CN118310984A - Quick measuring method for nitrite nitrogen in high salinity gradient environment by using spectrometry - Google Patents

Abstract

The invention relates to a rapid measurement method of nitrite nitrogen in seawater in a high salinity gradient environment by a spectrometry, which comprises the following steps: collecting ultraviolet visible absorption spectrum of a mixed solution of standard nitrate nitrogen and nitrite nitrogen, calculating absorbance and performing first-order differential treatment on the absorbance; screening out first-order differential spectrum data of a wave band required by model training in a wave band after 235nm by using a correlation coefficient search algorithm to form a training set; establishing an inversion model of the nitrite nitrogen concentration of the machine, and carrying out optimization training on model parameters by using a training set; and collecting an ultraviolet visible absorption spectrum of a seawater sample to be detected, calculating absorbance, subtracting turbidity interference, performing first-order differential treatment on the seawater sample, and inputting the seawater sample to be detected into a trained nitrite nitrogen concentration inversion model to obtain a nitrite nitrogen content prediction result. The method is suitable for the conditions of multi-salinity gradient of near-shore complex seawater and nitrate nitrogen interference, can rapidly determine the nitrite nitrogen content, and provides reference for instrument development.

Description

A rapid measurement method for nitrite nitrogen in seawater with high salinity gradient using spectroscopy

Technical Field

The invention belongs to the technical field of rapid detection of seawater quality parameters, and relates to a method for rapid measurement of nitrite nitrogen in seawater in a high-salinity gradient environment using a spectral method.

Background technique

Nitrite nitrogen is an important nitrogen source in coastal seawater. It is usually produced by nitrification and plays an important biogeochemical role in marine ecosystems. Under certain environmental conditions, such as high concentrations of pollutant emissions, eutrophication and hypoxia, the accumulation and transformation of nitrite nitrogen may lead to harmful effects. Monitoring and controlling the concentration of nitrite nitrogen in water bodies is one of the important environmental protection and water quality management measures.

The traditional method uses the naphthylethylenediamine spectrophotometry to measure nitrite nitrogen in seawater, which requires the preparation of a color developer, has cumbersome steps, is time-consuming, and is prone to secondary pollution. The absorption spectroscopy method is fast and reagent-free, but in the ultraviolet-visible absorption spectrum, the absorption peaks of nitrate nitrogen and nitrite nitrogen overlap, which greatly interferes with the measurement of nitrite nitrogen. In addition, in complex coastal seawater, the salinity gradient varies widely, and factors such as interference from multiple dissolved organic matter can lead to ambiguity in the measurement results. Therefore, it is necessary to consider a variety of interfering factors in complex coastal seawater to accurately measure the concentration of nitrite nitrogen.

Summary of the invention

In order to solve the above technical problems, the purpose of the present invention is to provide a method for rapid measurement of nitrite nitrogen in seawater in a high-salinity gradient environment using a spectral method. Based on ultraviolet-visible absorption spectroscopy, the first-order difference is used to remove interference, and a multi-layer perceptron modeling method with correlation coefficient search is used to quickly determine the nitrite nitrogen content.

The present invention provides a method for rapidly measuring nitrite nitrogen in seawater in a high-salinity gradient environment using a spectral method, comprising:

Step 1: Collect the UV-visible absorption spectrum of the standard nitrate nitrogen and nitrite nitrogen mixed solution, calculate the absorbance and perform first-order difference processing on it;

Step 2: Use the correlation coefficient search algorithm to select the first-order difference spectral data of the band required for model training in the band after 235nm to form a training set;

Step 3: Establish a multi-layer perceptron nitrite nitrogen concentration inversion model and use the training set to optimize the model parameters;

Step 4: Collect the UV-visible absorption spectrum of the seawater sample to be tested, calculate the absorbance, perform first-order difference processing on it after deducting turbidity interference, and then input it into the trained multi-layer perceptron nitrite nitrogen concentration inversion model to obtain the predicted result of nitrite nitrogen content.

Furthermore, the step 1 is specifically as follows:

Step 1.1: Prepare multiple sets of mixed solutions of nitrate nitrogen and nitrite nitrogen, collect UV-visible absorption spectra, calculate the absorbance using the Lambert-Beer law, and use the Savitzky-Golay convolution smoothing algorithm to remove random noise;

Step 1.2: Perform first-order difference processing on the absorbance of the denoised UV-visible absorption spectrum.

Furthermore, the absorbance is calculated using the Lambert-Beer law in step 1.1 as follows:

Where _Aλ is the absorbance value at wavelength λ, I is the measured light intensity at wavelength λ, _Iw is the light intensity of pure water at wavelength λ, and _Id is the dark spectrum intensity of the spectrometer at wavelength λ.

Furthermore, in step 1.1, the Savitzky-Golay convolution smoothing algorithm is used to remove random noise as follows:

The Savitzky-Golay convolution smoothing algorithm was used to fit the absorbance data within the filter window, with a window width of 5, a polynomial order of 2, and a moving step of 1;

The filter operator used is For five consecutive equally spaced absorbance values _An-2 , An _-1 , _An , _An+1 and An _+2, the following denoising process is performed:

Where A _n is the absorbance value of the sequence number n in the spectrum, is the absorbance value with serial number n in the denoised spectrum.

Furthermore, the step 1.2 is specifically as follows:

in, is the first-order difference spectrum value with sequence number n in the denoised spectrum, is the absorbance value of sequence number n in the denoised spectrum, is the absorbance value of sequence number n-1 in the denoised spectrum.

Furthermore, the step 2 is specifically as follows: calculating the Pearson correlation coefficient between the first-order difference spectrum value of each wavelength in the band after 235 nm and the concentration, and selecting the first-order difference spectrum data of several wavelengths with the strongest correlation to form a training set;

The Pearson correlation coefficient between the first-order difference spectrum value and the concentration was calculated according to the following formula:

in, is the denoised first-order difference spectrum value at wavelength λ in the i-th group of standard sample mixed solution, m is the number of groups of standard sample mixed solution, is the average value of the first-order difference spectrum after denoising at wavelength λ in all standard sample mixed solutions; is the average concentration of all standard sample mixed solutions, _Ci is the concentration of the ith group of standard sample mixed solutions, and p is the correlation coefficient between the first-order difference spectrum value and the concentration. When its absolute value is greater than 0.8, it is considered to be strongly correlated. The first-order difference spectrum data of the bands with strong correlation are screened out to form the training set.

Furthermore, the step 3 is specifically as follows:

A multi-layer perceptron nitrite nitrogen concentration inversion model was established, using the LeakyReLU activation function, the MSE loss function, and the Adam optimization algorithm for optimization; the model is specifically as follows:

C _pre =H(H(EW ₁ )W ₂ )

Among them, _Cpre is the nitrite nitrogen concentration vector predicted by the model, E is the input first-order difference spectral data, _W1 and _W2 are the weight matrices of the model, and H is the LeakyReLU activation function.

Furthermore, the step 4 is specifically as follows:

Step 4.1: Collect the UV-visible absorption spectrum of the seawater sample to be tested, calculate the absorbance, and perform smoothing and denoising;

Step 4.2: subtract the turbidity interference value from the absorbance of the smoothed and denoised UV-visible absorption spectrum to deduct the turbidity interference;

Step 4.3: After the absorbance after deducting the turbidity interference is processed by the first-order difference, it is input into the trained multi-layer perceptron nitrite nitrogen concentration inversion model to obtain the predicted result of nitrite nitrogen content.

Furthermore, the step 4.2 of deducting turbidity interference is specifically as follows:

Step 4.2.1: Use formazine to prepare turbidity standard solutions, with a total of 15 turbidity standard solution samples with concentration gradients of 0.1, 0.2, 0.4, 0.6, 0.8, 1, 2, 3, 4, 5, 6, 7, 8, 9, and 10 NTU;

Step 4.2.2: Select the denoised absorbance values of five wavelengths with 550 nm as the central wavelength in each turbidity standard solution sample to form a matrix F _train , and the turbidity values of each turbidity standard solution sample to form a vector TSS _train ;

Step 4.2.3: Use the normal equation to determine the coefficients M of the linear regression model, as follows:

Step 4.2.4: Select the denoised absorbance values of the five wavelengths with 550nm as the central wavelength of the seawater sample to be tested to form a vector F, and calculate the predicted turbidity value using the following formula:

TSS _pre = FM

Among them, TSS _pre is the predicted turbidity value;

Step 4.2.5: Let the predicted absorbance value vector at 235-400 nm be A _pre , and the fitting formula is as follows:

A _pre =K(TSS _pre )+B

Among them, A _pre is the predicted absorbance value vector, K and B are coefficient vectors, and the fitting is performed through the turbidity standard solution;

Step 4.2.6: After obtaining the predicted absorbance value vector A _pre , perform turbidity interference subtraction. The absorbance value vector of the absorption spectrum at 235nm-400nm after denoising is recorded as A, and the interference is subtracted according to the following formula:

A _after ＝AA _before

Among them, A _after is the absorbance value vector after deducting turbidity interference.

The method for rapidly measuring nitrite nitrogen in seawater in a high-salinity gradient environment using a spectral method of the present invention has at least the following beneficial effects:

1. The rapid measurement method of nitrite nitrogen of the present invention, based on the first-order difference method and the multi-layer perceptron, can effectively solve the problem of spectral aliasing interference between nitrate nitrogen and nitrite nitrogen, quickly determine the nitrite nitrogen content, save manpower and material costs, and improve monitoring efficiency;

2. The present invention can solve the problem that the interference of bromide ions and chloride ions caused by the wide range of salinity changes in complex nearshore seawater is difficult to accurately deduct, and improves the stability of nitrite nitrogen determination by spectrometry.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG1 is a flow chart of a method for rapidly measuring nitrite nitrogen in seawater in a high-salinity gradient environment using a spectral method according to the present invention.

Detailed ways

The present invention will be further described below in conjunction with the accompanying drawings.

As shown in FIG1 , a method for rapidly measuring nitrite nitrogen in seawater in a high salinity gradient environment using a spectral method of the present invention comprises:

Step 1: Collect the UV-visible absorption spectrum of the standard nitrate nitrogen and nitrite nitrogen mixed solution, calculate the absorbance and perform first-order difference processing on it. The specific steps of step 1 are:

Step 1.1: Prepare a mixed solution of nitrate nitrogen and nitrite nitrogen, collect UV-visible absorption spectra, calculate the absorbance using the Lambert-Beer law, and use the Savitzky-Golay convolution smoothing algorithm to remove random noise.

In the specific implementation, sodium nitrite is used as the standard substance for nitrite nitrogen, and potassium nitrate is used as the standard substance for nitrate nitrogen. The configuration concentration table is shown in Table 1. There are a total of 49 sets of data for the 2 factor 7 level combination.

Table 1 Nitrate nitrogen and nitrite nitrogen calibration solution sample concentration table

In a specific implementation, the range of the UV-visible absorption spectrum is 210-710 nm, the interval is 0.625 nm, and there are 800 data points in total.

In specific implementation, the absorbance is calculated using the Lambert-Beer law as follows:

Where _Aλ is the absorbance value at wavelength λ, I is the measured light intensity at wavelength λ, _Iw is the light intensity of pure water at wavelength λ, and _Id is the dark spectrum intensity of the spectrometer at wavelength λ.

In specific implementation, the Savitzky-Golay convolution smoothing algorithm is used to remove random noise as follows:

The Savitzky-Golay convolution smoothing algorithm was used to fit the absorbance data within the filter window, with a window width of 5, a polynomial order of 2, and a moving step of 1;

The filter operator used is For five consecutive equally spaced absorbance values _An-2 , An _-1 , _An , _An+1 and An _+2, the following denoising process is performed:

Where A _n is the absorbance value of the sequence number n in the spectrum, is the absorbance value with serial number n in the denoised spectrum.

Step 1.2: Perform first-order difference processing on the absorbance of the denoised UV-visible absorption spectrum. The step 1.2 specifically performs first-order difference processing according to the following formula:

in, is the first-order difference spectrum value with sequence number n in the denoised spectrum, is the absorbance value of sequence number n in the denoised spectrum, is the absorbance value of sequence number n-1 in the denoised spectrum.

Step 2: Use the correlation coefficient search algorithm to select the first-order difference spectral data of the band required for model training in the band after 235nm to form a training set. The specific steps of step 2 are as follows:

The Pearson correlation coefficient between the first-order difference spectral value and the concentration at each wavelength in the band after 235 nm was calculated, and the first-order difference spectral data of several bands with the strongest correlation were selected to form a training set.

The Pearson correlation coefficient between the first-order difference spectrum value and the concentration was calculated according to the following formula:

in, is the denoised first-order difference spectrum value at wavelength λ in the i-th group of standard sample mixed solution, m is the number of groups of standard sample mixed solution, is the average value of the first-order difference spectrum after denoising at wavelength λ in all standard sample mixed solutions; is the average concentration of all standard sample mixed solutions, _Ci is the concentration of the ith group of standard sample mixed solutions, and p is the correlation coefficient between the first-order difference spectrum value and the concentration. When its absolute value is greater than 0.8, it is considered to be strongly correlated. The first-order difference spectrum data of the bands with strong correlation are screened out to form the training set.

Step 3: Establish a multi-layer perceptron nitrite nitrogen concentration inversion model, and use the training set to optimize the model parameters. Step 3 is specifically as follows:

A multi-layer perceptron nitrite nitrogen concentration inversion model was established, using the LeakyReLU activation function, the MSE loss function, and the Adam optimization algorithm for optimization; the model is specifically as follows:

C _pre =H(H(EW ₁ )W ₂ )

Among them, _Cpre is the nitrite nitrogen concentration vector predicted by the model, E is the input first-order difference spectral data, _W1 and _W2 are the weight matrices of the model, and H is the LeakyReLU activation function.

Step 4: Collect the UV-visible absorption spectrum of the seawater sample to be tested, calculate the absorbance, perform first-order difference processing on it after deducting the turbidity interference, and then input it into the trained multi-layer perceptron nitrite nitrogen concentration inversion model to obtain the prediction result of nitrite nitrogen content. The specific step 4 is:

Step 4.1: Collect the UV-visible absorption spectrum of the seawater sample to be tested, calculate the absorbance, and perform smoothing and denoising. The specific operation process is the same as step 1.

Step 4.2: Subtract the turbidity interference value from the absorbance of the smoothed and denoised UV-visible absorption spectrum to deduct the turbidity interference, specifically:

Step 4.2.1: Use formazine to prepare turbidity standard solutions. A total of 15 turbidity standard solution samples with concentration gradients of 0.1, 0.2, 0.4, 0.6, 0.8, 1, 2, 3, 4, 5, 6, 7, 8, 9, and 10 NTU were prepared.

Step 4.2.2: Select the denoised absorbance values of five wavelengths with 550 nm as the central wavelength in each turbidity standard solution sample to form a matrix F _train , and the turbidity values of each turbidity standard solution sample to form a vector TSS _train .

Step 4.2.3: Use the normal equation to determine the coefficients M of the linear regression model, as follows:

Step 4.2.4: Select the denoised absorbance values of the five wavelengths with 550nm as the central wavelength of the seawater sample to be tested to form a vector F, and calculate the predicted turbidity value using the following formula:

TSS _pre = FM

Among them, TSS _pre is the predicted turbidity value.

Step 4.2.5: Let the predicted absorbance value vector at 235-400 nm be A _pre , and the fitting formula is as follows:

A _pre =K(TSS _pre )+B

Among them, A _pre is the predicted absorbance value vector, K and B are coefficient vectors, and fitting is performed through the turbidity standard solution.

Step 4.2.6: After obtaining the predicted absorbance value vector A _pre , perform turbidity interference subtraction. The absorbance value vector of the absorption spectrum at 235nm-400nm after denoising is recorded as A, and the interference is subtracted according to the following formula:

A _after ＝AA _before

Among them, A _after is the absorbance value vector after deducting turbidity interference.

Step 4.3: After the absorbance after deducting the turbidity interference is processed by first-order difference, it is input into the trained multi-layer perceptron nitrite nitrogen concentration inversion model to obtain the prediction result of nitrite nitrogen content. The first-order difference processing process is the same as the corresponding part in step 1.

The above description is only a preferred embodiment of the present invention and is not intended to limit the concept of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims (9)

1. A method for rapid measurement of nitrite nitrogen in seawater in a high salinity gradient environment using a spectral method, characterized by comprising: Step 1: Collect the UV-visible absorption spectrum of the standard nitrate nitrogen and nitrite nitrogen mixed solution, calculate the absorbance and perform first-order difference processing on it; Step 2: Use the correlation coefficient search algorithm to select the first-order difference spectral data of the band required for model training in the band after 235nm to form a training set; Step 3: Establish a multi-layer perceptron nitrite nitrogen concentration inversion model and use the training set to optimize the model parameters; Step 4: Collect the UV-visible absorption spectrum of the seawater sample to be tested, calculate the absorbance, perform first-order difference processing on it after deducting turbidity interference, and then input it into the trained multi-layer perceptron nitrite nitrogen concentration inversion model to obtain the predicted result of nitrite nitrogen content. 2. The method for rapid measurement of nitrite nitrogen in seawater in a high salinity gradient environment using spectroscopy as claimed in claim 1, wherein step 1 is specifically as follows: Step 1.1: Prepare multiple sets of mixed solutions of nitrate nitrogen and nitrite nitrogen, collect UV-visible absorption spectra, calculate the absorbance using the Lambert-Beer law, and use the Savitzky-Golay convolution smoothing algorithm to remove random noise; Step 1.2: Perform first-order difference processing on the absorbance of the denoised UV-visible absorption spectrum. 3. The method for rapid measurement of nitrite nitrogen in seawater in a high-salinity gradient environment using spectroscopy as claimed in claim 1, characterized in that the absorbance calculated using the Lambert-Beer law in step 1.1 is specifically: Where _Aλ is the absorbance value at wavelength λ, I is the measured light intensity at wavelength λ, _Iw is the light intensity of pure water at wavelength λ, and _Id is the dark spectrum intensity of the spectrometer at wavelength λ. 4. The method for rapid measurement of nitrite nitrogen in seawater in a high-salinity gradient environment using spectroscopy as claimed in claim 1, characterized in that the Savitzky-Golay convolution smoothing algorithm is used to remove random noise in step 1.1: The Savitzky-Golay convolution smoothing algorithm was used to fit the absorbance data within the filter window, with a window width of 5, a polynomial order of 2, and a moving step of 1; The filter operator used is For five consecutive equally spaced absorbance values _An-2 , An _-1 , _An , _An+1 and An _+2, the following denoising process is performed: Where A _n is the absorbance value of the sequence number n in the spectrum, is the absorbance value with serial number n in the denoised spectrum. 5. The method for rapid measurement of nitrite nitrogen in seawater in a high salinity gradient environment using spectroscopy as claimed in claim 4, wherein step 1.2 is specifically as follows: in, is the first-order difference spectrum value with sequence number n in the denoised spectrum, is the absorbance value of sequence number n in the denoised spectrum, is the absorbance value of sequence number n-1 in the denoised spectrum. 6. The method for rapid measurement of nitrite nitrogen in seawater in a high-salinity gradient environment using spectroscopy as claimed in claim 1, characterized in that the step 2 specifically comprises: calculating the Pearson correlation coefficient between the first-order difference spectral value and the concentration at each wavelength in the band after 235 nm, and selecting the first-order difference spectral data of several wavelengths with the strongest correlation to form a training set; The Pearson correlation coefficient between the first-order difference spectrum value and the concentration was calculated according to the following formula: in, is the denoised first-order difference spectrum value at wavelength λ in the i-th group of standard sample mixed solution, m is the number of groups of standard sample mixed solution, is the average value of the first-order difference spectrum after denoising at wavelength λ in all standard sample mixed solutions; is the average concentration of all standard sample mixed solutions, _Ci is the concentration of the ith group of standard sample mixed solutions, and p is the correlation coefficient between the first-order difference spectrum value and the concentration. When its absolute value is greater than 0.8, it is considered to be strongly correlated. The first-order difference spectrum data of the bands with strong correlation are screened out to form the training set. 7. The method for rapid measurement of nitrite nitrogen in seawater in a high salinity gradient environment using spectroscopy as claimed in claim 1, wherein step 3 is specifically: A multi-layer perceptron nitrite nitrogen concentration inversion model was established, using the LeakyReLU activation function, the MSE loss function, and the Adam optimization algorithm for optimization; the model is specifically as follows: C _pre =H(H(EW ₁ )W ₂ ) Among them, _Cpre is the nitrite nitrogen concentration vector predicted by the model, E is the input first-order difference spectral data, _W1 and _W2 are the weight matrices of the model, and H is the LeakyReLU activation function. 8. The method for rapid measurement of nitrite nitrogen in seawater in a high salinity gradient environment using spectroscopy as claimed in claim 1, wherein step 4 is specifically: Step 4.1: Collect the UV-visible absorption spectrum of the seawater sample to be tested, calculate the absorbance, and perform smoothing and denoising; Step 4.2: Subtract the turbidity interference value from the absorbance of the smoothed and denoised UV-visible absorption spectrum to deduct the turbidity interference; Step 4.3: After the absorbance after deducting the turbidity interference is processed by the first-order difference, it is input into the trained multi-layer perceptron nitrite nitrogen concentration inversion model to obtain the predicted result of nitrite nitrogen content. 9. The method for rapid measurement of nitrite nitrogen in seawater in a high-salinity gradient environment using spectroscopy as claimed in claim 8, wherein the step 4.2 of deducting turbidity interference is specifically as follows: Step 4.2.1: Use formazine to prepare turbidity standard solutions, with a total of 15 turbidity standard solution samples with concentration gradients of 0.1, 0.2, 0.4, 0.6, 0.8, 1, 2, 3, 4, 5, 6, 7, 8, 9, and 10 NTU; Step 4.2.2: Select the denoised absorbance values of five wavelengths with 550 nm as the central wavelength in each turbidity standard solution sample to form a matrix F _train , and the turbidity values of each turbidity standard solution sample to form a vector TSS _train ; Step 4.2.3: Use the normal equation to determine the coefficients M of the linear regression model, as follows: Step 4.2.4: Select the denoised absorbance values of the five wavelengths with 550nm as the central wavelength of the seawater sample to be tested to form a vector F, and calculate the predicted turbidity value using the following formula: TSS _pre = FM Among them, TSS _pre is the predicted turbidity value; Step 4.2.5: Let the predicted absorbance value vector at 235-400 nm be A _pre , and the fitting formula is as follows: A _pre =K(TSS _pre )+B Among them, A _pre is the predicted absorbance value vector, K and B are coefficient vectors, and the fitting is performed through the turbidity standard solution; Step 4.2.6: After obtaining the predicted absorbance value vector A _pre , perform turbidity interference subtraction. The absorbance value vector of the absorption spectrum at 235nm-400nm after denoising is recorded as A, and the interference is subtracted according to the following formula: A _after ＝AA _before Among them, A _after is the absorbance value vector after deducting turbidity interference.