A borehole transient electromagnetic multi-parameter three-dimensional fast inversion method and device

Abstract

The application discloses a kind of ground well transient electromagnetic multi-parameter three-dimensional fast inversion method and device, and main technical scheme is: first, the original transient electromagnetic response signal containing magnetic field intensity and magnetic field rate of change is collected, after being converted into magnetic field intensity data, high-precision magnetic field cumulative response data is obtained by time integration and error correction;Subsequently, using the inversion model of pre-built containing Cole-Cole model magnetic dipole forward equation, layered background field module and adaptive three-dimensional grid, extract only the observation magnetic anomaly data containing target geological body, build composite objective function and adopt steepest descent method iterative solution;Finally, through geological constraint, depth weight, variable regularization, multidimensional optimization is carried out to initial three-dimensional Cole-Cole parameter model, and the relative root mean square error of magnetic field cumulative response is used as convergence determination index, and iteration is carried out until convergence, and the final three-dimensional Cole-Cole parameter model is output as the inversion result.

Description

Technical Field

[0001] This application relates to the field of geophysical exploration technology, and in particular to a method and apparatus for rapid three-dimensional inversion of transient electromagnetic multi-parameters in wells. Background Technology

[0002] Borehole Transient Electromagnetic (BHTEM) technology, with its high vertical resolution and strong sensitivity to deep conductive geological bodies, has been widely used in the field of deep mineral resource exploration. This technology acquires the transient electromagnetic response signal of the subsurface medium through a single field excitation and reconstructs the distribution characteristics of electrical parameters in the subsurface space through inversion calculations. Its applications cover multiple scenarios, including deep mineral resource exploration, geological structure delineation, oil and gas resource exploration, and geological hazard investigation.

[0003] Currently, although transient electromagnetic (TEM) technology is widely used in wellbore drilling, traditional 3D TEM inversion requires solving complex Maxwell's equations, which is essentially a highly nonlinear inversion problem with extremely high computational demands. On conventional computer architectures, completing a single 3D inversion often takes hours or even days, making it difficult to meet the real-time processing and rapid decision-making needs of exploration sites. Therefore, overcoming this technical bottleneck and providing a rapid 3D TEM multi-parameter inversion solution for wellbore drilling is a key technical problem urgently needing to be solved in this field. Summary of the Invention

[0004] This application provides a rapid three-dimensional inversion method and apparatus for transient electromagnetic multi-parameter in geothermal wells. The main purpose is to significantly reduce the computational complexity of inversion by transforming the highly nonlinear time-domain electromagnetic response into a linear magnetic field accumulation response. At the same time, by relying on dual physical quantity data fusion, multi-parameter constraint optimization, and accurate convergence determination, the stability and resolution of the inversion are effectively balanced, ensuring that the output three-dimensional parameter model conforms to geological laws and has high-precision imaging effect. Thus, it provides a rapid three-dimensional inversion solution for transient electromagnetic multi-parameter in geothermal wells that "takes into account both inversion efficiency and high quality".

[0005] To achieve the above objectives, this application mainly provides the following technical solutions:

[0006] The first aspect of this application provides a rapid three-dimensional inversion method for transient electromagnetic multi-parameter induction of wells, the method comprising:

[0007] The raw transient electromagnetic response signal of the underground geological body is collected, and the physical quantities of the raw transient electromagnetic response signal include at least: magnetic field strength and magnetic field change rate;

[0008] Under the premise of uniformly converting the data of different physical quantity dimensions in the original transient electromagnetic response signal into magnetic field strength data of a single physical quantity magnetic field strength dimension, the magnetic field cumulative response data corresponding to the original transient electromagnetic response signal is obtained.

[0009] The accumulated magnetic field response data is processed using a pre-built inversion model to output an initial three-dimensional Cole-Cole parameter model. The inversion model includes an adaptive three-dimensional grid, a layered background field module, and the magnetic dipole forward modeling equations of the Cole-Cole model. The adaptive three-dimensional grid, the layered background field module, and the magnetic dipole forward modeling equations of the Cole-Cole model work together to establish the mapping relationship between subsurface parameters and electromagnetic response, thus obtaining the theoretical basis for inversion. The layered background field module removes interference and extracts effective anomalies, and the adaptive three-dimensional grid serves as the spatial carrier for inversion, together completing the inversion processing of the accumulated magnetic field response data.

[0010] The initial three-dimensional Cole-Cole parameter model is optimized by applying geological constraints, depth weights, and variable regularization to output the optimized first target three-dimensional Cole-Cole parameter model. The optimization process includes at least: correcting outlier solutions, compensating for deep attenuation, and refining the boundaries.

[0011] The relative root mean square error between the inverted calculated magnetic field cumulative response value and the observed magnetic field cumulative response value is used as the inversion convergence criterion. The three-dimensional Cole-Cole parameter model of the first target is verified to obtain the three-dimensional Cole-Cole parameter model of the second target, which serves as the result of the three-dimensional rapid inversion of transient electromagnetic multi-parameters in the well.

[0012] A second aspect of this application provides a rapid three-dimensional inversion device for transient electromagnetic multi-parameter inversion of a well, the device comprising:

[0013] The data input unit is used to acquire the original transient electromagnetic response signal of the underground geological body. The physical quantities of the original transient electromagnetic response signal include at least: magnetic field strength and magnetic field change rate.

[0014] The magnetic field response accumulation determination unit is used to obtain the magnetic field accumulation response data corresponding to the original transient electromagnetic response signal, provided that the data of different physical quantity dimensions in the original transient electromagnetic response signal are uniformly transformed into magnetic field intensity data of a single physical quantity magnetic field intensity dimension.

[0015] The inversion processing unit is used to process the magnetic field cumulative response data using a pre-built inversion model, and output an initial three-dimensional Cole-Cole parameter model. The inversion model includes an adaptive three-dimensional grid, a layered background field module, and the magnetic dipole forward modeling equation of the Cole-Cole model. The adaptive three-dimensional grid, the layered background field module, and the magnetic dipole forward modeling equation of the Cole-Cole model work together to establish the mapping relationship between subsurface parameters and electromagnetic response using the magnetic dipole forward modeling equation, thus obtaining the theoretical basis for inversion. The layered background field module removes interference and extracts effective anomalies, and the adaptive three-dimensional grid serves as the spatial carrier for inversion, jointly completing the inversion processing of the magnetic field cumulative response data.

[0016] The constrained optimization unit is used to optimize the initial three-dimensional Cole-Cole parameter model through geological constraints, depth weights, and variable regularization, and output the optimized first target three-dimensional Cole-Cole parameter model. The optimization process includes at least: correcting outlier solutions, compensating for deep attenuation, and refining the boundary.

[0017] The result output unit is used to verify the first target three-dimensional Cole-Cole parameter model by using the relative root mean square error between the inverted calculated magnetic field cumulative response value and the observed magnetic field cumulative response value as the inversion convergence criterion, and to obtain the second target three-dimensional Cole-Cole parameter model, which serves as the result of the ground well transient electromagnetic multi-parameter three-dimensional fast inversion.

[0018] A third aspect of this application provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described method for rapid three-dimensional inversion of transient electromagnetic parameters in wells.

[0019] A fourth aspect of this application provides an electronic device, the device including at least one processor, and at least one memory and bus connected to the processor;

[0020] The processor and the memory communicate with each other via the bus.

[0021] The processor is used to call program instructions in the memory to execute the above-described method for rapid three-dimensional inversion of transient electromagnetic parameters in wells.

[0022] The fifth aspect of this application provides a computer program product that, when executed on a data processing device, is adapted to execute a program that initializes the steps of the above-described transient electromagnetic multi-parameter three-dimensional fast inversion method for wellbore.

[0023] By employing the above-described technical solution, the technical solution provided in this application has at least the following advantages:

[0024] This application provides a method and apparatus for rapid three-dimensional inversion of transient electromagnetic multi-parameters in wells. It constructs a closed-loop technical architecture that integrates offline pre-modeling (inversion model) with measured data processing: First, raw transient electromagnetic response signals containing magnetic field strength and rate of change are acquired and uniformly converted into magnetic field strength data. High-precision cumulative magnetic field response data is then obtained through time integration and error correction. Subsequently, using a pre-built offline inversion model containing the Cole-Cole model magnetic dipole forward equation, a layered background field module, and an adaptive three-dimensional mesh, observed magnetic anomaly data containing only the target geological body are extracted. A composite objective function is constructed and iteratively solved using the steepest descent method. Finally, the initial three-dimensional Cole-Cole parameter model is optimized in multiple dimensions through geological constraints, depth weighting, and variable regularization. The relative root mean square error (RMS) of the cumulative magnetic field response is used as the convergence criterion, and the iteration continues until convergence. The final three-dimensional Cole-Cole parameter model is then output as the inversion result.

[0025] Compared to the inefficiency caused by the strong nonlinearity and huge computational load of solving Maxwell's equations in traditional well transient electromagnetic inversion, this application significantly reduces the computational complexity of inversion by converting the strong nonlinear time-domain electromagnetic response into a linear magnetic field accumulation response. Combined with offline pre-modeling, it avoids the time-consuming operation of real-time modeling during the inversion process, further improving computational efficiency. At the same time, relying on dual physical quantity data fusion, multi-parameter constraint optimization, and accurate convergence judgment, it effectively balances the stability and resolution of the inversion, ensuring that the output three-dimensional parameter model not only conforms to geological laws but also has high-precision imaging effect. Thus, it provides a fast three-dimensional inversion solution for well transient electromagnetic multi-parameters that "balances inversion efficiency and high quality".

[0026] The above description is only an overview of the technical solution of this application. In order to better understand the technical means of this application and to implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of this application more obvious and understandable, the following are specific embodiments of this application. Attached Figure Description

[0027] Various other advantages and benefits will become apparent to those skilled in the art upon reading the following detailed description of preferred embodiments. The accompanying drawings are for illustrative purposes only and are not intended to limit the scope of this application. Furthermore, the same reference numerals denote the same parts throughout the drawings. In the drawings:

[0028] Figure 1 A flowchart of a rapid three-dimensional inversion method for transient electromagnetic multi-parameter in a wellbore, provided in this application embodiment;

[0029] Figure 2 A comparison curve of the calculated and corrected cumulative magnetic field response values ​​provided in the embodiments of this application;

[0030] Figure 3 A schematic diagram of adaptive mesh generation (vertical section) provided for an embodiment of this application.

[0031] Figure 4 The Cole-Cole parameter mapping function relationship curve provided in the embodiments of this application;

[0032] Figure 5 Vertical cross-sectional view of the three-dimensional Cole-Cole parameter model obtained by inversion provided in the embodiments of this application;

[0033] Figure 6 A horizontal slice of the target body (depth 800m) provided for an embodiment of this application;

[0034] Figure 7 This is a block diagram of a transient electromagnetic multi-parameter three-dimensional rapid inversion device for wells provided in an embodiment of this application. Detailed Implementation

[0035] Exemplary embodiments of the present application will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present application are shown in the drawings, it should be understood that the present application may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this application will be thorough and complete, and will fully convey the scope of the present application to those skilled in the art.

[0036] This application provides a method for rapid three-dimensional inversion of transient electromagnetic multi-parameters in wells, such as... Figure 1 As shown, the following specific steps (including steps 101-105) are provided in this embodiment of the invention:

[0037] Step 101: Collect the original transient electromagnetic response signal of the underground geological body. The physical quantities of the original transient electromagnetic response signal include at least: magnetic field strength and magnetic field change rate.

[0038] Based on the law of electromagnetic induction, a ground-based transmitting coil transmits a step pulse current into the ground. After the current is turned off, eddy currents are induced in the underground conductive-polarized geological bodies (ore bodies, surrounding rocks, etc.). The eddy currents generate a secondary magnetic field that decays over time, which is the transient electromagnetic response signal.

[0039] In the context of transient electromagnetic exploration in a ground well, the raw transient electromagnetic response signal acquired in this embodiment includes at least "magnetic field strength". "Physical quantity and rate of change of magnetic field" "Physical quantities; among them, 'magnetic field strength' directly reflects the steady-state electromagnetic response of underground geological bodies and is more sensitive to low-frequency signals of deep targets, while 'magnetic field change rate' directly reflects the transient electromagnetic response of underground geological bodies and is more sensitive to high-frequency signals of shallow targets. The embodiments of this application realize dual-channel acquisition of raw transient electromagnetic response signals. If one channel is interfered with by noise, it can be corrected by the other channel. By ensuring the consistency of the two physical quantities, outliers can be eliminated, the data signal-to-noise ratio can be improved, and the ambiguity of inversion can be reduced from the source."

[0040] Furthermore, the embodiments of this application preprocess the acquired raw transient electromagnetic response signal (filtering and denoising), and the specific explanation includes the following:

[0041] In this embodiment of the application, the original transient electromagnetic response signal collected is an original mixed signal. In addition to the effective secondary magnetic field signal that reflects the real physical properties of the underground geological body, it also contains a large amount of electromagnetic interference from the field environment, such as those from two types of interference sources: one is power frequency interference with extremely high energy and concentrated at 50Hz±1Hz (from power grids, transformers and other power infrastructure); the other is non-single frequency interference such as broadband harmonics, switching pulses and electromagnetic radiation generated by power facilities, accompanied by environmental Gaussian white noise, etc.

[0042] These electromagnetic interferences can severely overwhelm weak and effective signals at depth, distorting the true attenuation characteristics of the signals. If directly used for subsequent inversion, this can lead to false anomalies, a significant increase in multiple solutions, and even failure to converge, completely destroying the accuracy and efficiency of multi-parameter three-dimensional fast inversion. Therefore, this application preprocesses the original transient electromagnetic response signal (filtering and denoising), for example, by using a dual system of "adaptive threshold method + notch filter" to accurately remove 50Hz power frequency interference and various power facility interferences, purifying the effective observation data and providing a reliable input with a high signal-to-noise ratio for subsequent magnetic field cumulative response calculation and multi-parameter inversion, ensuring the accuracy and stability of the inversion results from the source. Specific explanations include the following:

[0043] Among them, the adaptive threshold method (threshold formula is) The formula is applied as follows: In electromagnetic signals collected in the field, random noise (such as environmental electromagnetic noise) usually follows a Gaussian white noise distribution. The upper limit of its amplitude is determined by the number of sampling points N. The larger the number of sampling points N (the longer the corresponding signal duration), the higher the theoretical maximum amplitude of the noise. Therefore, the threshold θ is automatically adjusted upward as N increases. Conversely, the smaller N is, the threshold is automatically adjusted downward, thereby avoiding over-filtering that leads to distortion of the effective transient electromagnetic signal.

[0044] Specifically, the denoising implementation process implemented using the above adaptive threshold formula in this application embodiment includes the following:

[0045] The collected magnetic field strength and the rate of change of magnetic field Perform a wavelet transform to decompose the signal into wavelet coefficients of different frequencies; use the formula... An adaptive threshold is calculated, and wavelet coefficients with amplitudes less than the threshold are identified as noise and set to zero, while those greater than the threshold are considered valid signals and retained. Finally, an inverse wavelet transform is performed to obtain the denoised clean signal. Compared to traditional fixed thresholds, which cannot adapt to signals with different sampling lengths, this embodiment of the application uses such an adaptive threshold formula, which can automatically adjust the threshold according to the number of sampling points of the signal itself. This ensures thorough denoising while preserving the details of the valid transient signal to the greatest extent, making it a better adaptive solution for transient electromagnetic data preprocessing.

[0046] The adaptive threshold method, as the basic and general approach used in the embodiments of this application, can cover the vast majority of interferences, including: broadband harmonics, random pulses, electromagnetic radiation and other non-single-frequency interferences generated by power facilities; Gaussian white noise and random electromagnetic noise in the environment; and part of the energy of 50Hz power frequency interference (but cannot be completely and thoroughly eliminated).

[0047] 50Hz power frequency interference is the strongest and most concentrated narrowband single-frequency interference in field exploration. It cannot be completely filtered out by adaptive thresholding alone, leaving residual energy that affects subsequent inversion. Therefore, this study adopts a dual synergistic scheme of "adaptive thresholding + notch filter (2Hz stopband width)": using a notch filter (2Hz stopband) to specifically and accurately attenuate the narrowband signal from 49Hz to 51Hz, thoroughly eliminating the strong energy of the 50Hz power frequency and its adjacent harmonics. After notch filtering, the adaptive thresholding method is further used to clean up residual power frequency harmonics, broadband interference from power facilities, and environmental random noise, ultimately outputting clean and effective observation data.

[0048] Step 102: Under the premise of uniformly converting the data of different physical quantity dimensions in the original transient electromagnetic response signal into magnetic field strength data of a single physical quantity magnetic field strength dimension, obtain the magnetic field cumulative response data corresponding to the original transient electromagnetic response signal.

[0049] The data of different physical quantities (magnetic field strength, magnetic field change rate) in the original transient electromagnetic response signal are uniformly transformed into data of a single magnetic field strength dimension. Based on this, the cumulative magnetic field response data corresponding to the original signal is obtained through time integration and truncation error correction. Using this step 102, the differences in the dimensions and scales of different physical quantities can be eliminated, and the data format can be normalized, providing a unified and stable input basis for subsequent inversion. In addition, the traditional strongly nonlinear time-domain electromagnetic response is transformed into a linear cumulative magnetic field response, which greatly reduces the computational complexity of the inversion and fundamentally solves the core pain point of "low efficiency" in traditional inversion.

[0050] Specifically, step 102 can be further refined to include the following implementation steps: A1-A3;

[0051] A1. The original magnetic field rate of change component contained in the original transient electromagnetic response signal is converted into a standard magnetic field strength component through integration. This standard magnetic field strength component is then aligned with the original magnetic field strength component inherent in the original transient electromagnetic response signal to form a consistent magnetic field strength dataset. Based on the consistent magnetic field strength dataset, the total magnetic field strength data corresponding to the original transient electromagnetic response signal is generated.

[0052] The physical quantities of the raw transient electromagnetic response signal acquired in step 101 include at least: magnetic field strength and magnetic field change rate. For the physical quantity "magnetic field change rate", since it is represented by the induced electromotive force value, it is necessary to first integrate the induced electromotive force value within the acquisition window to obtain the true magnetic field strength through definite integration. Specifically, it can be achieved using the following formula (1): Formula (1); where, t 1i For the first i The start time of the response observation window. t 2i For the first i The response time is set to the end of the observation window, thus integrating the rate of change of the magnetic field within that specific window. This is equivalent to calculating the total change (difference) in the magnetic field strength during that time period. This is done to unify the dimensionality of the input data, ensuring that subsequent calculations only process uniform physical quantities. .

[0053] A2. Perform time integration on the total magnetic field strength data within the preset observation time window to obtain the first magnetic field cumulative response data.

[0054] It should be noted that, in order to distinguish between the "magnetic field cumulative response data" before and after the correction process, the embodiments of this application use "first" and "second" as identifiers: the data before the correction process is "first magnetic field cumulative response data", and the data after the correction process is "second magnetic field cumulative response data".

[0055] In this embodiment of the application, the cumulative response value of the magnetic field is calculated using the trapezoidal integral method. Magnetic field cumulative response value The definition of is (expressed using the following formula (2)). Formula (2) is essentially the calculation of the total area of ​​the magnetic field curve on the time axis: Formula (2); t 1 To ensure an effective observation start time (typically 0.52 μs for the Canadian PEM transient electromagnetic system), the goal is to eliminate strong transient interference caused by the transmitter shutting off the current in the early stages and select only the stable and effective portion of the subsequent signal as the integration start point. t N The observation end time (typically 10ms for the Canadian PEM transient electromagnetic system) is used to capture the attenuated tail of the effective signal and the integration endpoint. Used to represent the cumulative response of a magnetic field, it no longer focuses on a single moment. t It is not the instantaneous value, but a combination of factors from... t 1 arrive t N The total induced energy generated by the underground geological body throughout the entire time period.

[0056] Compared to traditional inversion methods that directly process... Due to the nonlinear decay characteristics of the magnetic field, the inversion equation is highly nonlinear and converges very slowly. Therefore, this application's embodiments introduce... Then, the inversion problem is transformed into "inversion area", which mathematically realizes the transformation from nonlinear to linear, greatly reducing the computational difficulty and convergence pressure of subsequent inversion. This is the fundamental premise for achieving "fast inversion".

[0057] A3. To address the truncation error caused by the time integration operation, the first magnetic field cumulative response data is corrected to obtain the second magnetic field cumulative response data, which serves as the magnetic field cumulative response data corresponding to the original transient electromagnetic response signal.

[0058] direct integration There is an inherent truncation error, for example: (with) t N (Taking 10ms as an example) Because the sampling only lasts for 10ms before stopping, the magnetic field still has a slight decay energy that is not collected after 10ms. Direct integration will lead to... The calculated value is too small.

[0059] Therefore, to compensate for the integral truncation error, this application embodiment indirectly corrects it through the rate of change of the magnetic field, satisfying the expression in formula (3), and achieving a corrected relative error ≤3%, as shown in the example. Figure 2As shown, this application provides a comparison curve between the calculated value and the corrected value of the magnetic field cumulative response.

[0060] Formula (3); using the mathematical transformation technique of integration by parts, the equation is transformed into... The integral is converted into a pair The integral here utilizes the rate of change of the magnetic field ( Characteristics of ) Although It decays rapidly in the later stages, but and its weighted terms It can often better approximate the integral effect at the tail of the signal.

[0061] This transformation indirectly compensates for the energy loss during the uncollected period, and the relative error of the cumulative response value of the corrected magnetic field is ≤3%, which is an extremely high level of accuracy. This proves the effectiveness of the correction method and ensures the accuracy of the data input into the subsequent inversion process.

[0062] Step 103: Process the magnetic field cumulative response data using the pre-built inversion model to output an initial three-dimensional Cole-Cole parameter model. The inversion model includes an adaptive three-dimensional grid, a layered background field module, and the magnetic dipole forward modeling equations of the Cole-Cole model. The adaptive three-dimensional grid, the layered background field module, and the magnetic dipole forward modeling equations of the Cole-Cole model work together to establish the mapping relationship between subsurface parameters and electromagnetic response using the magnetic dipole forward modeling equations, thus obtaining the theoretical basis for the inversion. The layered background field module removes interference and extracts effective anomalies, and the adaptive three-dimensional grid serves as the spatial carrier for the inversion, together completing the inversion processing of the magnetic field cumulative response data.

[0063] This application embodiment constructs an inversion model that balances accuracy and efficiency and adapts to multiple parameters (comprising three parts: an adaptive 3D mesh, a layered background field module, and the magnetic dipole forward modeling equations of the Cole-Cole model). These three parts work together, using the adaptive 3D mesh as the spatial basis, the layered background field module as the data support, and the magnetic dipole forward modeling equations as the inversion core. They are executed in the order of "mesh partitioning → background field stripping → forward and inversion" to jointly complete the processing of the accumulated magnetic field response data and output the initial 3D Cole-Cole parameter model.

[0064] Among them, the adaptive 3D mesh serves as the spatial computing carrier for inversion. First, it performs differentiated numerical discretization of the underground space to provide a 3D spatial framework for subsequent background field calculation and forward modeling. At the same time, it adapts to geological structures at different depths by refining the near-surface mesh and coarsening the far-field / deep mesh to balance the computational efficiency and imaging accuracy of the inversion.

[0065] The layered background field module calculates the magnetic field cumulative response data corresponding to the layered background field based on the pre-divided adaptive 3D grid. Then, it removes the background electromagnetic interference of non-target areas from the total magnetic field cumulative response data, extracts the observation magnetic anomaly data containing only the target geological body, eliminates the interference of non-target bodies on the inversion results, and provides clean, focused and effective input data for subsequent inversion.

[0066] The Cole-Cole model's magnetic dipole forward modeling equations use an adaptive three-dimensional grid as the spatial carrier and observed magnetic anomaly data as input. Based on the Cole-Cole model, it describes the conductive and polarization coupled electromagnetic response of underground geological bodies and establishes a quantitative mapping relationship between underground Cole-Cole parameters and observed magnetic anomaly data. By constructing a composite objective function and solving iteratively, the initial three-dimensional Cole-Cole parameter model is finally output, completing the core calculation of the inversion.

[0067] The inversion model constructed in this application will be explained in detail in the following three parts:

[0068] (I) Mesh Generation: Adaptive 3D Mesh Design

[0069] The computational cost of 3D inversion is positively correlated with the number of grids. Therefore, it is necessary to use adaptive meshing, with fine grids in the target area (for high-resolution requirements) and coarse grids in the far area (for low-resolution requirements), to minimize the computational cost while ensuring inversion accuracy and achieve "fast inversion".

[0070] This application's embodiments employ adaptive rectangular grids to partition underground space, adhering to the principle of "densification in the target area and sparser meshes in the distant area," balancing computational efficiency and resolution. The multi-dimensional grid design includes both vertical and horizontal grids. Specific explanations are as follows:

[0071] Vertical grid, dimensions Δz represents the vertical sampling interval of the sensor (2~5m), ensuring that the vertical resolution matches the sampling interval. The vertical grid size is taken as 1 / 2 of the sensor sampling interval, satisfying the Nyquist sampling theorem, ensuring that the vertical resolution perfectly matches the sampling accuracy of the observation data, and avoiding the loss of vertical information due to an overly coarse grid, or the redundant computation due to an overly fine grid.

[0072] Example: If the vertical sampling interval of the sensor is 2m, then the vertical grid size is 1m; if it is 5m, then the grid size is 2.5m.

[0073] The horizontal grid divides the underground space into three zones: Zone 1 (target area, 0~500m): Dimension h x =h y=5m; Region 2 (transition zone, 500~1500m): Dimension h x =h y =10m; Region 3 (far zone, >1500m): Dimension h x =h y =20m. For example, such as Figure 3 The diagram shown is a schematic diagram (vertical section) of adaptive mesh generation provided in this application embodiment.

[0074] Based on the distance from the target area, three levels of adaptive meshing are used: Target area (0~500m): 5m fine mesh to ensure high-resolution imaging of the target geological body; Transition area (500~1500m): 10m medium mesh to balance accuracy and computational load; Far area (>1500m): 20m coarse mesh. The far area has minimal impact on the response of the target area, and the coarse mesh greatly reduces the computational load.

[0075] In the "adaptive rectangular mesh partitioning of underground space", this application embodiment provides boundary conditions and infinity simulation, such as: the top of the mesh extends 100m above the ground surface, the bottom extends 500m below the deepest burial depth of the target body, the horizontal boundary is ≥2000m from the edge of the target area, and the boundary cells use a decay factor. (r is the distance from the boundary cell to the center of the target area), simulating an infinitely far boundary.

[0076] Numerical simulations require simulating infinity boundaries (to avoid boundary reflections interfering with forward modeling results). This application's embodiments achieve this through two methods: physical extension and mathematical attenuation. Physical extension involves significantly extending the mesh boundary (top +100m, bottom +500m, horizontal ≥2000m) to keep the boundary away from the target area and reduce its influence. Mathematical attenuation involves introducing an attenuation factor to the boundary cells. As the distance r increases, the response of the boundary element decays exponentially, perfectly simulating the electromagnetic characteristics of an infinitely far boundary and completely eliminating boundary reflection interference.

[0077] (II) Background Field Model: Layered Background Field Module

[0078] The essence of transient electromagnetic inversion is anomaly field inversion, and its core physical relationship is: total response = background field (response of layered homogeneous medium) + anomaly field (response of target geological body). Based on this, the embodiments of this application use a layered medium model as the background field. Specifically, the following formula (4) is used to achieve accurate calculation of the reference field:

[0079] Formula (4); where, Let be the time-domain background field response of the k-th layer medium; For the Green's function in the frequency domain of layered media containing the Cole-Cole parameter (which describes the electromagnetic response characteristics of layered media); Here is the frequency domain expression for the emission current. I 0 The peak value of the emission current; σk, ηk, τk, ck are the Cole-Cole model parameters of the k-th dielectric layer; ω is the angular frequency. It is the inverse Fourier transform kernel.

[0080] Based on the above formula (4), the processing flow implemented in this application includes: (1) Frequency domain forward modeling: In the frequency domain, the electromagnetic response of the layered medium is calculated by the Green function containing the Cole-Cole parameter. This method is compatible with the induced polarization effect of the geological body and realizes the accurate characterization of the conduction-polarization coupling characteristics; (2) Time domain transformation: The frequency domain response is converted into the time domain response by the inverse Fourier transform to obtain the time domain background field that is completely matched with the time dimension of the field measurement data; (3) Anomaly field extraction: The calculated background field is subtracted from the total observed field to extract the anomaly field containing only the target geological body, which is used as the input data for subsequent inversion. The response interference of non-target areas is avoided from the source, providing a stable benchmark for the "nonlinear to linear" inversion, while taking into account both inversion efficiency and accuracy.

[0081] (III) Forward modeling governing equations: Forward modeling equations of the magnetic dipole in the Cole-Cole model

[0082] When the observation time t is much greater than the electromagnetic diffusion time constant τ of the geological body m At this time, the transient electromagnetic response of the underground geological body satisfies the late-stage approximation. The complex full-space electromagnetic response can then be simplified to a magnetic dipole field, significantly reducing the computational complexity of forward modeling. This is a key physical prerequisite for achieving the "three-dimensional rapid inversion" in this application's embodiment. Traditional transient electromagnetic forward modeling only considers a single conductivity parameter. This application's embodiment introduces the Cole-Cole rock physics model, incorporating the induced polarization effect (polarization characteristics) of the geological body into the forward modeling system. This achieves simultaneous forward modeling of four multiple parameters: conductivity, polarizability, time constant, and dispersion coefficient, solving the technical pain point that traditional methods cannot invert induced polarization parameters. Based on the aforementioned late-stage time domain assumption ( , (where is the electromagnetic diffusion time constant of the geological body). In conjunction with the Cole-Cole rock physics model, the embodiments of this application equate the transient electromagnetic response of dense conductive-polarized geological bodies to a magnetic dipole with induced polarization, thereby constructing forward modeling control equations.

[0083] Specifically, this application provides a magnetic field response formula for a magnetic dipole, as shown in formula (5): Formula (5); where, V is the vacuum permeability, a physical constant. This is the unit magnetization vector corrected by the Cole-Cole model; The strength of the main magnetic field at the center of the unit; Let be the position vector between the observation point and the cell center. The distance between two points; This is a Cole-Cole parameter mapping function.

[0084] Furthermore, the Cole-Cole parameter mapping function can be represented by the following formula (6): Formula (6); where σ is conductivity (reflecting the conductivity of the geological body, a core electrical parameter), η is polarizability (reflecting the strength of the induced polarization effect of the geological body, a core induced polarization parameter), τ is time constant (reflecting the time decay characteristics of the induced polarization effect), c is dispersion coefficient (ranging from 0.1 to 0.9, reflecting the frequency dependence of the induced polarization effect), and t is the observation time.

[0085] Formula (5) is the classical expression for the magnetic field of a magnetic dipole, obtained through... By embedding Cole-Cole multi-parameter forward modeling, accurate forward modeling of the conduction-polarization coupling effect is achieved: when η=0, the formula degenerates into the magnetic dipole field of a traditional pure conductive medium, which is compatible with the traditional forward modeling model.

[0086] When η>0, the induced polarization effect is automatically introduced, which can forward model the transient electromagnetic response of polarized geological bodies (such as metal mines and aquifers), providing a physical basis for multi-parameter inversion.

[0087] Based on the explanations in (I), (II), and (III) above, this embodiment establishes a forward modeling model of a magnetic dipole with multiple Cole-Cole parameters through forward modeling control equations, achieving accurate forward modeling of the conduction-polarization coupling effect and providing physical support for multi-parameter inversion. The layered background field model provides a stable benchmark for the "nonlinear to linear" inversion, avoiding interference from non-target regions and improving inversion stability. Adaptive mesh partitioning balances inversion accuracy and computational efficiency, providing a key numerical guarantee for achieving "fast three-dimensional inversion."

[0088] Furthermore, this inversion model can be pre-built offline. Below, this embodiment utilizes such an inversion model to process the accumulated magnetic field response data: electromagnetic interference in non-target areas is removed by stripping away the layered background field, and the observed magnetic anomalies containing only the target geological body are extracted; a composite objective function is constructed based on the forward modeling equation, and the steepest descent method is used for iterative solution, ultimately outputting an initial three-dimensional Cole-Cole parameter model, thus completing the inversion solution from observation data to the three-dimensional distribution of multiple subsurface parameters (σ, η, τ, c). Specifically, step 103 can be further refined to include the following implementation steps: B1-B7;

[0089] B1. Call the adaptive three-dimensional mesh in the pre-built inversion model to numerically discretize the underground space and obtain each element of the adaptive three-dimensional mesh, providing a three-dimensional spatial computing carrier for subsequent background field calculation and forward modeling.

[0090] B2. Based on each element of the adaptive 3D mesh, call the layered background field module in the inversion model to calculate the cumulative response data of the background magnetic field corresponding to the layered background field.

[0091] B3. By subtracting the background magnetic field cumulative response data from the magnetic field cumulative response data corresponding to the original transient electromagnetic response signal, we obtain the observed magnetic anomaly data that only reflects the electromagnetic characteristics of the target geological body.

[0092] B4. Based on the magnetic dipole forward modeling equation of the Cole-Cole model in the inversion model, a mathematical mapping relationship between the electrical conductivity and polarization coupling electromagnetic response of underground geological bodies and the observed magnetic anomaly data is established in advance, providing a theoretical basis for subsequent inversion solutions;

[0093] like Figure 4 As shown in the embodiment of this application, a quantitative relationship curve between the Cole-Cole parameter and the magnetic field cumulative response Q is provided. The influence of conductivity σ, polarizability η, time constant τ, and frequency correlation coefficient c on the magnetic field cumulative response is verified by the controlled variable method, providing theoretical support for the calculation of the magnetic field cumulative response in step 102 and the inversion model involved in step 103.

[0094] B5. Establish spatial correlation between the magnetic dipole forward modeling equation of the Cole-Cole model, the observed magnetic anomaly data reflecting the electromagnetic characteristics of the target geological body, and each cell of the adaptive three-dimensional grid, and determine the solution domain of the Cole-Cole parameters corresponding to each grid cell.

[0095] The following examples will be used to explain B1-B5 above:

[0096] The total magnetic field cumulative response value obtained in step 102 Subtracting the cumulative magnetic field response corresponding to the layered background field The resulting difference is then used as the "observed magnetic anomaly" in the magnetic inversion, completely stripping away the electromagnetic response of non-target areas (surrounding rock, layered media) and retaining only the effective anomalies of the target body (ore body, anomalous body), thus eliminating non-target interference at the source and improving the relevance of the inversion. The definition of is expressed by the following formula (7): Formula (7);

[0097] By combining the forward control equations containing the Cole-Cole model, the problem of magnetic field cumulative response inversion is transformed into a linear magnetic inversion problem of solving the Cole-Cole parameters (σ, η, τ, c) of each grid element.

[0098] Compared to traditional transient electromagnetic inversion: directly dealing with the time-domain magnetic field Inversion is problematic because the electromagnetic response exhibits a highly nonlinear relationship with underground parameters, resulting in slow convergence, multiple solutions, and extremely high computational costs. This application's embodiment uses the accumulated magnetic field response from step 102. Replacement instantaneous By incorporating the integral effect of the time dimension into the inversion, a mathematical transformation from nonlinearity to linearity is achieved. Furthermore, by combining the Cole-Cole magnetic dipole forward model in the inversion model, the inversion problem is transformed into: given the observed magnetic anomaly The linear equations of the four Cole-Cole parameters (σ, η, τ, c) for each grid cell are obtained by inverse calculation. After linearization, the computational cost of inversion decreases by an order of magnitude, which provides a basis for realizing "fast three-dimensional inversion". At the same time, it greatly reduces the ambiguity and improves the stability of inversion.

[0099] B6. Based on the observed magnetic anomaly data reflecting the electromagnetic characteristics of the target geological body, the mathematical mapping relationship, and the solution domain of the Cole-Cole parameters corresponding to each grid cell, a composite objective function is constructed, which includes a data fitting term, a regularization constraint term, and a parameter correlation term. Among them, the data fitting term is used to characterize the fitting error between the observed magnetic anomaly data and the forward modeling response, the regularization constraint term is used to ensure the spatial continuity of the inversion model, and the parameter correlation term, based on the physical essence of the Cole-Cole model, sets inequality constraints on the value range of each parameter.

[0100] This step constructs a three-in-one composite objective function consisting of "data fitting term + regularization constraint term + parameter correlation term" to provide optimization criteria for multi-parameter magnetic inversion. The expression is as follows: (8): Formula (8); In Formula (8), σ, η, τ, and c are the conductivity, polarizability, time constant, and dispersion coefficient to be inverted, respectively; To observe the cumulative magnetic field response vector (i.e., the total cumulative magnetic field response calculated in step 102) ); The cumulative magnetic field response vector corresponding to the layered background field (obtained by integrating the layered background field Bb(t) in the inversion model); The cumulative magnetic field response vector calculated for forward modeling of the target body is obtained through the Cole-Cole parameter mapping function. Please provide a solution.

[0101] In equation (8) This is the data weight matrix (diagonal matrix), and its diagonal elements... in σi For the first i The observation error of each observation point is determined by the standard deviation of multiple repeated observations and is used to adaptively weight the fitting contribution of different observation points.

[0102] Furthermore, the data fitting term in equation (8) is as follows: The data weighting matrix is ​​used to measure the degree of fit between observed magnetic anomalies and forward-modeled magnetic anomalies. It is the core constraint of the inversion and aims to make the calculation results fit the measured data as closely as possible. Then, through the data weighting matrix, higher weights are given to observation points with small errors and lower weights are given to points with large errors, which automatically suppresses the influence of noise on the inversion and improves the fitting accuracy.

[0103] In equation (8) For L1 regularization terms: The scale weight matrix (diagonal matrix) has the following diagonal elements. , Vj For the first j The volume of each grid cell; this regularization term can achieve sparsity constraints, favoring targets with low volume and high physical property differences, and strengthening the ability to identify anomalous body boundaries. That is, by weighting the inverse of the cell volume, small volume cells (target area densified grid) are given higher weights, which strengthens the ability to identify small-scale targets and meets the need for "focusing on anomalous bodies" in geological exploration.

[0104] In equation (8) The L2 smoothing constraint term suppresses excessive model oscillations and ensures the spatial continuity of the inversion results. Specifically, ▽ represents the gradient operator. The spatial weight matrix is ​​used (the target area cell weight is 1.0, and the far area cell weight is 0.5). The target area uses a high weight (1.0) to ensure details, while the far area uses a low weight (0.5) to reduce the computational load. This ensures the resolution of the anomaly while stabilizing the inversion iteration process.

[0105] λ1 and λ2 are regularization parameters, adaptively determined using the L-curve method: the abscissa of the L-curve is... The vertical axis is The parameter value corresponding to the inflection point of the curve is the optimal regularization parameter, achieving a balance between fitting accuracy and model smoothness. The L-curve method is a classic method in the field of inversion. By plotting the curve of "fitting error - regularization term", it finds the optimal balance point between fitting accuracy and smoothness, avoiding overfitting (good fit but model oscillation) or underfitting (smooth model but poor fit), and achieving adaptive parameter selection without the need for manual parameter tuning.

[0106] The parameter correlation terms are based on the physical essence of the Cole-Cole model. Inequality constraints are constructed by binding reasonable value ranges for all core parameters to ensure the physical rationality of the inversion parameters. Based on the engineering measurement range of the Cole-Cole model, the parameter value constraints are set as follows:

[0107] Conductivity σ∈

[10] -3 [102S / m], polarizability η∈[0,0.6];

[0108] The dispersion coefficient c∈[0.1, 0.9] and the time constant τ∈[10-3, 10s].

[0109] Therefore, the range of each parameter is derived from the industry's measured range in geophysical exploration, ensuring the physical rationality of the inversion results. During the iterative solution process, if a parameter exceeds its corresponding value range, it is directly projected onto the range boundary as the input for the next iteration. This constraint not only avoids unrealistic and unreasonable parameter solutions but also stabilizes the inversion iteration process, significantly improving the geological reliability of the inversion results.

[0110] B7. The steepest descent method is used to iteratively solve the composite objective function. The iteration termination condition is that the relative change of the composite objective function between two iterations is less than a preset threshold or a preset number of iterations. After the iteration terminates, the obtained Cole-Cole parameters are assigned to each element of the adaptive three-dimensional mesh to generate an initial three-dimensional Cole-Cole parameter model, which is used as the processing result of the inversion model.

[0111] (1) The steepest descent method is used to solve for the minimum value of the multi-parameter objective function. The steepest descent method is a classic algorithm for unconstrained optimization. It iteratively updates the parameters along the negative gradient direction (▽F) of the objective function F, minimizing the objective function value at each step, and quickly converges to the global optimum. Specifically, the iterative update formula is as follows (9): Formula (9);

[0112] The gradient ▽F of the objective function consists of three parts: the data fitting term, the L1 regularization term, and the L2 smoothing constraint term. The expression is as follows: gradient calculation formula (10): Formula (10); where, is the Jacobian matrix, describing the effect of the changes in the parameters to be inverted (σ,η,τ,c) on the cumulative response of the magnetic field; Δ is the Laplace operator, used to calculate the spatial second derivative of the parameters to achieve smooth constraint of the model; sign() is the sign function, used to calculate the gradient of the L1 regularization term; the definitions of the remaining parameters are consistent with the multi-parameter objective function in step 103.

[0113] Furthermore, as in equation (10) The gradient of the data fitting term drives the parameters to be updated in the direction that minimizes the fitting error; The gradient of the L1 regularization term is used to implement sparse constraints and strengthen the anomalous body; The gradient of the L2 smoothing term is used to achieve spatial smoothing and suppress model oscillations.

[0114] (2) Fixed step size is prone to iterative oscillation (too large step size) or slow convergence (too small step size). The Armijo criterion achieves adaptive step size selection.

[0115] The step size β of the k-th iteration k The Armijo criterion is used to adaptively select the objective function value to ensure that the objective function value satisfies the inequality constraint (as shown in formula (11)). If the step size is too large and the function value increases, the step size is automatically reduced. If the step size is too small and the convergence is slow, the step size is automatically increased. This ensures that the objective function value decreases strictly in each iteration, avoids oscillation, and greatly improves the convergence speed and stability.

[0116] For example: βk is the step size of the k-th iteration, which is adaptively selected using the Armijo criterion, and ensures that the objective function satisfies the inequality constraints (as in formula (11)):

[0117] Formula (11); where γ=10 -4 The step size is the control parameter for the Armijo criterion. If the step size is too large and causes the function value to rise, the step size is automatically reduced; if the step size is too small and the convergence is slow, the step size is automatically increased. This ensures that the objective function value is strictly reduced in each iteration, avoids oscillations, and greatly improves the convergence speed and stability.

[0118] Furthermore, the iteration termination condition provided in this application embodiment is as follows: a dual termination criterion is adopted to balance inversion accuracy and computational efficiency, and iteration stops when any of the following conditions are met:

[0119] Relative error threshold: (That is, the relative change in the objective function value between the two iterations is less than) (assuming inversion convergence); and maximum number of iterations: the number of iterations is ≥50, to avoid infinite iterations due to local optima and ensure computational efficiency.

[0120] Combining B1-B4 above, step 103 transforms the traditional highly nonlinear transient electromagnetic inversion into a linear multi-parameter inversion through mapping logic. This is the core prerequisite for achieving "fast 3D inversion." A composite objective function of "data fitting + regularization + parameter constraints" is constructed, which balances fitting accuracy, model stability, and physical rationality, thus solving the problem of multiple solutions in multi-parameter inversion. The steepest descent method combined with Armijo adaptive step size is used to achieve rapid convergence, completing the 3D multi-parameter inversion within 50 iterations, meeting the needs of engineering applications.

[0121] Step 104: Optimize the initial 3D Cole-Cole parameter model by applying geological constraints, depth weights, and variable regularization, and output the optimized first target 3D Cole-Cole parameter model. The optimization process includes at least: correcting outlier solutions, compensating for deep attenuation, and refining the boundary.

[0122] The initial 3D Cole-Cole parameter model is optimized in multiple dimensions using three optimization methods: geological constraints, depth weights, and variable regularization. Geological constraints correct outlier solutions based on prior regional geological information and eliminate parameters that do not conform to geological laws. Depth weights are introduced to compensate for the attenuation of deep electromagnetic response and balance the imaging accuracy of shallow and deep parts. Variable regularization optimizes the model through a two-stage strategy (strong L1 range locking and strong L2 boundary refinement). The final output is the first target 3D Cole-Cole parameter model.

[0123] Therefore, step 104 is a step to improve the accuracy and reliability of the inversion results obtained in step 103. By utilizing three types of constraints—geological priors, depth compensation, and adaptive regularization—it addresses issues such as inversion ambiguity, deep signal attenuation, and boundary ambiguity, ultimately outputting a high-precision 3D model that conforms to geological laws. Specifically, step 104 is further refined into the following implementation steps C1-C3;

[0124] C1. Based on regional geological prior information and lithological property test data, set corresponding lithological value boundaries for the Cole-Cole parameters of each grid cell in the initial three-dimensional Cole-Cole parameter model; project parameters exceeding the value boundaries to the legal interval using the projection operator, correct the anomalous solutions generated during the inversion process, and obtain the three-dimensional Cole-Cole parameter model corrected by geological constraints.

[0125] Based on known regional geological data and physical property test results, different lithological domains are defined, and exclusive upper and lower boundaries of the Cole-Cole parameters are set for different lithological units. The projection operator corrects parameters that exceed the range during the iteration process to a reasonable range, constraining the physical rationality and geological consistency of the inversion solution from a geological prior level, and completely avoiding non-geological anomalies. The parameter constraint ranges for each lithological unit are shown in Table 1 below:

[0126] Table 1

[0127]

[0128] The projection operator corrects the parameters that exceed the constraint interval during the iteration process to a reasonable range. The expression of the projection operator is as shown in formula (12): Formula (12); where, x Let (σ, η, τ, c) be the parameters to be corrected during the iteration process. x min , x max These represent the lower and upper limits of the parameters for the corresponding lithological unit. This is a truncated projection operator that automatically projects the hyperboundary parameters to the boundary of the constraint interval. Its execution logic is as follows: if the parameters... x lie in[ x min , x max If the value is within the specified range, then the original value remains unchanged; if x < x min Then the parameters will be corrected to x min ;like x > x max Then the parameters will be corrected to x max By using such a projection operator, unreasonable parameter solutions that are not geologically relevant are avoided, and the inversion iteration process is stabilized.

[0129] C2. Introduce a depth weight function to apply differentiated weights to grid cells at different depths in the geologically constrained 3D Cole-Cole parameter model, compensating for the attenuation of deep electromagnetic response, and obtain a depth-weighted 3D Cole-Cole parameter model.

[0130] The transient electromagnetic response signal exhibits a geometric attenuation effect: the deeper the detection depth, the weaker the electromagnetic response signal, and the resolution of the inversion for deep targets is significantly reduced. This application introduces a depth weighting function to apply differentiated weights to grid cells at different depths, enhancing the inversion contribution of deep cells, offsetting the effect of signal attenuation, and achieving balanced imaging of targets at both shallow and deep depths. Specifically, the depth weighting function is expressed as in formula (13): Formula (13); where, z 0 The initial depth of the target body (determined by geological survey, i.e., the burial depth of the top of the target body); H is the attenuation coefficient (valued between 500 and 800 m, used to control the attenuation rate of the weight with depth). z These are the depth coordinates of the grid cells.

[0131] Based on this function, differentiated weights can be set: near-surface unit ( z ≤ 100m), setting w z =1.0, maintaining the original weights to avoid near-surface inversion distortion; deep elements (z>1000 m), set w z =0.1~0.3, and apply higher weights to deep units through an exponential function (in formula (13)). z The larger the value, the smaller the absolute value of the exponent. w z The larger the value, the more effective the deep weight compensation becomes.

[0132] The attenuation coefficient H, ranging from 500 to 800 m, is an empirical value used to balance the weights of shallow and deep regions, avoiding overcompensation that could lead to false anomalies in deep areas. This method numerically compensates for the geometric attenuation of electromagnetic signals, solving the problem of "clear images in shallow areas but blurry images in deep areas" in traditional inversion methods, and effectively improving the imaging accuracy of deep targets (such as deep ore bodies).

[0133] C3. A preset two-stage variable regularization strategy is adopted to perform boundary optimization on the three-dimensional Cole-Cole parameter model after depth weight compensation to obtain the first target three-dimensional Cole-Cole parameter model. The preset two-stage variable regularization strategy includes: the first stage strengthens L1 regularization to lock the target volume range, and the second stage weakens L1 regularization and strengthens L2 regularization to constrain the boundary optimization.

[0134] For different stages of the inversion, the weight ratio of L1 / L2 regularization (λ1 / λ2) is dynamically adjusted to achieve a two-stage optimization of "focusing first, then refining": the first stage strengthens L1 regularization to quickly screen targets with high material property differences; the second stage weakens L1 regularization to finely adjust the target boundary, balancing inversion efficiency and imaging accuracy. Specific explanations are as follows:

[0135] Phase 1: Screening of Targets with High Physical Property Differences: L1 regularization is enhanced by setting λ1 / λ2 = 10. Utilizing the sparsity constraint of L1 regularization, priority is given to units with significant physical property differences (especially electrical conductivity and polarizability parameters). Units with parameters exceeding three times the mean of the surrounding rock are retained, while interference signals with low physical property differences are eliminated, initially identifying high-probability target areas.

[0136] L1 regularization has sparse constraint characteristics. After strengthening the L1 weight, the inversion will prioritize focusing on units with significant differences in physical properties (such as ore bodies with high conductivity and high polarization), automatically eliminating low-difference interference from surrounding rocks, and quickly locking in high-probability target body regions, significantly improving inversion efficiency. It quickly filters out candidate regions of target bodies from massive grids, narrowing the computational scope for the second-stage fine optimization and avoiding redundant calculations in global iteration.

[0137] Phase 2: Refined adjustment of target body boundary: Weaken L1 regularization, set λ1 / λ2=1, enhance the spatial smoothness constraint of L2 regularization, and refine the Cole-Cole parameter of the target body boundary unit to match the target body shape with geological laws (such as lithological contact relationship and structural distribution direction) and improve the spatial characterization accuracy of the target body.

[0138] After weakening L1 and strengthening L2 regularization, the smoothness constraint of the inversion is enhanced, which can make fine adjustments to the target body boundary, making the target body shape more in line with geological laws (such as the continuity of lithological contact zones and the orientation of structures), solving the problem of "rough boundaries and abrupt shapes" in the first stage, and improving the spatial characterization accuracy of the target body.

[0139] Based on the first and second stages mentioned above, a progressive inversion from "coarse positioning to fine characterization" was achieved, which not only ensured the accuracy of target identification but also improved the detail precision of imaging, ultimately outputting a three-dimensional model that conforms to geological understanding.

[0140] Combining C1-C3 above, and for example Figure 5(Vertical profile of the three-dimensional Cole-Cole parameter model obtained by inversion provided in this application embodiment) This application embodiment uses lithological parameter constraints to ensure that the inversion results strictly conform to the regional geological laws and completely eliminate non-geological anomaly solutions; it uses depth weight constraints to offset the geometric attenuation of electromagnetic signals and achieve balanced imaging of target bodies at both deep and shallow depths; it uses a two-stage variable regularization strategy to balance rapid target body screening and fine boundary characterization, greatly improving the engineering practicality of the inversion.

[0141] Step 105: Using the relative root mean square error between the inverted calculated magnetic field cumulative response value and the observed magnetic field cumulative response value as the inversion convergence criterion, the three-dimensional Cole-Cole parameter model of the first target is verified to obtain the three-dimensional Cole-Cole parameter model of the second target, which serves as the result of the three-dimensional rapid inversion of transient electromagnetic multi-parameters in the well.

[0142] The relative root mean square error (RMS) between the inverted calculated magnetic field cumulative response value and the observed magnetic field cumulative response value is used as the convergence criterion to verify the three-dimensional Cole-Cole parameter model of the first target: if the RMS is less than the preset threshold (e.g., 5%), the inversion is determined to be converged, and the three-dimensional Cole-Cole parameter model of the second target is directly output as the final result; if it is not converged, return to step 104 to re-optimize until the convergence condition is met.

[0143] Step 105 establishes quantitative convergence criteria to ensure the accuracy and reliability of the inversion results, avoiding insufficient inversion or excessive iteration, forming a closed-loop process of "optimization-verification-iteration." This ensures that the final output 3D model conforms to geological laws and meets accuracy requirements. The final output second objective 3D Cole-Cole parameter model can be directly used for the 3D physical property interpretation of underground geological bodies, providing data support for exploration, engineering, and other applications. Specifically, step 105 can be further refined to include the following implementation steps:

[0144] D1. Call the pre-established mathematical mapping relationship and the magnetic dipole forward modeling equation of the Cole-Cole model to perform forward modeling calculation on the three-dimensional Cole-Cole parameter model of the first target, and obtain the inverse calculation magnetic field cumulative response value corresponding to the three-dimensional Cole-Cole parameter model of the first target;

[0145] D2. Obtain the cumulative response value of the observed magnetic field corresponding to the observed magnetic anomaly data that reflects the electromagnetic characteristics of the target geological body;

[0146] D3. Substitute the observed magnetic field cumulative response value and the inverted magnetic field cumulative response value corresponding to the three-dimensional Cole-Cole parameter model of the first target into the preset relative root mean square error calculation formula to obtain the relative root mean square error value of the current inversion result.

[0147] D4. Compare the relative root mean square error value with the preset convergence threshold to determine whether the inversion has converged.

[0148] D51. If so, then there is no need to optimize the three-dimensional Cole-Cole parameter model of the first target; it can be directly used as the three-dimensional Cole-Cole parameter model of the second target.

[0149] D52. If the inversion is determined to be convergent, discard the first objective 3D Cole-Cole parameter model; iteratively execute the optimization process based on the initial 3D Cole-Cole parameter model to obtain a new first objective 3D Cole-Cole parameter model, and verify the new first objective 3D Cole-Cole parameter model until the inversion is determined to be convergent. Then, directly use the new first objective 3D Cole-Cole parameter model as the second objective 3D Cole-Cole parameter model.

[0150] Furthermore, for D1-D52 above, the embodiments of this application will be explained and illustrated with the following examples:

[0151] Step 105 is the final closed-loop step in the entire methodology. Its core is to determine the inversion convergence by verifying the fitting accuracy, and finally output a three-dimensional multi-parameter geological model that meets the engineering requirements, thus completing the entire closed-loop process from observation data to the characterization of underground geological bodies.

[0152] The relative root mean square error (RMS) between the calculated and observed cumulative magnetic field responses is used as the convergence criterion for inversion. Iteration stops when the relative RMS is less than 5%, and the final three-dimensional Cole-Cole parameter model is output, containing the three-dimensional spatial distribution of conductivity σ, polarizability η, time constant τ, and dispersion coefficient c for each grid cell. Specific explanations are as follows:

[0153] Root mean square error (RMS): It is the most critical indicator for evaluating the fitting accuracy in the field of geophysical inversion. It is used to quantify the degree of deviation between the "cumulative magnetic field response calculated by inversion and forward modeling" and the "cumulative magnetic field response observed in reality".

[0154] This application uses an iterative termination logic. When the calculated relative RMS meets the condition of <5% during the inversion iteration process, the inversion is determined to be converged and the iteration is stopped immediately. If the threshold is not reached, the iterative solution in step 103 and the multi-parameter constraint optimization in step 104 are continued until the accuracy requirements are met.

[0155] Furthermore, the advantage of setting relative RMS < 5% as the iteration termination condition is that 5% is a recognized high-precision inversion threshold in the field of transient electromagnetic exploration. At this error level, the inversion model closely matches the measured data, effectively avoiding the problem of underfitting (large deviation between the model and the data). Setting the threshold to 5% rather than infinitely approaching 0 is to avoid overfitting the observation noise during inversion, which could lead to false anomalies in the model and ensure the geological rationality of the model. This threshold balances inversion accuracy and computational efficiency, avoiding meaningless iterative calculations while meeting exploration requirements, and meeting the requirements of engineering applications.

[0156] Furthermore, the "output result" implemented in this embodiment is a three-dimensional Cole-Cole parameter model, which includes four core Cole-Cole parameters for each grid cell in the underground three-dimensional space. Each parameter corresponds to different physical properties of the underground geological body, as shown in Table 2 below:

[0157] Table 2

[0158]

[0159] The final output of this application's embodiment is a three-dimensional Cole-Cole multi-parameter model. This model uses an adaptive three-dimensional mesh as its spatial carrier and assigns four core Cole-Cole parameters (conductivity σ, polarizability η, time constant τ, and dispersion coefficient c) to each mesh cell in the underground three-dimensional space, fully characterizing the electrical conductivity and polarization coupling electromagnetic properties of the underground geological body. This model can be visualized through horizontal slicing, vertical cross-sectioning, and three-dimensional volume rendering, intuitively presenting the three-dimensional spatial distribution and physical properties of the target geological body (such as a metallic ore body), providing accurate data support for mineral exploration and geological interpretation.

[0160] For example, such as Figure 6 As shown in the figure, the horizontal slice diagram of the target body provided in this embodiment of the scheme shows the visualization result of the conductivity σ parameter in the three-dimensional Cole-Cole parameter model of the second target at a depth of 800m.

[0161] To further explain steps 101-105 above in more detail, this application provides the following specific example, "Application in Deep Copper-Nickel Deposit Exploration in a Certain Mining Area":

[0162] (1) Data import and preprocessing:

[0163] Import transient electromagnetic observation data from the ground well, including magnetic field strength. Magnetic field change rate The number of sampling points N=158, and the vertical sampling interval of the sensor Δz=5m.

[0164] Preprocessing employs db4 wavelet 3-level decomposition filtering with adaptive thresholding. To eliminate interference from power facilities, valid data within the time period of 0.54μs to 1ms is retained, while noise data with a response value less than 1nT / s is removed.

[0165] (2) Calculation of cumulative magnetic field response:

[0166] The cumulative response of the magnetic field was calculated using the trapezoidal integral method. Using the rate of change of magnetic field Correcting the integral truncation error, after correction The relative error is 3.3%.

[0167] (3) The inversion model construction can be, but is not limited to, offline pre-construction; the background field model and mesh generation of the inversion model are explained in detail below, including the following:

[0168] (3.1) Background field model: Based on the borehole geological data of the working area, a three-layered background field model was established, and the parameters are shown in Table 3 below:

[0169] Table 3

[0170]

[0171] (3.2) Mesh generation: The underground space is numerically discretized using an adaptive 3D mesh: vertical mesh size h z =5m; the horizontal direction is divided into 3 regions with different sizes: 0~500m depth range is 10m×10m, 500~1500m depth range is 20m×20m, and >1500m depth range is 40m×40m; the overall grid range is x∈[-2000, 2000]m, y∈[-2000, 2000]m, z∈[-1700, 100]m.

[0172] (4) Magnetic field cumulative response mapping and multi-parameter magnetic inversion, specifically including the following: objective function parameter setting, regularization parameter determination and iterative solution.

[0173] (4.1) Objective function parameter settings: data weight matrix The diagonal element is 1 / σ i ( σ i (Mean value: 0.5 pT·ms / A); Model weight matrix The diagonal element is 1 / V j (Target area unit volume) V j=62500m 3 ); Geometric weight matrix Set the weight of the target area cell to 1.0 and the weight of the far area cell to 0.5.

[0174] (4.2) Determination of regularization parameter: The regularization parameter λ1 = 0.01 was determined using the L-curve method.

[0175] λ2=0.1.

[0176] (4.3) Iterative solution: The steepest descent method is used to iteratively solve the composite objective function. After 58 iterations, the termination condition is met (relative root mean square error RMS ≤ 5%), and the total time is 9047s.

[0177] (5) Multi-parameter constraint optimization, specifically including the following geological constraints, depth weight compensation, and two-stage variable regularization optimization:

[0178] (5.1) Geological constraints: Based on the lithological and physical characteristics of mineralized zones, surrounding rocks, and fault fracture zones, reasonable boundary values ​​are set for the Cole-Cole parameters of each grid unit to correct the inversion anomaly solutions.

[0179] (5.2) Depth weight compensation: Set depth weight parameters z0=575m, H=600m, apply a compensation weight of 0.25 to the deep units with depth z>1000m to offset the attenuation of deep electromagnetic response and balance the imaging accuracy of deep and shallow parts.

[0180] (5.3) Two-stage variable regularization optimization:

[0181] Phase 1 (High-Difference Target Body Screening): Retain units whose parameters exceed the average value of the surrounding rock by more than 3 times, eliminate interference from low physical property differences, and quickly locate the spatial range of the target geological body;

[0182] The second stage (refined adjustment of the target body boundary): weaken L1 regularization and strengthen L2 regularization smoothing constraints to refine the target body boundary so that the target body shape is completely matched with the geological laws.

[0183] (6) Inversion Result Output: After multi-parameter constraint optimization and convergence verification, the relative root mean square error (RMS) of the inversion decreased from the initial 27% to 3.8%, and the final three-dimensional Cole-Cole parameter model was output:

[0184] The parameters of the mineralized body are as follows: electrical conductivity σ = 180~277 S / m, polarizability η = 0.06~0.09, time constant τ = 0.08~0.12 s, and dispersion coefficient c = 0.4~0.6. The spatial range of the delineated target body is: strike 320 degrees, dip northeast, and depth 575~1100 m. The results show good agreement with the VMS mineralized body verified by drilling, thus verifying the effectiveness and accuracy of this method.

[0185] Furthermore, as a response to the above Figure 1 The implementation of the method shown in this application provides a rapid three-dimensional inversion device for transient electromagnetic multi-parameter inversion of wells. This device embodiment corresponds to the aforementioned method embodiment. For ease of reading, this device embodiment will not repeat the details of the aforementioned method embodiment, but it should be clear that the device in this embodiment can implement all the contents of the aforementioned method embodiment. This device is applied to provide a rapid three-dimensional inversion solution for transient electromagnetic multi-parameter inversion of wells that "balances inversion efficiency and high quality," specifically as follows... Figure 7 As shown, the device includes:

[0186] The data input unit 21 is used to collect the original transient electromagnetic response signal of the underground geological body. The physical quantities of the original transient electromagnetic response signal include at least: magnetic field strength and magnetic field change rate.

[0187] The magnetic field response accumulation determination unit 22 is used to obtain the magnetic field accumulation response data corresponding to the original transient electromagnetic response signal, provided that the data of different physical quantity dimensions in the original transient electromagnetic response signal are uniformly converted into magnetic field intensity data of a single physical quantity magnetic field intensity dimension.

[0188] The inversion processing unit 23 is used to process the accumulated magnetic field response data using a pre-built inversion model and output an initial three-dimensional Cole-Cole parameter model. The inversion model includes the magnetic dipole forward modeling equation of the Cole-Cole model, a layered background field module, and an adaptive three-dimensional mesh. The magnetic dipole forward modeling equation is used to describe the conductive and polarization coupled electromagnetic response of the underground geological body. The layered background field module is used to remove electromagnetic interference from non-target areas and extract the observed magnetic anomalies containing the target geological body. The adaptive three-dimensional mesh is used to partition the underground space.

[0189] The constraint optimization unit 24 is used to optimize the initial three-dimensional Cole-Cole parameter model through geological constraints, depth weights, and variable regularization, and output the optimized first target three-dimensional Cole-Cole parameter model. The optimization process includes at least: correcting outlier solutions, compensating for deep attenuation, and refining the boundary.

[0190] The result output unit 25 is used to verify the first target three-dimensional Cole-Cole parameter model by using the relative root mean square error between the inverted calculated magnetic field cumulative response value and the observed magnetic field cumulative response value as the inversion convergence criterion, and to obtain the second target three-dimensional Cole-Cole parameter model, which serves as the result of the three-dimensional fast inversion of transient electromagnetic multi-parameters in the well.

[0191] Furthermore, in this embodiment, the magnetic field response accumulation determination unit 22 is also specifically applied to perform the detailed implementation steps A1-A3 in step 102 above, which will not be repeated here.

[0192] Furthermore, in this embodiment, the inversion processing unit 23 is also specifically applied to execute the detailed implementation steps B1-B6 in step 103 above, which will not be repeated here.

[0193] Furthermore, in this embodiment, the constraint optimization unit 24 is also specifically applied to execute the detailed implementation steps C1-C3 in step 104 above, which will not be repeated here.

[0194] Furthermore, in this embodiment, the result output unit 25 is also specifically used to execute the detailed implementation steps D1-D52 in step 105 above, which will not be repeated here.

[0195] As described above, the transmission line hidden danger prediction device based on big data analysis includes a processor and a memory. The data input unit, magnetic field response accumulation determination unit, inversion processing unit, constraint optimization unit and result output unit are all stored in the memory as program units. The processor executes the above program units stored in the memory to realize the corresponding functions.

[0196] The processor contains a kernel, which retrieves the corresponding program units from memory. One or more kernels can be configured. By adjusting kernel parameters, the computational complexity of the inversion is significantly reduced by transforming the highly nonlinear time-domain electromagnetic response into a linear magnetic field accumulation response. Simultaneously, relying on dual-physical quantity data fusion, multi-parameter constraint optimization, and precise convergence determination, the stability and resolution of the inversion are effectively balanced. This ensures that the output 3D parameter model conforms to geological laws while possessing high-precision imaging effects, thus providing a rapid 3D inversion solution for transient electromagnetic multi-parameters in wells that balances inversion efficiency and high quality.

[0197] This application provides a computer-readable storage medium storing a computer program. When the computer program is executed by a processor, it implements the above-described method for rapid three-dimensional inversion of transient electromagnetic parameters in wells.

[0198] This application provides an electronic device, which includes at least one processor, at least one memory and a bus connected to the processor; wherein the processor and the memory communicate with each other through the bus; the processor is used to call program instructions in the memory to execute the above-described transient electromagnetic multi-parameter three-dimensional fast inversion method for wells.

[0199] This application also provides a computer program product that, when executed on a data processing device, is suitable for executing a program that initializes the steps of a rapid three-dimensional inversion method for transient electromagnetic multi-parameters in upper ground wells.

[0200] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

[0201] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0202] The above are merely embodiments of this application and are not intended to limit the scope of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of the claims of this application.

Claims

1. A rapid three-dimensional inversion method for transient electromagnetic multi-parameter wellbore drilling, characterized in that, The method includes: The raw transient electromagnetic response signal of the underground geological body is collected, and the physical quantities of the raw transient electromagnetic response signal include at least: magnetic field strength and magnetic field change rate; Under the premise of uniformly converting the data of different physical quantity dimensions in the original transient electromagnetic response signal into magnetic field strength data of a single physical quantity magnetic field strength dimension, the magnetic field cumulative response data corresponding to the original transient electromagnetic response signal is obtained. The accumulated magnetic field response data is processed using a pre-built inversion model to output an initial three-dimensional Cole-Cole parameter model. The inversion model includes an adaptive three-dimensional grid, a layered background field module, and the magnetic dipole forward modeling equations of the Cole-Cole model. The adaptive three-dimensional grid, the layered background field module, and the magnetic dipole forward modeling equations of the Cole-Cole model work together to establish the mapping relationship between subsurface parameters and electromagnetic response, thus obtaining the theoretical basis for inversion. The layered background field module removes interference and extracts effective anomalies, and the adaptive three-dimensional grid serves as the spatial carrier for inversion, together completing the inversion processing of the accumulated magnetic field response data. The initial three-dimensional Cole-Cole parameter model is optimized by applying geological constraints, depth weights, and variable regularization to output the optimized first target three-dimensional Cole-Cole parameter model. The optimization process includes at least: correcting outlier solutions, compensating for deep attenuation, and refining the boundaries. The relative root mean square error between the inverted calculated magnetic field cumulative response value and the observed magnetic field cumulative response value is used as the inversion convergence criterion. The three-dimensional Cole-Cole parameter model of the first target is verified to obtain the three-dimensional Cole-Cole parameter model of the second target, which serves as the result of the three-dimensional rapid inversion of transient electromagnetic multi-parameters in the well.

2. The method according to claim 1, characterized in that, The step of obtaining the cumulative magnetic field response data corresponding to the original transient electromagnetic response signal, under the premise of uniformly converting the data of different physical quantity dimensions in the original transient electromagnetic response signal into magnetic field strength data of a single physical quantity dimension, includes: The original magnetic field rate of change component contained in the original transient electromagnetic response signal is converted into a standard magnetic field strength component through integration. This standard magnetic field strength component is then aligned with the original magnetic field strength component inherent in the original transient electromagnetic response signal to form a consistent magnetic field strength dataset. Based on the consistent magnetic field strength dataset, generate the total magnetic field strength data corresponding to the original transient electromagnetic response signal; The total magnetic field strength data is integrated over a time within a preset observation time window to obtain the first magnetic field cumulative response data. To address the truncation error caused by the time integration operation, the first magnetic field cumulative response data is corrected to obtain the second magnetic field cumulative response data, which serves as the magnetic field cumulative response data corresponding to the original transient electromagnetic response signal.

3. The method according to claim 1, characterized in that, The process of using a pre-built inversion model to process the accumulated magnetic field response data and output an initial three-dimensional Cole-Cole parameter model includes: The adaptive 3D mesh in the pre-built inversion model is called to numerically discretize the underground space and obtain each element of the adaptive 3D mesh, which provides a 3D spatial computing carrier for subsequent background field calculation and forward modeling. Based on each element of the adaptive three-dimensional mesh, the layered background field module in the inversion model is called to calculate the cumulative response data of the background magnetic field corresponding to the layered background field. By subtracting the background magnetic field cumulative response data from the magnetic field cumulative response data corresponding to the original transient electromagnetic response signal, we obtain the observed magnetic anomaly data that only reflects the electromagnetic characteristics of the target geological body. Based on the magnetic dipole forward modeling equation of the Cole-Cole model in the inversion model, a mathematical mapping relationship between the electrical conductivity and polarization coupling electromagnetic response of underground geological bodies and the observed magnetic anomaly data is established in advance, providing a theoretical basis for subsequent inversion solutions; The magnetic dipole forward modeling equation of the Cole-Cole model, the observed magnetic anomaly data reflecting the electromagnetic characteristics of the target geological body, and each cell of the adaptive three-dimensional grid are spatially correlated to determine the Cole-Cole parameter solution domain corresponding to each grid cell. Based on the observed magnetic anomaly data reflecting the electromagnetic characteristics of the target geological body, the mathematical mapping relationship, and the Cole-Cole parameter solution domain corresponding to each grid cell, a composite objective function is constructed, including a data fitting term, a regularization constraint term, and a parameter correlation term. The data fitting term characterizes the fitting error between the observed magnetic anomaly data and the forward modeling response; the regularization constraint term ensures the spatial continuity of the inversion model; and the parameter correlation term, based on the physical essence of the Cole-Cole model, sets inequality constraints on the value ranges of each parameter. The composite objective function is solved iteratively using the steepest descent method. The iteration termination condition is that the relative change of the composite objective function between two iterations is less than a preset threshold or a preset number of iterations. After the iteration terminates, the solved Cole-Cole parameters are assigned to each element of the adaptive three-dimensional mesh to generate an initial three-dimensional Cole-Cole parameter model, which is used as the processing result of the inversion model.

4. The method according to claim 1, characterized in that, The optimization process, which optimizes the initial 3D Cole-Cole parameter model through geological constraints, depth weights, and variable regularization, outputs an optimized first target 3D Cole-Cole parameter model, including: Based on prior regional geological information and lithological property test data, the Cole-Cole parameters of each grid cell in the initial three-dimensional Cole-Cole parameter model are set with corresponding lithological value boundaries. By projecting parameters that exceed the value boundaries to the legal range using the projection operator, the abnormal solutions generated during the inversion process are corrected, and a three-dimensional Cole-Cole parameter model corrected by geological constraints is obtained. A depth weighting function is introduced to apply differentiated weights to grid cells at different depths in the geologically constrained 3D Cole-Cole parameter model, compensating for the attenuation of deep electromagnetic response, and thus obtaining a depth-weighted 3D Cole-Cole parameter model. A preset two-stage variable regularization strategy is adopted to perform boundary optimization processing on the depth-weighted compensated three-dimensional Cole-Cole parameter model to obtain the first target three-dimensional Cole-Cole parameter model. The preset two-stage variable regularization strategy includes: the first stage strengthens L1 regularization to lock the target volume range, and the second stage weakens L1 regularization and strengthens L2 regularization to constrain the boundary optimization.

5. The method according to claim 3, characterized in that, The relative root mean square error between the inverted calculated cumulative magnetic field response value and the observed cumulative magnetic field response value is used as the inversion convergence criterion to verify the three-dimensional Cole-Cole parameter model of the first target, thereby obtaining the three-dimensional Cole-Cole parameter model of the second target, which serves as the result of the rapid three-dimensional inversion of transient electromagnetic parameters in the wellbore, including: The pre-established mathematical mapping relationship and the magnetic dipole forward modeling equation of the Cole-Cole model are invoked to perform forward modeling calculation on the three-dimensional Cole-Cole parameter model of the first target, so as to obtain the inverse calculation magnetic field cumulative response value corresponding to the three-dimensional Cole-Cole parameter model of the first target; Obtain the cumulative response value of the observed magnetic field corresponding to the observed magnetic anomaly data reflecting the electromagnetic characteristics of the target geological body; Substitute the observed magnetic field cumulative response value and the inverted magnetic field cumulative response value corresponding to the three-dimensional Cole-Cole parameter model of the first target into the preset relative root mean square error calculation formula to obtain the relative root mean square error value of the current inversion result; The relative root mean square error value is compared with a preset convergence threshold to determine whether the inversion has converged. If so, there is no need to optimize the three-dimensional Cole-Cole parameter model of the first target; it can be directly used as the three-dimensional Cole-Cole parameter model of the second target.

6. The method according to claim 5, characterized in that, If the inversion is determined to be convergent, the method further includes: Discard the three-dimensional Cole-Cole parameter model of the first target; The operation of iteratively performing optimization processing based on the initial three-dimensional Cole-Cole parameter model to obtain a new first target three-dimensional Cole-Cole parameter model is performed, and the new first target three-dimensional Cole-Cole parameter model is verified until the inversion convergence is determined. Then, the new first target three-dimensional Cole-Cole parameter model is directly used as the second target three-dimensional Cole-Cole parameter model.

7. A rapid three-dimensional inversion device for transient electromagnetic multi-parameter modeling of wells, characterized in that, The device includes: The data input unit is used to acquire the original transient electromagnetic response signal of the underground geological body. The physical quantities of the original transient electromagnetic response signal include at least: magnetic field strength and magnetic field change rate. The magnetic field response accumulation determination unit is used to obtain the magnetic field accumulation response data corresponding to the original transient electromagnetic response signal, provided that the data of different physical quantity dimensions in the original transient electromagnetic response signal are uniformly transformed into magnetic field intensity data of a single physical quantity magnetic field intensity dimension. The inversion processing unit is used to process the magnetic field cumulative response data using a pre-built inversion model, and output an initial three-dimensional Cole-Cole parameter model. The inversion model includes an adaptive three-dimensional grid, a layered background field module, and the magnetic dipole forward modeling equation of the Cole-Cole model. The adaptive three-dimensional grid, the layered background field module, and the magnetic dipole forward modeling equation of the Cole-Cole model work together to establish the mapping relationship between subsurface parameters and electromagnetic response using the magnetic dipole forward modeling equation, thus obtaining the theoretical basis for inversion. The layered background field module removes interference and extracts effective anomalies, and the adaptive three-dimensional grid serves as the spatial carrier for inversion, jointly completing the inversion processing of the magnetic field cumulative response data. The constrained optimization unit is used to optimize the initial three-dimensional Cole-Cole parameter model through geological constraints, depth weights, and variable regularization, and output the optimized first target three-dimensional Cole-Cole parameter model. The optimization process includes at least: correcting outlier solutions, compensating for deep attenuation, and refining the boundary. The result output unit is used to verify the first target three-dimensional Cole-Cole parameter model by using the relative root mean square error between the inverted calculated magnetic field cumulative response value and the observed magnetic field cumulative response value as the inversion convergence criterion, and to obtain the second target three-dimensional Cole-Cole parameter model, which serves as the result of the ground well transient electromagnetic multi-parameter three-dimensional fast inversion.

8. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the rapid three-dimensional inversion method for transient electromagnetic multi-parameter wellbore as described in any one of claims 1-6.

9. An electronic device, characterized in that, The device includes at least one processor, and at least one memory and bus connected to the processor; The processor and the memory communicate with each other via the bus. The processor is used to call program instructions in the memory to execute the transient electromagnetic multi-parameter three-dimensional fast inversion method for wells as described in any one of claims 1-6.

10. A computer program product, characterized in that, When executed on a data processing device, it is suitable to execute a program that initializes the well transient electromagnetic multi-parameter three-dimensional fast inversion method steps as described in any one of claims 1-6.

Images