CO2 storage area intelligent monitoring method based on multi-source data combination

Abstract

The invention discloses a CO2 storage area intelligent monitoring method based on multi-source data combination, and the method comprises the steps: obtaining a hyperspectral image of a CO2 storage area through the macroscopic monitoring of the hyperspectral image, and extracting a spectral index which is high in CO2 sensitivity and correlation; the method comprises the following steps: inverting underground resistivity distribution of a CO2 sequestration area through a high-density electrical method, then measuring carbonate contents in soil at different positions, respectively determining a relationship between the carbonate contents and spectral indexes to establish a spectram-CO2 sequestration correlation model, determining a relationship between the carbonate contents and resistivity, and establishing a resistivity-CO2 sequestration correlation model; and finally, fusing various parameter data to construct a random forest joint monitoring model, and realizing CO2 storage area leakage prediction. According to the method, single data dimension limitation is broken through through spectrum-electrical characteristic coupling, monitoring of the full-area storage state of the CO2 storage area from the earth surface to the deep part is achieved, long-term stability of the CO2 storage area is guaranteed, and technical support is provided for safe implementation of coal mine carbon waste storage engineering.

Description

Technical Field

[0001] The present invention belongs to the technical field of underground carbon sequestration monitoring, and specifically relates to an intelligent monitoring method for CO2 sequestration areas based on multi-source data integration. Background Art

[0002] During the entire cycle of CO2 injection to long-term stable sequestration in underground coal mine spaces, there are many uncertainties and potential risks. If the sealing performance is insufficient, acidic CO2 fluids may penetrate and contaminate groundwater or damage surface vegetation, causing irreversible harm to the environment. At this time, it is particularly important to conduct precise monitoring of the carbon sequestration site to understand its underground sequestration status and migration trend. Traditional single methods are difficult to achieve large-scale and high-efficiency monitoring, and it is difficult to timely warn of possible CO2 leakage risks.

[0003] As an important method for identifying surface vegetation and soil conditions, remote sensing image technology is based on the principle that CO2 leakage will change the spectral characteristics of soil and vegetation. It can achieve large-area synchronous observation, quickly obtain macroscopic data, and qualitatively or quantitatively infer the volume and scope of CO2 leakage from spectral differences. However, this technology mainly focuses on the surface and near-surface, and it is difficult to reach deep underground. In addition, surface vegetation, buildings, water bodies, etc. seriously interfere with remote sensing signals, and their spectral characteristics are confused with each other, greatly reducing the accuracy of identification and inversion.

[0004] In high-density electrical sounding in geophysical exploration, by injecting current into the ground and measuring resistivity information at different depths, the underground electrical properties can be directly obtained. Since the change in soil CO2 content will reduce the regional resistivity, it can be used to monitor whether CO2 leaks. Compared with remote sensing images, high-density electrical sounding can clearly present the vertical distribution state of the CO2 sequestration area in depth detection. However, the signals of electrical prospecting are extremely vulnerable to interference from surface environmental factors, such as terrain undulation, soil moisture changes, surface metal objects, etc. These interference factors will cause noise and errors in the collected data, seriously affecting the accuracy and resolution of the electrical prospecting results. It is difficult to accurately define the CO2 sequestration status and scope only relying on this method.

[0005] Therefore, how to provide a new monitoring method that can monitor the sequestration status of the entire area from the surface to the deep part of the CO2 sequestration area, accurately judge whether CO2 leakage occurs, ensure the long-term stability of the CO2 sequestration area, and provide technical support for the safe implementation of the coal mine carbon waste sequestration project is the research direction required by the present invention. Summary of the Invention

[0006] In view of the problems existing in the above-mentioned prior art, the present invention provides an intelligent monitoring method for CO2 storage areas based on the combination of multi-source data, which can realize the monitoring of the storage state of the entire area from the surface to the deep part of the CO2 storage area, accurately judge whether CO2 leakage occurs, ensure the long-term stability of the CO2 storage area, and provide technical support for the safe implementation of the coal mine carbon waste storage project.

[0007] In order to achieve the above object, the technical solution adopted by the present invention is: an intelligent monitoring method for CO2 storage areas based on the combination of multi-source data, including the following steps:

[0008] Step 1: Obtain remote sensing data of the CO2 storage area: Determine the CO2 storage area and obtain the hyperspectral image and digital elevation model data of the surface of this area;

[0009] Step 2: Obtain real spectral information: After preprocessing the hyperspectral image obtained in Step 1, obtain the real spectral information of the CO2 storage area;

[0010] Step 3: Remote sensing data processing: After processing the real spectral information processed in Step 2 through remote sensing image processing software, extract the vegetation index and texture features of the surface of the CO2 storage area, and then use the band operation tool to screen and calculate the spectral indices with high sensitivity and correlation to CO2;

[0011] Step 4: Determine the resistivity distribution of the CO2 storage area: Determine the layout position of the high-density electrical sounding lines in the detection area according to the real spectral information and digital elevation model data, and then collect resistivity data at different depths in the CO2 storage area. The resistivity in the CO2 accumulation area will decrease. Regularize and invert the measurement data to determine the resistivity distribution of different positions in the CO2 storage area;

[0012] Step 5: Measure the carbonate content of the CO2 storage area: According to the real spectral information and the resistivity distribution determined in Step 4, collect soil samples at different positions in the CO2 storage area and measure the carbonate content in the soil at different positions;

[0013] Step 6: Construct a spectral-CO2 storage correlation model: One-to-one correspondence between the carbonate content at different positions in Step 5 and the spectral indices at the corresponding positions, and use partial least squares regression to establish the relationship between the carbonate content in the soil and the spectral indices, so as to construct a spectral-CO2 storage correlation model;

[0014] Step 7: Construct a resistivity-CO2 storage correlation model: One-to-one correspondence between the carbonate content at different positions in Step 5 and the resistivity conditions at the corresponding positions, determine the change relationship between the carbonate content and the resistivity in the CO2 storage area, so as to construct a resistivity-CO2 storage correlation model;

[0015] Step VIII. Continuous monitoring of the CO2 storage area: Obtain the carbonate content and spectral index distribution at all locations in the CO2 storage area according to the model constructed in Step VI, and obtain the carbonate content and resistivity distribution at all locations in the CO2 storage area according to the model constructed in Step VII, so as to obtain the corresponding carbonate content, spectral index and resistivity at each location; construct a random forest joint inversion model, use part of the data at all locations as the training set and the other part as the validation set, and after training the model, use the model to monitor the CO2 migration situation in the CO2 storage area.

[0016] Further, the preprocessing in Step II includes geometric correction, radiometric correction and atmospheric correction.

[0017] Further, the spectral indices with high CO2 sensitivity and correlation in Step III include the Normalized Difference Vegetation Index (NDVI) and the Vegetation Stress Index.

[0018] Further, in Step IV, the regularization inversion determines the resistivity distribution at different locations in the CO2 storage area, and the specific formula is:

[0019] Φ=‖W d (d obs -d pred )|| 2 +α||W m m|| 2

[0020] where d obs is the observed data, d pred is the predicted data, W d is the weighting matrix, W m is the model weighting matrix, m is the model parameter vector, and α is the regularization parameter.

[0021] Further, in Step V, the carbonate content W in the soil at different locations is measured, and the specific formula is:

[0022]

[0023] where C x is the sample determination concentration, V is the fixed volume after digestion, and m is the mass of the soil sample.

[0024] Further, the construction of the spectral-CO2 storage correlation model in Step VI is specifically:

[0025] Use partial least squares regression (PLSR) to establish the relationship between the carbonate content y in the soil and the spectral indices x1, x2, x3, x4,... x p . Assume that there are n soil samples, and generate the independent variable matrix X=(x ij) nxp For the independent variable X and the dependent variable y, after standardizing them, we get X * and y * , then, through iterative calculation, extract the principal components t1, t2…, t m , and establish a regression equation:

[0026]

[0027] After denormalizing the predicted value y* of the soil carbonate content, we get the predicted value Finally, obtain the spectral - CO2 sequestration correlation model.

[0028] Furthermore, the specific method for constructing the resistivity - CO2 sequestration correlation model in step seven is as follows:

[0029] Use the linear regression method to determine the variation relationship between the carbonate content and the resistivity in the CO2 sequestration area. Among them, the soil carbonate content is y, and the corresponding resistivity obtained by regularization inversion is Z. Through the analysis of multiple sampling points, the linear regression equation is obtained:

[0030] y = AZ + B

[0031] where A and B are regression coefficients. By taking the derivative of the sum of the squared differences between the predicted value y* and the observed value y to be zero, the regression coefficients A and B are solved. Finally, the resistivity - CO2 sequestration correlation model is obtained.

[0032] Furthermore, in step eight, 90% of the data is used as the training set, 10% of the sample data is used as the validation set, and the model is trained with the spectral index and resistivity as the inputs and the carbonate content as the output. After completing the training, a carbonate content threshold is set. During the subsequent monitoring of the CO2 sequestration area, the spectral index and resistivity at different positions are collected in real - time and input into the model. The model outputs the predicted value of the carbonate content at the corresponding position. The predicted value is compared with the set threshold, and all positions where the predicted value exceeds the threshold are marked, so as to delineate the CO2 migration range and realize the continuous monitoring of the CO2 sequestration area.

[0033] Compared with the prior art, the present invention proposes a method combining multi-source data fusion and deep learning, which utilizes the advantages of hyperspectral image macroscopic monitoring and high-density electrical method underground detection to achieve three-dimensional monitoring from the surface to a certain depth underground. Hyperspectral images of the CO2 sequestration area are obtained respectively, and spectral indices with high sensitivity and correlation to CO2 are extracted, including information such as surface vegetation indices and soil spectral characteristics. The underground resistivity distribution and CO2 plume migration path in the CO2 sequestration area are inverted through the high-density electrical method. Then, the carbonate content in the soil at different positions is measured, and the relationship between the carbonate content and the spectral index is determined to establish a spectral-CO2 sequestration correlation model, and the relationship between the carbonate content and the resistivity is determined to establish a resistivity-CO2 sequestration correlation model. Finally, the remote sensing spectral characteristics, resistivity parameters, and terrain data are fused to construct a random forest joint monitoring model to achieve the prediction of CO2 sequestration space leakage. This method breaks through the limitation of a single data dimension through spectral-electrical feature coupling, combines machine learning algorithms to establish a model to improve the accuracy of identifying the sequestration boundary, and reduces the on-site sampling volume by means of preliminary remote sensing screening. Combined with targeted detection by the high-density electrical method, the cost can be greatly reduced, the detection cycle can be shortened, and a basis for decision-making can be quickly provided. Therefore, the present invention can realize the monitoring of the entire area sequestration state of the CO2 sequestration area from the surface to the deep part, and can accurately judge whether CO2 leakage occurs, ensuring the long-term stability of the CO2 sequestration area and providing technical support for the safe implementation of the coal mine carbon waste sequestration project. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Figure 1 is a schematic diagram of the overall process of the present invention;

[0035] Figure 2 is a resistivity distribution map of the CO2 sequestration area in the embodiment of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0036] The present invention will be further described below.

[0037] As Figure 1 shown, the present invention includes the following steps:

[0038] Step 1: Obtain remote sensing data of the CO2 sequestration area: Determine the CO2 sequestration area and obtain the hyperspectral data of the GF-5 satellite on the surface of this area and the digital elevation model data (DEM data) of this area at the China Resources Satellite Application Center;

[0039] Step 2: Obtain real spectral information: After performing geometric correction, radiometric correction, and atmospheric correction on the hyperspectral image obtained in Step 1, obtain the real spectral information of the CO2 sequestration area;

[0040] Step 3: Remote sensing data processing: After processing the real spectral information from step 2 using remote sensing image processing software, vegetation indices and texture characteristics of the surface of the CO2 storage area are extracted. Band operation tools are then used to screen and calculate spectral indices with high sensitivity and correlation with CO2, including the Normalized Difference Vegetation Index (NDVI) and the Vegetation Stress Index.

[0041] Step 4: Determine the resistivity distribution of the CO2 storage area: Determine the layout of the high-density electrical method lateral line in the detection area based on the real spectrum information and digital elevation model data. Then, collect resistivity data at different depths in the CO2 storage area. The resistivity in the CO2 accumulation area will decrease. Regularize and invert the measured data to determine the resistivity distribution at different locations in the CO2 storage area. Figure 2 As shown, the specific formula is:

[0042] Φ=||W d (d obs -d pred )|| 2 +α||W m m|| 2

[0043] where d obs is the observed data, d pred is the predicted data, W d is the weight matrix, W m is the model weight matrix, m is the model parameter vector, and α is the regularization parameter.

[0044] Step 5: Determine the carbonate content in the CO2 storage area: Based on the real spectral information and the resistivity distribution determined in step 4, soil samples are collected at different locations in the CO2 storage area. The carbonate content W in the soil at different locations is determined in the laboratory using atomic absorption spectroscopy (AAS). The specific formula is:

[0045]

[0046] Among them C x is the concentration of the sample, V is the volume after digestion, and m is the mass of the soil sample.

[0047] Step 6: Construct a spectrum-CO2 storage correlation model: By mapping the carbonate content at different locations in step 5 to the spectral index at the corresponding location, partial least squares regression is used to establish the relationship between the carbonate content in the soil and the spectral index, thereby constructing a spectrum-CO2 storage correlation model. Specifically:

[0048] Partial least squares regression (PLSR) was used to establish the relationship between the carbonate content y in soil and the spectral indices x1, x2, x3, x4, ... x pRegarding the relationship, assuming there are n soil samples, an independent variable matrix X = (x ij ) nxp and the dependent variable y are generated from the n soil samples. After standardizing them, we get X * and y * . Then, the principal components t1, t2,..., t m are extracted through iterative calculations, and a regression equation is established:

[0049]

[0050] The predicted value y* of the soil carbonate content is de-standardized to obtain the predicted value Finally, a spectral - CO2 sequestration correlation model is obtained.

[0051] Step 7: Construct a resistivity - CO2 sequestration correlation model: By corresponding the carbonate content at different positions in Step 5 with the resistivity at the corresponding positions one by one, the variation relationship between the carbonate content and the resistivity in the CO2 sequestration area is determined, and thus a resistivity - CO2 sequestration correlation model is constructed. Specifically:

[0052] The linear regression method is used to determine the variation relationship between the carbonate content and the resistivity in the CO2 sequestration area. Among them, the soil carbonate content is y, and the corresponding resistivity obtained by regularization inversion is Z. The linear regression equation is obtained through the analysis of multiple sampling points:

[0053] y = AZ + B

[0054] where A and B are regression coefficients. The regression coefficients A and B are solved by setting the derivative of the difference square between the predicted value y* and the observed value y to zero. Finally, a resistivity - CO2 sequestration correlation model is obtained.

[0055] Step Eight: Continuous Monitoring of the CO2 Sequestration Area: Obtain the carbonate content and spectral index distribution at all locations in the CO2 sequestration area according to the model constructed in Step Six, and obtain the carbonate content and resistivity distribution at all locations in the CO2 sequestration area according to the model constructed in Step Seven, so as to obtain the corresponding carbonate content, spectral index, and resistivity at each location; construct a random forest joint inversion model, use 90% of the data at all locations as the training set and 10% of the sample data as the validation set, and use the spectral index and resistivity as inputs and the carbonate content as the output to train the model. After the training is completed, set the carbonate content threshold. During the subsequent monitoring of the CO2 sequestration area, collect the spectral index and resistivity at different locations in real time and input them into the model. The model outputs the predicted carbonate content value at the corresponding location. Compare the predicted value with the set threshold, and calibrate all locations where the predicted value exceeds the threshold, so as to delineate the CO2 migration range and achieve continuous monitoring of the CO2 sequestration area; in addition, in the CO2 migration range, further determine whether there is an area with a resistivity < 100 Ω·m. If it exists, it is judged as a high CO2 pollution area, which is convenient for taking corresponding measures in a timely manner.

[0056] The above are only the preferred embodiments of the present invention. It should be pointed out that for those of ordinary skill in the art, without departing from the principle of the present invention, several improvements and refinements can be made, and these improvements and refinements should also be regarded as the protection scope of the present invention.

Claims

1. An intelligent monitoring method for CO2 storage areas based on the combination of multi-source data, characterized in that The following steps are involved: Step 1: Obtain remote sensing data of the CO2 storage area: Determine the CO2 storage area and obtain hyperspectral images and digital elevation model data of the surface of the area; Step 2: Obtaining real spectral information: After pre-processing the hyperspectral image obtained in step 1, obtain the real spectral information of the CO2 storage area; Step 3: Remote sensing data processing: After processing the real spectral information from step 2 using remote sensing image processing software, the vegetation index and texture characteristics of the surface of the CO2 storage area are extracted. Then, band operation tools are used to screen and calculate the spectral index with high sensitivity and correlation with CO2; Step 4: Determine the resistivity distribution of the CO2 storage area: Determine the placement of high-density electrical lateral lines in the detection area based on real spectral information and digital elevation model data. Then, collect resistivity data at different depths in the CO2 storage area. Perform regularized inversion on the measured data to determine the resistivity distribution at different locations in the CO2 storage area. Step 5: Determine the carbonate content in the CO2 storage area: Based on the true spectral information and the resistivity distribution determined in step 4, collect soil samples at different locations in the CO2 storage area and measure the carbonate content in the soil at different locations; Step 6: Construct a spectrum-CO2 storage correlation model: By mapping the carbonate content at different locations in step 5 to the spectral index at the corresponding location, partial least squares regression is used to establish the relationship between the carbonate content in the soil and the spectral index, thereby constructing a spectrum-CO2 storage correlation model; Step 7: Construct a resistivity-CO2 storage correlation model: By mapping the carbonate content at different locations in step 5 to the resistivity at the corresponding locations, the relationship between the carbonate content and resistivity in the CO2 storage area is determined, thereby constructing a resistivity-CO2 storage correlation model; Step 8. Continuous monitoring of the CO2 storage area: Obtain the carbonate content and spectral index distribution of all locations in the CO2 storage area based on the model constructed in step 6, and obtain the carbonate content and resistivity distribution of all locations in the CO2 storage area based on the model constructed in step 7, so as to obtain the carbonate content, spectral index and resistivity corresponding to each location; construct a random forest joint inversion model, use part of the data of all locations as a training set and the other as a validation set, and after training the model, use the model to monitor the CO2 migration in the CO2 storage area.

2. The intelligent monitoring method for CO2 storage areas based on multi-source data integration according to claim 1, characterized in that, The preprocessing in step 2 includes geometric correction, radiation correction and atmospheric correction.

3. The CO2 storage area intelligent monitoring method based on multi-source data integration according to claim 1 is characterized in that: The spectral indices with high sensitivity and correlation with CO2 in step three include normalized vegetation index and vegetation stress index.

4. The method for intelligent monitoring of CO2 storage areas based on multi-source data integration according to claim 1 is characterized in that: In step 4, the resistivity distribution at different locations in the CO2 storage area is determined by regularized inversion. The specific formula is: Φ=||W d (d obs -d pred )|| 2 +a||W m m|| 2 where d obs is the observed data, d pred is the predicted data, W d is the weighting matrix, W m is the model weighting matrix, m is the model parameter vector, and α is the regularization parameter.

5. The method for intelligent monitoring of CO2 storage areas based on multi-source data integration according to claim 1 is characterized in that: In step 5, the carbonate content W in the soil at different locations is measured, and the specific formula is: Among them, C x is the measured concentration of the sample, V is the volume after digestion and constant volume, and m is the mass of the soil sample.

6. The intelligent monitoring method for CO2 storage areas based on the combination of multi-source data according to claim 1, characterized in that The specific steps of constructing the spectrum-CO2 storage correlation model in step 6 are as follows: Using partial least squares regression to establish the relationship between the carbonate content y in the soil and the spectral indices x1, x2, x3, x4, … x p Let the number of soil samples be n. Generate the independent variable matrix X = (x ij ) nxp and the dependent variable y from the n soil samples, and standardize them to obtain X[[ID=?]] * and y * . Then, extract the principal components t1, t2, …, t m through iterative calculations and establish the regression equation: It should be noted that there seems to be an unclear tag " * " in the original text which might be a formatting or encoding issue. The translation is done as accurately as possible based on the available clear text. The predicted value is obtained by denormalizing the predicted value y* of the soil carbonate content. Finally, a spectral-CO2 sequestration correlation model is obtained.

7. The method for intelligent monitoring of CO2 storage areas based on multi-source data integration according to claim 1 is characterized in that: The specific steps of constructing the resistivity-CO2 storage correlation model in step 7 are as follows: The linear regression method is used to determine the relationship between the carbonate content and resistivity in the CO2 storage area, where the soil carbonate content is y and the corresponding resistivity obtained by regularized inversion is Z. The linear regression equation is obtained by analyzing multiple sampling points: y=AZ+B Among them, A and B are regression coefficients. The regression coefficients A and B are solved by taking the square derivative of the difference between the predicted value y* and the observed value y to zero, and finally the resistivity-CO2 storage correlation model is obtained.

8. The intelligent monitoring method for CO2 sequestration area based on multi-source data combination according to claim 1, characterized in that, In step eight, 90% of the data is used as a training set and 10% of the sample data is used as a validation set. The model is trained with spectral index and resistivity as input and carbonate content as output. After the training is completed, a carbonate content threshold is set. In the subsequent monitoring of the CO2 storage area, the spectral index and resistivity at different locations are collected in real time and input into the model. The model outputs a predicted value of the carbonate content at the corresponding location, and the predicted value is compared with the set threshold. All locations where the predicted value exceeds the threshold are calibrated, thereby delineating the CO2 migration range and realizing continuous monitoring of the CO2 storage area.

Images