Resource electromagnetic exploration efficient probability inversion method, medium and electronic equipment

Abstract

The invention relates to the technical field of electromagnetic prospecting, and discloses a resource electromagnetic prospecting efficient probability inversion method, a medium and electronic equipment, the method adopts a Fourier neural network in machine learning, an electromagnetic forward modeling response is obtained through training by inputting a large number of conductivity models, and the network is connected with the electromagnetic forward modeling response through a cascade structure of multiple layers of neurons. Features are automatically extracted from a large amount of magnetotelluric response data, mapping from a conductivity model to magnetotelluric response is constructed, traditional forward modeling calculation of the conductivity model obtained through sampling each time is avoided, and the forward modeling calculation time in the sampling process is shortened; according to the method, the prior constraint of a related structure is constructed in advance and is added into inversion prior distribution, a Metropolis-Hasting acceptance criterion is reconstructed, posterior probability density distribution of a model is quickly searched by using prior information, the search time in an inversion space during sampling is shortened, and quantitative evaluation is performed on an obtained inversion result.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of electromagnetic exploration, in particular to a resource electromagnetic exploration efficient probability inversion method, medium and electronic equipment. BACKGROUND

[0002] Electromagnetic exploration has important application value in mineral exploration, geological exploration, deep structure research and underground geothermal resource exploration. Through electromagnetic sounding technology, we can obtain electromagnetic response curves at each measurement point for describing underground structure. Based on these electromagnetic responses, inversion can reveal the underground conductivity structure, so as to infer the distribution of underground structure. Therefore, electromagnetic inversion has important significance in practical application.

[0003] However, in the process of electromagnetic inversion, the measurement is greatly affected by interference, and the noise has a significant impact on the stability of the inversion result. This makes the inversion result may have multiple models fitting with the measured data, resulting in non-uniqueness of inversion. Therefore, it is necessary to quantitatively evaluate the inversion result. In order to solve this nonlinear problem, a probability inversion method based on Bayesian framework can be used. In this framework, all possible solutions are contained in the posterior probability density distribution, and through statistical analysis of this distribution, the inversion result can be effectively quantitatively evaluated. At present, this method has made significant progress.

[0004] In order to obtain an unbiased posterior probability density distribution, the existing algorithm needs to perform a large number of sampling calculations, resulting in a very time-consuming calculation process. Especially when inverting high-dimensional models, the amount of calculation increases sharply, causing "dimension disaster". Therefore, how to improve the sampling efficiency, shorten the calculation time, and quickly obtain the posterior probability density distribution of the model conductivity has important practical significance. SUMMARY

[0005] The purpose of the present application is to provide a resource electromagnetic exploration efficient probability inversion method, medium and electronic equipment to solve the problems of low sampling efficiency, long calculation time and inability to quickly obtain the posterior probability density distribution of the model conductivity in the prior art. The specific technical solutions are as follows: A resource electromagnetic exploration efficient probability inversion method, comprising the following steps: S1, obtaining the underground electromagnetic field response data of the research area by using an electromagnetic measuring instrument; constructing different conductivity models; using a Fourier neural network to train the generated conductivity models according to different sampling frequencies; obtaining the mapping output from the conductivity model to the electromagnetic response; S2, constructing a tower structure model required for inversion; constructing a Markov chain from the first sampling, and in the second sampling During the next sampling, corresponding candidate tower structure models are obtained according to different strategies based on the current tower structure model. The proposed model for the conductivity structure is obtained by performing inverse wavelet transform on the candidate tower structure model. At the same time, for the first Tower structure model during secondary sampling The corresponding resistivity model is obtained by performing inverse wavelet transform. ; S3. The suggested model obtained in S2 and resistivity model The input is fed into a pre-trained network at each frequency point, and the forward modeling calculation is performed in parallel on the GPU to obtain the corresponding electromagnetic response; S4. Apply zero-mean normal prior constraint distributions to the resistivity structures in the vertical and horizontal directions; construct the Metropolis-Hasting acceptance criterion based on the wavelet domain tower structure model with added prior constraints on the model structure, and calculate the acceptance rate of the proposed model. S5. Perform a judgment: if the conditions are met, accept the tower structure candidate model, thus obtaining the first... Tower structure model at +1 sampling for: Otherwise, reject the candidate tower structure model, thus obtaining the first... Tower structure model at +1 sampling for: , This is the previous tower structure model; take = +1, perform a second check: if If the number of samples is less than or equal to the maximum number of samples designed, return to S2; otherwise, proceed to the next step. S6. Perform statistical calculations on all suggested models to obtain the probability density distribution of underground electrical conductivity in the study area.

[0006] Preferably, in S2: constructing the required tower structure model for inversion includes designing the maximum depth of the tower structure model. The maximum number of nodes in the tower structure model was calculated. : ; Before constructing the Markov chain from the first sampling, the number of permutations and combinations at different node depths is calculated. The formula for calculating permutations and combinations in S2 is as follows: ; in: Indicates in Depth and Number of arrangements of a four-node tower structure model under each node; Indicates in depth and the number of nodes under the arrangement combination; , , counting when the algorithm recurses; representing the number of tower nodes; representing the depth of the tower structure model.

[0007] Preferably, in S2: the nodes of the tower structure model required for constructing the inversion are classified into existing nodes , generatable nodes and eliminable nodes , representing empty nodes; optionally, one type of node is changed in the tower structure model to form a tower structure candidate model ; If the generatable node is selected, i.e. the generation strategy, then the tower structure candidate model node , and the tower structure candidate model is: ; If the eliminable node is selected, i.e. the elimination strategy, then the tower structure candidate model node , and the tower structure candidate model is: ; If one of the existing nodes is selected, i.e. the perturbation strategy, then only the value of the selected node is changed , and the tower structure candidate model is: ; wherein: represents the specific tower structure model arrangement combination of the tower structure candidate model under depth and the number of nodes; represents the wavelet coefficient value when the number of nodes is , and represents the value of the selected thnode after the change; In S4: the inverse transform adopts a wavelet basis function.

[0008] Preferably, in S4: based on the thsampling, a zero-mean normal prior distribution is imposed on the vertical and horizontal resistivity structures as follows: ; wherein: represents the structure prior constraint in the direction under two dimensions; represents the structure prior constraint in the direction under two dimensions; represents rank of the number of model parameters in the direction; denotes rank of the number of model parameters in the direction; denotes different difference operator matrices; denotes different difference operator matrices; denotes regularization factor given by the structural constraint in the direction; denotes regularization factor given by the structural constraint in the direction; denotes transpose; by the wavelet inverse transform operator , the wavelet domain model is inverse transformed into the conductivity model , the structural prior distribution of the current th sampling needs to calculate the structural prior constraint and the structural prior constraint ; is expressed as: ; when , then: ; wherein: ; , the regularization factor is regarded as a hyperparameter, .

[0009] Preferably, the Metropolis-Hasting acceptance criterion based on the wavelet domain tower structure model with added model prior constraint is constructed, the acceptance rate of the proposed model is calculated, which specifically includes the following steps: S4.1, calculate the prior distribution of the current th sampling under the given tower structure model and the prior distribution of the tower structure candidate model ; S4.2, use the electromagnetic response obtained in S3 to calculate the likelihood function , wherein the calculation of the likelihood function is to compare the predicted electromagnetic response with the electromagnetic data collected by the electromagnetic measuring instrument, and calculate the fitting difference between them: ; ; wherein: represents the acquired electromagnetic data; represents the transpose; represents the covariance matrix of the data; the tower structure candidate model of the (n-1)th sampling is calculated in the same way the resistivity model obtained by inverse transformation the likelihood function and the fitting difference , then the ratio of the likelihood function value of the existing conductivity model and the new candidate conductivity model is calculated; ; S4.3, the proposal distribution probability from the tower structure model to the tower structure candidate model and the proposal distribution from the tower structure candidate model to the tower structure model are calculated ; ; S4.4, at the nth sampling, the acceptance rate of the proposed model is calculated based on the obtained prior distribution, likelihood function and proposal distribution according to the following formula acceptance criterion : ; wherein: is the Jacobian matrix, which is taken as .

[0010] Preferably, the prior distribution in S6.1 is: ; wherein: represents the wavelet sparse value prior distribution, ; is the prior distribution of the number of nodes, ; is the permutation distribution under the condition of depth and nodes, .

[0011] Preferably, the proposal distribution includes the proposal distribution of the perturbation strategy , the proposal distribution of the generation strategy and the proposal distribution of the extinction strategy , as follows: Proposal distribution of the perturbation strategy: wherein:​ represents the change of an arbitrary node value selected from the set, and the specific operation is Gaussian perturbation of the current value; represents the perturbation set distribution, which is a uniform distribution, represents the change of a wavelet coefficient.

[0012] The suggestion distribution of the generation strategy is: wherein: represents the change of an arbitrary node value selected from the set; and the suggestion distribution of the new node value ; represents the generation set distribution, which is a uniform distribution; The suggestion distribution of the extinction strategy is: wherein: represents the extinction set distribution, which is a uniform distribution.

[0013] The expression of the suggestion model acceptance rate of selecting the existing node for the perturbation step is: ;The expression of the suggestion model acceptance rate of selecting the generatable node is: ; ;wherein: represents the candidate extinction node number of the tower structure candidate model ;The expression of the suggestion model acceptance rate of selecting the generatable node is: ; ; wherein: represents the candidate generation node number of the tower structure candidate model .

[0014] The application further discloses a computer readable storage medium, wherein computer execution instructions are stored in the computer readable storage medium, and the computer execution instructions are used for realizing the resource electromagnetic exploration efficient probability inversion method.

[0015] The application further discloses an electronic device, which comprises a processor and a memory connected with the processor in communication; The memory stores computer execution instructions; The processor executes the computer execution instructions stored in the memory, so as to realize the resource electromagnetic exploration efficient probability inversion method.

[0016] By applying the technical scheme of the application, the following effects are achieved:​​ The application provides a resource electromagnetic exploration efficient probability inversion method, which adopts a Fourier neural network in machine learning, trains an electromagnetic forward response by inputting a large number of conductivity models, and can automatically extract features from a large amount of magnetotelluric response data through a cascade structure of multiple neurons, construct a mapping from a conductivity model to a magnetotelluric response, avoid traditional forward calculation of the conductivity model obtained each time sampling, greatly accelerate the forward calculation time in the sampling process, construct prior constraints related to the structure in advance, add to the inversion prior distribution, reconstruct the Metropolis-Hasting acceptance criterion, use the prior information to quickly search the posterior probability density distribution of the model, effectively reduce the search time in the inversion space during sampling, and quantitatively evaluate the inversion result, and quantify the uncertainty of the inversion result.

[0017] In addition to the purposes, features and advantages described above, the application has other purposes, features and advantages. The application will be further described below with reference to the drawings. BRIEF DESCRIPTION OF DRAWINGS

[0018] In order to more clearly illustrate the technical solutions of the embodiments of the application or the prior art, the drawings needed in the embodiment or prior art description will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the application, and those skilled in the art can also obtain other drawings according to these drawings without creative labor.

[0019] Figure 1 is the algorithm flow in the embodiment; Figure 2 is a schematic diagram of a tower structure model when the maximum depth is 3 in the embodiment (1-16 in the figure are serial numbers); Figure 3 is a schematic diagram of a tower structure model that can be disturbed, generated and disappeared in the embodiment (1-16 in the figure are serial numbers); Figure 4 is a block diagram of an electronic device in the embodiment; Figure 5 is a schematic diagram of a theoretical model designed in the embodiment; Figure 6 is a comparison of mean models obtained by inversion under the condition of having or not having a model constraint, wherein (a) is a posterior mean model obtained by one-chain sampling based on the image compression magnetotelluric resource electromagnetic exploration efficient probability inversion method; (b) is a posterior mean model obtained by 5-chain parallel tempering sampling based on the image compression magnetotelluric resource electromagnetic exploration efficient probability inversion method; (c) is a posterior mean model obtained by one-chain sampling in the embodiment of the application; Figure 7is the comparison of the inversion results with and without model constraints, wherein: (a) is a center position profile conductivity posterior probability density distribution map obtained by one chain sampling of the image compression-based high efficiency probability inversion method for georesources electromagnetic exploration; (b) is a center position profile conductivity posterior probability density distribution map obtained by 5-chain parallel annealing sampling of the image compression-based high efficiency probability inversion method for georesources electromagnetic exploration; (c) is a center position profile conductivity posterior probability density distribution map obtained by one chain sampling in the embodiment of the application; Figure 8 is the comparison of the fitting difference with and without model constraints in the embodiment; Figure 9 is the comparison of the node number with and without model constraints, wherein: (a) is the node number calculated by one chain of the image compression-based high efficiency probability inversion method for georesources electromagnetic exploration; (b) is the node number calculated by 5-chain parallel annealing of the image compression-based high efficiency probability inversion method for georesources electromagnetic exploration; (c) is the node number calculated by one chain in the embodiment of the application; Figure 10 is the comparison of the regularization coefficient distribution map in the embodiment; Figure 11 is the comparison of the running time of 10,000 times (indicated as 1) of the application and the running time of 10,000 times (indicated as 2) of the image compression-based high efficiency probability inversion method for georesources electromagnetic exploration in the embodiment; Figure 12 is a parallel computing schematic diagram of the Fourier neural network in the embodiment; Figure 13 is a prediction result map of a certain input model during training in the embodiment, wherein: (a) indicates a certain input conductivity model; (b) indicates the error between the predicted electromagnetic response and the real electromagnetic response; (c) indicates the real electromagnetic response; (d) indicates the predicted electromagnetic response. DETAILED DESCRIPTION

[0020] In order for those skilled in the art to better understand the present application, the application will be further described in detail below in combination with the drawings and specific embodiments. Obviously, the described embodiments are only part of the embodiments of the present application, but not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0021] Embodiment: A resource electromagnetic exploration high efficiency probability inversion method, the flow chart is shown in Figure 1 , and specifically includes the following steps: Step one, use electromagnetic measuring instrument to obtain the underground electromagnetic field response data of the research area, and through the analysis of the sampling data frequency range, the measured underground depth (i.e. electromagnetic induction depth) is calculated; according to different geological data, different conductivity models are constructed; using Fourier neural network, the generated conductivity model is trained according to different sampling frequency points, so as to construct the forward calculation network under different frequency points, and the mapping output from the conductivity model to the electromagnetic response is obtained.

[0022] In the preferred embodiment of the present embodiment, in order to obtain the conductivity model for training, the present embodiment defines the geometric structure of the grid according to the electromagnetic properties, uniformly divides the horizontal direction, and uses incremental grid in the vertical direction, so as to better simulate the longitudinal variation of the underground electrical structure. And through the conductivity generation function, the resistivity value of the underground block anomaly body is simulated. For example, in the range of 10-1000 Ω·m, the value is changed arbitrarily in the vertical direction of 20-25 layers, the horizontal direction of 9-16 columns and 17-24 columns, so as to simulate the layered or block geological structure and represent the non-uniformity of the geological background and the anomaly body.

[0023] The Fourier neural network adopted by the present application is a 5-layer Fourier neural network architecture, each layer contains a double-branch structure: a frequency domain processing branch, which performs fast Fourier transform (FFT) on the input data, filters out high-frequency components (retains the first 20% low-frequency components), and then performs inverse Fourier transform (IFFT) to extract global frequency domain features; a spatial feature branch, which captures local electrical structure details through a 2x2 convolution kernel. After weighted fusion, the two branches are linearly connected to the next layer, and finally a 2x2 dimensional result is output through the fully connected layer. The network input includes three channels of data: two-dimensional resistivity model, real part and imaginary part of electromagnetic response. The generated model is divided into training set, validation set and test set in the ratio of 6:2:2, and is trained according to the collected frequency, so as to achieve one frequency corresponding to one trained Fourier neural network.

[0024] Step two, construct the required tower structure model for inversion, obtain the tower structure model; calculate the number of tower arrangement combinations under different depth nodes, and start constructing the Markov chain from the first sampling.

[0025] In the preferred embodiment of the present embodiment, the details are shown in Figure 2 , the construction of the required tower structure model for inversion includes the design of the maximum depth of the tower structure model . Figure 2 The maximum depth is 3 as shown in the figure. The maximum number of nodes of the tower structure model is calculated as follows: ; The permutation and combination calculation formula in S2 is as follows: ; wherein: represents the number of permutations under depth and nodes; represents the number of permutations and combinations under depth and nodes; , , represents the count when the algorithm recurses; represents the number of tower nodes; represents the depth of the tower structure model.

[0026] Step three, based on the tower structure model, the tower structure candidate model under the tower structure model at the first sampling time is obtained by the generation strategy, the elimination strategy and the perturbation strategy in the algorithm .

[0027] In the preferred embodiment of the present application, the nodes of the tower structure model are classified into existing nodes , generatable nodes and eliminable nodes , represents an empty node, see Figure 3 ; optionally, one type of node is changed to form a tower structure candidate model ; If the generatable node is selected, i.e. the generation strategy, it represents the tower structure candidate model node , and the tower structure candidate model is: ; If the eliminable node is selected, i.e. the elimination strategy, it represents the tower structure candidate model node , and the tower structure candidate model is: ; If one of the existing nodes is selected, i.e. the perturbation strategy, it represents only the value of the selected node is changed , and the tower structure candidate model is: ; wherein: represents the specific permutation and combination of the tower structure candidate model under depth and nodes; represents the wavelet coefficient value when the number of nodes is , and represents the value of the selected th node after the change.

[0028] Step four, the tower structure candidate model under the tower structure model at the first sampling time is obtained by the generation strategy, the elimination strategy and the perturbation strategy in the algorithm Wavelet inverse transform to obtain the suggested model of conductivity structure ; the first sampling tower structure model inverse transform to obtain the resistivity model The inverse transform preferred in this embodiment uses a wavelet basis function.

[0029] Step five, the suggested model of conductivity structure obtained in step four and the resistivity model respectively, forward calculation is carried out using the trained network at different frequencies. In this step, the CUDA computing library is used to call the GPU to distribute the model in parallel to the trained Fourier neural network at different frequencies, and output the corresponding electromagnetic response.

[0030] Step six, impose a zero-mean normal prior constraint distribution on the vertical and horizontal resistivity structures; construct a Metropolis-Hasting acceptance criterion based on the wavelet domain tower structure model with added model structure prior constraints, and calculate the acceptance rate of the suggested model .

[0031] In this embodiment, based on the first sampling, a zero-mean normal prior distribution is imposed on the vertical and horizontal resistivity structures as follows: ; Where: represents the structure prior constraint in the direction under two dimensions; represents the structure prior constraint in the direction under two dimensions; represents the rank of the number of model parameters in the direction; represents the rank of the number of model parameters in the direction; represents a difference operator matrix with a rank of ; represents a difference operator matrix with a rank of ; represents the regularization factor given by the structure constraint in the direction; represents the regularization factor given by the structure constraint in the direction; represents transposition; Through the wavelet inverse transform operator , the wavelet domain model is inverse transformed into a conductivity model , and for the structure prior distribution of the first sampling, the calculation of Structural prior constraint and Structural prior constraint ; is expressed as: ; Here, since we think that the structural constraint term in the y direction should have the same effect as the structural constraint term in the z direction, the regularization should be consistent, that is, take When ; Where: ; .

[0032] In this embodiment, in order to avoid the influence of the uncertainty of the inversion result caused by the artificial selection of the regularization factor , the regularization factor is regarded as a hyperparameter in the present application, a relatively wide uniform prior range is given, and is taken, so that it is adaptively constrained to a certain smaller range in the inversion process.

[0033] In this embodiment, the Metropolis-Hasting acceptance criterion based on the wavelet domain tower structure model with added model prior constraint is constructed, and the acceptance rate of the proposed model is calculated, which specifically includes the following steps: Step 6.1, calculate the prior distribution of the tower structure model under the given tower structure model under the current sampling times and the prior distribution of the tower structure candidate model ; Step 6.2, use the electromagnetic response obtained in S5 to calculate the likelihood function , wherein the calculation of the likelihood function is to compare the predicted electromagnetic response with the electromagnetic data collected by the electromagnetic measuring instrument, and calculate the fitting difference between them: ; ; Wherein: represents the electromagnetic data collected; represents transposition; represents the covariance matrix of the data; The resistivity model obtained by inverse transformation of the tower structure candidate model under the sampling times likelihood function and the fitting difference , then calculate the existing conductivity model and the likelihood function value of the new candidate conductivity model ; ; Step 6.3, calculate the proposal distribution probability from the tower structure model to the tower structure candidate model and the proposal distribution from the tower structure candidate model to the tower structure model ; ; Step 6.4, at the first sampling, based on the obtained prior distribution, likelihood function and proposal distribution, construct the following formula acceptance criterion to calculate the acceptance rate of the proposed model : ; wherein: is the Jacobian matrix, and is taken .

[0034] Further preferred in the embodiment, the prior distribution is: ; wherein: indicates the wavelet sparse value prior distribution, and in the application, we limit the wavelet coefficient prior value distribution to a uniform distribution of [-1.5, 1.5], ; is the prior distribution of the number of nodes, ; is the permutation number distribution under the condition of depth and nodes, .

[0035] Further preferred in the embodiment, the proposal distribution includes the proposal distribution of the perturbation strategy , the proposal distribution of the generation strategy and the proposal distribution of the extinction strategy , as follows:

[0036] The proposal distribution of the perturbation strategy: wherein: indicates the change of any one node value selected from the set , and the specific operation is Gaussian perturbation of the current value; indicates the perturbation set distribution, which is a uniform distribution, indicates the change amount of the wavelet coefficient.​​

[0037] The proposal distribution of the generation strategy: wherein: represents the change of an arbitrary node value selected from the set The proposal distribution of the new node value ; represents the generation set distribution, which is a uniform distribution. The proposal distribution of the extinction strategy: wherein: represents the extinction set distribution, which is a uniform distribution.

[0038] In this embodiment, the expression of the proposal model acceptance rate of the existing node selected for the perturbation step is: ; The expression of the proposal model acceptance rate of the generatable node is: ; wherein: represents the number of candidate extinction nodes of the tower structure candidate model The expression of the proposal model acceptance rate of the extinguishable node is: ; wherein: represents the number of candidate generation nodes of the tower structure candidate model . Step seven, a determination is made, specifically: if the acceptance rate of the proposal model

[0039] satisfies greater than 0 and less than 1, the tower structure candidate model is accepted, and the tower structure model at the n+1th sampling is obtained as: ; otherwise, the tower structure candidate model is rejected, and the tower structure model at the n+1th sampling is obtained as: , ; wherein M represents the maximum sampling number designed. , is the previous tower structure model; n is taken as = n+1, and a second determination is made: if is less than or equal to the designed maximum sampling number, step three is returned to, otherwise the next step is entered.

[0040] ​​​​​​​Step eight, statistical calculation is made to all the proposed models to obtain the underground conductivity probability density distribution in the research area.

[0041] The schematic diagram of the theoretical model designed in the embodiment is shown in Figure 5 , the detection range is 20km*20km, the background conductivity is , and the conductivity of the anomaly body is . The frequency used is 8 frequency values: 0.05, 0.20, 0.5, 1.0, 5.0, 20, 50.0, 100 Hz, and the station mark is represented by a red triangle.

[0042] The comparison of the mean model obtained by inversion with and without model constraints is shown in Figure 6 , wherein: (a) is the posterior mean model obtained by one-chain sampling based on the image compression of the high-efficiency probability inversion method of the earth resource electromagnetic exploration; (b) is the posterior mean model obtained by 5-chain parallel tempering sampling based on the image compression of the high-efficiency probability inversion method of the earth resource electromagnetic exploration; and (c) is the posterior mean model obtained by one-chain sampling in the embodiment of the application. As can be seen from the results, when the image compression-based geoelectric probability inversion is used, multiple Markov chains need to be used in parallel tempering to avoid the model from falling into a local extremum, and after the model constraint term is added in the application, only one Markov chain can be used to invert the underground anomaly position and size.

[0043] The comparison of the uncertainty of the inversion results obtained with and without model constraints is shown in Figure 7 , wherein: (a) is the center position profile conductivity posterior probability density distribution diagram obtained by one-chain sampling based on the image compression of the high-efficiency probability inversion method of the earth resource electromagnetic exploration; (b) is the center position profile conductivity posterior probability density distribution diagram obtained by 5-chain parallel tempering sampling based on the image compression of the high-efficiency probability inversion method of the earth resource electromagnetic exploration; and (c) is the center position profile conductivity posterior probability density distribution diagram obtained by one-chain sampling in the embodiment of the application. As can be seen from the results, after the model constraint term is added, the model uncertainty constraint ability is strengthened, and the conductivity uncertainty in the deep part is small.

[0044] The comparison of the fitting difference with and without model constraints in the embodiment of the application is shown in Figure 8Wherein: the orange line represents the fitting difference of 5 parallel tempering chains without model constraints; the yellow line represents the fitting difference of 1 chain with model constraints; and the blue line represents the fitting difference of 1 chain without model constraints. As can be seen from the results, the yellow line in the application converges to a static sampling process after 0.6 million samplings, while the blue line needs to be sampled to 10,000 times, and the orange line needs to be sampled by 5 chains in parallel.

[0045] The comparison results of the number of nodes with and without model constraints are shown in the following table: Figure 9 Wherein: (a) is the number of nodes calculated by the one-chain of the efficient probability inversion method for earth resource electromagnetic exploration based on image compression; (b) is the number of nodes calculated by the 5-chain parallel tempering of the efficient probability inversion method for earth resource electromagnetic exploration based on image compression; and (c) is the number of nodes calculated by the one-chain in the embodiment of the application. As can be seen from the number of nodes, the number of nodes will decrease rapidly after the model constraint term is added, thereby reducing the burden of parameter dimension during inversion.

[0046] The regularization coefficient in the embodiment of the application The distribution diagram is shown in the following table: Figure 10 By giving the statistical distribution diagram of the regularization coefficient It can be seen that the regularization factor of the application can be adaptively constrained in a certain range during the inversion step, and the range is much narrower than the given prior range , which indicates that the effect is obvious, avoids the deviation caused by artificial giving, and saves the time of repeatedly adjusting the value to obtain better results.

[0047] The comparison diagram of the running time of 10,000 times (represented as 1) of the application and the running time of 10,000 times (represented as 2) of the efficient probability inversion method for earth resource electromagnetic exploration based on image compression is shown in the following table: Figure 11 Through the comparison of the two, it can be seen that the calculation time of the application is greatly reduced, and it only needs 3.18 minutes to run 10,000 times, which greatly improves the efficiency of inversion.

[0048] The parallel computing schematic diagram of the Fourier neural network in the embodiment of the application is shown in the following table: Figure 12 In the application, we use the Fourier neural network to predict the model response. In the model loading part, the trained neural networks of different frequencies are loaded into the inversion algorithm, and the neural networks of different frequencies are called in parallel during the forward calculation to calculate the electromagnetic response at all required frequencies at one time, thereby efficiently calculating the electromagnetic response of the model.

[0049] The prediction result diagram of a certain input model during the training of the network is shown in the following table:Figure 13 Wherein: (a) represents the conductivity model of a certain input; (b) represents the absolute error between the predicted electromagnetic response and the real electromagnetic response; (c) represents the real electromagnetic response; (d) represents the predicted electromagnetic response. By comparing (c) and (d), only the deep prediction effect is poor while the other shapes are basically consistent, and the maximum absolute error value is only 0.00088 in terms of error value of (b), which shows that the network trained by the application is accurate in calculation.

[0050] In addition, as shown in Figure 4 The embodiment further provides an electronic device 200, which comprises a processor 201 and a memory 202 connected with the processor 201 in communication; The memory 202 stores computer execution instructions; The processor 201 executes the computer execution instructions stored in the memory 202, so as to realize the electromagnetic probabilistic inversion method as described above.

[0051] In addition, the embodiment further provides a computer readable storage medium, which stores computer execution instructions, and the computer execution instructions are used to realize the file generation method as described above when executed by a processor.

[0052] It should be understood that the division of each module of the above system is only a logical functional division, and all or part of the modules can be integrated into one physical entity, or can be physically separated. The modules can all be implemented in the form of software called by a processing element; all can be implemented in the form of hardware; or some modules can be implemented in the form of software called by a processing element, and some modules can be implemented in the form of hardware. Each module can be a separate processing element, or can be integrated in a chip of the above device, in addition, the functions of each module can be stored in the form of program code in the memory of the above device, and called and executed by a processing element of the above device. In addition, all or part of the modules can be integrated together, or can be independently implemented. The processing element herein can be an integrated circuit having a signal processing capability. In the implementation process, each step of the above method or each module can be completed by the integrated logic circuit of hardware or the instruction of software in the processing element.

[0053] It should be understood that the above device embodiment is only illustrative, and the device of the present application can also be implemented in other ways. For example, the division of units / modules in the above embodiment is only a logical functional division, and another division mode can be used in actual implementation. For example, a plurality of units, modules or components can be combined, or can be integrated into another system, or some features can be ignored or not executed.

[0054] In addition, each functional unit / module in the various embodiments of the present application can be integrated in one unit / module, or each unit / module can exist physically, or two or more units / modules can be integrated together. The integrated unit / module can be realized in the form of hardware or in the form of a software program module.

[0055] If implemented in hardware, the integrated units / modules can be digital circuitry, analog circuitry, or a combination of both. The physical implementation of the hardware structure includes, but is not limited to, transistors, memristors, and the like. Unless specifically stated otherwise, a processor can be any suitable hardware processor, such as a central processing unit (CPU), a graphics processing unit (GPU), an application specific integrated circuit (ASIC), a digital signal processor (DSP), a programmable logic device (PLD), a field programmable gate array (FPGA), a controller, a microcontroller, a microprocessor, or other electronic device. Unless specifically stated otherwise, a memory can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as a U disk, a random-access memory (RAM), a static random-access memory (SRAM), a dynamic random-access memory (DRAM), an enhanced dynamic random-access memory (EDRAM), an electrically erasable programmable read-only memory (EEPROM), an erasable programmable read-only memory (EPROM), a programmable read-only memory (PROM), a read-only memory (ROM), a resistive random access memory (RRAM), a high-bandwidth memory (HBM), a hybrid memory cube (HMC), a magnetic memory, a flash memory, a magnetic disk, an optical disk, a mobile hard disk, or various media capable of storing program codes.

[0056] The integrated unit / module, if implemented in the form of a software program module and sold or used as an independent product, can be stored in a computer readable memory. Based on such understanding, the technical solutions of the present application essentially or the part of the prior art that contributes to the present application or the whole or part of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a memory and includes a plurality of instructions for causing a computer device (which can be a personal computer, a server or a network device, etc.) to execute all or part of the steps of the embodiments of the present application.

[0057] The above merely describes the preferred embodiments of the present application and is not intended to limit the present application. For those skilled in the art, the present application can have various modifications and changes. Any modification, equivalent replacement, improvement, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.

Claims

1. A method for efficient probabilistic inversion of resource electromagnetic exploration, characterized in that, The method comprises the following steps: S1, obtaining underground electromagnetic field response data of a study area by using an electromagnetic measuring instrument; constructing different conductivity models; training the generated conductivity models according to different sampling frequencies by using a Fourier neural network; obtaining mapping output from the conductivity model to the electromagnetic response; S2, constructing the tower structure model required by inversion; constructing the Markov chain from the first sampling, and according to the current tower structure model, obtaining the corresponding tower structure candidate model according to different strategies at the second sampling . ​ Wavelet inverse transform is performed on the tower structure candidate model to obtain a suggested model of the conductivity structure , and the tower structure model at the second sampling time is wavelet inverse transformed to obtain a corresponding resistivity model ;​​ S3, the suggestion model obtained in S2 and the resistivity model The trained network at each frequency point is input, forward calculation is completed in parallel on the GPU, and the corresponding electromagnetic response is obtained; S4, imposing a zero-mean normal prior constraint distribution on the vertical and horizontal resistivity structures; constructing a Metropolis-Hasting acceptance criterion based on the wavelet domain tower structure model with added model structure prior constraints, and calculating the acceptance rate of the proposed model; S5, make a decision, specifically: if satisfied, accept the tower structure candidate model, that is, get the tower structure model at the time of the first sampling +1 is: ; otherwise, reject the tower structure candidate model, that is, get the tower structure model at the time of the first sampling +1 is: , is the previous tower structure model; take = +1, make a second decision: if is less than or equal to the designed maximum sampling number, return to S2, otherwise proceed to the next step; S6, statistically calculating all proposed models to obtain the underground conductivity probability density distribution in the study area.

2. The resource electromagnetic survey efficient probabilistic inversion method of claim 1, wherein, In S2: constructing the required tower structure model for inversion includes designing the maximum depth of the tower structure model ; calculating the maximum number of nodes of the tower structure model : ; The arrangement combination number under the node at different depths is calculated before the Markov chain is constructed from the first sampling, and the arrangement combination calculation formula is as follows: ； wherein: represents the number of permutations at depth and nodes under a four-node tower structure model; represents the number of permutations at depth and nodes under a four-node tower structure model; , , represents the count when the algorithm recurses; represents the number of tower nodes; represents the depth of the tower structure model.

3. The resource electromagnetic survey efficient probabilistic inversion method of claim 2, wherein, In S2: the nodes of the tower model required for the inversion are classified as existing nodes , nodes that can be generated , and nodes that can be annihilated , are represented as empty nodes; Optionally, the nodes of one class are subjected to a change in the tower structure model, forming a tower structure candidate model ; If the node generation strategy is selected, it means that the tower structure candidate model node is generated , and the tower structure candidate model is ; If the vanishing node is selected, i.e. the vanishing strategy, it means that the tower structure candidate model node , and the tower structure candidate model is: ; If one of the existing nodes is selected, i.e. the perturbation strategy, it means that only the value of the selected node is changed, , the tower structure candidate model is: ; wherein: represents in depth and specific tower structure model arrangement of the tower structure candidate model under the represents the wavelet coefficient value when the number of nodes is represents the value after the change of the selected node.

4. The resource electromagnetic survey efficient probabilistic inversion method of claim 3, wherein, In S4: Based on the first Sub-sampling, a zero-mean normal prior is imposed on the vertical and horizontal resistivity structures as follows: ； wherein: represents the structure prior constraint in the direction of represents the structure prior constraint in the direction of represents the rank of the number of model parameters in the direction of represents the rank of the number of model parameters in the direction of represents the difference operator matrix with rank represents the difference operator matrix with rank represents the regularization factor given by the structure constraint in the direction of represents the regularization factor given by the structure constraint in the direction of represents the transpose;​​​​​​​​ by a wavelet inverse transform operator inverse transforming the wavelet domain model to the conductivity model the structure prior distribution for the first sampled structure needs to be computed the structure prior constraint and the structure prior constraint ; The performance is shown as: ； The structure constraint term in the y direction has the same effect as the structure constraint term in the z direction, i.e. take then we have: ； where: ; , the regularization factor is considered as a hyper-parameter, .

5. The resource electromagnetic survey high efficiency probabilistic inversion method of claim 4, wherein, constructing a Metropolis-Hasting acceptance criterion based on the wavelet domain tower structure model with added model prior constraints, and calculating the acceptance rate of the proposed model, which specifically comprises the following steps: S4.1, compute the current first prior distribution of the tower structure model under the second sampling and the prior distribution of the tower structure candidate model ; S4.2, computing a likelihood function using the electromagnetic response obtained in S3 , wherein the computation of the likelihood function is a comparison of the predicted electromagnetic response with the electromagnetic data acquired with the electromagnetic measuring instrument, computing a fit difference between them : ； ； wherein: represents the acquired electromagnetic data; represents the transpose; represents the covariance matrix of the data; Calculate the first in the same way Candidate models of tower structures obtained through subsampling Resistivity model obtained through inverse transformation likelihood function and fit difference Then, the existing conductivity model was calculated. With the new candidate conductivity model The ratio of the likelihood function values; ； S4.3, computing a proposal distribution probability from a tower structure model to a tower structure candidate model and a proposal distribution from a tower structure candidate model to a tower structure model ;​​ S4.4、in the fourth step At the time of sub-sampling, the acceptance rate of the suggestion model is calculated based on the obtained prior distribution, likelihood function and suggestion distribution according to the following formula acceptance criterion : ； where: is the Jacobian matrix, taken as .

6. The resource electromagnetic survey efficient probabilistic inversion method of claim 5, wherein, The prior distribution is: ； where: denotes the wavelet sparsity prior distribution, ; is the prior distribution of the number of nodes, ; is the depth and the number of nodes, .

7. The resource electromagnetic survey high efficiency probabilistic inversion method of claim 5, wherein, suggested distribution suggested distribution including perturbation strategies suggested distribution of generating strategies suggested distribution of vanishing strategies as follows: The suggestion distribution of the perturbation strategy is: Wherein: represents the change of an arbitrary node value selected from the set , and the specific operation is Gaussian perturbation of the current value; represents the perturbation set distribution, which is a uniform distribution, represents the change of the wavelet coefficient; Generating a proposal distribution for the strategy: where: represents the change in any one node value selected from the set ; the proposal distribution for the new node value ; represents generating a set distribution, which is a uniform distribution; The proposed distribution of extinction strategies is: where: denotes the extinction set distribution, which is a uniform distribution.

8. The resource electromagnetic survey efficient probabilistic inversion method of claim 5, wherein, Suggested model acceptance rate of selecting existing nodes for perturbation step The expression is: ； Selecting a suggested model acceptance rate for a generatable node The expression is: ； wherein: candidate model represented as a tower structure candidate dead nodes; Selecting a suggested model acceptance rate for evolvable nodes The expression is: ； wherein: is a tower structure candidate model is the number of candidate generation nodes.

9. A computer-readable storage medium, characterized in that, The computer readable storage medium stores computer execution instructions, and the computer execution instructions are executed by the processor to implement the resource electromagnetic exploration efficient probability inversion method in any one of claims 1-8.

10. An electronic device, comprising: It comprises: a processor and a memory connected in communication with the processor; the memory stores computer execution instructions; the processor executes the computer execution instructions stored in the memory to implement the resource electromagnetic exploration efficient probability inversion method in any one of claims 1-8.

Images