A monitoring device and method for carbon emission flux of sewage pipe well

Abstract

The present invention discloses a monitoring device for carbon emission flux of a sewage pipe well and a monitoring method based on the device. The device comprises a first carbon emission concentration monitoring device installed on the inner wall of the well, a flux box suspended above the water surface in the well, one side of the flux box is provided with an air vent connected to the gas in the well, and the other side introduces the gas in the box into a second carbon emission concentration measuring device outside the well through an air pipe; the method comprises obtaining carbon emission concentration data measured by the first and second carbon emission concentration monitoring devices within a unit time period, calculating the carbon emission flux within the time period according to the carbon emission concentration data measured by the second carbon emission concentration monitoring device, performing data-driven modeling based on a data set consisting of the carbon emission concentration data measured by the first carbon emission concentration monitoring device and the carbon emission flux data of step 2 to construct a carbon emission prediction model, and using the carbon emission prediction model to perform real-time carbon emission flux prediction on new carbon emission concentration measurement data provided by the first carbon emission concentration monitoring device deployed on the inner wall of the well. The present invention uses a carbon emission prediction model constructed using data collected in the early stage to accurately predict the carbon emission flux of the new carbon emission concentration measured by the first carbon emission concentration monitoring equipment. The first carbon emission concentration monitoring equipment has good portability and can be widely deployed in various sewage pipe wells, thereby realizing real-time monitoring of the carbon emission flux of regional sewage wells.

Description

Technical Field

[0001] The present invention relates to the technical field of carbon emission monitoring, and in particular to a monitoring device and a monitoring method for carbon emission flux of a sewage pipe well. Background Art

[0002] As a key component of urban infrastructure, the carbon emissions of sewage collection systems (mainly sewage pipe networks) are directly related to the future sustainable development of cities. Carbon emissions from sewage collection systems mainly come from anaerobic fermentation reactions of organisms, that is, a large amount of organic matter is decomposed by microbial communities attached to pipe walls and pipe sediments through anaerobic fermentation reactions in the anaerobic environment formed during pipeline transmission, thereby producing carbon emissions mainly in the form of methane (CH4). If the carbon emissions in sewage pipe networks can be accurately, reliably and conveniently monitored and estimated (measured in flux, in units of kg CO2eq / (m 2 *d)), from the perspective of "source, site and factory integration", it can provide important support for reducing pollution and carbon emissions as well as improving quality and efficiency of the entire urban sewage treatment system.

[0003] Although carbon emissions are actually produced in sewage pipes, they will flow into sewage wells with the flow of sewage and accumulate in sewage wells (also known as inspection wells), which then become the main place where carbon emissions from urban sewage systems are released into the atmosphere. In addition, from the perspective of monitoring feasibility, it is easier to operate and safer to arrange the equipment in sewage wells (such as attached to the well wall) than to install the relevant monitoring equipment on the wall of the sewage pipe.

[0004] However, the current monitoring methods of carbon emission flux in sewage wells still have the following problems:

[0005] (1) Conventional flux analysis monitoring methods are accurate but difficult to deploy in real time: The mainstream flux monitoring method for carbon emission analysis in the current sewage field is the flux box method (also known as the static box method), which is to float the box above the water surface in the well through a tripod to collect and analyze gases. The flux box is usually a relatively sealed container that can be placed anywhere in the ecosystem. When the box is placed above the gas source, the gas concentration in the box will change due to the gas exchange inside and outside the box. The gas flux is calculated by comparing the changes in gas concentration inside and outside the box over a certain period of time. Although the flux box method is recognized as an accurate method for obtaining carbon emission flux, its steps are cumbersome, there are many parts, and it relies on manual well drilling on-site operations. As a result, this time-consuming and labor-intensive on-site monitoring form cannot be applied to large-scale sewage pipe well monitoring operations in urban sewage pipe network areas, nor can it realize the construction and management of long-term online intelligent monitoring systems in the future.

[0006] (2) Current status of online monitoring with concentration as the main and flux as the auxiliary: The mainstream carbon emission monitoring equipment in the sewage industry (such as infrared spectroscopy equipment) on the market mainly monitors the concentration of carbon emissions (unit: ppm). It is rare to find equipment that directly monitors flux. In addition, flux monitoring equipment is generally larger in size and more complicated to maintain than concentration monitoring equipment.

[0007] (3) The current online monitoring equipment for well concentration cannot be used directly and accurately for flux calculation: The sewage well is a typical semi-closed system due to the continuous flow of water in the upstream and downstream pipes connected to it. The concentration curve in the semi-closed system will produce concentration drops or large fluctuations due to factors such as uncontrollable air flow disturbances in the well (affecting gas mass transfer) and hydraulic turbulence (affecting gas release), which will cause the gas concentration change rate to appear negative or its absolute value to be disturbed and reduced. At this time, if the flux formula based on the gas concentration change rate in a conventional closed system is directly used for calculation, either an unreasonable negative flux or an inaccurate flux with a large deviation will be obtained. Summary of the invention

[0008] Purpose of the invention: The purpose of the present invention is to provide a monitoring device for the carbon emission flux of a sewage pipe well, and to provide a monitoring method for the carbon emission flux of a sewage pipe well based on the monitoring device, so as to realize the intelligent and accurate prediction of the carbon emission flux in the well.

[0009] Technical solution: To achieve the above-mentioned purpose, a device for monitoring carbon emission flux of a sewage pipe well described in the present invention comprises a first carbon emission concentration monitoring device installed on the inner wall of the well and a flux box suspended above the water surface in the well, wherein one side of the flux box is provided with an air vent connected to the gas in the well, and the other side is provided with an air pipe, and the gas in the box is introduced into the second carbon emission concentration measuring device outside the well through the air pipe.

[0010] Wherein, a gas stirring mechanism is provided in the flux box, and the length of the gas pipe is less than or equal to 1.2 times of the well depth.

[0011] The monitoring method based on the above monitoring device comprises the following steps:

[0012] Step 1: Obtain carbon emission concentration data measured by the first carbon emission concentration monitoring device and the second carbon emission concentration monitoring device within a unit time period;

[0013] Step 2: Calculate the carbon emission flux during the period according to the carbon emission concentration data measured by the second carbon emission concentration monitoring device;

[0014] Step 3: Perform data-driven modeling based on the data set consisting of the carbon emission concentration data measured by the first carbon emission concentration monitoring device and the carbon emission flux data in step 2 to construct a carbon emission prediction model;

[0015] Step 4: Use the carbon emission prediction model to perform real-time prediction of carbon emission flux based on the new carbon emission concentration measurement data provided by the first carbon emission concentration monitoring device deployed on the inner wall of the well.

[0016] Among them, the method for obtaining data described in step 1 is: randomly select multiple sewage wells in a certain area to deploy the monitoring device, and within a unit time period, at intervals of t seconds, synchronously start the first carbon emission concentration monitoring device and the second carbon emission concentration monitoring device in all monitoring devices to directly measure the carbon emission concentration in the well and the carbon emission concentration in the gas drawn out through the flux box, respectively, and each measurement lasts for t' seconds.

[0017] The method for calculating the carbon emission flux in step 2 is: first, plot the carbon emission concentration change rate in the box according to the measured carbon emission concentration data, and then calculate the emission flux according to the concentration change rate:

[0018]

[0019] Where J is the carbon emission flux, unit: g GHG / (m 2 *d); m is the rate of change of carbon emission concentration in the flux box, unit: mgGHG / (m 3 *min); V is the effective volume of the flux box, unit: m 3 ; A is the bottom area covered by the flux box, unit: m 2 .

[0020] Among them, the data set described in step 3 is a number of data pairs collected by repeating steps 1-2 several times, and each data pair consists of a feature and a target, wherein the feature is concentration curve data composed of carbon emission concentration data measured by the first carbon emission concentration monitoring device in a group of monitoring devices within a unit time period, and the target is carbon emission flux data calculated from the carbon emission concentration data measured by the first carbon emission concentration monitoring device in a certain group of monitoring devices within the unit time period.

[0021] The method for constructing the carbon emission prediction model in step 3 is:

[0022] Step 301: Data processing: After preprocessing, data enhancement, and standardization, several groups of data pairs are randomly divided into training sets and validation sets;

[0023] Step 302: Model training: Use the neural network model to learn the relationship between the carbon emission concentration change curve and the carbon emission flux in the training set data, and establish a carbon emission prediction model based on the relationship between concentration and flux. This model is a non-explicit black box model. The black box model can output a carbon emission flux prediction for the input carbon emission concentration data using the concentration curve direct flux method, specifically:

[0024] Jp =F(c)=F(c1,c2,...);

[0025] In the formula, J p is the carbon emission flux of a sewage well, unit: g GHG / (m 2 *d); c1, c2, ... are the concentration curve data obtained by the first carbon emission concentration monitoring device deployed in the sewage well, that is, a column of time series data, in ppm; where F(.) is a non-explicit black box model;

[0026] Step 303: Model optimization: Use cross-validation based on the validation set to evaluate the generalization ability of the carbon emission prediction model and avoid overfitting; use evaluation indicators including mean square error, R 2 etc. to evaluate the performance of carbon emission prediction models.

[0027] The preprocessing in step 301 refers to identifying abnormal values ​​for several groups of data pairs based on the 3-times standard deviation method. The specific process is as follows:

[0028]

[0029]

[0030] In the formula, X and Y represent the eigenvalue and target value in the data pair respectively; i represents the sequence number of the data pair, j represents the sequence number of the eigenvalue in the data pair; n represents the total number of data pairs, and k represents the total number of eigenvalues ​​in a data pair; Respectively represent the mean of the characteristic value and the target value; S x With S y Represent the sample standard deviation of the eigenvalue and the target value respectively, where for any data pair i, if there is or Then the data pair is eliminated as a whole.

[0031] The data enhancement in step 301 refers to performing noise enhancement on both the feature value and the target value in each pair of data to reduce the absolute dependence of the model on the original data. The specific process is as follows:

[0032]

[0033]

[0034] In the formula, Y i * They respectively represent the new values ​​of the eigenvalue and target value in the original data after noise enhancement; They represent the noises taken for the eigenvalue and the target value, respectively, and both obey the normal distribution, where the mean and variance μ,σ2 The selection of parameters is based on sampling each X i,j ,Y i When the equipment has inherent analytical accuracy, multiple parallel sampling bias, and No more than X i,j ,Y i 15% of the value.

[0035] The step 301 of standardization refers to using the Z-Score method to synchronously scale the eigenvalues ​​and target values ​​in the data pair to a form with zero mean and unit variance. The specific process is:

[0036]

[0037] In the formula, S x With S y represent the sample standard deviation of the eigenvalue and the target value respectively, Y i z They represent the characteristic value and target value of the data after Z-Score standardization.

[0038] Among them, the neural network model described in step 3 is a one-dimensional convolutional neural network, which includes a structure of a convolutional layer, a pooling layer, a fully connected layer, a Dropout layer, and an output layer; wherein the convolutional layer adopts a filter dimension of 2, the number of filters is 3, and the stride is 1; the pooling layer adopts an average pooling method, the kernel size is 2, and the stride is 1; the fully connected layer adopts a single-layer neural network with 12 neurons, and uses the ReLU function as the activation function; the Dropout rate adopted by the Dropout layer is 50%.

[0039] Beneficial effects: The present invention has the following advantages: 1. The present invention uses the flux box to extract gas to accurately calculate the carbon emission flux in the sewage well, providing a high-quality data basis for the construction of the subsequent carbon emission prediction model and improving the accuracy of the model prediction;

[0040] 2. The present invention uses the existing neural network model to learn the relationship between the carbon emission flux and the carbon emission concentration curve to construct a carbon emission prediction model. The operation is simple and overcomes the problems of the existing carbon emission flux measurement method, which is difficult to deploy, time-consuming and inaccurate in measurement;

[0041] 3. The present invention utilizes a carbon emission prediction model to predict the carbon emission flux of the carbon emission concentration measured by the first carbon emission concentration monitoring device, wherein the first carbon emission concentration monitoring device has good portability and can be widely deployed in various sewage pipe wells, thereby being able to realize real-time monitoring of the carbon emission flux of regional sewage wells. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1It is a schematic diagram of the structure of the monitoring device;

[0043] Figure 2 Schematic diagram of the monitoring method flow chart. DETAILED DESCRIPTION

[0044] The technical solution of the present invention is described in detail below in conjunction with the embodiments and drawings.

[0045] like Figure 1 As shown, a device for monitoring carbon emission flux in a sewage pipe well includes a first carbon emission concentration monitoring device C1 installed on the inner wall of the well, and a flux box C2 suspended above the water surface in the well, wherein one side of the flux box C2 is provided with an air vent connected to the gas in the well, and the other side is provided with an air pipe, and the gas in the box is introduced into the second carbon emission concentration measuring device C3 outside the well through the air pipe.

[0046] The flux box C2 is made of organic glass material, cylindrical, with an inner diameter of 150mm, a height of 150mm, no bottom surface, and a bottom support with an inner diameter of about 300mm at the bottom of the box, which is equipped with a gas stirring mechanism. In order to ensure the minimum interference with the flux collection in the flux box C2, the length of the gas pipe is less than or equal to 1.2 times the well depth. Considering that the general well depth is 4m to 5m, the length of the connecting gas pipe used is about 5.5m.

[0047] like Figure 2 As shown, the monitoring method based on the above monitoring device includes the following steps:

[0048] Step 1: Obtain carbon emission concentration data measured by the first carbon emission concentration monitoring device C1 and the second carbon emission concentration monitoring device within a unit time period;

[0049] Step 2: Calculate the carbon emission flux during the period according to the carbon emission concentration data measured by the second carbon emission concentration monitoring device;

[0050] Step 3: Perform data-driven modeling based on the data set consisting of the carbon emission concentration data measured by the first carbon emission concentration monitoring device C1 and the carbon emission flux data in step 2 to construct a carbon emission prediction model;

[0051] Step 4: Use the carbon emission prediction model to perform real-time prediction of carbon emission flux based on the new carbon emission concentration measurement data provided by the first carbon emission concentration monitoring device C1 deployed on the inner wall of the well.

[0052] Among them, the method for obtaining data described in step 1 is: randomly select multiple sewage wells in a certain area to deploy the monitoring device, and within a unit time period, at intervals of t seconds, synchronously start the first carbon emission concentration monitoring device C1 and the second carbon emission concentration monitoring device in all monitoring devices to directly measure the carbon emission concentration in the well and the carbon emission concentration in the gas drawn out through the flux box C2, respectively, and each measurement lasts for t' seconds.

[0053] The method for calculating the carbon emission flux in step 2 is: first, plot the carbon emission concentration change rate in the box according to the measured carbon emission concentration data, and then calculate the emission flux according to the concentration change rate:

[0054]

[0055] Where J is the carbon emission flux, unit: g GHG / (m 2 *d); m is the carbon emission concentration change rate in the flux box C2, unit: mgGHG / (m 3 *min); V is the effective volume of flux box C2, unit: m 3 ; A is the bottom area covered by flux box C2, unit: m 2 .

[0056] Among them, the data set described in step 3 is a number of data pairs collected by repeating steps 1-2 several times, and each data pair consists of a feature and a target, wherein the feature is concentration curve data composed of carbon emission concentration data measured by the first carbon emission concentration monitoring device C1 in a group of monitoring devices within a unit time period, and the target is carbon emission flux data calculated from the carbon emission concentration data measured by the first carbon emission concentration monitoring device C1 in a certain group of monitoring devices within the unit time period.

[0057] The method for constructing the carbon emission prediction model in step 3 is:

[0058] Step 301: Data processing: After preprocessing, data enhancement, and standardization, several groups of data pairs are randomly divided into training sets and validation sets;

[0059] Step 302: Model training: Use the neural network model to learn the relationship between the carbon emission concentration change curve and the carbon emission flux in the training set data, and establish a carbon emission prediction model based on the relationship between concentration and flux. This model is a non-explicit black box model. The black box model can output a carbon emission flux prediction for the input carbon emission concentration data using the concentration curve direct flux method, specifically:

[0060] J p =F(c)=F(c1,c2,...);

[0061] In the formula, Jp is the carbon emission flux of a sewage well, unit: g GHG / (m 2 *d); c1, c2, ... are the concentration curve data obtained by the first carbon emission concentration monitoring device C1 deployed in the sewage well, that is, a column of time series data, in ppm; where F(.) is a non-explicit black box model;

[0062] Step 303: Model optimization: Use cross-validation based on the validation set to evaluate the generalization ability of the carbon emission prediction model and avoid overfitting; use evaluation indicators including mean square error, R 2 etc. to evaluate the performance of carbon emission prediction models.

[0063] The preprocessing in step 301 refers to identifying abnormal values ​​for several groups of data pairs based on the 3-times standard deviation method. The specific process is as follows:

[0064]

[0065]

[0066] In the formula, X and Y represent the eigenvalue and target value in the data pair respectively; i represents the sequence number of the data pair, j represents the sequence number of the eigenvalue in the data pair; n represents the total number of data pairs, and k represents the total number of eigenvalues ​​in a data pair; Respectively represent the mean of the characteristic value and the target value; S x With S y Represent the sample standard deviation of the eigenvalue and the target value respectively, where for any data pair i, if there is or Then the data pair is eliminated as a whole.

[0067] The data enhancement in step 301 refers to performing noise enhancement on both the feature value and the target value in each pair of data to reduce the absolute dependence of the model on the original data. The specific process is as follows:

[0068]

[0069]

[0070] In the formula, Y i * They respectively represent the new values ​​of the eigenvalue and target value in the original data after noise enhancement; They represent the noises taken for the eigenvalue and the target value, respectively, and both obey the normal distribution, where the mean and variance μ,σ 2 The selection of parameters is based on sampling each X i,j ,Y iWhen the equipment has inherent analytical accuracy, multiple parallel sampling bias, and No more than X i,j ,Y i 15% of the value.

[0071] The step 301 of standardization refers to using the Z-Score method to synchronously scale the eigenvalues ​​and target values ​​in the data pair to a form with zero mean and unit variance. The specific process is:

[0072]

[0073] In the formula, S x With S y represent the sample standard deviation of the eigenvalue and the target value respectively, Y i z They represent the characteristic value and target value of the data after Z-Score standardization.

[0074] Among them, the neural network model described in step 3 is a one-dimensional convolutional neural network, which includes a structure of a convolutional layer, a pooling layer, a fully connected layer, a Dropout layer, and an output layer; wherein the convolutional layer adopts a filter dimension of 2, the number of filters is 3, and the stride is 1; the pooling layer adopts an average pooling method, the kernel size is 2, and the stride is 1; the fully connected layer adopts a single-layer neural network with 12 neurons, and uses the ReLU function as the activation function; the Dropout rate adopted by the Dropout layer is 50%.

[0075] The monitoring device described in the present invention mainly includes a first carbon emission concentration monitoring device C1, a flux box C2, and a second carbon emission concentration measuring device C3 outside the well. Among them, the flux box C2 and the second carbon emission concentration measuring device C3 outside the well can measure the carbon emission flux more accurately, but due to the complex structure, it is not easy to be widely deployed. The first carbon emission concentration monitoring device C1 that simultaneously collects carbon concentration is affected by the well environment, and the accuracy is not high, but due to its good portability, it is conducive to wide deployment. Therefore, the present invention combines the three. In the data collection stage, the first carbon emission concentration monitoring device C1, the flux box C2, and the second carbon emission concentration measuring device C3 are deployed to construct a carbon emission prediction model, and the model is trained. In actual application, it is only necessary to input the data collected by the first carbon emission concentration monitoring device C1 deployed in the well into the trained carbon emission prediction model. The model can accurately predict the carbon emission flux, thereby replacing the flux box C2 and the second carbon emission concentration measuring device C3, and further realizing the real-time online intelligent monitoring of the carbon emission flux of sewage wells in a large area.

Claims

1. A monitoring method based on a monitoring device for carbon emission flux of a sewage pipe well, characterized in that: The following steps are involved: Step 1: Obtain carbon emission concentration data measured by the first carbon emission concentration monitoring device and the second carbon emission concentration monitoring device within a unit time period; Step 2: Calculate the carbon emission flux during the period according to the carbon emission concentration data measured by the second carbon emission concentration monitoring device; Step 3: Perform data-driven modeling based on the data set consisting of the carbon emission concentration data measured by the first carbon emission concentration monitoring device and the carbon emission flux data in step 2 to construct a carbon emission prediction model; The method for constructing the carbon emission prediction model is: Step 301: Data processing: After preprocessing, data enhancement, and standardization, several groups of data pairs are randomly divided into training sets and validation sets; Step 302: Model training: Use the neural network model to learn the relationship between the carbon emission concentration change curve and the carbon emission flux in the training set data, and establish a carbon emission prediction model based on the relationship between concentration and flux. This model is a non-explicit black box model. The black box model can output a carbon emission flux prediction for the input carbon emission concentration data using the concentration curve direct flux method, specifically: J p =F(c)=F(c1,c2,...) In the formula, J p is the carbon emission flux of a sewage well, unit: g GHG / (m 2 *d); c1, c2, ... are the concentration curve data obtained by the first carbon emission concentration monitoring device deployed in the sewage well, that is, a column of time series data, in ppm; where F(.) is a non-explicit black box model; Step 4: Use the carbon emission prediction model to perform real-time prediction of carbon emission flux based on new carbon emission concentration measurement data provided by the first carbon emission concentration monitoring device deployed on the inner wall of the well; The monitoring device includes a first carbon emission concentration monitoring device installed on the inner wall of the well, and a flux box suspended above the water surface in the well. One side of the flux box is provided with an air vent connected to the gas in the well, and the other side is provided with an air pipe, and the gas in the box is introduced into the second carbon emission concentration measuring device outside the well through the air pipe.

2. The monitoring method based on the monitoring device of carbon emission flux of sewage pipe well according to claim 1 is characterized in that: The method for obtaining data described in step 1 is: randomly select multiple sewage wells in a certain area to deploy the monitoring device, and within a unit time period, synchronously start the first carbon emission concentration monitoring device and the second carbon emission concentration monitoring device in all monitoring devices at intervals of t seconds to respectively measure the carbon emission concentration directly in the well and the carbon emission concentration in the gas drawn out through the flux box, and each measurement lasts for t' seconds.

3. The monitoring method based on the monitoring device of carbon emission flux of sewage pipe well according to claim 1 is characterized in that: The method for calculating the carbon emission flux in step 2 is: first, plot the carbon emission concentration change rate in the box according to the measured carbon emission concentration data, and then calculate the emission flux according to the concentration change rate: Where J is the carbon emission flux, unit: g GHG / (m 2 *d); m is the rate of change of carbon emission concentration in the flux box, unit: mgGHG / (m 3 *min); V is the effective volume of the flux box, unit: m 3 ; A is the bottom area covered by the flux box, unit: m 2 .

4. The monitoring method based on the monitoring device of carbon emission flux of sewage pipe well according to claim 1 is characterized in that: The data set described in step 3 is a number of data pairs collected by repeating steps 1-2 several times, and each data pair consists of a feature and a target, wherein the feature is concentration curve data composed of carbon emission concentration data measured by the first carbon emission concentration monitoring device in a group of monitoring devices within a unit time period, and the target is carbon emission flux data calculated from the carbon emission concentration data measured by the first carbon emission concentration monitoring device in a certain group of monitoring devices within the unit time period.

5. The monitoring method based on the monitoring device of carbon emission flux of sewage pipe well according to claim 1 is characterized in that: The neural network model in step 302 is a one-dimensional convolutional neural network, which includes a structure of a convolutional layer, a pooling layer, a fully connected layer, a Dropout layer, and an output layer; wherein the convolutional layer adopts a filter dimension of 2, a number of filters of 3, and a stride of 1; the pooling layer adopts an average pooling method, a kernel size of 2, and a stride of 1; the fully connected layer adopts a single-layer neural network with 12 neurons, and uses a ReLU function as an activation function; the Dropout rate adopted by the Dropout layer is 50%; The carbon emission prediction model in step 302 is optimized by using cross-validation based on the validation set to evaluate the generalization ability of the carbon emission prediction model and avoid overfitting; at the same time, the evaluation indicators include mean square error, R 2 etc. to evaluate the performance of carbon emission prediction models.

6. The monitoring method based on the monitoring device of carbon emission flux of sewage pipe well according to claim 1 is characterized in that: The preprocessing in step 301 refers to identifying outliers for several groups of data pairs based on the 3-times standard deviation method. The specific process is as follows: In the formula, X and Y represent the eigenvalue and target value in the data pair respectively; i represents the sequence number of the data pair, j represents the sequence number of the eigenvalue in the data pair; n represents the total number of data pairs, and k represents the total number of eigenvalues ​​in a data pair; Represent the means of the eigenvalue and target value respectively; S x With S y Represent the sample standard deviation of the eigenvalue and the target value respectively, where for any data pair i, if there is or Then the data pair is eliminated as a whole.

7. The monitoring method based on the monitoring device of carbon emission flux of sewage pipe well according to claim 1 is characterized in that: The data enhancement in step 301 refers to performing noise enhancement on both the feature value and the target value in each pair of data to reduce the absolute dependence of the model on the original data. The specific process is as follows: In the formula, Y i * They respectively represent the new values ​​of the eigenvalue and target value in the original data after noise enhancement; They represent the noises taken for the eigenvalue and the target value, respectively, and both obey the normal distribution, where the mean and variance μ,σ 2 The selection of parameters is based on sampling each X i,j ,Y i When the equipment has inherent analytical accuracy, multiple parallel sampling bias, and No more than X i,j ,Y i 15% of the value.

8. The monitoring method based on the monitoring device of carbon emission flux of sewage pipe well according to claim 1 is characterized in that: The step 301 of standardization refers to using the Z-Score method to synchronously scale the eigenvalues ​​and target values ​​in the data pair to a form with zero mean and unit variance. The specific process is: In the formula, S x With S y represent the sample standard deviation of the eigenvalue and the target value respectively, Y i z They represent the characteristic value and target value of the data after Z-Score standardization.

9. The monitoring method based on the monitoring device of carbon emission flux of sewage pipe well according to claim 1 is characterized in that: A gas stirring mechanism is provided in the flux box in the monitoring device, and the length of the gas pipe is less than or equal to 1.2 times of the well depth.

Images