Construction model adjustment method and system based on artificial intelligence and medium

Abstract

The invention provides a construction model adjustment method and system based on artificial intelligence and a medium, and relates to the technical field of geological exploration, and the method comprises the steps: obtaining an initial construction model, and generating an adjustment parameter set based on a seismic data volume; constructing an adjustment function and a integrating degree evaluation function; combining the two functions to generate a target optimization function; iteratively updating and adjusting the parameter set through a preset optimization algorithm, optimizing a target optimization function, obtaining an optimized seismic axis cumulative continuous length after each iteration, and judging whether a preset convergence condition is met or not; and when a preset convergence condition is met, outputting the optimized seismic axis cumulative continuous length as an optimal seismic axis cumulative continuous length, and outputting a corresponding optimal adjustment parameter set to obtain an optimized construction model. According to the method and the device, the technical problem that the precision and the reliability of the construction model are insufficient due to the fact that the adjustment parameters are selected according to expert experience and subjective judgment in the prior art is solved, and the adjustment efficiency and the model precision of the construction model are improved.

Description

Technical Field

[0001] The present application relates to the field of geological exploration technology, and specifically to a structural model adjustment method, system and medium based on artificial intelligence. Background Art

[0002] The structural model is a mathematical model used to describe underground rock formations, geological structures and their physical characteristics in the fields of geological exploration and oil and gas exploration. It has important application value in the exploration and exploitation of underground resources and early warning of geological disasters. In order to ensure the accuracy and reliability of the structural model, it must be continuously adjusted to adapt to changes in different geological conditions.

[0003] Existing structural model adjustment methods usually rely on expert experience and subjective judgment to select adjustment parameters. Experts use geological data such as seismic data and drilling data, combined with experience, to modify the model. Although this method can achieve certain results under simple geological conditions, it still faces many challenges under complex geological characteristics. First, expert experience and subjective judgment have strong limitations. When dealing with complex or unknown geological conditions, such as multi-stage fault development areas, it is difficult to make accurate adjustments. Secondly, since the adjustment process relies on manual intervention, it is easy to cause human bias, resulting in the adjusted model not fitting the actual geological structure well enough. In addition, the process of manually adjusting parameters is time-consuming and inefficient, and it is difficult to find the optimal combination of model parameters in a short time. Therefore, under complex geological conditions, it is difficult for existing methods to effectively improve the accuracy and reliability of structural models, affecting the accuracy of exploration and development decisions. Summary of the invention

[0004] The present application provides an artificial intelligence-based structural model adjustment method, system and medium, which solves the technical problem that the prior art relies on expert experience and subjective judgment to select adjustment parameters, resulting in insufficient accuracy and reliability of the structural model under complex geological features, and achieves the technical effect of improving the structural model adjustment efficiency and model accuracy.

[0005] In view of the above problems, on the one hand, the present application provides a structural model adjustment method based on artificial intelligence, the method comprising: obtaining an initial structural model, and generating an adjustment parameter set based on a seismic data body, wherein the seismic data body includes multiple geological feature data; constructing an adjustment function, wherein the adjustment function receives the adjustment parameter set as input, adjusts the initial structural model, and generates an adjusted structural model; constructing a fit evaluation function, wherein the fit evaluation function receives the adjusted structural model as input, calculates the cumulative continuous length of the seismic axis based on the adjusted structural model, and uses the cumulative continuous length of the seismic axis as a fit index; combining the adjustment function and the fit evaluation function to generate a target optimization function, wherein the optimization goal of the target optimization function is to maximize the cumulative continuous length of the seismic axis; iteratively updating the adjustment parameter set through a preset optimization algorithm, optimizing the target optimization function, obtaining the optimized cumulative continuous length of the seismic axis after each round of iteration, and judging whether the preset convergence condition is met; when the preset convergence condition is met, outputting the optimized cumulative continuous length of the seismic axis as the optimal cumulative continuous length of the seismic axis, and outputting the corresponding optimal adjustment parameter set to obtain an optimized structural model.

[0006] On the other hand, the present application also provides an artificial intelligence-based structural model adjustment system, the system comprising: a model initialization module, used to obtain an initial structural model, and generate an adjustment parameter set based on a seismic data body, wherein the seismic data body includes a plurality of geological feature data; an initial model adjustment module, used to construct an adjustment function, wherein the adjustment function receives the adjustment parameter set as input, adjusts the initial structural model, and generates an adjusted structural model; a fit evaluation module, used to construct a fit evaluation function, wherein the fit evaluation function receives the adjusted structural model as input, calculates the cumulative continuous length of the seismic axis based on the adjusted structural model, and uses the cumulative continuous length of the seismic axis as is a fit index; a target optimization function generation module, which is used to combine the adjustment function and the fit evaluation function to generate a target optimization function, wherein the optimization goal of the target optimization function is to maximize the cumulative continuous length of the seismic axis; an adjustment parameter optimization module, which is used to iteratively update the adjustment parameter set through a preset optimization algorithm to optimize the target optimization function, and after each round of iteration, obtain the optimized cumulative continuous length of the seismic axis, and determine whether the preset convergence condition is met; an optimized adjustment parameter output module, which is used to output the optimized cumulative continuous length of the seismic axis as the optimal cumulative continuous length of the seismic axis when the preset convergence condition is met, and output the corresponding optimal adjustment parameter set to obtain an optimized structural model.

[0007] In a third aspect, the present application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps in the above-mentioned artificial intelligence-based construction model adjustment method.

[0008] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0009] This application realizes the automatic adjustment and optimization of the structural model through automated and intelligent means. First, by automatically generating a set of adjustment parameters from seismic data, the reliance on manual experience is reduced, and the accuracy and consistency of parameter selection are improved. Then, the constructed adjustment function and fit evaluation function transform the model adjustment problem into an optimization problem, clarify the optimization goals and evaluation criteria, and make the model adjustment process more systematic and scientific. Through the iterative optimization of the preset optimization algorithm, the optimal combination of adjustment parameters is automatically found, which significantly improves the efficiency and accuracy of the model adjustment and avoids the blindness and inefficiency of traditional manual adjustment. The optimized structural model finally obtained has higher accuracy and reliability, can better reflect the actual characteristics of the underground structure, and provides strong support for the decision-making of geological exploration and petroleum engineering.

[0010] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] Figure 1 A schematic flow chart of a construction model adjustment method based on artificial intelligence provided in an embodiment of the present application.

[0012] Figure 2 A schematic diagram of the structure of an artificial intelligence-based construction model adjustment system provided in an embodiment of the present application.

[0013] Explanation of the reference numerals: model initialization module 10 , initial model adjustment module 20 , fit evaluation module 30 , target optimization function generation module 40 , adjustment parameter optimization module 50 , optimization adjustment parameter output module 60 . DETAILED DESCRIPTION

[0014] The embodiments of the present application provide an artificial intelligence-based structural model adjustment method, system and medium, thereby solving the technical problem that the prior art relies on expert experience and subjective judgment to select adjustment parameters, resulting in insufficient accuracy and reliability of the structural model under complex geological features, thereby achieving the technical effect of improving the structural model adjustment efficiency and model accuracy.

[0015] Embodiment 1, as Figure 1 As shown, the embodiment of the present application provides a construction model adjustment method based on artificial intelligence, the method comprising:

[0016] Step S1: Acquire an initial structural model, and generate an adjustment parameter set based on a seismic data volume, wherein the seismic data volume includes a plurality of geological characteristic data.

[0017] Specifically, the initial structural model refers to a model established through preliminary geological exploration or known data, which is used to describe the characteristics of underground rock formations, geological structures, etc., including the preliminary inferred stratigraphic distribution, fault locations and other geological features. The seismic data body is a collection of seismic reflection data collected through seismic exploration technology, which represents the information of underground structures. The seismic data body contains multiple geological feature data, such as stratigraphic interfaces, faults, lithology changes, etc. The adjustment parameter set is used to modify the values ​​or parameters of the initial structural model, such as fault locations, fault geometry, etc.

[0018] Obtain an initial structural model, which can be derived from early exploration data or geological inference results obtained by other means. Then, use seismic data processing software to read and analyze the seismic data body, and extract the geological characteristic data therein, which include the distribution of horizons, faults, inclination angles of strata, etc. For example, the amplitude, frequency and other information of seismic reflection waves are used to identify the characteristics of stratum interfaces and faults. Based on these geological characteristic data, a set of adjustment parameters for adjusting the structural model is calculated. For example, the stratum thickness parameters are calculated using the stratum reflection time difference in the seismic data body, and the depth adjustment range of the stratum interface, the offset of the fault position and other parameters are determined. These adjustment parameter sets provide a basis for improving the initial model for subsequent model adjustments.

[0019] Step S2: constructing an adjustment function, wherein the adjustment function receives the adjustment parameter set as input, adjusts the initial construction model, and generates an adjusted construction model.

[0020] Specifically, the adjustment function is a mathematical function used to adjust the initial structural model. The adjustment function receives a set of adjustment parameters as input, adjusts elements such as stratum interfaces and faults in the model according to these parameters, and generates an adjusted structural model.

[0021] In this step, according to different geological conditions and adjustment requirements, a suitable function form is selected, such as a linear function, a polynomial function or a function based on a neural network, and then a mathematical modeling software is used to construct an adjustment function. The adjustment parameter set obtained in step S1 is input into the adjustment function, and the adjustment function adjusts the initial structural model according to the mathematical relationship defined inside it. For example, if the adjustment function is a linear function, and the formation thickness parameter in the adjustment parameter has a linear relationship with the formation thickness in the initial structural model, then the formation thickness is changed according to the input parameter value, thereby generating an adjustment structural model.

[0022] By adjusting the function, the initial structural model can be effectively adjusted according to the adjustment parameter set to make it closer to the actual geological structure.

[0023] Step S3: construct a fit evaluation function, wherein the fit evaluation function receives the adjusted structural model as input, calculates the cumulative continuous length of the seismic axis based on the adjusted structural model, and uses the cumulative continuous length of the seismic axis as a fit indicator.

[0024] Specifically, the fit evaluation function is a function used to evaluate the fit between the adjusted structural model and the seismic data. The function receives the adjusted structural model as input, calculates the cumulative continuous length of the seismic axis based on the adjusted structural model, and uses it as a fit indicator. Among them, the cumulative continuous length of the seismic axis refers to the cumulative continuous length of the seismic axis (such as stratigraphic interface, fault, etc.) of the adjusted structural model in the seismic data body, which reflects the degree of match between the model and the seismic data. The longer the length, the higher the fit.

[0025] Construct a fit evaluation function, which receives the adjusted structural model generated in step S2 as input, identifies and calculates the relevant information of the seismic axis in the adjusted structural model, and then cumulatively calculates the continuous length of the seismic axis, and uses this length as a fit indicator. The calculation process can be completed using geological modeling and analysis software, which has the function of identifying the seismic axis and calculating the relevant length. For example, in a three-dimensional adjusted structural model, the seismic axis is identified along the seismic wave propagation path, the continuous length of the seismic axis in each stratum is calculated and accumulated, and the cumulative continuous length of the seismic axis is obtained, and it is used as a fit indicator. The fit index provides a quantitative indicator for evaluating the rationality of the adjusted structural model, which can intuitively judge the degree of fit between the adjusted model and the actual situation in terms of seismic characteristics.

[0026] Step S4: combining the adjustment function and the fit evaluation function to generate a target optimization function, wherein the optimization goal of the target optimization function is to maximize the cumulative continuous length of the earthquake axis.

[0027] Specifically, the target optimization function is a function composed of an adjustment function and a fit evaluation function, which is used to optimize the set of adjustment parameters. Its optimization goal is to maximize the cumulative continuous length of the seismic axis, that is, to improve the fit between the adjusted structural model and the seismic data. With maximizing the cumulative continuous length of the seismic axis as the optimization goal, the adjustment function and the fit evaluation function are combined to form the target optimization function. The target optimization function combines the adjustment process with the evaluation criteria, realizes the organic combination of model adjustment and evaluation, and provides a clear goal for the entire adjustment optimization process, that is, maximizing the cumulative continuous length of the seismic axis. At the same time, it also improves the systematicness and consistency of the optimization process.

[0028] Step S5: Iteratively update the adjustment parameter set through a preset optimization algorithm to optimize the target optimization function. After each round of iteration, obtain the cumulative continuous length of the optimized seismic axis and determine whether the preset convergence condition is met.

[0029] Specifically, the preset optimization algorithm is an algorithm used to iteratively update and adjust a set of parameters, such as the gradient descent method, Newton's downhill method, etc. The appropriate optimization algorithm is selected according to the characteristics of the target optimization function. For example, the gradient descent method can be selected for most cases; for large-scale data sets or complex models, the stochastic gradient descent (SGD) algorithm can be used; when the model is complex and the linear conditions are poor, the Newton's downhill method can be used. The preset convergence condition is the condition for the end of the optimization process, which can be set to reach the maximum number of iterations, the optimization increment is less than the threshold, etc., to determine whether the optimization goal has been achieved.

[0030] Select a preset optimization algorithm and iteratively update the adjustment parameter set according to the target optimization function. In each round of iteration, calculate the cumulative continuous length of the optimized seismic axis and determine whether the preset convergence condition is met. In the specific implementation process, the optimization process can be implemented using deep learning frameworks such as TensorFlow or PyTorch in Python. These frameworks provide a wealth of optimization algorithms and automatic derivation functions, which can facilitate gradient calculation and parameter update.

[0031] Through the iterative update of the preset optimization algorithm, the automatic optimization of the adjustment parameter set is achieved, which reduces manual intervention and improves the efficiency and accuracy of the adjustment parameter optimization.

[0032] Step S6: When the preset convergence condition is reached, the optimized seismic axis cumulative continuous length is output as the optimal seismic axis cumulative continuous length, and the corresponding optimal adjustment parameter set is output to obtain an optimized structural model.

[0033] Specifically, the optimal cumulative continuous length of the seismic axis is the maximum cumulative continuous length of the seismic axis achieved during the optimization process, indicating the highest degree of fit between the adjusted structural model and the seismic data. The optimal adjustment parameter set is the adjustment parameter set that maximizes the cumulative continuous length of the seismic axis, which is used to generate the final optimized structural model.

[0034] When the preset convergence condition is reached, the calculated cumulative continuous length of the optimized seismic axis is output as the optimal cumulative continuous length of the seismic axis, and the corresponding optimal adjustment parameter set is output at the same time. Then, the initial structural model is finally adjusted using the optimal adjustment parameter set and the adjustment function to obtain the optimized structural model. The optimized structural model can better reflect the actual characteristics of the underground structure, improve the decision-making accuracy of exploration and development, and reduce risks and costs.

[0035] Furthermore, the objective optimization function is as follows: F(c0, c1, ..., c i )=l; where F(c0,c1,…,c i ) represents the target optimization function, which indicates the relationship between the cumulative continuous length of the earthquake axis and the set of adjustment parameters, c0, c1, …, c i Characterize the set of adjustment parameters, c i represents the i-th parameter in the adjustment parameter set, and l represents the cumulative continuous length of the earthquake axis.

[0036] Specifically, the objective optimization function F(c0, c1, ..., c i )=l establishes the relationship between the adjustment parameter set and the cumulative continuous length of the earthquake axis. The adjustment parameter set c0, c1, ..., c i It is the core variable of model adjustment, and each parameter represents a different adjustment item of the structural model, such as the location of the fault, the depth of the rock layer or the shape of the layer. l is the cumulative continuous length of the seismic axis, which is used to indicate the degree of match between the seismic characteristics in the model and the actual seismic data, so as to quantify the adjustment effect of the structural model. The larger the cumulative continuous length of the seismic axis, the higher the fit between the structural model and the seismic data. When the parameters c0, c1, ..., c in the adjustment parameter set are i When changes occur, the function F(c0, c1, ..., c i ) can calculate the corresponding cumulative continuous length l of the earthquake axis.

[0037] In the optimization process, the parameters in the adjustment parameter set are constantly changed, such as the position offset of the fault, the adjustment of the layer depth, the correction of the lithology distribution, etc., and then the adjusted parameters are substituted into the target optimization function F (c0, c1, ..., c i )=l, calculate the cumulative continuous length of the new earthquake axis, and judge whether the optimization effect is achieved. By optimizing this function, find a set of adjustment parameters (c0, c1, ..., ci ), so that l is maximized, that is, the fit between the structural model and the seismic data is optimal.

[0038] This objective optimization function provides a clear mathematical basis for the optimization process of the entire structural model, allowing the optimization algorithm to be adjusted according to the impact of the adjustment parameter set on the cumulative continuous length of the seismic axis, thereby optimizing in the direction of maximizing the cumulative continuous length of the seismic axis, and ultimately obtaining an optimized structural model that is more in line with actual geological conditions.

[0039] Furthermore, in step S5, the adjustment parameter set is iteratively updated by a preset optimization algorithm, including:

[0040] Step S54A-1: Calculate the updated gradient of the target optimization function according to the influence relationship between the adjustment parameter set and the cumulative continuous length of the seismic axis.

[0041] Step S54A-2: Based on the update gradient, perform iterative update of the adjustment parameter set according to a preset learning rate, and the formula is as follows: Among them, t represents the number of iterations, represents the value of the parameter in the tth iteration, Characterizes the parameter c in the t+1th iteration i The value of α represents the preset learning rate. Characterizes the updated gradient of the target optimization function, which represents the target optimization function F with respect to the parameter c i The partial derivative of .

[0042] Specifically, in a specific implementation, a gradient descent algorithm is used to iteratively update the adjustment parameter set. First, the updated gradient of the target optimization function is calculated based on the influence relationship between the adjustment parameter set and the cumulative continuous length of the earthquake axis. In the actual operation process, an automatic differentiation tool, such as TensorFlow or PyTorch in Python, is used to differentiate the target optimization function and calculate the partial derivative of the target optimization function with respect to each adjustment parameter, thereby obtaining the updated gradient The updated gradient provides the direction of parameter adjustment, that is, adjusting the parameters in the opposite direction of the gradient can increase the objective function value the fastest.

[0043] Get updated gradients After that, the adjustment parameter set is iteratively updated in combination with the preset learning rate α. The preset learning rate α is a preset constant used to control the step size of the adjustment parameter set update at each iteration, and determines the speed of adjusting the adjustment parameter set according to the update gradient during the optimization process. And the preset learning rate α is substituted into the formula Iterate and update each parameter in the adjustment parameter set. In each iteration, the current parameter value (In the tth iteration, parameter c i The value of) is multiplied by the preset learning rate α and the updated gradient The product of , and then subtract this result from the current value to get the value of the next iteration (In the tth iteration, parameter c i ).

[0044] Through iterative updates, the adjustment parameter set can be gradually adjusted in the direction of making the target optimization function optimal (i.e. maximizing the cumulative continuous length of the earthquake axis). The existence of the preset learning rate α can reasonably control the speed of adjustment and avoid adjusting too fast or too slow, thereby effectively improving the efficiency and accuracy of optimization.

[0045] Furthermore, before iteratively updating the adjustment parameter set by using a preset optimization algorithm, the following steps are included:

[0046] Step S51: based on the seismic data volume, extract the sample adjustment parameter set of the sample structural model, and extract the corresponding sample seismic axis cumulative continuous length set.

[0047] Step S52: Based on a neural network, the sample adjustment parameter set is used as input, and the sample seismic axis cumulative continuous length set is used as output to construct an adjustment parameter influence relationship analyzer.

[0048] Step S53: Based on the adjustment parameter influence relationship analyzer, an influence relationship between the adjustment parameter set and the cumulative continuous length of the seismic axis is generated.

[0049] Specifically, the sample structural model is a representative structural model selected from the seismic data body, which is used to train and analyze the relationship between the adjustment parameters and the cumulative continuous length of the seismic axis. The sample adjustment parameter set is the set of adjustment parameters corresponding to the sample structural model, including sample values ​​of parameters such as stratum depth and fault location. The sample seismic axis cumulative continuous length set is the sample value of the cumulative continuous length of the seismic axis corresponding to the sample structural model, reflecting the fit between the sample model and the seismic data.

[0050] First, multiple representative sample structural models are selected from the seismic data body. For example, structural models of different geological feature areas can be selected, such as oil and gas structures, fault structures, etc. Then, the adjustment parameter set of each sample structural model is extracted, such as the sample values ​​of parameters such as the depth of the stratigraphic interface and the location of the fault. At the same time, the cumulative continuous length of the seismic axis of each sample structural model is calculated to form a set of cumulative continuous lengths of the sample seismic axis. The sample structural model and the corresponding cumulative continuous length of the seismic axis can reflect the fit between the model and the seismic data under different geological characteristics, and provide rich input and output data for subsequent neural network training.

[0051] The adjustment parameter impact relationship analyzer is an analyzer built on a neural network, which is used to analyze the relationship between the adjustment parameter set and the cumulative continuous length of the earthquake axis, and predict the impact of the adjustment parameter on the cumulative continuous length of the earthquake axis. To build the adjustment parameter impact relationship analyzer, we must first select a suitable neural network structure, such as a multi-layer perceptron (MLP) or a convolutional neural network (CNN), and determine the number of layers and the number of neurons in each layer according to the characteristics and complexity of the sample data. Then, the sample adjustment parameter set is used as input and the sample earthquake axis cumulative continuous length set is used as output to train the neural network. During the training process, the weights and biases of the network are adjusted by the back propagation algorithm, so that the network learns the mapping relationship between the input (sample adjustment parameter set) and the output (sample earthquake axis cumulative continuous length set). The training process of the adjustment parameter impact relationship analyzer can be implemented using a deep learning framework such as TensorFlow or PyTorch. For example, a simple multi-layer perceptron neural network is used, with 2 neurons in the input layer (corresponding to the depth of formation A and the position of fault C), 3 neurons in the hidden layer, and 1 neuron in the output layer (corresponding to the cumulative continuous length of the earthquake axis). The above sample data is used for training, and the weights and biases of the network are adjusted so that the predicted cumulative continuous length of the earthquake axis output by the network is as close as possible to the value in the actual set of cumulative continuous lengths of the sample earthquake axis.

[0052] Different adjustment parameter combinations are input into the trained adjustment parameter influence relationship analyzer, and the adjustment parameter combinations are analyzed and processed using the forward propagation process of the neural network to predict the corresponding cumulative continuous length prediction value of the seismic axis. Based on the prediction results, the change in the cumulative continuous length of the seismic axis when a certain adjustment parameter changes is analyzed, thereby generating the influence relationship between the adjustment parameter set and the cumulative continuous length of the seismic axis.

[0053] The neural network can capture the complex nonlinear relationship between the adjustment parameters and the cumulative continuous length of the seismic axis, thereby accurately predicting the impact of the adjustment parameters on the cumulative continuous length of the seismic axis, and providing a reliable basis for the influence relationship for calculating the updated gradient of the target optimization function in the subsequent parameter optimization process.

[0054] Furthermore, in step S5, determining whether a preset convergence condition is met includes:

[0055] Step S55: After each round of iteration, obtain the current round adjustment parameter set and the corresponding current round cumulative continuous length of the earthquake axis.

[0056] Step S56: Compare the cumulative continuous length of the earthquake axis in this round with the cumulative continuous length of the earthquake axis in the previous round, and calculate the optimized increment.

[0057] Step S57: If the optimization increment is less than or equal to a preset optimization increment threshold, or the number of iterations reaches a preset number of iterations threshold, it is determined that the preset convergence condition is met.

[0058] Specifically, after each iteration, the adjustment parameter set of this round and the cumulative continuous length of the seismic axis of this round calculated based on these adjustment parameter sets are obtained from the intermediate results of the optimization algorithm. For example, when using the gradient descent algorithm for iterative optimization, each time a round of updating of the adjustment parameter set is completed, the updated adjustment parameter set is recorded, and then these adjustment parameters are substituted into the fit evaluation function to calculate the cumulative continuous length of the seismic axis of this round.

[0059] The optimization increment is the difference between the cumulative continuous length of the seismic axis in this round and the cumulative continuous length of the seismic axis in the previous round, which indicates the degree of improvement in the model fit in the current iteration. Compare the cumulative continuous length of the seismic axis in this round with the cumulative continuous length of the seismic axis in the previous round to calculate the optimization increment. For example, if the cumulative continuous length of the seismic axis in the previous round is l1 and the cumulative continuous length of the seismic axis in this round is l2, then the optimization increment = l2-l1. By calculating the optimization increment, the degree of improvement in the model fit in each round of iteration can be quantified, providing a quantitative basis for judging the effectiveness of the optimization process. At the same time, the size of the optimization increment can be used as an important indicator to judge whether the optimization process is close to convergence, helping to determine whether to continue iteration or stop optimization.

[0060] The preset optimization increment threshold is the minimum value of the optimization increment pre-set in the optimization algorithm. When the optimization increment is less than or equal to the threshold, the optimization process is considered to have reached convergence, that is, the model fit is no longer significantly improved. The preset iteration number threshold is the maximum number of iterations pre-set in the optimization algorithm. When the number of iterations reaches this threshold, the optimization process is stopped regardless of the optimization increment to prevent infinite iterations. After each round of iterations, the calculated optimization increment is compared with the preset optimization increment threshold, and the current cumulative number of iterations is compared with the preset iteration number threshold. If the optimization increment is less than or equal to the preset optimization increment threshold, or the number of iterations reaches the preset iteration number threshold, it is determined that the preset convergence condition is met and the optimization process is stopped.

[0061] By setting the preset optimization increment threshold and iteration number threshold, we can effectively prevent over-optimization and avoid continuing to perform a large number of iterations when the model fit is not significantly improved, saving computing resources and time, and ensuring the effectiveness of the optimization results and the efficiency of the optimization process.

[0062] In a specific implementation, in step S5, iteratively updating the adjustment parameter set by a preset optimization algorithm includes:

[0063] Step S54B-1: In each round of iteration, a number of first adjustment parameters are randomly selected from the adjustment parameter set.

[0064] Step S54B-2: Calculate a first update gradient of the target optimization function based on the first adjustment parameters.

[0065] Step S54B-3: Based on the first update gradient, perform iterative update of the adjustment parameter set according to a preset learning rate, and the formula is as follows: Among them, s represents the number of iterations, Characterizes the parameter c in the sth iteration i The value of Characterizes the parameter c in the s+1th iteration i The value of α represents the preset learning rate. Characterize the first update gradient.

[0066] Specifically, in addition to the gradient descent method described above, the stochastic gradient descent (SGD) algorithm can also be used to iteratively update the adjustment parameter set. First, in each round of iteration, a random number generator or a random sampling algorithm is used to select a random number from the entire adjustment parameter set (c0, c1, ..., c i ) randomly selects several parameters as the first adjustment parameters. This random selection method introduces randomness into the optimization process. Different from the conventional calculation of update gradients based on all adjustment parameters, this method may explore some optimization directions that are not easy to find in conventional methods, which helps to avoid the optimization process from falling into the local optimal solution and increase the possibility of finding the global optimal solution.

[0067] According to the randomly selected first adjustment parameters, the partial derivatives of the target optimization function with respect to these first adjustment parameters are calculated according to the partial derivative method, and then these partial derivatives are combined to obtain the first update gradient For example, if the first adjustment parameters are c2 and c4, then we need to calculate F(c0,c1,...,c i) with respect to the partial derivatives of c2 and c4, and then according to certain rules (such as vector combination), the first updated gradient is obtained. Similarly, in actual operation, automatic differentiation tools such as TensorFlow or PyTorch in Python can be used to complete the first updated gradient The calculation process.

[0068] According to the first update gradient Combined with the preset learning rate α, the adjustment parameter set is iteratively updated, and the first update gradient And the preset learning rate α is substituted into the formula Each parameter in the adjustment parameter set is iteratively updated in . The iterative process is similar to that in the gradient descent method. The specific process can be referred to the above related description.

[0069] The stochastic gradient descent algorithm reduces the amount of computation in each iteration and improves optimization efficiency by randomly extracting some parameters. It is particularly suitable for large-scale model adjustment problems and significantly improves the flexibility and efficiency of the optimization process.

[0070] In a specific implementation, in step S5, iteratively updating the adjustment parameter set by a preset optimization algorithm includes:

[0071] Step S54C: Calculate the Hessian matrix of the target optimization function, and iteratively update the adjustment parameter set using a second-order optimization strategy. The formula is as follows: Among them, r represents the number of iterations, Characterizes the parameter c in the rth iteration i The value of Characterizes the parameter c in the r+1th iteration i The value of Represents the updated gradient of the target optimization function, H(F) represents the Hessian matrix of the target optimization function, and represents the second-order derivative information of F, [H(F)] -1 It is the inverse matrix of the Hessian matrix, which is used to adjust the gradient direction.

[0072] Specifically, the Newton downhill method can also be used to optimize the set of adjustment parameters for iterative update. First, the Hessian matrix of the target optimization function is calculated. The Hessian matrix is ​​the target optimization function F(c0, c1, ..., c i). This matrix contains the curvature information of the target optimization function with respect to the adjustment parameters, which helps to more accurately determine the adjustment direction and amplitude of the parameters. In actual operation, mathematical analysis software (such as Mathematica, Maple, etc.) or programming languages ​​(such as SciPy library in Python) can be used to calculate the target optimization function F(c0, c1, ..., c i ) all the second-order mixed partial derivatives to construct the Hessian matrix. For example, for a binary function F(c0,c1), we need to calculate These four second-order mixed partial derivatives are then arranged in matrix form into a 2×2 Hessian matrix, and the inverse matrix of the Hessian matrix [H(F)] is calculated -1 , use this inverse matrix to adjust the update gradient Among them, the updated gradient The updated gradient calculation method is the same as that in the aforementioned gradient descent method. The detailed calculation process can refer to the explanations in the aforementioned steps.

[0073] Will update the gradient and the inverse of the Hessian matrix [H(F)] -1 Substitute into the formula For each parameter c in the adjustment parameter set i Perform iterative updates. For each iteration, the current parameter value (In the rth iteration, parameter c i value) multiplied by [H(F)] -1 and The product of , and then subtract this result from the current value to get the value of the next iteration (In the r+1th iteration, parameter c i ).

[0074] The second-order optimization strategy uses the inverse matrix of the Hessian matrix to adjust the gradient direction, and can more accurately adjust the parameter set according to the curvature of the target optimization function. Compared with the first-order optimization, this optimization algorithm can converge to the optimal solution in fewer iterations, especially when the target function has complex curvature, which can improve the speed and accuracy of optimization and make the adjustment parameter set reach the state of optimizing the target optimization function (i.e., the maximum cumulative continuous length of the seismic axis) more quickly.

[0075] Furthermore, the method described in the embodiment of the present application also includes:

[0076] The target optimization function and the preset optimization algorithm are encapsulated to generate an automated structural model adjustment module, wherein the automated structural model adjustment module automatically imports the seismic data volume and the initial structural model and outputs an optimized structural model.

[0077] Specifically, the target optimization function and the code and related logic of the preset optimization algorithm are integrated and packaged into a module or class using a programming language, making it an independent and reusable unit, namely, the automatic structural model adjustment module. The input and output interfaces are defined for the automatic structural model adjustment module. The input interface is set to receive the seismic data body and the initial structural model, and the output interface is set to output the optimized structural model. The module can automatically execute the entire process of structural model adjustment, including the steps of importing the seismic data body and the initial structural model, performing optimization adjustment, and outputting the optimized structural model.

[0078] The encapsulated automatic structural model adjustment module has higher reusability and ease of use. For different seismic data bodies and initial structural models, there is no need to rewrite the code of the target optimization function and the preset optimization algorithm. You only need to call this module and pass in the corresponding input data to get the optimized structural model, which greatly improves work efficiency, reduces the duplication of code writing, and makes the entire structural model adjustment process more modular and standardized.

[0079] In summary, the artificial intelligence-based structural model adjustment method provided in the embodiment of the present application has the following technical effects:

[0080] The embodiment of the present application introduces an adjustment method based on artificial intelligence. Through the combination of an automated adjustment process and a target optimization function, the structural model can be accurately optimized to ensure that it is highly consistent with geological features such as key horizons and faults in the seismic data body. The iterative update mechanism of the optimization algorithm makes the structural model adjustment process more intelligent and automated, greatly improving the adjustment accuracy and efficiency. At the same time, the method reduces manual deviations and ensures the consistency of the structural model with the actual geological conditions, thereby significantly improving the accuracy and reliability of the structural model in complex geological environments.

[0081] Embodiment 2, as Figure 2 As shown, the embodiment of the present application provides a construction model adjustment system based on artificial intelligence, and the system includes:

[0082] The model initialization module 10 is used to obtain an initial structural model and generate an adjustment parameter set based on a seismic data volume, wherein the seismic data volume includes a plurality of geological characteristic data.

[0083] The initial model adjustment module 20 is used to construct an adjustment function, wherein the adjustment function receives the adjustment parameter set as input, adjusts the initial construction model, and generates an adjusted construction model.

[0084] The fit evaluation module 30 is used to construct a fit evaluation function, wherein the fit evaluation function receives the adjusted structural model as input, calculates the cumulative continuous length of the seismic axis based on the adjusted structural model, and uses the cumulative continuous length of the seismic axis as a fit indicator.

[0085] The target optimization function generation module 40 is used to combine the adjustment function and the fit evaluation function to generate a target optimization function, wherein the optimization goal of the target optimization function is to maximize the cumulative continuous length of the seismic axis.

[0086] The adjustment parameter optimization module 50 is used to iteratively update the adjustment parameter set through a preset optimization algorithm, optimize the target optimization function, obtain the cumulative continuous length of the optimized seismic axis after each round of iteration, and determine whether the preset convergence condition is met.

[0087] The optimization adjustment parameter output module 60 is used to output the optimized seismic axis cumulative continuous length as the optimal seismic axis cumulative continuous length when the preset convergence condition is reached, and output the corresponding optimal adjustment parameter set to obtain the optimized structural model.

[0088] Furthermore, the objective optimization function is as follows: F(c0, c1, ..., c i )=l; where F(c0,c1,...,c i ) represents the target optimization function, which represents the relationship between the cumulative continuous length of the earthquake axis and the set of adjustment parameters, c0, c1, ..., c i Characterize the set of adjustment parameters, c i represents the i-th parameter in the adjustment parameter set, and l represents the cumulative continuous length of the earthquake axis.

[0089] Furthermore, the adjustment parameter optimization module 50 iteratively updates the adjustment parameter set through a preset optimization algorithm, and the execution steps include:

[0090] According to the influence relationship between the adjustment parameter set and the cumulative continuous length of the seismic axis, the update gradient of the target optimization function is calculated; based on the update gradient, the adjustment parameter set is iteratively updated according to the preset learning rate, and the formula is as follows: Among them, t represents the number of iterations, represents the value of the parameter in the tth iteration, Characterizes the parameter c in the t+1th iteration i The value of α represents the preset learning rate. Characterizes the updated gradient of the target optimization function, which represents the target optimization function F with respect to the parameter c i The partial derivative of .

[0091] Further, according to the influence relationship between the adjustment parameter set and the cumulative continuous length of the seismic axis, the update gradient of the target optimization function is calculated. Before that, the adjustment parameter optimization module 50 is also used to perform the following steps:

[0092] Based on the seismic data body, a sample adjustment parameter set of the sample structural model is extracted, and the corresponding sample seismic axis cumulative continuous length set is extracted; based on a neural network, the sample adjustment parameter set is used as input, and the sample seismic axis cumulative continuous length set is used as output to construct an adjustment parameter influence relationship analyzer; based on the adjustment parameter influence relationship analyzer, the influence relationship between the adjustment parameter set and the seismic axis cumulative continuous length is generated.

[0093] Further, the parameter optimization module 50 is adjusted to determine whether a preset convergence condition is reached, and the execution steps include:

[0094] After each round of iteration, obtain the adjustment parameter set of this round and the corresponding cumulative continuous length of the seismic axis of this round; compare the cumulative continuous length of the seismic axis of this round with the cumulative continuous length of the seismic axis of the previous round, and calculate the optimization increment; if the optimization increment is less than or equal to the preset optimization increment threshold, or the number of iterations reaches the preset iteration number threshold, it is determined that the preset convergence condition is met.

[0095] In a specific implementation, the adjustment parameter optimization module 50 iteratively updates the adjustment parameter set through a preset optimization algorithm, and the execution steps include:

[0096] In each round of iteration, a number of first adjustment parameters are randomly selected from the adjustment parameter set; based on the first adjustment parameters, a first update gradient of the target optimization function is calculated; based on the first update gradient, the adjustment parameter set is iteratively updated according to a preset learning rate, and the formula is as follows: Among them, s represents the number of iterations, Characterizes the parameter c in the sth iteration i The value of Characterizes the parameter c in the s+1th iteration i The value of α represents the preset learning rate. Characterize the first update gradient.

[0097] In a specific implementation, the adjustment parameter optimization module 50 iteratively updates the adjustment parameter set through a preset optimization algorithm, and the execution steps include:

[0098] The Hessian matrix of the target optimization function is calculated, and the adjustment parameter set is iteratively updated through a second-order optimization strategy. The formula is as follows: Among them, r represents the number of iterations, Characterizes the parameter c in the rth iteration i The value of Characterizes the parameter c in the r+1th iteration i The value of Represents the updated gradient of the target optimization function, H(F) represents the Hessian matrix of the target optimization function, and represents the second-order derivative information of F, [H(F)] -1 It is the inverse matrix of the Hessian matrix, which is used to adjust the gradient direction.

[0099] Furthermore, the system described in the embodiment of the present application is also used to perform the following steps:

[0100] The target optimization function and the preset optimization algorithm are encapsulated to generate an automated structural model adjustment module, wherein the automated structural model adjustment module automatically imports the seismic data volume and the initial structural model and outputs an optimized structural model.

[0101] Through the above detailed description of the construction model adjustment method based on artificial intelligence in this specification, those skilled in the art can clearly understand the construction model adjustment system based on artificial intelligence in this embodiment. For the system disclosed in Example 2, since it corresponds to the method disclosed in Example 1 and has corresponding functional modules and beneficial effects, the relevant parts can be referred to the description of the method part.

[0102] Embodiment three, based on the same inventive concept of the construction model adjustment method based on artificial intelligence in the aforementioned embodiment one, the present application also provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by the processor, the various steps of the above-mentioned construction model adjustment method based on artificial intelligence are implemented, and the same technical effect can be achieved. To avoid repetition, it will not be repeated here.

[0103] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present application. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to the embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A structural model adjustment method based on artificial intelligence, characterized in that: The method comprises: Acquire an initial structural model, and generate an adjustment parameter set based on a seismic data volume, wherein the seismic data volume includes a plurality of geological characteristic data; Constructing an adjustment function, wherein the adjustment function receives the adjustment parameter set as input, adjusts the initial construction model, and generates an adjusted construction model; Constructing a fit evaluation function, wherein the fit evaluation function receives the adjusted structural model as an input, calculates the cumulative continuous length of the seismic axis based on the adjusted structural model, and uses the cumulative continuous length of the seismic axis as a fit indicator; Combining the adjustment function and the fit evaluation function to generate a target optimization function, wherein the optimization goal of the target optimization function is to maximize the cumulative continuous length of the earthquake axis; Iteratively updating the adjustment parameter set through a preset optimization algorithm to optimize the target optimization function, obtaining the cumulative continuous length of the optimized seismic axis after each round of iteration, and judging whether a preset convergence condition is met; When the preset convergence condition is reached, the optimized seismic axis cumulative continuous length is output as the optimal seismic axis cumulative continuous length, and the corresponding optimal adjustment parameter set is output to obtain an optimized structural model.

2. The artificial intelligence-based structural model adjustment method according to claim 1, characterized in that: The objective optimization function is as follows: F(c0,c1,...,c i )=l; Among them, F(c0,c1,...,c i ) represents the target optimization function, which represents the relationship between the cumulative continuous length of the earthquake axis and the set of adjustment parameters, c0, c1, ..., c i Characterize the set of adjustment parameters, c i represents the i-th parameter in the adjustment parameter set, and l represents the cumulative continuous length of the earthquake axis.

3. The artificial intelligence-based structural model adjustment method according to claim 1, characterized in that: The iterative updating of the adjustment parameter set by using a preset optimization algorithm includes: Calculating the updated gradient of the target optimization function according to the influence relationship between the adjustment parameter set and the cumulative continuous length of the seismic axis; Based on the update gradient, the adjustment parameter set is iteratively updated according to a preset learning rate, and the formula is as follows: Among them, t represents the number of iterations, represents the value of the parameter in the tth iteration, Characterizes the parameter c in the t+1th iteration i The value of α represents the preset learning rate. Characterizes the updated gradient of the target optimization function, which represents the target optimization function F with respect to the parameter c i The partial derivative of .

4. The artificial intelligence-based structural model adjustment method according to claim 3, characterized in that: The step of calculating the updated gradient of the target optimization function according to the influence relationship between the adjustment parameter set and the cumulative continuous length of the seismic axis comprises: Based on the seismic data volume, a sample adjustment parameter set of the sample structural model is extracted, and a corresponding sample seismic axis cumulative continuous length set is extracted; Based on a neural network, the sample adjustment parameter set is used as input, and the sample earthquake axis cumulative continuous length set is used as output to construct an adjustment parameter influence relationship analyzer; Based on the adjustment parameter influence relationship analyzer, an influence relationship between the adjustment parameter set and the cumulative continuous length of the seismic axis is generated.

5. The artificial intelligence-based structural model adjustment method according to claim 3, characterized in that: The determining whether a preset convergence condition is met includes: After each round of iteration, the adjustment parameter set of this round and the corresponding cumulative continuous length of the earthquake axis of this round are obtained; Compare the cumulative continuous length of the earthquake axis in the current round with the cumulative continuous length of the earthquake axis in the previous round, and calculate the optimized increment; If the optimization increment is less than or equal to a preset optimization increment threshold, or the number of iterations reaches a preset number of iterations threshold, it is determined that the preset convergence condition is met.

6. The artificial intelligence-based structural model adjustment method according to claim 1, characterized in that: The iterative updating of the adjustment parameter set by a preset optimization algorithm further includes: In each round of iteration, a number of first adjustment parameters are randomly selected from the adjustment parameter set; Based on the first adjustment parameters, calculating a first update gradient of the objective optimization function; Based on the first update gradient, the adjustment parameter set is iteratively updated according to a preset learning rate, and the formula is as follows: Among them, s represents the number of iterations, Characterizes the parameter c in the sth iteration i The value of Characterizes the parameter c in the s+1th iteration i The value of α represents the preset learning rate. Characterize the first update gradient.

7. The artificial intelligence-based structural model adjustment method according to claim 1, characterized in that: The iterative updating of the adjustment parameter set by a preset optimization algorithm further includes: The Hessian matrix of the target optimization function is calculated, and the adjustment parameter set is iteratively updated through a second-order optimization strategy. The formula is as follows: Among them, r represents the number of iterations, Characterizes the parameter c in the rth iteration i The value of Characterizes the parameter c in the r+1th iteration i The value of Represents the updated gradient of the target optimization function, H(F) represents the Hessian matrix of the target optimization function, and represents the second-order derivative information of F, [H(F)] -1 It is the inverse matrix of the Hessian matrix, which is used to adjust the gradient direction.

8. The artificial intelligence-based structural model adjustment method according to claim 1, characterized in that: The method further comprises: The target optimization function and the preset optimization algorithm are encapsulated to generate an automated structural model adjustment module, wherein the automated structural model adjustment module automatically imports the seismic data volume and the initial structural model and outputs an optimized structural model.

9. The structural model adjustment system based on artificial intelligence is characterized by: The system is used to execute the artificial intelligence-based construction model adjustment method according to any one of claims 1 to 7, comprising: A model initialization module, used to obtain an initial structural model and generate an adjustment parameter set based on a seismic data body, wherein the seismic data body includes a plurality of geological characteristic data; An initial model adjustment module, used for constructing an adjustment function, wherein the adjustment function receives the adjustment parameter set as input, adjusts the initial construction model, and generates an adjusted construction model; A fit evaluation module, used to construct a fit evaluation function, wherein the fit evaluation function receives the adjusted structural model as input, calculates the cumulative continuous length of the seismic axis based on the adjusted structural model, and uses the cumulative continuous length of the seismic axis as a fit indicator; A target optimization function generation module, used for combining the adjustment function and the fit evaluation function to generate a target optimization function, wherein the optimization target of the target optimization function is to maximize the cumulative continuous length of the seismic axis; An adjustment parameter optimization module is used to iteratively update the adjustment parameter set through a preset optimization algorithm, optimize the target optimization function, obtain the cumulative continuous length of the optimized seismic axis after each round of iteration, and determine whether a preset convergence condition is met; The optimized adjustment parameter output is used to output the optimized seismic axis cumulative continuous length as the optimal seismic axis cumulative continuous length when the preset convergence condition is reached, and output the corresponding optimal adjustment parameter set to obtain the optimized structural model.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the construction model adjustment method based on artificial intelligence as described in any one of claims 1 to 8 are implemented.

Images