Neural network-based multi-parameter imaging method for induced polarization effect of electrical source

Abstract

The application relates to a kind of neural network-based electric source induction-polarization symbiotic effect multi-parameter imaging method.According to the fractional order model of polarized medium, the conductivity formula is substituted into Maxwell equation, and the electric source induction-polarization symbiotic effect formula is derived.According to the geological data of the experimental area, the polarized medium model of different conductivity, polarization rate, frequency dispersion coefficient and time constant parameters is constructed and numerically simulated to construct a sample set.The structure and activation function of the neural network are optimized, the neural network is trained and the performance is optimized.The neural network is applied to extract the multi-parameter of polarized medium from the measured data of electric source induction-polarization symbiotic effect, and multi-parameter-depth imaging is realized.The purpose of the application is to extract the multi-parameter information of polarized medium, and the conductivity and polarization rate imaging results are more accurate compared with the traditional resistivity imaging method.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of geophysical signal processing, and particularly relates to an electrical source induction-polarization coexistence effect inversion method. BACKGROUND

[0002] In the field of geophysical exploration, the electrical property information of underground medium is obtained by inversion of measured geophysical data. The actual underground medium containing polymetallic ore and other media is a polarized medium, and the resistivity is a complex number related to frequency. Therefore, under the excitation of alternating field, electromagnetic induction and polarization effect are generated at the same time. Electromagnetic induction can be used to distinguish the lithology information of stratum, and polarization effect may cause the electromagnetic response to produce a "reverse sign" phenomenon, which contains the information of polarized medium such as metal ore. For the complex geological structure of underground and the polarized medium model, the conductivity, polarization rate, frequency dispersion coefficient and time constant of the induction-polarization coexistence effect need to be extracted.

[0003] CN111413738B discloses a time domain induced polarization spectrum analysis method and system of porous medium. The relaxation time distribution is estimated by joint inversion of apparent polarization rate data calculated by multiple different charging times, and the pore size distribution can be obtained.

[0004] CN110133733A discloses a conductivity-polarizability multi-parameter imaging method based on particle swarm optimization algorithm. The zero-frequency conductivity and time constant are obtained by early and late data, and the parameter extraction of the Cole-Cole complex conductivity model of polarized medium is realized by using particle swarm algorithm.

[0005] CN110673218A discloses a method for extracting polarization medium parameter information in transient electromagnetic response of grounding conductor source. The underground resistivity information is obtained by inversion of vertical magnetic field which is less affected by polarization effect. The electric field response is obtained by forward calculation, and the pure polarization response in the measured data is obtained and the polarization parameters are obtained by inversion. However, this method solves the parameters such as resistivity and polarization rate respectively, so it is of great significance to study the multi-parameter imaging method for the fractional order model of polarized medium. SUMMARY

[0006] The technical problem to be solved by the present application is to provide an electrical source induction-polarization coexistence effect multi-parameter imaging method based on neural network, which can quickly and accurately predict the electrical property structure of actual underground medium.

[0007] The present application is implemented in the following way,

[0008] The electrical source induction-polarization coexistence effect multi-parameter imaging method based on neural network comprises the following steps:

[0009] 1) According to the fractional order model of polarized medium, the fractional order complex conductivity formula is substituted into Maxwell equation, and the formula of electric source induction-polarization coexistence effect is derived;

[0010] 2) According to the geological data of the experimental area, the underground medium model parameter information is obtained, the polarization medium model with different conductivity, polarization rate, frequency dispersion coefficient and time constant is designed, the electric source induction-polarization coexistence effect formula of step 1 is applied to carry out numerical simulation, and the sample set is constructed;

[0011] 3) The neural network structure and the activation function are optimized and selected, the sample set of step 2 is applied to training, the performance of the neural network is optimized, and the neural network is established;

[0012] 4) The measured data of the electric source induction-polarization coexistence effect are preprocessed, the neural network of step 3 is applied, and the polarization medium model parameters of the measured data are extracted, wherein the parameters include zero-frequency conductivity, polarization rate, frequency dispersion coefficient, time constant and depth;

[0013] 5) The results of step 4 are applied to carry out conductivity-depth and polarization rate-depth imaging, and the underground medium information is obtained.

[0014] Further, the step 2 comprises calculating the electric source response according to the electric source induction-polarization coexistence effect formula, and analyzing the influence of conductivity, polarization rate, frequency dispersion coefficient and time constant on the electric source induction-polarization coexistence effect; according to the analysis result and the geological data of the experimental area, the polarization medium model with different conductivity, polarization rate, frequency dispersion coefficient and time constant is designed, and the electric source induction-polarization coexistence effect of each model is calculated.

[0015] Further, the step 3 comprises the following steps:

[0016] The sample set of different polarization medium models and the electric source induction-polarization coexistence effect thereof is constructed;

[0017] According to the sample set, the neural network structure is optimized and designed, and the activation function is optimized and selected;

[0018] The sample set is input to train the neural network;

[0019] The performance of the neural network is optimized, and the loss function of the neural network is calculated;

[0020] It is judged whether the loss function reaches a threshold value, if greater than the threshold value, the step of optimizing and designing the structure of the neural network and optimizing and selecting the activation function is carried out;

[0021] The trained neural network is saved.

[0022] Compared with the prior art, the present application has the beneficial effects that:

[0023] The present application aims at the electric source induction-polarization symbiotic effect, adopts neural network inversion to accurately obtain conductivity, polarization rate, frequency dispersion coefficient, time constant and other parameter results, is favorable for practicalization of electric source detection technology, and the method provides new technical support for developing electromagnetic detection to find mineral resources. BRIEF DESCRIPTION OF DRAWINGS

[0024] Figure 1 It is a neural network-based electric source induction-polarization symbiotic effect multi-parameter imaging method flow chart;

[0025] Figure 2 It is the resistivity-depth effect graph and the polarization rate-depth effect graph of one embodiment of the present application. DETAILED DESCRIPTION

[0026] In order to make the purpose, technical scheme and advantages of the present application more clear, the present application is further described in detail below in combination with embodiments. It should be understood that the specific embodiments described here are only used to explain the present application, and are not used to limit the present application.

[0027] EMBODIMENT

[0028] Taking the Cole-Cole model as an example

[0029] In combination with Figure 1 As shown in the figure, a neural network-based electric source induction-polarization symbiotic effect multi-parameter imaging method comprises the following steps:

[0030] 1) According to the fractional order model of the polarization medium, the fractional order complex conductivity formula is substituted into the Maxwell equation, and the electric source induction-polarization symbiotic effect formula is derived;

[0031] The Cole-Cole model complex resistivity expression is:

[0032]

[0033] Where σ0 is the zero-frequency resistivity, η is the polarization rate, τ is the time constant, c is the frequency dispersion coefficient, and ω is the angular frequency. The fractional order complex resistivity expression is substituted into the Maxwell equation, it is assumed that the center of the electric source coincides with the coordinate origin, and extends to-L and L along the x-axis to the two sides, respectively, and the vertical component expression of the magnetic field of the electric source induction-polarization symbiotic effect is:

[0034]

[0035] Where I is the transmitting current, μ0 is the magnetic permeability, J1 is the first-order expression of the Bessel function, r is the transmitting-receiving distance, y is the y coordinate of the receiving point, z is the z coordinate of the receiving point, λ, x' are the integrated variables. γ TE The reflection coefficient is γ,

[0036] 2) According to the geological data of the experimental area, obtain the parameter information of the underground medium model, design polarization medium models with different conductivity, polarization rate, dispersion coefficient and time constant, and apply the electrical source induction-polarization coexistence effect formula in step 1 to carry out numerical simulation to build a sample set;

[0037] According to the electrical source response calculated by the electrical source induction-polarization coexistence effect formula, analyze the influence of conductivity, polarization rate, dispersion coefficient and time constant on the electrical source induction-polarization coexistence effect; according to the analysis results and the geological data of the experimental area, design polarization medium models with different conductivity, polarization rate, dispersion coefficient and time constant, and calculate the electrical source induction-polarization coexistence effect of each model.

[0038] 3) Optimize the selection of neural network structure and activation function, train the neural network with the sample set in step 2, optimize the performance of the neural network and build the neural network;

[0039] Comprising the following steps:

[0040] I. Constructing different polarization medium models and their sample sets of electrical source induction-polarization coexistence effect;

[0041] II. According to the sample set, optimize the design of neural network structure and optimize the selection of activation function;

[0042] III. Input the sample set for neural network training;

[0043] IV. Optimize the performance of the neural network and calculate the loss function of the neural network;

[0044] V. Determine whether the loss function reaches the threshold value, if greater than the threshold value, then optimize the design of the structure of the neural network and optimize the selection of the activation function in step II;

[0045] VI. Save the trained neural network.

[0046] 4) Preprocess the measured data of electrical source induction-polarization coexistence effect, apply the neural network in step 3 to extract the polarization medium model parameters of the measured data, including zero-frequency conductivity, polarization rate, dispersion coefficient, time constant and depth;

[0047] 5) Apply the results of step 4 to conduct conductivity-depth and polarization rate-depth imaging to obtain underground medium information.

[0048] Steps 4) and 5) comprise the following steps:

[0049] A. According to the detection requirements, apply the superconducting quantum sensor to conduct actual electrical source detection and collect induction-polarization coexistence effect measured data;

[0050] B, the measured data are preprocessed, including superposition, denoising and data sampling;

[0051] C, the preprocessed measured data in step B are subjected to parameter extraction by applying a neural network, to extract zero-frequency conductivity, polarizability, frequency dispersion coefficient, time constant and depth;

[0052] D, conductivity-depth imaging and polarizability-depth imaging are drawn and the imaging results are analyzed to obtain underground medium information.

[0053] Figure 2 For adopting Figure 1 The resistivity-depth and polarizability-depth effect diagrams of one embodiment of the application shown in the results conform to the theoretical model of the embodiment, and provide a new idea and method for high-precision and high-efficiency inversion of electrical source induction-polarization effect data.

[0054] The above only describes the preferred embodiments of the application and is not intended to limit the application, and any modification, equivalent replacement and improvement made within the spirit and principle of the application shall be included in the protection scope of the application.

Claims

1. A neural network-based multi-parameter imaging method for electric source induced polarization co-effect, characterized in that, It comprises the following steps: 1) According to the fractional order model of polarized medium, the fractional order complex conductivity formula is substituted into Maxwell equation, and the formula of electric source induction-polarization coexistence effect is derived; 2) According to the geological data of the experimental area, the underground medium model parameter information is obtained, the polarized medium model with different conductivity, polarization rate, frequency dispersion coefficient and time constant is designed, the electric source induction-polarization coexistence effect formula of step 1 is applied to numerical simulation, and the sample set is constructed; 3) The neural network structure and activation function are optimized and selected, the sample set of step 2 is applied to training, the performance of the neural network is optimized, and the neural network is established; The step 3) comprises the following steps: Constructing sample set of different polarized medium model and its electric source induction-polarization coexistence effect; According to the sample set, the neural network structure is optimized and designed, and the activation function is optimized and selected; Input sample set for neural network training; The performance of the neural network is optimized, and the loss function of the neural network is calculated; Determine whether the loss function reaches the threshold value, if greater than the threshold value, then optimize the structure of the neural network and optimize the selection of the activation function step; Save the trained neural network; 4) The measured data of electric source induction-polarization coexistence effect is preprocessed, the neural network of step 3 is applied, and the polarized medium model parameters of the measured data are extracted, wherein the parameters include zero frequency conductivity, polarization rate, frequency dispersion coefficient, time constant and depth; 5) The results of step 4 are applied to conductivity-depth and polarization rate-depth imaging to obtain underground medium information.

2. The method of claim 1, wherein, The step 2 comprises calculating the electric source response according to the electric source induction-polarization coexistence effect formula, and analyzing the influence of conductivity, polarization rate, frequency dispersion coefficient and time constant on the electric source induction-polarization coexistence effect; According to the analysis result and the geological data of the experimental area, the polarized medium model with different conductivity, polarization rate, frequency dispersion coefficient and time constant is designed, and the electric source induction-polarization coexistence effect of each model is calculated.

Images