Nonlinear inversion method and system of apparent resistivity based on EConvNet-C algorithm

Abstract

The present invention discloses a nonlinear inversion method and system of apparent resistivity based on the EConvNet‑C algorithm, including: obtaining apparent resistivity data of a set landfill leachate water level through high-density electrical measurement; inputting the apparent resistivity data into a trained EConvNet‑C leachate water level recognition model to obtain inverted landfill resistivity data. The present invention integrates a multi-layer perceptron and an EConvNet‑T into an EConvNet‑C leachate water level recognition model. The multi-layer perceptron can transform an undefinable potential space into a low-resolution feature image space, thereby realizing nonlinear inversion of potential differential data carrying leachate information to true resistivity; providing guidance for later pollution control and soil remediation of landfills.

Description

Technical Field

[0001] The invention relates to the technical field of landfill leachate level measurement technology, and in particular to an apparent resistivity nonlinear inversion method and system based on an EConvNet-C algorithm. Background Art

[0002] The statements in this section merely provide background information related to the present invention and do not necessarily constitute prior art.

[0003] Landfills are the main disposal sites for solid waste and are also the places where solid waste environmental risks, such as soil and groundwater pollution, occur. Excessively high leachate levels in landfills not only increase the risk of leachate leakage, but may also affect the stability of solid waste piles, easily causing safety risks such as landslides and collapses, and pose both safety and environmental risks.

[0004] Leachate water level detection is a prerequisite for leachate water level control. The prior art discloses the use of electrical resistivity tomography (ERT) to detect the leachate water level in landfills. Since the solid waste below the saturated level of the leachate is in a water-saturated state, there is an obvious difference in resistivity between the solid waste and the solid waste above the saturated level. This is the physical basis for the ERT method to achieve leachate water level detection.

[0005] The ERT method usually includes two steps. The first is to obtain the apparent resistivity data through electrode layout, and the second is to obtain the real resistivity data of the underground medium through inversion. The traditional ERT method usually uses the traditional Wenner (AMNB) device and the dipole (AM) device to obtain the apparent resistivity, and then obtains the real resistivity data through the least squares inversion method, but there are often the following technical problems:

[0006] (1) The data collected by the traditional ERT method only contains the leachate information of one vertical section, and the deeper the depth, the less information it carries;

[0007] (2) The least squares inversion method depends on the selection of the initial model and requires repeated calculation of the Jacobian matrix. If the prior knowledge and information are insufficient and the initial model is not properly selected, it is easy to fall into local extreme values ​​when using the linearization method for resistivity imaging inversion, resulting in non-optimal solutions or even wrong solutions. Summary of the invention

[0008] To solve the above problems, the present invention proposes a method and system for non-linear inversion of apparent resistivity based on the EConvNet-C algorithm. Based on the potential difference data collected by the "chuan"-shaped measurement device, it can identify the leachate levels in different regions under deep garbage; and uses the EConvNet-C model to achieve non-linear inversion of the potential difference data carrying leachate information to the true resistivity.

[0009] To achieve the above object, in some embodiments, the following technical solutions are adopted:

[0010] A method for non-linear inversion of apparent resistivity based on the EConvNet-C algorithm, comprising:

[0011] Obtaining apparent resistivity data of the leachate level of a set landfill through high-density electrical method measurement;

[0012] Inputting the apparent resistivity data into a trained EConvNet-C leachate level identification model to obtain the inverted resistivity data of the landfill.

[0013] As a further solution, the EConvNet-C leachate level identification model includes:

[0014] Reconstructing the data volume of each channel through a multi-layer perceptron to obtain a corresponding low-resolution feature image;

[0015] Inputting the low-resolution feature image into the EConvNet-T network structure to obtain the final prediction result of resistivity data.

[0016] As a further solution, the EConvNet-T network structure is based on the U-Net network as the main body.

[0017] As a further solution, the encoding structure of the EConvNet-T network structure first extracts features through several convolutional layers and pooling layers to obtain a high-dimensional feature map with low resolution to obtain context information; the decoding structure propagates the context information to the high-resolution layer to predict position information; and fills the bottom layer information through skip connections to improve the segmentation accuracy.

[0018] As a further solution, the training process of the EConvNet-C leachate level identification model includes:

[0019] Simulating a landfill through simulation, setting an HDPE anti-seepage membrane on the inner surface of the landfill, and setting an inverted trapezoidal hexahedron in the landfill to simulate the geometric shape of the leachate;

[0020] By changing the landfill, the leachate level, and the material parameters of the landfill material and the leachate, simulating the resistivity intervals represented by different sites and leachate levels;

[0021] The apparent resistivity data at different landfill heights and different leachate levels are obtained through high-density electrical method measurement, and a sample data set for the EConvNet-C leachate level identification model is constructed;

[0022] Based on the sample data set, the EConvNet-C leachate level identification model is trained.

[0023] In some other embodiments, the following technical solution is adopted:

[0024] An apparent resistivity non-linear inversion system based on the EConvNet-C algorithm, comprising:

[0025] A data acquisition module for obtaining the apparent resistivity data of the set landfill leachate level through high-density electrical method measurement;

[0026] A resistivity data inversion module for inputting the apparent resistivity data into the trained EConvNet-C leachate level identification model to obtain the inverted landfill resistivity data.

[0027] In some other embodiments, the following technical solution is adopted:

[0028] A terminal device, which includes a processor and a memory. The processor is used to implement each instruction; the memory is used to store multiple instructions, and the instructions are suitable for being loaded and executed by the processor to perform the above-mentioned apparent resistivity non-linear inversion method based on the EConvNet-C algorithm.

[0029] In some other embodiments, the following technical solution is adopted:

[0030] A computer-readable storage medium, in which multiple instructions are stored, and the instructions are suitable for being loaded and executed by the processor of the terminal device to perform the above-mentioned apparent resistivity non-linear inversion method based on the EConvNet-C algorithm.

[0031] Compared with the prior art, the beneficial effects of the present invention are:

[0032] (1) Through the "川"-shaped measuring device of the present invention, the current can pass through more media, and the collected data carries a large amount of leachate information, which helps to identify the leachate levels in different areas under the deep garbage.

[0033] (2) Since there is no one-to-one correspondence between the apparent resistivity of the "川"-shaped measurement device and the resistivity model, it is impossible to use the EConvNet-T based on the U-Net structure for feature learning; the EConvNet-C leachate level identification model that combines a multi-layer perceptron and EConvNet-T in the present invention, where the multi-layer perceptron can transform the indefinable potential space into a low-resolution feature image space, thereby realizing the non-linear inversion of the potential difference data carrying leachate information to the true resistivity; the EConvNet-C leachate level identification model of the present invention can accurately invert the leachate level, providing guidance for the later pollution control and soil remediation of landfills.

[0034] Other features and advantages of the additional aspects of the present invention will be given in part in the following description, become apparent in part from the following description, or be understood through the practice of this aspect. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] FIG. 1(a)-(b) are respectively schematic diagrams of the traditional Wenner device and the secondary device;

[0036] Figure 2 is a schematic diagram of the "川"-shaped measurement device in the embodiment of the present invention;

[0037] FIG. 3(a)-(c) are respectively schematic diagrams of the first group, the second group, and the third group of the electrode arrangement modes of the "川"-shaped measurement device in the embodiment of the present invention;

[0038] Figure 4 is a schematic diagram of the non-linear inversion method of apparent resistivity based on the EConvNet-C algorithm in the embodiment of the present invention;

[0039] FIG. 5(a)-(b) are respectively a schematic diagram and a side view of the three-dimensional theoretical model of the landfill in the embodiment of the present invention;

[0040] Figure 6 is a schematic diagram of the EConvNet-T network structure in the embodiment of the present invention;

[0041] Figure 7 is a schematic diagram of the EConvNet-C network structure in the embodiment of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0042] It should be noted that the following detailed description is illustrative and is intended to provide further explanation of the present application. Unless otherwise specified, all technical and scientific terms used in the present invention have the same meaning as commonly understood by those of ordinary skill in the technical field to which the present application belongs.

[0043] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit the exemplary embodiments according to the present application. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that when the terms "comprise" and / or "include" are used in this specification, it indicates the presence of features, steps, operations, devices, components and / or combinations thereof.

[0044] Embodiment 1

[0045] The deep convolutional neural network (CNN) has a deeper network layer and stronger parameter processing capabilities, and has the characteristics of local perception field, weight sharing and feature extraction. Its convergence speed and prediction accuracy are better than those of the back propagation neural network (BPNN). U-Net is a classic model used for image semantic segmentation in deep neural networks. U-Net adopts a symmetrical structure design, integrates the low-dimensional and high-dimensional features in the CNN network, and uses a small amount of data sets for training and testing. It plays an important role in the field of image segmentation.

[0046] Based on this, in one or more embodiments, a nonlinear inversion method of apparent resistivity based on the EConvNet-C algorithm is disclosed, which specifically includes the following process:

[0047] (1) Obtain the apparent resistivity data of the set landfill leachate water level through high-density electrical measurement;

[0048] The apparent resistivity data can be obtained by using conventional high-density electrical measurement devices (Wenner device and diode device), or by using a "Chuan"-shaped measurement device.

[0049] HDPE membrane is laid around and on the bottom of the landfill, with a resistivity of 10 17 Ω·m, its high resistance makes the current unable to pass through the HDPE membrane, but flows along the surface of the HDPE membrane. The traditional high-density electrical measurement device is shown in Figure 1(a)-(b), which runs poles on a measuring line, resulting in the current only flowing through a section of the leachate area, and the data obtained only carries a small amount of leachate information.

[0050] Therefore, in this embodiment, a "C"-shaped measuring device is used to obtain apparent resistivity data. The "C"-shaped measuring device is powered by electrodes on two measuring lines that are not in the same section, and the established current field can cover the entire leachate area, so that most of the current flows through the leachate area. The obtained data carries a large amount of leachate information, which improves the recognition rate of the leachate water level.

[0051] Figure 2The structural schematic diagram of the Sichuan-shaped measuring device is given, specifically including: a number of power supply electrodes A arranged on the first measuring line, a number of measuring electrodes arranged on the second measuring line, and a number of power supply electrodes B arranged on the third measuring line; the first measuring line, the second measuring line and the third measuring line are respectively arranged on different profile lines.

[0052] In this embodiment, combined with Figure 4 , the first measuring line, the second measuring line and the third measuring line are arranged in parallel, and the second measuring line is located between the first measuring line and the third measuring line.

[0053] The electrode spacing on the first measuring line and the third measuring line is the same, and the spacing of the measuring electrodes on the second measuring line is half of the electrode spacing on the first measuring line or the third measuring line; the power supply electrode A on the first measuring line and the power supply electrode B on the third measuring line are arranged in a staggered manner; on the second measuring line, there are measuring electrodes respectively corresponding to the positions of each power supply electrode A and power supply electrode B.

[0054] The power supply electrode A is connected to the positive pole of the power supply through the first measuring line, and the power supply electrode B is connected to the negative pole of the power supply through the third measuring line; all electrodes are connected to the controller, and the controller selects the power supply electrodes A, B and the measuring electrodes M, N among all electrodes according to the set arrangement method for data acquisition.

[0055] Figure 2 Among them, 32 electrodes are used as the power supply electrode A on the first measuring line, 32 electrodes are used as the power supply electrode B on the third measuring line, and 64 electrodes are used as the measuring electrodes M, N on the second measuring line. By supplying current I through the A electrode and -I through the B electrode, the potential (potential difference) ΔV between the measuring electrodes M and N is obtained, and the apparent resistivity ρ = K·ΔV / I at this point (the midpoint of M and N) is obtained, and its device coefficient The recording point is the midpoint of MN, and the depth is r AB / 6.

[0056] The measuring method of the Sichuan-shaped measuring device is specifically as follows:

[0057] The arrangement of the electrodes is divided into 31 layers in total, and the depth of each layer is the same; the arrangement of the electrodes in each layer is measured in sequence according to the set order;

[0058] Assume that the A i electrode on the first measuring line is vertically extended to the third measuring line and is located at the midpoint position between B i and B i+1 ;

[0059] The arrangement of the electrodes in the first layer is: every time the power supply electrode A moves to an A i , the power supply electrode B correspondingly selects B i and B i+1 electrodes to supply negative electricity;

[0060] Specifically, referring to Figure 3(a), the first layer supplies current I to power supply electrode A1, and supplies current -I to power supply electrodes B1 and B2 respectively, and records the potential of the midpoints of measuring electrodes M1 and M2; then power supply electrode A2 supplies current I, and power supply electrodes B2 and B3 supply current -I respectively, and records the potential of the midpoints of measuring electrodes M2 and M3; and so on, finally power supply electrode A32 supplies current I, and power supply electrode B32 supplies current -I, and records the potential of the midpoints of measuring electrodes M63 and M64.

[0061] The arrangement of the second layer of electrodes is as follows: each time the power supply electrode A moves to an A i , the power supply electrode B is selected accordingly i-1 and B i+2 The electrode is negatively charged;

[0062] Specifically, referring to Figure 3(b), in the second layer, the power supply electrode A1 is supplied with current I, the power supply electrode B3 is supplied with current -I, and the potential of the midpoints of the measuring electrodes M2 and M5 is recorded; then the power supply electrode A2 is supplied with current I, the power supply electrodes B1 and B4 are supplied with current -I respectively, and the potential of the midpoints of the measuring electrodes M1 and M4 is recorded; and so on, finally the power supply electrode A32 is supplied with current I, the power supply electrode B31 is supplied with current -I, and the potential of the midpoints of the measuring electrodes M61 and M64 is recorded.

[0063] The arrangement of the third layer of electrodes is as follows: each time the power supply electrode A moves to an A i , the power supply electrode B is selected accordingly i-2 and B i+3 The electrode is negatively charged;

[0064] Specifically, referring to Figure 3(c), the third layer supplies current I to power electrode A1 and current -I to power electrode B4, and records the potential of the midpoints of measuring electrodes M2 and M7; then power electrode A2 supplies current I and current -I to power electrode B5, and records the potential of the midpoints of measuring electrodes M4 and M9; and so on, finally power electrode A32 supplies current I and current -I to power electrode B30, and records the potential of the midpoints of measuring electrodes M59 and M64.

[0065] The arrangement of the electrodes in other layers is similar, until the arrangement of the electrodes in the 31st layer: each time the power supply electrode A moves to an A i , the power supply electrode B is selected accordingly i-30 and B i+31 The electrode is negatively charged.

[0066] (2) The apparent resistivity data is input into the trained EConvNet-C leachate level recognition model to obtain the inverted landfill resistivity data.

[0067] In this embodiment, the multi-layer perceptron MLP and the EConvNet-T network structure are integrated to obtain the EConvNet-C leachate water level recognition model.

[0068] Among them, the EConvNet-T network structure is obtained by adjusting the depth and width on the basis of the U-Net network, which can better adapt to the scale of apparent resistivity data.

[0069] The EConvNet-T network uses the encoding-decoding structure of the U-Net network. The EConvNet-T network structure is as follows: Figure 6 As shown in the figure, the encoding structure first extracts features through several convolutional layers and pooling layers to obtain a high-dimensional feature map with lower resolution, which can effectively capture context information. The decoding structure propagates the context information to the high-resolution layer to better predict the position information. In addition, the underlying information is filled in through jump connections to improve the segmentation accuracy.

[0070] The parameters of the EConvNet-T network are set as follows: the activation function of the middle layer is relu, the last layer is sigmoid, and the learning rate is 10 -3 , the number of training rounds is 5000, and the optimizer is Adam.

[0071] The operating environment of EConvNet-T is as follows: the CPU is Inter Core i7-10870HQ, the running memory is 16GB, and the system is 64-bit win10 system.

[0072] EConvNet-C leachate level recognition model Figure 7 As shown in the figure, the obtained apparent resistivity data is first reorganized into 992-channel input of the EConvNet-C leachate water level recognition model. Since the amount of input data per channel is large, taking two survey lines with 32 electrodes per survey line as an example, the amount of data per channel can reach 2016. When the inversion area is 64m×12.7m, for a large amount of data that lacks spatial correspondence, a multi-layer perceptron MLP is used to reconstruct the 992-channel input using a 3-times fully connected network to obtain 992 corresponding small-scale (32m×6m) low-resolution feature images. On this basis, the reconstructed low-resolution feature image is feature reorganized (1024×32×6) through the EConvNet-T network structure.

[0073] In this embodiment, the process of training the EConvNet-C leachate level recognition model is as follows: Figure 4 As shown, it includes two processes: generating representative and high-quality sample data, and training the sample data so that the mean square error between the network output and the expected output vector meets the target requirement.

[0074] Among them, the quality, quantity, and authenticity of the sample dataset determine the accuracy and practicality of the model. Due to the complex geoelectric structure in actual landfill sites, it is difficult to obtain training set data from actual sites containing leachate. Using a simulation model, apparent resistivity and potential difference data of leachate with complex compositions, different leachate positions, and different landfill materials can be obtained as training set data and test data, providing a data basis for model learning.

[0075] The complex compositions of leachate and landfill materials lead to resistivity anomalies in the site. Different landfill site heights and different leachate water levels and positions will cause abnormal distributions of resistivity in the landfill. In this embodiment, a simplified ideal model is established by the simulation software COMSOL to simulate scenarios with a site height of 6 - 10 meters and a leachate water level of 0.1 - 4 meters, for studying the resistivity differences between two representative sites where leachate accumulates at the bottom of the HDPE membrane and leachate accumulates in the middle of the landfill materials; as shown in Figures 5(a)-(b), the simulation process is as follows: First, a three-dimensional cuboid is established to simulate the earth, with a volume of 200m * 200m * 30m, and inverted trapezoidal hexahedrons with different heights are established inside it to simulate the landfill; Second, small inverted trapezoidal hexahedrons with a length of 200m, a width of 200m, and different heights are placed in the simulated site to simulate the geometric shape of the leachate, and an HDPE anti-seepage membrane with a resistivity of 10 17 Ω·m is set on the inner surface of the simulated landfill. The liquid levels of the leachate accumulated at the bottom of the HDPE membrane and the leachate accumulated in the landfill materials change with the site environment; Finally, a traditional device or a "川" - shaped measuring device is used to collect data.

[0076] By changing the landfill height, leachate water level, and material parameters (resistivity and relative permittivity) of the landfill materials and leachate, the resistivity intervals represented by different sites and leachate water levels are simulated.

[0077] For the geometric model of the representative landfill site, the electric field distribution generated by the current source S during the detection process of the high-density electrical method measurement device follows the Poisson distribution:

[0078]

[0079] Among them, ρ is the resistivity, Ω·m; is the electric potential, V, and j c is the charge density, A / m 2 . Given the resistivity in the geometric model, the electric potential at any point in the medium can be calculated.

[0080] At the surface boundary, air is an insulator, satisfying the Neumann boundary condition:

[0081]

[0082] For the boundary at infinity, to maintain the physical properties of the potential and eliminate the electrical mutation, a mixed boundary condition is selected for use:

[0083]

[0084] Based on the established simulation model, the "川字型" measurement device is used to simulate the pairwise subtraction of electric potentials to process them into electric potential difference data. Finally, 200 groups of data are obtained as the sample dataset D of the inversion model, and D is divided into a training set and a test set according to a ratio of 7:3 for the training of the EConvNet-C leachate level identification model; the input of the EConvNet-C leachate level identification model is the electric potential difference data (apparent resistivity data) of two measurement electrodes, and the output is the true resistivity value.

[0085] The apparent resistivity data is input into the trained EConvNet-C leachate level identification model to obtain the inverted landfill resistivity data.

[0086] In this embodiment, the EConvNet-C leachate level identification model is verified through the measured data of a landfill in Jiangxi; for a certain hazardous waste landfill in Jiangxi, the total area of the storage area is about 196393 m 2 , and the total storage capacity is about 935,000 m 3 . The landfill uses a straight-through pipe gravity flow for diversion and there is rainwater infiltration, resulting in a large amount of leachate accumulated in the landfill. The detection devices used are the Wenner device, the dipole device, and the "川字型" measurement device. The on-site survey line layout: for the traditional Wenner device and the dipole device, the survey line length is 50.4 meters, the electrode spacing is 0.8 meters, and there are a total of 64 electrodes; in the "川字型" measurement device, the two power supply survey lines of the power supply electrodes are both 64 meters long, the electrode spacing is 2 meters, the acquisition survey line length is 64 meters, and the electrode spacing is 1 meter.

[0087] The apparent resistivity data obtained by the above devices is inverted using the EConvNet-T algorithm, the EConvNet-C algorithm, and the traditional least squares (LS) algorithm respectively. The mean square error (MSE) is used to evaluate the performance of the two inversion algorithms, which is defined as follows:

[0088]

[0089] In the formula, represents the predicted value of the i-th training sample; zi represents the true value of the i-th training sample, and N represents the total number of samples. MSE describes the prediction deviation of the algorithm, and the smaller its value, the smaller the inversion error of the algorithm.

[0090] Finally, the comparison results of the inversion results are shown in Table 1:

[0091] Table 1 Comparison results of MSE of two algorithms

[0092]

[0093] From the data in Table 1, it can be seen that in the theoretical model testing phase, the MSE value of the EConvNet-T algorithm of each model is smaller than the least squares method, and the MSE of the EConvNet-C algorithm is smaller than that of EConvNet-T, which shows that the prediction error of the EConvNet-C algorithm is small and the accuracy is better.

[0094] In addition, the theoretical model results of leachate accumulation on HDPE membrane are shown in Table 2:

[0095] Table 2 Results of leachate level height for model 1 and model 2

[0096]

[0097] (1) For the representative leachate scenario 1, the leachate level calculated by the EConvNet-T inversion algorithm is slightly lower than the theoretical model, the result in the second pole is slightly higher than the theoretical model, and the result in the EConvNet-C is closest to the theoretical model; the recognition accuracy rates are 81.0%, 83%, and 90%, respectively. However, the LS cannot find the leachate and cannot determine the leachate level.

[0098] (2) For the representative leachate scenario 2, the leachate level calculated by the EConvNet-T inversion algorithm is slightly lower than the theoretical model, the result in the second pole is slightly higher than the theoretical model, and the result in the EConvNet-C is closest to the theoretical model; the recognition accuracy rates are 80.6%, 84.7%, and 93.3%, respectively. However, LS cannot find the leachate and cannot determine the leachate level.

[0099] Combining the above two points, the EConvNet-C inversion algorithm can identify the leachate level attached to the HDPE membrane, and can identify the depth position of the HDPE membrane; however, the least squares inversion method can only identify the HDPE membrane, and the position of the membrane has a large error compared with the theoretical model, and the positioning is inaccurate; the EConvNet-C inversion algorithm describes the leachate level more clearly and accurately, and can suppress the influence of data noise on the inversion results. In addition, the EConvNet-C inversion algorithm is more accurate than LS in predicting the true resistivity value.

[0100] The theoretical model results of leachate accumulation in the middle of landfill material are shown in Table 3:

[0101] Table 3 Leachate level results for model 3 and model 4

[0102]

[0103] (1) For the representative leachate scenario 3, the leachate level calculated by the EConvNet-T inversion algorithm is slightly lower than the theoretical model, the result in the second pole is slightly higher than the theoretical model, and the result in the EConvNet-C is closest to the theoretical model; the recognition accuracy rates are 74.6%, 75.3%, and 80%, respectively. However, LS cannot find the leachate and cannot determine the leachate level.

[0104] (2) For representative leachate scenario 4, the leachate level calculated by the EConvNet-T inversion algorithm is slightly lower than the theoretical model, the result in the second pole is slightly higher than the theoretical model, and the result in the EConvNet-C is closest to the theoretical model; the recognition accuracy rates are 70.5%, 72.5%, and 75%, respectively. However, LS cannot find the leachate and cannot determine the leachate level.

[0105] Combining the above two points, the EConvNet-C inversion algorithm can identify the leachate level accumulated in the middle of the landfill material, and can identify the depth position of the HDPE membrane; however, the least squares inversion method can only identify the HDPE membrane, and the position of the membrane has a large error compared with the theoretical model, and the positioning is inaccurate; the EConvNet-C inversion algorithm describes the leachate level more clearly and accurately, and can suppress the influence of data noise on the inversion results. In addition, the EConvNet-C inversion algorithm is more accurate than LS in predicting the true resistivity value.

[0106] Although the above describes the specific implementation mode of the present invention in conjunction with the accompanying drawings, it is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art on the basis of the technical solution of the present invention without creative work are still within the scope of protection of the present invention.

Claims

1. A nonlinear inversion method for apparent resistivity based on the EConvNet-C algorithm, characterized in that: Including: Obtaining apparent resistivity data of the leachate level of a set landfill through high-density electrical method measurement, specifically including: obtaining apparent resistivity data of the leachate level of the landfill through a "Sichuan character" measurement device; wherein, the "Sichuan character" measurement device specifically includes: a plurality of power supply electrodes A arranged on the first measurement line, a plurality of measurement electrodes arranged on the second measurement line, and a plurality of power supply electrodes B arranged on the third measurement line; the first measurement line, the second measurement line, and the third measurement line are respectively arranged on different profile lines; the first measurement line, the second measurement line, and the third measurement line are arranged in parallel, and the second measurement line is located between the first measurement line and the third measurement line; the electrode spacing on the first measurement line and the third measurement line is the same, and the spacing of the measurement electrodes on the second measurement line is half of the electrode spacing on the first measurement line or the third measurement line; the power supply electrodes A on the first measurement line and the power supply electrodes B on the third measurement line are arranged in a staggered manner; measurement electrodes corresponding to the positions of each power supply electrode A and power supply electrode B are arranged on the second measurement line; the power supply electrode A is connected to the positive pole of the power supply through the first measurement line, and the power supply electrode B is connected to the negative pole of the power supply through the third measurement line; all electrodes are connected to the controller, and the controller selects the power supply electrodes A, B and the measurement electrodes M, N from all electrodes according to the set arrangement method for data acquisition. Inputting the apparent resistivity data into the trained EConvNet-C leachate level recognition model to obtain the inverted landfill resistivity data; the EConvNet-C leachate level recognition model includes: reconstructing the data volume of each channel through a multi-layer perceptron to obtain the corresponding low-resolution feature image; inputting the low-resolution feature image into the EConvNet-T network structure to perform feature recombination on the reconstructed low-resolution feature image to obtain the final resistivity data prediction result; the EConvNet-T network structure is based on the U-Net network; the encoding structure of the EConvNet-T network structure first extracts features through several convolutional layers and pooling layers and obtains a high-dimensional feature map with low resolution to obtain context information; the decoding structure propagates the context information to the high-resolution layer to predict position information; the bottom layer information is filled through skip connections to improve the segmentation accuracy.

2. The nonlinear inversion method of apparent resistivity based on the EConvNet-C algorithm according to claim 1, characterized in that: Select power supply electrode A to provide current according to the set arrangement , the power supply electrode B provides current ; Obtaining the potential difference between the measurement electrodes M and N corresponding to the power supply electrodes A and B, and then obtaining the apparent resistivity at the midpoint of the measurement electrodes M and N; the electrode arrangement method is multiple groups, and the depth of each group is the same; the electrode arrangement methods of each group are measured in sequence according to the set order.

3. The nonlinear inversion method of apparent resistivity based on the EConvNet-C algorithm according to claim 1, characterized in that: The training process of the EConvNet-C leachate level recognition model includes: Simulating a landfill through simulation, setting an HDPE anti-seepage membrane on the inner surface of the landfill, and setting an inverted trapezoidal hexahedron in the landfill to simulate the geometric shape of the leachate. By changing the landfill, the leachate level, and the material parameters of the landfill material and the leachate, simulating the resistivity intervals represented by different sites and leachate levels. The apparent resistivity data of different landfill heights and different leachate levels were obtained through high-density electrical measurement, and the sample data set of the EConvNet-C leachate level recognition model was constructed. The EConvNet-C leachate level recognition model is trained based on the sample data set.

4. An apparent resistivity nonlinear inversion system based on the EConvNet-C algorithm, characterized in that: include: The data acquisition module is used to obtain the apparent resistivity data of the leachate water level of the set landfill through high-density electrical measurement, specifically comprising: obtaining the apparent resistivity data of the leachate water level of the landfill through a "C"-shaped measuring device; wherein the "C"-shaped measuring device specifically comprises: a plurality of power supply electrodes A arranged on a first measuring line, a plurality of measuring electrodes arranged on a second measuring line, and a plurality of power supply electrodes B arranged on a third measuring line; the first measuring line, the second measuring line and the third measuring line are arranged on different profile lines respectively; The first measuring line, the second measuring line and the third measuring line are arranged in parallel, and the second measuring line is located between the first measuring line and the third measuring line; the electrode spacings on the first measuring line and the third measuring line are the same, and the spacing of the measuring electrodes on the second measuring line is half of the electrode spacing on the first measuring line or the third measuring line; the power supply electrode A on the first measuring line and the power supply electrode B on the third measuring line are arranged in a staggered manner; the second measuring line is provided with measuring electrodes corresponding to the positions of each power supply electrode A and power supply electrode B respectively; the power supply electrode A is connected to the positive electrode of the power supply through the first measuring line, and the power supply electrode B is connected to the negative electrode of the power supply through the third measuring line; all the electrodes are connected to the controller, and the controller selects the power supply electrodes A, B and the measuring electrodes M, N from all the electrodes according to the set arrangement to collect data; The resistivity data inversion module is used to input the apparent resistivity data into the trained EConvNet-C leachate water level recognition model to obtain the inverted landfill resistivity data; the EConvNet-C leachate water level recognition model comprises: reconstructing the data volume of each channel through a multi-layer perceptron to obtain a corresponding low-resolution feature image; the low-resolution feature image is input into the EConvNet-T network structure, and the reconstructed low-resolution feature image is reorganized to obtain the final resistivity data prediction result; The EConvNet-T network structure is based on the U-Net network; the encoding structure of the EConvNet-T network structure first extracts features through several convolutional layers and pooling layers to obtain a low-resolution high-dimensional feature map to obtain context information; the decoding structure propagates the context information to the high-resolution layer to predict the position information; and the underlying information is filled in through jump connections to improve the segmentation accuracy.

5. A terminal device, comprising a processor and a memory, wherein the processor is used to implement each instruction; and the memory is used to store multiple instructions, characterized in that: The instructions are suitable for being loaded by a processor and executing the apparent resistivity nonlinear inversion method based on the EConvNet-C algorithm as described in any one of claims 1 to 3.

6. A computer-readable storage medium storing a plurality of instructions, characterized in that: The instructions are suitable for being loaded by a processor of a terminal device and executing the apparent resistivity nonlinear inversion method based on the EConvNet-C algorithm as described in any one of claims 1 to 3.

Images