows:

[0047]

[0048] In the preferred solution, in step P2, the screening conditions are that the influent COD of the ozone oxidation advanced treatment process > 30 mg / L, and the effluent COD of the process is 20 - 30 mg / L, so as to meet the effluent standard and not overdose ozone, resulting in too low COD value.

[0049] In the preferred solution, in step P4, 8.7 g of ozone is added per cubic meter of sewage during the operation of the ozone oxidation advanced treatment system, the effluent flow rate of the ozone oxidation advanced treatment system is measured, and the total ozone dosage per unit time can be obtained by multiplying the ozone dosage by the effluent flow rate.

[0050] In the preferred solution, in step P6, the specific process of performing multiple linear regression analysis on the data set E includes:

[0051] Given an example X = (x1, x2, x3, …, x d ), where x i are all the values of x on the i-th attribute. The linear model attempts to obtain a function for prediction through a linear combination of attributes, that is

[0052]

[0053] In vector form, it is expressed as:

[0054]

[0055] where

[0056] Since intuitively expresses the importance of each attribute in prediction, the linear model has good interpretability; determining and b can determine the formula;

[0057] We attempt to obtain

[0058]

[0059] Its matrix expression and its expansion are

[0060]

[0061] y iRepresents the predicted value; in ON / OFF control, it represents the value of the 0 / 1 column; in ozone dosage prediction, it represents the corresponding ozone dosage. ij This represents the various factors affecting the load, which in this paper are the spectral absorption peak variables, β0 represents the constant term, and β i (i = 1, 2, ..., n) represents the regression coefficient, b' i Indicates random perturbation;

[0062] The above formula can be simplified as follows:

[0063] Y = Xβ + ε (2.5)

[0064] In the formula: Y is the multivariate load matrix, v is the random error matrix, X is the influencing factor matrix, and β is the regression coefficient matrix;

[0065] The regression parameters are estimated using the least squares method, and the regression function is obtained, which is the prediction model; its formula is:

[0066]

[0067] In the preferred scheme, the quantitative control formula for ozone addition determined in step P6 is as follows:

[0068]

[0069] Based on the results of the multivariate regression analysis, the equation for the linear LDA analysis model is obtained, where A... xxx , representing the absorbance value at a wavelength of XXXnm, k i The fitting parameters of the equation are represented by y (i = 1, 2, ..., 10); y represents the ozone dosage flow rate (g / h).

[0070] The formula includes 10 variables. Under the daily operating condition of ozone dosage of 8.7 g / m³, the corresponding ozone dosage (g / h) is calculated based on the corresponding wastewater flow rate. Multiple linear regression analysis is then performed on the ozone dosage and the selected spectral absorbance variables to establish the formula. The final established formula is as follows:

[0071]

[0072] The beneficial effects achieved by this invention are as follows:

[0073] The ozone dosing control method proposed in this invention, based on real-time ultraviolet spectroscopy monitoring, establishes a relationship between ultraviolet spectral data and ozone dosing control. This helps to precisely control the ozone dosage, ensuring its effective utilization in wastewater treatment. This precise control not only helps avoid overdosing but also significantly saves energy and reduces the ineffective operating time of the ozone dosing system in the daily operation of wastewater treatment plants, thereby further improving wastewater treatment efficiency. This not only enhances treatment efficiency but also helps ensure that effluent quality meets predetermined standards, while simultaneously reducing the overall operating costs of wastewater treatment.

[0074] In application, only the absorbance value at a specific wavelength needs to be extracted, and then the method according to this invention can determine whether ozone needs to be added and the corresponding ozone dosage. This invention achieves precise control of ozone addition to secondary treated effluent through a combination of ultraviolet-visible spectroscopy and chemometrics, laying the foundation for the application of ozone dosing control systems in municipal integrated wastewater treatment plants. Attached Figure Description

[0075] Figure 1 The results of linear discriminant analysis (LDA) were used to determine the ozone dosing ON / OFF control step for this invention.

[0076] Figure 2 Flowchart for constructing the linear discriminant analysis dataset for ozone dosing ON / OFF control in this invention;

[0077] Figure 3 This is a diagram showing the composition of the ozone dosing control system of the present invention; Detailed Implementation

[0078] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings. In addition, the forms of the various structures described in the following embodiments are merely illustrative. The present invention is not limited to the structures described in the following embodiments. All other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0079] Reference Figures 1-3 To address the problem of prolonged ineffective operation of ozone dosing systems during the daily operation of wastewater treatment plants, this invention proposes a method for establishing an ozone dosing control system model in advanced wastewater treatment. This method includes a step S for determining whether ozone dosing is ON / OFF and a step P for establishing a precise quantitative control model for ozone dosing.

[0080] Step S specifically includes:

[0081] Step 1: Use a 24-hour automatic sampler to continuously obtain influent and effluent samples from the ozone oxidation deep treatment process.

[0082] Step 2: Use a UV-Vis full-wavelength scanner to obtain the absorbance of organic matter in the influent and effluent of the ozone oxidation deep treatment process in the UV range.

[0083] Step 3: Plot the ultraviolet absorption spectrum with wavelength as the x-axis and absorbance as the y-axis and determine the absorption peaks.

[0084] Step 4: Extract the absorption peak wavelength and corresponding absorbance value, preprocess the extracted data to improve data accuracy, and ensure that the number of variables is much smaller than the number of data groups, that is, the number of columns is much smaller than the number of rows, and that each data group has an absorption peak.

[0085] Step 5: Determine COD using the "Determination of Chemical Oxygen Demand in Water - Dichromate Method" (HJ828-2017).

[0086] Step 6: Construct a new dataset for linear discriminant analysis (LDA), determine whether to add ozone based on COD values, select data greater than 10% for effect verification, and determine the corresponding formula.

[0087] Step 3 specifically includes: importing data from a UV-Vis full-wavelength scanner into Origin software, and plotting a spectrum with wavelength as the x-axis and absorbance as the y-axis. Identifying absorption peaks in the spectrum; an absorption peak is defined as the location of maximum absorbance on the curve. The peak finding is set to the five maximum values ​​around the peak, and the baseline is set to y=0. Extracting the absorption peak wavelengths and corresponding absorbance data to create dataset A.

[0088] Step 4 specifically includes the following process: In order to reduce the impact of accidental errors in equipment or monitoring processes on the dataset, dataset A should be preprocessed by deleting wavelengths with an absorption peak frequency of 12 times or less in all observations, and finally only 10 variables should be retained to form a new dataset B.

[0089] Step 6 includes the following specific steps:

[0090] S6-1: Add a "Ozone Dosing Switch" variable to dataset B to obtain dataset C. The "Ozone Dosing Switch" variable uses numbers to represent whether ozone needs to be added. According to the switch control principle, "1" represents yes and "0" represents no. "1" and "0" are distinguished based on COD values. Samples with COD values ​​> 27 mg / L are assigned a value of "1" for "Ozone Dosing Switch," indicating that ozone needs to be added. Samples with COD values ​​≤ 27 mg / L are assigned a value of "0" for "Ozone Dosing Switch," indicating that ozone does not need to be added. There are 87 sets of data with known types. 70 sets are selected to build a model, of which 25 sets are "0," meaning no ozone needs to be added, 45 sets are "1," meaning ozone needs to be added, and the remaining 17 sets of observations are used for validation.

[0091] S6-2: First, import dataset C into Unscrambler software. Use the 70 observations in dataset C for modeling. Click File-Import-Excel and select the corresponding document. After importing the document, select the "Ozone Dosing Switch" variable and change the data category option to set the "Ozone Dosing Switch" variable as a category column. Next, start LDA. Click Analyze-LDA under the Task section and make the following settings: select 10 variables in the Predictors column and select the "Ozone Dosing Switch" category column in the classification column. All 10 variables have equal weights (1), and the distance calculation method is set to linear. Then perform LDA analysis.

[0092] Predictions are made using the remaining 17 observations in dataset C, and the accuracy of the prediction set is typically used as the model evaluation metric. Click the LDA module in the Predict-Classification option under the task section, and select the LDA model built using one of the three methods mentioned above for each prediction model.

[0093] S6-3: Perform multiple linear regression on dataset C to determine the regression equation for the linear LDA model. Click the Analyze-MLR option under the Task section and make the following settings: select 10 variables in the Predictors column and select the "Ozone Dosing Switch" category column in the classification column.

[0094] S6-4: Principle of Linear Discriminant Analysis;

[0095] Let the entire sample dataset be Z = {(x1,y1),(x2,y2),…,(x...} N ,y N )}, where x i Let y be a d-dimensional vector, representing the number of samples and the features of the samples. iLet N be the class label for the corresponding sample, and N be the number of samples. Assuming there are K classes in total, the dataset for each class can be represented as follows:

[0096] This dataset x i y is a 10-dimensional vector. i In this paper, the value is 0 or 1, and there are two types of samples.

[0097] The goal of LDA is to map the original high-dimensional samples to have the minimum intra-class distance d in the low-dimensional feature space. w and the maximum inter-class distance d b d w and d b It is expressed as follows:

[0098]

[0099] In the formula, Let be the basis vector matrix in the low-dimensional feature space, where m is the dimension of the low-dimensional space, and m < d. Combining the above two objectives, the following optimization objective can be derived as the loss function of LDA:

[0100]

[0101] Define the within-class scatter matrix S respectively w and the inter-class scatter matrix S b As shown in the following formula:

[0102]

[0103]

[0104] The loss function can then be transformed into:

[0105]

[0106] Since the solution of the loss function depends only on the direction of W and not on its magnitude, let W... T When W=1, the loss function becomes a constrained form, i.e.:

[0107]

[0108] The solution can be obtained using the Lagrange multiplier method; the solution to W is a matrix. If x is the matrix formed by the eigenvectors corresponding to the first m largest eigenvalues, then x i The projection, i.e., the dimensionality-reduced sample features, can be represented as x. i =W T x i

[0109] The specific process of LDA is as follows:

[0110] Step 1: Calculate the mean vector of all samples. and the mean vector of each class of samples

[0111] Step 2: Calculate the within-class scatter matrix S w and the inter-class scatter matrix S b ;

[0112] Step 3: Find the matrix eigenvalues ​​and eigenvectors;

[0113] Step 4: Take the eigenvectors corresponding to the first m largest eigenvalues ​​to obtain the projection matrix ∈ W;

[0114] Step 5: Calculate the dimensionality-reduced sample features x i .

[0115] Step P specifically includes:

[0116] Step 1: Obtain influent and effluent samples from the ozone oxidation deep treatment process and determine the COD values.

[0117] Step 2: Screening COD data and corresponding UV absorption spectra. (Data screening criteria: influent COD > 30 mg / L for ozone oxidation advanced treatment process, and effluent COD 20-30 mg / L. The premise of this study is that an effluent COD of 20-30 mg / L from the ozone oxidation advanced treatment process meets effluent standards and does not lead to excessive ozone addition (excessive ozone addition increases energy consumption, resulting in excessively low COD values). Based on these criteria, 46 sets of spectral data for the influent of the ozone oxidation advanced treatment process were obtained.)

[0118] Step 3: Filter the data in dataset B' obtained in step S1 according to the screening conditions in step 2, and retain the corresponding ultraviolet absorption spectrum values ​​(process influent COD > 30 mg / L, process effluent COD 20-30 mg / L).

[0119] Step 4: Calculate the total ozone dosage based on on-site operating conditions. (During wastewater treatment plant operation, 8.7g of ozone is added per cubic meter of wastewater; the effluent flow rate of the ozone oxidation deep treatment process is 960-1256m³.) 3 The total hourly ozone dosage can be obtained by multiplying the ozone dosage by the flow rate.

[0120] Step 5: After filtering and reducing dataset B, 46 sets of observations were retained. A "Total Ozone Dosage" variable was added to dataset B' to construct dataset E.

[0121] Step 6: Perform multiple linear regression analysis on dataset E. The specific process includes:

[0122] To determine the regression equation for ozone dosage, perform multiple linear regression on dataset E (P6-1). Import dataset E into the software by clicking File-Import-Excel and selecting the corresponding dataset E. After selecting the 10 spectral variables, set the ranges for the independent and dependent variables by clicking Edit-Define range and selecting the corresponding columns. Click Analyze-MLR under the Task section and make the following settings: select 10 variables for the Predictors column and select the total ozone dosage category column for the classification column.

[0123] P6-2 Multiple Linear Regression Analysis Procedure: Given an example X = (x1, x2, x3, ..., x...) described by d attributes. d ), where x i All are values ​​of x on the i-th attribute. The linear model attempts to obtain a function that makes predictions through a linear combination of attributes, i.e.

[0124]

[0125] It is generally written in vector form.

[0126]

[0127] in

[0128] because The linear model intuitively expresses the importance of each attribute in the prediction, thus exhibiting good interpretability. The formula can be determined by b'.

[0129] We are trying to get

[0130]

[0131] The matrix expression and expansion of MLR

[0132]

[0133] y i Representing the predicted value, in ON / OFF control, it represents the value of the 0 / 1 column; in ozone dosage prediction, it represents the corresponding ozone dosage x. ij This represents the various factors affecting the load, which in this paper are the spectral absorption peak variables, β0 represents the constant term, and β i (i = 1, 2, ..., n) represents the regression coefficient, b' i This indicates a random disturbance.

[0134] The above formula can be simplified as follows:

[0135] Y = Xβ + ε (2.5)

[0136] In the formula: Y is the multivariate load matrix, ε is the random error matrix, X is the influencing factor matrix, and β is the regression coefficient matrix.

[0137] The regression parameters are estimated using the least squares method, and the regression function is obtained, which is the prediction model. Its formula is

[76] :

[0138]

[0139] Example 1

[0140] This embodiment provides a modeling method for ozone dosing ON / OFF control based on ultraviolet spectral data, such as... Figure 1 As shown, it includes the following steps:

[0141] (1) Taking the secondary treated effluent of a wastewater treatment plant as the research object, the absorbance of organic matter in the influent of the ozone deep treatment process in the ultraviolet spectral range was obtained by using an ultraviolet-visible full-wavelength scanner.

[0142] (2) The data from the full-wavelength scanner is transmitted to the control system, and absorption peaks are extracted based on the recorded data. The absorption peak data is preprocessed to reduce the impact of accidental errors in the equipment or monitoring process on the dataset, retaining those with high frequency of occurrence, and ensuring that absorption peak data is retained in each data set.

[0143] (3) Measure the COD of the influent and effluent of the ozone treatment process, assign values ​​to the corresponding ultraviolet spectral data category variables according to the COD value of the influent, and perform linear discriminant analysis.

[0144] (4) Based on the recorded data, establish the following ozone dosing discrimination control formula:

[0145] y = b + k1A 228 +k2A 229.5 +k3A 237 -k4A 253.5 -k5A 254 +k6A 255.5 -k7A 266 -

[0146] k8A 272.5 -k9A 362.5 -k 10 A 371

[0147] Based on the results of the multivariate regression analysis, the equation for the linear LDA analysis model is obtained, where A... xxx , representing the absorbance value at a wavelength of XXX nm, k iThe fitting parameters of the equation are represented by (i = 1, 2, ..., 10). y is dimensionless. When y ≥ 0.5, it is considered y = 1, meaning ozone needs to be added. When y < 0.5, it is considered y = 0, meaning ozone does not need to be added.

[0148] like Figure 1 As shown, the horizontal axis is 0, the vertical axis is 1, and the straight line is the target line for discrimination. Points above the straight line are those where y = 1 is determined, and ozone injection is required. Points below the straight line are those where y = 0 is determined, and ozone injection is not required. The closer the points are to the straight line, the worse the discrimination effect. The bolded points are points where discrimination is incorrect.

[0149] The final formula is as follows:

[0150] y = 0.6876 + 0.2923A 228 +0.1332A 229.5 +0.3998A 237 -0.017A 253.5

[0151] -0.824A 254 +0.1039A 255.5 -0.6209A 266 -0.7626A 272.5 -0.1383A 362.5

[0152] -0.0775A 371

[0153] Example 2

[0154] In this embodiment,

[0155] (1) Taking the secondary treated effluent of a wastewater treatment plant as the research object, ozone was added. The absorbance of organic matter in the influent of the ozone advanced treatment process in the ultraviolet spectral range was obtained using a full-spectrum scanner.

[0156] (2) The data from the full-wavelength scanner is transmitted to the control system, and the absorption peaks are extracted based on the recorded data. The absorption peak data is preprocessed to reduce the impact of accidental errors in the equipment or monitoring process on the dataset, retaining the most frequent ones, and ensuring that absorption peak data is retained in each set of data.

[0157] (3) COD was measured in the effluent from the ozone deep treatment process, and the influent UV spectral data of the ozone oxidation deep treatment process with COD between 20-30 mg / L were selected to meet the requirements.

[0158] (4) Combine the screening equation data with the total ozone dosage to form a dataset for multiple linear regression analysis.

[0159] (5) Based on the recorded data, establish the following precise quantitative formula for ozone dosing:

[0160] y = b + k1A 228 +k2A 229.5 +k3A 237 -k4A 253.5 -k5A 254 +k6A 255.5 -k7A 266 -

[0161] k8A 272.5 -k9A 362.5 -k 10 A 371

[0162] Based on the results of the multivariate regression analysis, the equation for the linear LDA analysis model is obtained, where A... xxx , representing the absorbance value at the corresponding wavelength, k i The fitting parameters of the equation are represented by y (i = 1, 2, ..., 10). y represents the ozone dosage flow rate (g / h).

[0163] The final formula is as follows:

[0164] y = 8643.2 - 3358A 228 +2234.1A 229.5 -1206.9A 237 -11470A 253.5 +2460.5A 254 -

[0165] 274.9A 255.5 +4114.7A 266 +19751.3A 272.5 -76949.1A 362.5 +58869.92A 371 .

[0166] Example 3,

[0167] The control process in this embodiment is as follows: Figure 3 As shown, the ozone dosing control process based on real-time ultraviolet spectroscopy monitoring consists of two steps: qualitative and quantitative. The qualitative step involves determining whether ozone dosing is necessary based on the formula in Example 1, corresponding to the ON / OFF control section of this invention. The quantitative step involves determining the specific time period for ozone dosing based on the formula in Example 2, and then precisely dosing the ozone.

[0168] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for controlling ozone dosing based on real-time monitoring of ultraviolet spectrum, characterized in that, It includes the judgment step S of whether the ozone needs to be added and the quantitative control step P of the ozone addition, and the specific process is as follows: The specific process of step S includes: S1, continuously obtaining water sample in the ozone oxidation advanced treatment system; S2, selecting a plurality of samples, and using a UV-visible light full wavelength scanner to obtain the absorbance of the organic matter in the ozone oxidation advanced treatment system in the ultraviolet light range; S3, drawing the ultraviolet absorption spectrum with wavelength as the abscissa and absorbance as the ordinate, and determining the absorption peak; S4, extracting the absorption peak wavelength and the corresponding absorbance value, establishing a data set A, and preprocessing the extracted data so that the number of variables is much smaller than the number of data sets, that is, the number of columns is much smaller than the number of rows, and ensuring that each group of data has an absorption peak, establishing a data set B; each row of the data set B represents different observation values, and each column represents a spectral variable, that is, the absorbance value at a specific wavelength; S5, determining the COD of all samples; S6, constructing a new data set for LDA linear discriminant analysis to determine the ozone addition discriminant formula; The specific process of step P includes: P1, continuously obtaining water sample in the ozone oxidation advanced treatment system and determining the COD value; P2, selecting a plurality of samples, setting a screening condition, screening the COD data, and establishing a data set B'; P3, determining the ultraviolet absorption spectrum value corresponding to the screened COD data; P4, determining the total ozone addition amount according to the operation condition of the ozone oxidation advanced treatment system; P5, adding a column of "total ozone addition amount" variable in the data set B', and constructing a data set E; P6, performing multivariate linear regression analysis on the data set E to determine the quantitative control formula of the ozone addition.

2. The method according to claim 1, wherein the method is characterized by: In step S4, the preprocessing process is to delete the wavelengths with an absorption peak frequency of 12 times or less in all observation values. 3.The method of claim 1, wherein the method comprises: In step S3, the absorption peak of the spectrum is determined, and the definition of the absorption peak is the place with the maximum absorbance on the curve, and the peak search is set to the maximum value of the surrounding 5 values, and the peak search baseline is set to y=0.

4. The method according to claim 1, wherein the method is characterized by: The specific process of step S6 includes: S61: determining the training set data: adding a column of "ozone addition switch" variable to the data set B to obtain a data set C; the "ozone addition switch" variable is represented by a number to represent whether the ozone needs to be added, according to the switch control principle, "1" represents yes, and "0" represents no, "1" and "0" are distinguished according to the COD value, the sample with COD value>27mg / L is assigned a value of "1", representing that the ozone needs to be added; the sample with COD value≤27mg / L is assigned a value of "0", representing that the ozone does not need to be added; from the data set C, a plurality of groups are selected to establish a model, and the proportion of "1" and "0" in the plurality of groups is (8-10):(4-6); S62: Performing LDA is linear discriminant analysis: Let the entire sample dataset be Z = {(x1,y1),(x2,y2),…,(x...}. N ,y N )}, where x i Let y be a d-dimensional vector, representing the number of samples and the features of the samples. i Let N be the class label for the corresponding sample, and N be the number of samples. Assuming there are K classes in total, the dataset for each class can be represented as follows: The dataset x i is a 10-dimensional vector, y i is either 0 or 1, for a total of two classes of samples; The goal of LDA is to map the original high-dimensional samples in a low-dimensional feature space such that the within-class distances d w and the between-class distances d b are minimized w and maximized b respectively, where d w and d b are defined as follows: wherein is a matrix of basis vectors in the low-dimensional feature space, and m is the dimension of the low-dimensional space, and m < d; in view of the above two objectives, the following optimization objective can be obtained as the loss function of LDA: The within-class scatter matrix S w and the between-class scatter matrix S b are defined as follows: Then the loss function can be changed to: Since the solution of the loss function is only related to the direction of W, but not its magnitude, we can let W T W = 1, the loss function becomes a constrained form, i.e. The solution of W can be solved by Lagrange multiplier method, which is a matrix composed of eigenvectors corresponding to the first m largest eigenvalues of matrix The projection of x i , i.e., the sample features after dimension reduction, can be expressed as x i = W T x i ; the specific process is as follows: S621: Calculate the mean vector of all samples and the mean vector of each class of samples S622: Calculate the within-class scatter matrix S w and the between-class scatter matrix S b ; S623: find eigenvalues and eigenvectors of matrix S625: compute the matrix S624: taking the eigenvectors corresponding to the first m largest eigenvalues to obtain the projection matrix ∈W; S625: Calculate the sample feature x after dimension reduction i .

5. The method according to claim 1, wherein the method is characterized by: In step S6, the determined ozone addition discriminant formula is: y = b + k1A 228 + k2A 229.5 + k3A 237 - k4A 253.5 - k5A 254 + k6A 255.5 - k7A 266 - k8A 272.5 - k9A 362.5 - k 10 A 371 wherein A xxx represents the absorbance value at XXX nm wavelength, k i represents the fitting parameters (i = 1, 2, …, 10); y is unitless, y > 0.5 is considered y = 1, i.e. ozone needs to be dosed; y < 0.5 is considered y = 0, i.e. no ozone needs to be dosed. The finally established formula is as follows: 6.The method of claim 1, wherein the method further comprises: In step P2, the screening condition is that the COD of the influent of the ozone oxidation advanced treatment process is greater than 30 mg / L, and the COD of the effluent is 20-30 mg / L, so as to meet the effluent standard and not to cause excessive addition of ozone, which leads to too low COD value.

7. The method according to claim 1, wherein the method is characterized by: In step P4, 8.7 g of ozone is added per cubic meter of sewage during the operation of the ozone oxidation advanced treatment system, the effluent flow of the ozone oxidation advanced treatment system is measured, and the total ozone addition amount per unit time can be obtained by multiplying the ozone addition amount by the effluent flow. 8.The method of claim 1, wherein the method further comprises: determining a concentration of the ozone in the water using the UV spectrum. In step P6, the specific process of the multiple linear regression analysis on the data set E includes: Given an example X = (x1, x2, x3,..., x d d) described by d attributes, where xi is the value of x at the i-th attribute, a linear model tries to get a function that predicts by a linear combination of the attributes, i.e. i ​ The vector form is: wherein Because Linear models have good interpretability because the importance of each attribute in the prediction is directly expressed; determining And b can determine the formula; We try to get The matrix expression and its expansion are y i represent the predicted value, in ON / OFF control, the value of 0 / 1 number column, in ozone dosage prediction, the corresponding ozone dosage; x ij represent the various factors affecting the load, in this paper, namely the spectral absorption peak variable, β0 represents the constant term, β i (i = 1, 2,..., n) represents a regression coefficient, b i represents a random disturbance; The above formula is simply written as: Y = Xβ + ε (2.5) In the formula, Y is a multivariate load matrix, ε is a random error matrix, X is an influencing factor matrix, and β is a regression coefficient matrix; The regression function is obtained by estimating the regression parameters by using the least square method, that is, the prediction model; the formula is: 9.The method of claim 1, wherein the method further comprises: determining a concentration of the ozone in the water using the UV spectrum. In step P6, the quantitative control formula of the determined ozone addition is: According to the results of multiple regression analysis, the equation of linear LDA analysis model is obtained, in which A xxx represents the absorbance value at XXX nm wavelength, k i represents the fitting parameter of equation (i=1, 2, …, 10); y represents the ozone dosage flow (g / h); the formula contains 10 variables, and under the daily operation condition of the ozone dosage of 8.7 g / m 3 , the corresponding ozone dosage flow (g / h) is calculated combined with the corresponding sewage flow, the ozone dosage flow is subjected to multiple linear regression analysis with the screened spectral absorbance variable, and the formula is established; the finally established formula is as follows:

Images