Water inlet abnormity early warning system and method for sewage treatment plant

Abstract

The invention discloses an inflow abnormity early warning system and method for a sewage treatment plant, and the method comprises the steps: setting a real-time monitoring device, and collecting and testing the data of an inflow sample; arranging a three-dimensional fluorescence spectrometer, and collecting three-dimensional fluorescence spectrum data of the inflow sample; preprocessing the collected data of the inlet water sample and the three-dimensional fluorescence spectrum data; establishing a standard library, and collecting data of inflow samples at different moments to fill the standard library; performing early warning analysis on the water sample condition of the water inlet sample, and judging the water sample condition of the water inlet sample; updating the data of the standard library according to the water sample condition of the water inlet sample; the method comprises the following steps: combining conventional physicochemical indexes with three-dimensional fluorescence spectrum data, performing dimension reduction analysis on the three-dimensional fluorescence spectrum data, extracting three-dimensional fluorescence characteristics, performing clustering analysis on an inflow sample and a standard library sample in combination with a competitive neural network, and calculating a matching rate of the inflow sample and the standard library sample so as to judge whether the inflow sample is abnormal or not. And the accuracy of three-dimensional fluorescence spectrum early warning is improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of abnormal warning of sewage treatment plants, and in particular relates to a system and method for warning abnormal water inflow of sewage treatment plants. Background Art

[0002] At present, domestic sewage treatment plants usually adopt a treatment process that combines pretreatment, biological treatment, and deep treatment. The biological system is the core link of the entire sewage treatment system. However, due to the characteristics of microorganisms, high concentrations of toxic and harmful substances in the influent water quality will seriously affect the stable operation of the sewage treatment plant and the compliance with emission standards.

[0003] Online monitoring of sewage treatment plant influent is mainly based on monitoring of conventional parameters (COD Cr , total phosphorus, and ammonia nitrogen). Monitoring conventional parameters sometimes fails to promptly and effectively reflect changes in toxic and hazardous substances in sewage treatment plant influent. In recent years, three-dimensional fluorescence spectroscopy has been widely used in water pollution early warning and source tracing due to its rich spectral information, high sensitivity, rapid response, simple analysis, and lack of secondary pollution. However, this technique has a limited scope of application and has a good match rate for most fluorescent compounds with conjugated and rigid planar structures. However, it is not very sensitive to laundry wastewater containing detergents or restaurant wastewater containing animal and vegetable oils. High concentrations of phosphorus in detergents and high concentrations of animal and vegetable oils in restaurant wastewater also have a certain impact on the biochemical treatment of urban sewage treatment plants. In addition, three-dimensional fluorescence spectroscopy data is often processed and analyzed using peak extraction and peak similarity comparison. However, due to the complex management of urban sewage treatment plants and the large differences in influent quality over different time periods, relying solely on peak extraction to determine influent quality and similarity to determine whether water quality is abnormal is often prone to misjudgment.

[0004] Therefore, the present invention combines the monitoring of conventional parameters with three-dimensional fluorescence spectroscopy technology to form a rapid and effective sewage treatment plant inlet water quality early warning system and method, so as to quickly activate the emergency plan when the water quality is abnormal, and protect the sewage treatment plant biological system from being impacted by toxic and harmful substances. Summary of the Invention

[0005] In response to the problems existing in the existing technology, the present invention provides a sewage treatment plant influent abnormality early warning system and method, which combines the measurement results of conventional parameters of sewage treatment plant influent water quality with three-dimensional fluorescence spectrum data, establishes a standard database, sets early warning values, and combines competitive neural networks to compare and analyze influent samples with standard samples to determine whether the influent samples are abnormal, thereby improving the accuracy of sewage treatment plant influent water quality abnormality early warning.

[0006] The technical solution of the present invention is achieved as follows:

[0007] A method for warning abnormal water inflow in a sewage treatment plant, comprising the following steps:

[0008] S1. Install a real-time monitoring device at the water inlet of the sewage treatment plant to collect data from influent samples and conduct tests;

[0009] S2. Setting a three-dimensional fluorescence spectrometer to regularly collect three-dimensional fluorescence spectrum data of the influent sample;

[0010] S3, preprocessing the three-dimensional fluorescence spectrum data collected from the influent sample, removing and filling the fluorescence values in the Raman scattering and Rayleigh scattering regions in the three-dimensional fluorescence spectrum data, and then performing Gaussian smoothing processing;

[0011] S4, normalizing the pre-processed three-dimensional fluorescence spectrum data, converting it into a grayscale image, and performing gain processing;

[0012] S5. Establish a standard library of conventional parameters and three-dimensional fluorescence spectrum data of influent samples, and set early warning values;

[0013] S6. Perform early warning analysis on the water sample condition of the influent sample, using conventional parameter test results and three-dimensional fluorescence spectrum characteristic data as input, and perform competitive neural network analysis. Based on whether the conventional parameter test results and the competitive neural network analysis results exceed the early warning value, determine whether the influent sample is abnormal;

[0014] S7. Update the data in the standard library according to the water sample conditions of the influent sample.

[0015] Furthermore, in S1, the real-time monitoring device is a conventional parameter real-time monitoring device, and the data of the test influent sample are conventional parameters, including pH value, chemical oxygen demand, ammonia nitrogen content, and total phosphorus content.

[0016] Furthermore, in said S2, a three-dimensional fluorescence spectrometer is set in the sewage treatment plant, and the three-dimensional fluorescence spectrum data of the influent sample is collected once every time t, and the collection parameters are set, including the scanning range, the width of the emission and excitation wavelengths, and the scanning speed.

[0017] Furthermore, in S3, three-dimensional fluorescence spectrum data of pure water is collected every 12 hours, and the three-dimensional fluorescence spectrum data of pure water is used as the three-dimensional fluorescence spectrum data of the blank sample; for the three-dimensional fluorescence spectrum data of the influent sample, the three-dimensional fluorescence spectrum data of the influent sample is collected once every time t, and the three-dimensional fluorescence spectrum data of the most recent blank sample is deducted from the collected three-dimensional fluorescence spectrum data of the influent sample, and then Rayleigh scattering and Raman scattering are deducted, and then the Delaunay triangle interpolation method is used to fill the fluorescence value of the deducted area, and the data processed by the Delaunay triangle interpolation method is Gaussian smoothed.

[0018] Furthermore, in S4, the three-dimensional fluorescence spectrum data is normalized, and the specific formula is:

[0019]

[0020] Among them, FI (x，m) represents the fluorescence intensity after linear normalization at the excitation wavelength x nm and emission wavelength m nm, fi (x，m) It represents the fluorescence intensity obtained by scanning with a fluorescence spectrophotometer at an excitation wavelength of x nm and an emission wavelength of m nm. min is the minimum fluorescence intensity after removing the scattered light, fi max is the maximum fluorescence intensity after excluding scattering;

[0021] The normalized three-dimensional fluorescence spectrum data is converted into grayscale data with a pixel size of 534×609, and the grayscale data is subjected to gain processing. The specific formula is:

[0022]

[0023] Where X is the fluorescence intensity F after linear normalization (n,m) value; f(X) is the fluorescence intensity value obtained after transformation by the S-shaped growth curve deformation function, k is the gain coefficient, which is a positive real number assigned to control the degree of fluorescence intensity scaling. Through testing of multiple influent samples, the k value of the sewage treatment plant influent water quality sample is 0.5.

[0024] Furthermore, in S5, a standard library of conventional parameters and three-dimensional fluorescence spectrum data of influent samples is established, and an early warning value is set; specifically, the following steps are performed:

[0025] Set the standard parameter range of chemical oxygen demand, ammonia nitrogen content, and total phosphorus content,

[0026] If the conventional parameters of chemical oxygen demand, ammonia nitrogen content, and total phosphorus content obtained by S4 fall within the standard parameter range, they are classified as normal data;

[0027] If the conventional parameters of chemical oxygen demand, ammonia nitrogen content, and total phosphorus content obtained by S4 exceed the standard parameter range, they will be listed as abnormal data;

[0028] For normal data that meet the standard parameter range, the p-value statistical method is used to determine whether the three-dimensional fluorescence spectrum data is normal data;

[0029] The conventional parameters of chemical oxygen demand, ammonia nitrogen content, total phosphorus content and three-dimensional fluorescence spectrum data that are normal after testing are placed in the normal database;

[0030] The conventional parameters of chemical oxygen demand, ammonia nitrogen content, total phosphorus content and three-dimensional fluorescence spectrum data that are abnormal data after inspection are placed in the abnormal database;

[0031] Fill the normal database so that the number of influent samples corresponding to the normal data in the normal database reaches Z, and use the normal database formed by Z normal data as the standard database.

[0032] Furthermore, the warning values of conventional parameters and three-dimensional fluorescence spectrum data are set;

[0033] The warning value settings for conventional parameters specifically include:

[0034] Calculate the average and standard deviation of the chemical oxygen demand, ammonia nitrogen content, and total phosphorus content corresponding to all influent samples in the normal database respectively. Add three times the corresponding standard deviation to the average values of the calculated chemical oxygen demand, ammonia nitrogen content, and total phosphorus content. The final corresponding values are used as the warning values of chemical oxygen demand, ammonia nitrogen content, and total phosphorus content respectively;

[0035] Collect the conventional parameter measurement results and three-dimensional fluorescence spectrum data of M normal influent samples, and process the three-dimensional fluorescence spectrum data according to S3 and S4;

[0036] Each 3D fluorescence spectrum data set in the M processed data is combined with the Z 3D fluorescence spectrum data sets in the normal database to form Z+1 3D fluorescence spectrum data sets of influent samples. For each 3D fluorescence spectrum data set, the principal component analysis method is used to extract the first S principal component values in the 3D fluorescence spectrum data set as the eigenvalues of the 3D fluorescence spectrum data set.

[0037] The characteristic values of the three-dimensional fluorescence spectrum data and the corresponding conventional parameters of Z+1 influent samples are used as the input of the competitive neural network. After the competitive neural network analysis, the output item is obtained and the output item is set to 0 or 1.

[0038] If the output result is 1, divide the Z+1 influent samples into Z influent samples and +1 influent samples, where the Z influent samples are the corresponding Z influent samples in the normal database, and the +1 influent sample is one of the M influent samples; then calculate the number of influent samples in the output results of the Z influent samples that have the same output result as the +1 influent sample;

[0039] Perform Q competitive neural network analyses on each Z+1 influent sample; calculate the average number of influent samples that have the same output as the +1 influent sample;

[0040] Divide the average of the number of influent samples that have the same output as the +1 influent sample by Z to obtain the matching rate of one of the influent samples;

[0041] The matching rate of each of the M influent samples is calculated in sequence, and then the average matching rate of the M influent samples is calculated. The average matching rate of the M influent samples is used as the warning value of the competitive neural network analysis result.

[0042] Furthermore, the conventional parameter test results and three-dimensional fluorescence spectrum data of the influent sample are collected. If the pH value of the influent sample is greater than 9 or less than 6, the influent sample is considered an abnormal sample.

[0043] If the pH value of the influent sample is normal, determine whether the measured values of the chemical oxygen demand, ammonia nitrogen content, and total phosphorus content of the influent sample are greater than the corresponding warning values of the chemical oxygen demand, ammonia nitrogen content, and total phosphorus content. If the measured result of any of the conventional parameters is greater than the corresponding warning value of the parameter, the sample is determined to be an abnormal sample;

[0044] If the measurement results of the pH value, chemical oxygen demand, ammonia nitrogen content, and total phosphorus content of the influent sample are all normal, a competitive neural network analysis is performed to calculate the matching rate; if the matching rate of the influent sample is less than the warning value of the competitive neural network analysis result, the influent sample is an abnormal sample; otherwise, the influent sample is a normal sample.

[0045] Furthermore, in S7, the data of the standard library is updated:

[0046] If the influent sample is a normal sample, the data of the corresponding influent sample is saved in the normal database, and the data of the influent sample with the longest time from the currently saved influent sample is replaced to form a new standard database;

[0047] If the influent sample is an abnormal sample, the data of the corresponding influent sample is saved in the abnormality database, and the data of the corresponding three-dimensional fluorescence spectrum data is saved at the same time.

[0048] A sewage treatment plant water inlet abnormality early warning system applies a sewage treatment plant water inlet abnormality early warning method as described in any one of the above.

[0049] Compared with the prior art, the present invention achieves the following beneficial effects:

[0050] The present invention provides a sewage treatment plant influent anomaly warning system and method. This system processes collected data, performs dimensionality reduction analysis on three-dimensional fluorescence spectral data through principal component analysis of influent samples, extracts three-dimensional fluorescence features, and uses a competitive neural network to analyze influent samples against standard library samples. The system calculates the matching rate between influent and standard library samples, thereby determining whether the influent sample is abnormal and improving the accuracy of three-dimensional fluorescence spectral warnings. This system effectively reflects changes in influent water quality and preserves water quality information, providing a basis for rapidly identifying pollution emission sources, and for timely response and accountability.

[0051] By preprocessing the data, the gain coefficient for processing the three-dimensional fluorescence grayscale image of the water quality of urban sewage treatment plants is confirmed. The three-dimensional fluorescence characteristic images of different types of water quality samples have their own characteristic responses. Based on the three-dimensional fluorescence spectrum characteristics of urban sewage treatment plants, the present invention confirms the gain coefficient for the influent water quality of sewage treatment plants, adds a p-value statistical judgment mechanism for abnormal data, and improves the accuracy of subsequent competitive neural network early warning results.

[0052] The physicochemical parameters and three-dimensional fluorescence spectral data are combined as the input of the competitive neural network. The results of physicochemical parameters such as COD, ammonia nitrogen, and total phosphorus are used to assist in monitoring the changes in the influent water quality of urban sewage treatment plants, effectively solving the problem that three-dimensional fluorescence spectroscopy cannot effectively monitor slaughterhouse wastewater containing oil and fat and washing wastewater containing detergents. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] Figure 1 This is a flow chart of a method for early warning of abnormal water inflow in a sewage treatment plant provided by an embodiment of the present invention;

[0054] Figure 2 This is an original image of three-dimensional fluorescence spectrum data of a water sample in a sewage treatment plant water inlet abnormality early warning method provided by an embodiment of the present invention;

[0055] Figure 3 This is a three-dimensional fluorescence blank-subtracted spectrum of a water sample in a sewage treatment plant inlet abnormality early warning method provided by an embodiment of the present invention;

[0056] Figure 4 This is a three-dimensional fluorescence subtraction scattering spectrum of a water sample in a sewage treatment plant water inlet abnormality early warning method provided by an embodiment of the present invention;

[0057] Figure 5 This is a spectrum of three-dimensional fluorescent triangle interpolation and Gaussian smoothing processing of a water sample in a sewage treatment plant water inlet abnormality early warning method provided by an embodiment of the present invention;

[0058] Figure 6 This is a three-dimensional fluorescent grayscale image of a water sample in a sewage treatment plant water inlet abnormality early warning method provided by an embodiment of the present invention;

[0059] Figure 7 The invention provides a method for warning abnormal water inflow to a sewage treatment plant provided by an embodiment of the present invention, which is a grayscale image of a three-dimensional fluorescent grayscale image of a water sample processed by an S-shaped growth curve. DETAILED DESCRIPTION

[0060] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts shall fall within the scope of protection of the present invention.

[0061] Example

[0062] like Figures 1 to 7 A method for warning abnormal water inflow in a sewage treatment plant comprises the following steps:

[0063] S1. Install a real-time monitoring device at the water inlet of the sewage treatment plant to collect data from influent samples and conduct tests;

[0064] The real-time monitoring device is a conventional parameter real-time monitoring device, and the data of the test influent sample are conventional parameters, including pH value, chemical oxygen demand, ammonia nitrogen content, and total phosphorus content.

[0065] Chemical oxygen demand (COD) Cr , ammonia nitrogen is NH3-N, and total phosphorus is TP.

[0066] S2. Setting a three-dimensional fluorescence spectrometer to regularly collect three-dimensional fluorescence spectrum data of the influent sample;

[0067] A three-dimensional fluorescence spectrometer is set up in the sewage treatment plant, and the three-dimensional fluorescence spectrum data of the influent sample is collected every time t. The collection parameters are set, including the scanning range, the width of the emission and excitation wavelengths, and the scanning speed.

[0068] Specifically, the three-dimensional fluorescence spectrum data of the water sample is collected every 20 minutes, and the data collection conditions are as follows: the excitation wavelength (E x ) scanning range 220~600nm, emission wavelength (E m The scanning range was 230–650 nm, with emission and excitation wavelength widths of 5 nm and a scanning speed of 12,000 nm / min. The collected three-dimensional fluorescence spectral data had a dimension of 77 × 85, with 77 representing the number of excitation wavelengths and 85 representing the number of emission wavelengths.

[0069] S3. Preprocessing the three-dimensional fluorescence spectrum data collected from the influent sample by removing and filling the fluorescence values in the Raman scattering and Rayleigh scattering regions in the three-dimensional fluorescence spectrum data, and then performing Gaussian smoothing processing; specifically, collecting three-dimensional fluorescence spectrum data of pure water every 12 hours, and using the three-dimensional fluorescence spectrum data of pure water as the three-dimensional fluorescence spectrum data of the blank sample;

[0070] For the three-dimensional fluorescence spectrum data of the influent sample, the three-dimensional fluorescence spectrum data of the influent sample is collected once every time t, and the three-dimensional fluorescence spectrum data of the most recent blank sample is deducted from the collected three-dimensional fluorescence spectrum data of the influent sample, and then Rayleigh scattering and Raman scattering are deducted;

[0071] Rayleigh scattering is divided into first-order Rayleigh scattering and second-order Rayleigh scattering. First-order Rayleigh scattering generally occurs in the region where the excitation wavelength and emission wavelength differ by ±10-15nm, while second-order Rayleigh scattering generally occurs in the region where the emission wavelength is equal to twice the excitation wavelength ±10-15nm. Raman scattering generally occurs near Rayleigh scattering. Rayleigh scattering and Raman scattering will interfere with subsequent data analysis results. Therefore, Rayleigh scattering and Raman scattering need to be eliminated in the preprocessing of three-dimensional fluorescence data. In this embodiment, the first-order Rayleigh scattering is deducted within 10nm and the lower 25nm; the second-order Rayleigh scattering is deducted within 10nm and the lower 30nm; and the first-order Raman scattering and the second-order Raman scattering are deducted within 10nm and the lower 20nm.

[0072] Delaunay triangle interpolation is then used to fill in the fluorescence values in the subtracted areas. After subtracting the Rayleigh and Raman scattering data, the data values are empty (NA), making subsequent data analysis impossible. Therefore, the Delaunay triangle interpolation algorithm is used to fill in the subtracted data. The Delaunay triangle interpolation algorithm divides the three-dimensional fluorescence spectral data matrix into a series of connected, non-overlapping triangular meshes, and then performs the required interpolation within the empty circumcircle of each triangle.

[0073] The data processed by Delaunay triangle interpolation is Gaussian smoothed. Gaussian filtering is typically implemented using a convolution operation. First, a Gaussian kernel is created, which is a two-dimensional Gaussian function matrix. This kernel is then applied to each pixel in the image, and the smoothed pixel value is obtained by calculating the weighted sum of the kernel and the local image region. Gaussian smoothing eliminates image noise introduced by Delaunay triangle interpolation, providing a better image foundation for subsequent feature analysis.

[0074] like Figures 2 to 5 , demonstrating the process of processing three-dimensional fluorescence spectrum raw data, subtracting Rayleigh scattering, triangular interpolation and Gaussian smoothing.

[0075] S4, normalizing the pre-processed three-dimensional fluorescence spectrum data, converting it into a grayscale image, and performing gain processing;

[0076] Specifically, the three-dimensional fluorescence spectrum data after deducting Rayleigh scattering and Raman scattering is normalized, and the specific formula is:

[0077]

[0078] Among them, FI(x，m) represents the fluorescence intensity after linear normalization at the excitation wavelength x nm and emission wavelength m nm, fi (x，m) It represents the fluorescence intensity obtained by scanning with a fluorescence spectrophotometer at an excitation wavelength of x nm and an emission wavelength of m nm. min is the minimum fluorescence intensity after removing scattered light, fi max is the maximum fluorescence intensity after excluding scattering;

[0079] The normalized three-dimensional fluorescence spectrum data is converted into grayscale data with a pixel size of 534×609, which is divided into 256 grayscale levels (0-255), where 0 represents complete black and 255 represents complete white. The data is then subjected to gain processing. The specific formula for gain processing is:

[0080]

[0081] Where X is the fluorescence intensity F after linear normalization (n,m) f(X) is the fluorescence intensity value obtained after transformation by the S-shaped growth curve deformation function. k is the gain coefficient, a positive real number assigned to control the degree of fluorescence intensity scaling. Through testing of multiple influent samples, the k value for the sewage treatment plant influent water quality sample is 0.5. k is the gain coefficient, a positive real number assigned to control the degree of fluorescence intensity scaling. After verification, the k value for the sewage treatment plant influent water quality sample is 0.5.

[0082] The original three-dimensional fluorescence spectrum data of 95 sewage treatment plant influent samples and the three-dimensional fluorescence spectrum data of water samples diluted 5 times were collected. The data were preprocessed according to the above S3 and S4 steps. Different k values were input during the preprocessing process. The k value ranged from 0.1 to 100. The top 7 PC data with a cumulative variance contribution rate of 85% were used as the competitive neural network input vectors. Five competitive neural network analyses were carried out. The number of sample pairs Q with the same competitive neural network output results in the 95 pairs of data was counted. The MR value was calculated according to the formula MR = Q / 95*100%. The k value corresponding to the highest MR value was the optimal k value for this project. The results showed that when the k value was 0.5, the MR value was the highest, so the k value for this project was 0.5. The following is the source table of k values:

[0083]

[0084] S5. Establish a standard library of conventional parameters and three-dimensional fluorescence spectrum data of influent samples, and set early warning values;

[0085] Set the standard parameter range of chemical oxygen demand, ammonia nitrogen content, and total phosphorus content,

[0086] If the conventional parameters of chemical oxygen demand, ammonia nitrogen content, and total phosphorus content obtained by S4 fall within the standard parameter range, they are classified as normal data;

[0087] If the conventional parameters of chemical oxygen demand, ammonia nitrogen content, and total phosphorus content obtained by S4 exceed the standard parameter range, they will be listed as abnormal data;

[0088] For normal data that meet the standard parameter range, the p-value statistical method is used to determine whether the three-dimensional fluorescence spectrum data is normal data;

[0089] The conventional parameters of chemical oxygen demand, ammonia nitrogen content, total phosphorus content and three-dimensional fluorescence spectrum data that are normal after testing are placed in the normal database;

[0090] The conventional parameters of chemical oxygen demand, ammonia nitrogen content, total phosphorus content and three-dimensional fluorescence spectrum data that are abnormal data after inspection are placed in the abnormal database;

[0091] Fill the normal database so that the number of influent samples corresponding to the normal data in the normal database reaches Z, and use the normal database formed by Z normal data as the standard database.

[0092] In said S5, Z water quality samples are collected, if the conventional parameter measurement results of individual samples show pH>9 or <6, or COD Cr >500mg / L, or TP>8mg / L, or NH3-N>50mg / L, or pH value, COD Cr If the measured values of TP and NH3-N exceed the design range of the sewage treatment plant's influent, they are abnormal data, and the collected data of this part of the samples are placed in the abnormal database (database 1); if the measured values of the conventional parameters are not abnormal data, the three-dimensional fluorescence spectrum data of the collected samples are processed according to S3, and the p-value method is used to determine whether the three-dimensional fluorescence spectrum data is abnormal data according to the group number of 4. If the p-value analysis results of the three-dimensional fluorescence spectrum data of individual samples exceed the threshold and there are few other similar samples, then this part of the samples are abnormal samples, and the collected data of this part of the samples are placed in the abnormal database (database 1); if both the conventional parameters and the three-dimensional fluorescence spectrum data are not abnormal data, then this part of the sample data is placed in the normal database (database 2); cyclically collect the data of the influent samples at different times, pre-process the collected data, and determine whether it is abnormal data until the normal data corresponding to the normal database (database 2) is filled to Z (Z value ≥ 216).

[0093] Specifically, data from 216 influent samples were collected, of which 9 samples had TP determination results > 8 mg / L and were listed as abnormal data. After p-value judgment, 6 samples were listed as abnormal data. The influent sample data were continued to be collected until 216 normal data were collected in the standard library.

[0094] Set warning values for conventional parameters and three-dimensional fluorescence spectrum data;

[0095] The warning value settings for conventional parameters specifically include:

[0096] Calculate the average and standard deviation of the chemical oxygen demand, ammonia nitrogen content, and total phosphorus content corresponding to all influent samples in the normal database respectively. Add three times the corresponding standard deviation to the average values of the calculated chemical oxygen demand, ammonia nitrogen content, and total phosphorus content. The final corresponding values are used as the warning values of chemical oxygen demand, ammonia nitrogen content, and total phosphorus content respectively;

[0097] Statistical analysis was performed on the results of conventional parameter measurements collected in the normal database, and the conventional parameter COD was calculated respectively. Cr , TP, NH3-N warning values, COD Cr The warning values of the three conventional parameters, TP, and NH3-N, are the average value of the parameter measurement results + 3 times the standard deviation. The specific formula is:

[0098]

[0099] Where, is the average value of the measurement results, s i is the standard deviation, X n is the measurement result, n is the number of measurements, i refers to the conventional parameter COD Cr ,TP,NH3-N.

[0100] Collect the conventional parameter measurement results and three-dimensional fluorescence spectrum data of M normal influent samples, and process the three-dimensional fluorescence spectrum data according to S3 and S4;

[0101] Each 3D fluorescence spectrum data set in the M processed data is combined with the Z 3D fluorescence spectrum data sets in the normal database to form Z+1 3D fluorescence spectrum data sets of influent samples. For each 3D fluorescence spectrum data set, the principal component analysis method is used to extract the first S principal component values in the 3D fluorescence spectrum data set as the eigenvalues of the 3D fluorescence spectrum data set.

[0102] The characteristic values of the three-dimensional fluorescence spectrum data and the corresponding conventional parameters of Z+1 influent samples are used as the input of the competitive neural network. After the competitive neural network analysis, the output item is obtained and the output item is set to 0 or 1.

[0103] If the output result is 1, divide the Z+1 influent samples into Z influent samples and +1 influent samples, where the Z influent samples are the corresponding Z influent samples in the normal database, and the +1 influent sample is one of the M influent samples; then calculate the number of influent samples in the output results of the Z influent samples that have the same output result as the +1 influent sample;

[0104] Perform Q competitive neural network analyses on each Z+1 influent sample; calculate the average number of influent samples that have the same output as the +1 influent sample;

[0105] Divide the average of the number of influent samples that have the same output as the +1 influent sample by Z to obtain the matching rate of one of the influent samples;

[0106] Calculate the matching rate of each of the M influent samples in sequence, then calculate the average of the matching rates of the M influent samples, and use the average of the matching rates of the M influent samples as the warning value of the competitive neural network analysis results; the specific formula is as follows:

[0107]

[0108] Where, P i It represents the matching rate of the single competitive neural network analysis results of each sample; N represents the number of samples in the normal database whose competitive neural network output results are consistent with the output results of the influent sample; Z represents the total number of samples in the standard library, The average matching rate of the influent sample obtained by Q competitive neural network analysis is the matching rate of the sample; WT EEMs is the warning value of three-dimensional fluorescence spectrum data, that is, the warning value of competitive neural network analysis results.

[0109] Specifically, the conventional parameter measurement results collected in the normal database (database 2) are statistically analyzed to calculate the conventional parameter COD Cr , TP, NH3-N warning values, COD Cr The warning values of three conventional parameters, TP, and NH3-N, are the average value of the parameter measurement results + 3 times the standard deviation. The warning values are calculated using the measurement results of 216 normal influent samples. After calculation, WT CODCr =471, WT NH3-N =48.9, WT TP =6.9, WT indicates the warning value.

[0110] The conventional parameter measurement results and three-dimensional fluorescence spectrum data of 83 normal influent samples were collected. The three-dimensional fluorescence spectrum data were preprocessed according to S3 and S4 above. The preprocessed data were then combined with the 216 preprocessed three-dimensional fluorescence spectrum data from the normal database to form a three-dimensional fluorescence spectrum data set of 217 samples. The principal component analysis method was used to extract the three-dimensional fluorescence spectrum data characteristics of the data set, and the first four principal components were used as the three-dimensional fluorescence spectrum characteristic values. The three-dimensional fluorescence spectrum characteristic values and the test results of the conventional parameters were used as the input of the competitive neural network. The output of the competitive neural network was set to 0 or 1. The competitive neural network analysis was performed five times for each sample. The number of samples in the normal database whose competitive neural network output results were consistent with the output results of the influent samples was counted, and the matching rate was calculated. The average matching rate of 83 normal samples, 82.7%, was used as the warning value of the competitive neural network analysis results. The following is the source of the warning value data of the competitive neural network analysis:

[0111]

[0112]

[0113] S6. Perform early warning analysis on the water sample condition of the influent sample, using conventional parameter test results and three-dimensional fluorescence spectrum characteristic data as input, and perform competitive neural network analysis. Based on whether the conventional parameter test results and the competitive neural network analysis results exceed the early warning value, determine whether the influent sample is abnormal;

[0114] Collect the test results of routine parameters and three-dimensional fluorescence spectrum data of the influent sample. If the pH value of the influent sample is greater than 9 or less than 6, the influent sample is considered an abnormal sample.

[0115] If the pH value of the influent sample is normal, determine whether the measured values of the chemical oxygen demand, ammonia nitrogen content, and total phosphorus content of the influent sample are greater than the corresponding warning values of the chemical oxygen demand, ammonia nitrogen content, and total phosphorus content. If the measured result of any of the conventional parameters is greater than the corresponding warning value of the parameter, the sample is determined to be an abnormal sample;

[0116] If the measurement results of the pH value, chemical oxygen demand, ammonia nitrogen content, and total phosphorus content of the influent sample are all normal, a competitive neural network analysis is performed to calculate the matching rate; if the matching rate of the influent sample is less than the warning value of the competitive neural network analysis result, the influent sample is an abnormal sample; otherwise, the influent sample is a normal sample.

[0117] Specifically, the data of diluted industrial wastewater and normal influent samples were collected, including the pH value, COD CrThe measurement results of parameters such as TP and NH3-N were normal. Competitive neural network analysis was carried out. The three-dimensional fluorescence spectrum data of the influent sample were preprocessed according to the above S3 and S4. The matching rate of each sample was calculated according to the output results of the competitive neural network. The matching rate of the diluted industrial wastewater was lower than the warning value WT of the competitive neural network analysis results. EEMs , which is abnormal data; the matching rate of normal water samples is higher than 82.7%.

[0118]

[0119]

[0120] S7. Update the data in the standard library according to the water sample conditions of the influent sample.

[0121] Specifically, if the water sample of the influent sample is a normal sample, the data of the corresponding influent sample will be saved in the normal database, and the data of the influent sample with the longest time distance from the currently saved influent sample will be replaced; that is, the test data at the earliest time in database 2 will be replaced; and a new standard library will be formed.

[0122] If the water sample is an abnormal sample, the corresponding water sample data is saved to the abnormality database, and the corresponding three-dimensional fluorescence spectrum data is also saved. That is, it is saved to database 1, and the three-dimensional fluorescence data at the time of the abnormality is retained for further pollution source tracing and accountability.

[0123] A sewage treatment plant water inlet abnormality early warning system applies the sewage treatment plant water inlet abnormality early warning method described above.

[0124] The present invention provides a sewage treatment plant water inlet abnormality early warning system and method, which sets Rayleigh scattering and Raman scattering, adds the function of automatically searching the area where Rayleigh scattering and Raman scattering are located, and automatically removing Rayleigh scattering and Raman scattering. Traditional scattering subtraction requires automatic input of 15, 15, 10, 10 of the center line position of the first-level Raman scattering, second-level Raman scattering, first-level Rayleigh scattering, and second-level Rayleigh scattering. However, due to the difference in three-dimensional fluorescence spectra, it may result in that the input of 15, 15, 10, 10 cannot completely subtract the scattering. Therefore, in the present invention, all E x The data is scanned and the peak of each column is found. The position of the scattering and the position of the peak are matched. After the corresponding peak is matched, it is cut off according to the peak width.

[0125] Based on the three-dimensional fluorescence characteristics of the influent water quality of urban sewage treatment plants, the gain coefficient for processing the three-dimensional fluorescence grayscale image of the water quality of urban sewage treatment plants is determined to be 0.5. The three-dimensional fluorescence characteristic images of different types of water quality samples have their own characteristic responses, so each type of water quality sample has its own unique k value response. The present invention determines the k value for the influent water quality of sewage treatment plants based on the three-dimensional fluorescence spectrum characteristics of urban sewage treatment plants, thereby improving the accuracy of subsequent competitive neural network early warning results.

[0126] Since the output result of the subsequent competitive neural network is 1 or 2, if there is data with large differences in database 2, it will affect the classification of the competitive neural network. Therefore, the p-value statistical judgment of abnormal data is added in the present invention to ensure the accuracy of the subsequent competitive neural network warning results.

[0127] The physicochemical parameters and three-dimensional fluorescence spectral data are combined as the input of the competitive neural network. The results of physicochemical parameters such as COD, ammonia nitrogen, and total phosphorus are used to assist in monitoring the changes in the influent water quality of urban sewage treatment plants, effectively solving the problem that three-dimensional fluorescence spectroscopy cannot effectively monitor slaughterhouse wastewater containing oil and fat and washing wastewater containing detergents.

[0128] In other technical solutions, the output result of the competitive neural network is 1 to the number of sample categories, while in this patent, database 2 and the samples to be analyzed are actually samples of the same category 1. According to other technical solutions, the data result is only 1, but in order to realize the application of the competitive neural network in the early warning system, this patent adopts a hypothesis method. If an abnormality occurs, the abnormal water sample is a category 2 sample, and its output result is different from the Z samples in database 2, then the output result of the competitive neural network is set to 1 or 2, and the basis for judging whether the output result of the sample to be analyzed is the same as the majority of standard samples in data 2. At the same time, based on the actual sample test results of the urban sewage treatment plant, the early warning threshold of this method is set to 82.2%.

[0129] Other technical solutions generally collect batches of data and then update the data information in the database. However, this patent provides an iterative update function for the standard sample data of the standard library. Every time a normal water sample is tested, a sample in database 2 is replaced. If the water quality sample has always been a normal sample, a new sample database can be updated every 3 days, thereby solving the impact of periodic changes in the influent water quality of urban sewage treatment plants due to seasons, economic activities, etc.

[0130] This invention provides a real-time and effective early warning system for changes in the influent quality of urban sewage treatment plants, providing technical support for ensuring stable operation of sewage treatment plants. By retaining key water quality fingerprint information, this system can help quickly identify pollution sources, provide timely response measures, and hold people accountable when abnormal influent quality occurs.

[0131] Based on the disclosure and teachings of the above description, those skilled in the art may also make changes and modifications to the above embodiments. Therefore, the present invention is not limited to the specific embodiments disclosed and described above, and modifications and variations of the present invention should also fall within the scope of protection of the claims of the present invention. In addition, although certain specific terms are used in this description, these terms are only for convenience of description and do not constitute any limitation to the present invention.

Claims

1. A method for warning abnormal water inflow in a sewage treatment plant, characterized in that: The following steps are involved: S1. Install a real-time monitoring device at the water inlet of the sewage treatment plant to collect data from influent samples and conduct tests; S2. Setting a three-dimensional fluorescence spectrometer to regularly collect three-dimensional fluorescence spectrum data of the influent sample; S3, preprocessing the three-dimensional fluorescence spectrum data collected from the influent sample, removing and filling the fluorescence values in the Raman scattering and Rayleigh scattering regions in the three-dimensional fluorescence spectrum data, and then performing Gaussian smoothing processing; S4, normalizing the pre-processed three-dimensional fluorescence spectrum data, converting it into a grayscale image, and performing gain processing; S5. Establish a standard library of conventional parameters and three-dimensional fluorescence spectrum data of influent samples, and set early warning values; S6. Perform early warning analysis on the water sample condition of the influent sample, using conventional parameter test results and three-dimensional fluorescence spectrum characteristic data as input, and perform competitive neural network analysis. Based on whether the conventional parameter test results and the competitive neural network analysis results exceed the early warning value, determine whether the influent sample is abnormal; S7. Update the data in the standard library according to the water sample conditions of the influent sample.

2. A sewage treatment plant water inlet abnormality early warning method according to claim 1, characterized in that: In S1, the real-time monitoring device is a conventional parameter real-time monitoring device, and the data of the test influent sample are conventional parameters, including pH value, chemical oxygen demand, ammonia nitrogen content, and total phosphorus content.

3. A sewage treatment plant water inlet abnormality early warning method according to claim 2, characterized in that: In S2, a three-dimensional fluorescence spectrometer is set up in the sewage treatment plant to collect three-dimensional fluorescence spectrum data of the influent sample every time t, and the collection parameters are set, including scanning range, emission and excitation wavelength width, and scanning speed.

4. A sewage treatment plant water inlet abnormality early warning method according to claim 3, characterized in that: In the S3, three-dimensional fluorescence spectrum data of pure water is collected every 12 hours, and the three-dimensional fluorescence spectrum data of pure water is used as the three-dimensional fluorescence spectrum data of the blank sample; for the three-dimensional fluorescence spectrum data of the influent sample, the three-dimensional fluorescence spectrum data of the influent sample is collected once every time t, and the three-dimensional fluorescence spectrum data of the most recent blank sample is deducted from the collected three-dimensional fluorescence spectrum data of the influent sample, and then Rayleigh scattering and Raman scattering are deducted, and then the fluorescence value of the deducted area is filled by the Delaunay triangle interpolation method, and the data processed by the Delaunay triangle interpolation method is Gaussian smoothed.

5. A sewage treatment plant water inlet abnormality early warning method according to claim 4, characterized in that: In S4, the three-dimensional fluorescence spectrum data is normalized, and the specific formula is: Among them, FI (x，m) represents the fluorescence intensity after linear normalization at the excitation wavelength x nm and emission wavelength m nm, fi (x，m) It represents the fluorescence intensity obtained by scanning with a fluorescence spectrophotometer at an excitation wavelength of x nm and an emission wavelength of m nm. min is the minimum fluorescence intensity after removing the scattered light, fi max is the maximum fluorescence intensity after excluding scattering; The normalized three-dimensional fluorescence spectrum data is converted into grayscale data with a pixel size of 534×609, and the grayscale data is subjected to gain processing. The specific formula is: Where X is the fluorescence intensity F after linear normalization (n,m) value; f(X) is the fluorescence intensity value obtained after transformation by the S-shaped growth curve deformation function, k is the gain coefficient, which is a positive real number assigned to control the degree of fluorescence intensity scaling. Through testing of multiple influent samples, the k value of the sewage treatment plant influent water quality sample is 0.

5.

6. A sewage treatment plant water inlet abnormality early warning method according to claim 5, characterized in that: In said S5, a standard library of conventional parameters and three-dimensional fluorescence spectrum data of influent samples is established, and an early warning value is set; specifically, the following steps are performed: Set the standard parameter range of chemical oxygen demand, ammonia nitrogen content, and total phosphorus content, If the conventional parameters of chemical oxygen demand, ammonia nitrogen content, and total phosphorus content obtained by S4 fall within the standard parameter range, they are classified as normal data; If the conventional parameters of chemical oxygen demand, ammonia nitrogen content, and total phosphorus content obtained by S4 exceed the standard parameter range, they will be listed as abnormal data; For normal data that meet the standard parameter range, the p-value statistical method is used to determine whether the three-dimensional fluorescence spectrum data is normal data; The conventional parameters of chemical oxygen demand, ammonia nitrogen content, total phosphorus content and three-dimensional fluorescence spectrum data that are normal after testing are placed in the normal database; The conventional parameters of chemical oxygen demand, ammonia nitrogen content, total phosphorus content and three-dimensional fluorescence spectrum data that are abnormal data after inspection are placed in the abnormal database; Fill the normal database so that the number of influent samples corresponding to the normal data in the normal database reaches Z, and use the normal database formed by Z normal data as the standard database.

7. A sewage treatment plant water inlet abnormality early warning method according to claim 6, characterized in that: Set warning values for conventional parameters and three-dimensional fluorescence spectrum data; The warning value settings for conventional parameters specifically include: Calculate the average and standard deviation of the chemical oxygen demand, ammonia nitrogen content, and total phosphorus content corresponding to all influent samples in the normal database respectively. Add three times the corresponding standard deviation to the average values of the calculated chemical oxygen demand, ammonia nitrogen content, and total phosphorus content. The final corresponding values are used as the warning values of chemical oxygen demand, ammonia nitrogen content, and total phosphorus content respectively; Collect the conventional parameter measurement results and three-dimensional fluorescence spectrum data of M normal influent samples, and process the three-dimensional fluorescence spectrum data according to S3 and S4; Each 3D fluorescence spectrum data set in the M processed data is combined with the Z 3D fluorescence spectrum data sets in the normal database to form Z+1 3D fluorescence spectrum data sets of influent samples. For each 3D fluorescence spectrum data set, the principal component analysis method is used to extract the first S principal component values in the 3D fluorescence spectrum data set as the eigenvalues of the 3D fluorescence spectrum data set. The characteristic values of the three-dimensional fluorescence spectrum data and the corresponding conventional parameters of Z+1 influent samples are used as the input of the competitive neural network. After the competitive neural network analysis, the output item is obtained and the output item is set to 0 or 1. If the output result is 1, divide the Z+1 influent samples into Z influent samples and +1 influent samples, where the Z influent samples are the corresponding Z influent samples in the normal database, and the +1 influent sample is one of the M influent samples; then calculate the number of influent samples in the output results of the Z influent samples that have the same output result as the +1 influent sample; Perform Q competitive neural network analyses on each Z+1 influent sample; calculate the average number of influent samples that have the same output as the +1 influent sample; Divide the average of the number of influent samples that have the same output as the +1 influent sample by Z to obtain the matching rate of one of the influent samples; The matching rate of each of the M influent samples is calculated in sequence, and then the average matching rate of the M influent samples is calculated. The average matching rate of the M influent samples is used as the warning value of the competitive neural network analysis result.

8. A sewage treatment plant water inlet abnormality early warning method according to claim 7, characterized in that: Collect the test results of routine parameters and three-dimensional fluorescence spectrum data of the influent sample. If the pH value of the influent sample is greater than 9 or less than 6, the influent sample is considered an abnormal sample. If the pH value of the influent sample is normal, determine whether the measured values of the chemical oxygen demand, ammonia nitrogen content, and total phosphorus content of the influent sample are greater than the corresponding warning values of the chemical oxygen demand, ammonia nitrogen content, and total phosphorus content. If the measured result of any of the conventional parameters is greater than the corresponding warning value of the parameter, the sample is determined to be an abnormal sample; If the measurement results of the pH value, chemical oxygen demand, ammonia nitrogen content, and total phosphorus content of the influent sample are all normal, a competitive neural network analysis is performed to calculate the matching rate; if the matching rate of the influent sample is less than the warning value of the competitive neural network analysis result, the influent sample is an abnormal sample; otherwise, the influent sample is a normal sample.

9. A sewage treatment plant water inlet abnormality early warning method according to claim 8, characterized in that: In S7, the data of the standard library is updated: If the influent sample is a normal sample, the data of the corresponding influent sample is saved in the normal database, and the data of the influent sample with the longest time from the currently saved influent sample is replaced to form a new standard database; If the influent sample is an abnormal sample, the data of the corresponding influent sample is saved in the abnormality database, and the data of the corresponding three-dimensional fluorescence spectrum data is saved at the same time.

10. A sewage treatment plant water inlet abnormality early warning system, characterized in that: A sewage treatment plant water inlet abnormality early warning method as described in any one of claims 1 to 9 is applied.

Images