Model preprocessing method and system for lithology identification model

Abstract

The invention provides a model preprocessing method and system for a lithology identification model, and the method comprises the steps: obtaining a hyper-parameter range of a plurality of hyper-parameters in a pre-constructed lithology identification model, constructing a first preset number of individuals based on the parameter range of the plurality of hyper-parameters, completing the initialization of a population, updating the initialized population in a preset round; each update round comprises global optimization and local optimization; in the global optimization step, the position of the individual is updated based on the distance between the individual and the target individual, and the updated position of the individual is obtained; in the step of local optimization, the position of each individual in the global optimization group is optimized again based on the individual influence parameter group, and local optimization is completed; and after updating of a preset round, screening a final individual through the fitness, performing model training on the lithology identification model by applying hyper-parameters of the individual, and performing lithology identification after training is completed.

Description

Technical Field

[0001] The present invention relates to the technical field of lithology identification, and particularly to a method and system for preprocessing a lithology identification model. Background Art

[0002] Using machine learning models for lithology identification plays a crucial role in modern geological exploration and oil and gas resource development. Its importance is reflected in multiple dimensions, deeply affecting the efficiency, accuracy, and technological innovation of the industry. First of all, from the perspective of efficiency, traditional lithology identification mainly relies on the experience and expertise of geologists. This process is often time-consuming and laborious, and may be limited by personal subjective judgment and knowledge scope. Machine learning models, on the other hand, can automatically extract features from a large amount of logging data and, through algorithm optimization and learning, achieve rapid and accurate identification of different rock types. This not only greatly shortens the identification cycle, improves work efficiency, but also reduces the uncertainty brought by human factors, making the results of lithology identification more objective and reliable.

[0003] Secondly, in terms of accuracy, machine learning models can learn the complex relationships between rock types and logging responses by training a large amount of logging data with known lithologies, including the subtle differences that are difficult to capture by traditional methods. This powerful learning ability enables machine learning models to show higher accuracy in lithology identification, especially in the face of complex geological environments and the coexistence of multiple rock types, where its advantages are even more significant. This not only helps geologists understand the underground rock distribution more accurately, but also provides a solid foundation for subsequent oil and gas resource evaluation and development.

[0004] In the model construction stage, it is necessary to initially determine the model parameters. Most processing methods, when dealing with data with a hierarchical structure and certain natural inspiration, due to the mismatch between the parameter changes and the iteration process, the results tend to emphasize local search and ignore global search, resulting in the emergence of local optimum problems, thereby reducing the robustness of the entire model, directly affecting the parallelism and efficiency of model execution, and delaying the convergence speed of the model. Summary of the Invention

[0005] In view of this, embodiments of the present invention provide a method for preprocessing a lithology identification model to eliminate or improve one or more defects existing in the prior art.

[0006] One aspect of the present invention provides a method for preprocessing a lithology identification model. The steps of the method include: Obtain the hyperparameter ranges of multiple hyperparameters in a pre-constructed lithology identification model, construct a first preset number of individuals based on the parameter ranges of the multiple hyperparameters to complete the initialization of the population, and update the initialized population for a preset number of rounds; Include global optimization and local optimization in each update round; In the step of global optimization, calculate the fitness of individuals in the current population, screen a second preset number of individuals as target individuals based on the fitness of the individuals, and update the position of the individual based on the distance between the individual and the target individual to obtain the updated position of the individual; In the step of local optimization, calculate the individual influence parameter group of each individual after global optimization. The individual influence parameter group includes at least one individual influence parameter, and re-optimize the position of each individual in the globally optimized population based on the individual influence parameter group to complete the local optimization; After the update of the preset number of rounds, screen the final individuals through fitness. The lithology recognition model uses the hyperparameters of these individuals for model training, and performs lithology recognition after the training is completed.

[0007] Adopting the above solution, this solution can pre-select the hyperparameters of the lithology recognition model, and include two position optimization processes in each update round, namely global optimization and local optimization. In the process of global optimization, determine the target individual through fitness, and can continuously iteratively track the position of the target. Through the observation and analysis of the target, the purpose of accurately pursuing the target is achieved; in the process of local optimization, introduce the individual influence parameters of the individual, consider various aspects of the influence of the individual, and realize the optimization process of the state by appropriately adjusting these parameters, improve the behavior efficiency of the population, and improve the convergence speed.

[0008] In some embodiments of the present invention, the step of global optimization further includes: Taking the update of the position of the individual based on the distance between the individual and the target individual to obtain the updated position of the individual as the first updated position; Calculate the Gaussian distribution, calculate the second updated position of the individuals in the population based on the first updated position of the individuals in the population and the Gaussian distribution, and take the second updated position of the individual as the final position of the individual in the step of global optimization.

[0009] In some embodiments of the present invention, in the step of updating the position of the individual based on the distance between the individual and the target individual to obtain the updated position of the individual, calculate the step size of each individual relative to a target individual based on the position of the individual and the position of the target individual, and calculate the updated position of the individual based on the step size of the individual relative to each target individual.

[0010] In some embodiments of the present invention, in the step of calculating the step size of each individual relative to a target individual based on the position of the individual and the position of the target individual, and calculating the updated position of the individual based on the step size of the individual relative to each target individual, calculate the step size of the individual relative to a target individual based on the following formula: ; Among them, represents the step size of individual 1 relative to the target individual ; represents the position of the target individual ; represents the coefficient vector of individual 1; represents the distance between individual 1 and the target individual ;

[0011] Based on the following formula, calculate the updated position of an individual based on the step size of the individual relative to each target individual: ; Among them, represents the updated position of the individual, represents the superimposed value of the step sizes of the individual relative to each target individual, represents the number of target individuals.

[0012] In some embodiments of the present invention, in the step of calculating the second updated position of an individual in the population based on the first updated position of the individual in the population and the Gaussian distribution, and taking the second updated position of the individual as the final position of the individual in the step of global optimization, the following formula is used to calculate the second updated position: ; Among them, represents the second updated position, represents the first updated position, represents the preset maximum number of iterations, , , and γ are all preset constants, D represents the number of dimensions of the position vector, represents the value of a dimension in the position vector of the first updated position, represents the value of the corresponding dimension in the position vector of the target individual, represents the preset Gaussian mutation constant, represents the position vector of the target individual; represents the position vector of the first updated position; represents the Gaussian distribution.

[0013] In some embodiments of the present invention, in the step of calculating the second updated position of an individual in the population based on the first updated position of the individual in the population and the Gaussian distribution, and taking the second updated position of the individual as the final position of the individual in the step of global optimization, calculate the second updated position of an individual based on each target individual respectively, and calculate the average value of multiple second updated positions as the final position of the individual.

[0014] In some embodiments of the present invention, the individual influence parameters include a group attraction value, a dynamic inertia weight value, a group cohesion value, a target attraction value, and a hypothetical individual interference value. In the step of calculating the individual influence parameter group of each individual after global optimization, the group attraction value, the dynamic inertia weight value, the group cohesion value, the target attraction value, and the hypothetical individual interference value are calculated according to the following formulas: ; ; ; ; ; where S, A, C, F, and E respectively represent the group attraction value, the dynamic inertia weight value, the group cohesion value, the target attraction value, and the hypothetical individual interference value for an individual; X represents the current position of the calculated individual, represents the current position of any individual j other than the calculated individual in the current population; N represents the number of individuals other than the calculated individual in the current population; represents the current velocity of individual j; represents the attraction of the target individual to the calculated individual; represents the interference value of the hypothetical individual to the calculated individual.

[0015] In some embodiments of the present invention, in the step of re-optimizing the position of each individual in the globally optimized group based on the individual influence parameter group, the movement amount of the individual's re-movement is calculated based on the individual influence parameter group, and the target position of the individual is calculated based on the movement amount of the individual's re-movement.

[0016] In some embodiments of the present invention, in the step of calculating the movement amount of the individual's re-movement based on the individual influence parameter group, the movement amount of the individual's re-movement is calculated according to the following formula: ; where, represents the movement amount of the individual's re-movement; s, a, c, f, and e represent the calculation weights corresponding to the group attraction value, the dynamic inertia weight value, the group cohesion value, the target attraction value, and the hypothetical individual interference value; represents the distance between the current position of the individual and the final position of the previous round, represents the corresponding calculation weight.

[0017] The second aspect of the present invention further provides a model preprocessing system for a lithology identification model. The system includes a computer device, which includes a processor and a memory. Computer instructions are stored in the memory, and the processor is configured to execute the computer instructions stored in the memory. When the computer instructions are executed by the processor, the system implements the steps implemented by the method described above.

[0018] The third aspect of the present invention further provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, it implements the steps implemented by the aforementioned model preprocessing method for the lithology identification model.

[0019] The additional advantages, objectives, and features of the present invention will be partially elaborated in the following description, and will become partially apparent to those of ordinary skill in the art after studying the following text, or can be learned from the practice of the present invention. The objectives and other advantages of the present invention can be pointed out and obtained specifically in the specification and the accompanying drawings.

[0020] Those skilled in the art will understand that the objectives and advantages that can be achieved by the present invention are not limited to the above specific descriptions, and the above and other objectives that the present invention can achieve will be more clearly understood according to the following detailed description. Description of the Drawings

[0021] The drawings described herein are used to provide a further understanding of the present invention, form a part of this application, and do not limit the present invention.

[0022] Figure 1 It is a schematic diagram of an implementation manner of the model preprocessing method for the lithology identification model of the present invention; Figure 2 It is a schematic diagram of the result comparison between the scheme (GWO_FA) of Experimental Group 1 of the present invention and the Grey Wolf Optimization (GWO) algorithm of the prior art using Test Function 1; Figure 3 It is a schematic diagram of the result comparison between the scheme (GWO_FA) of Experimental Group 1 of the present invention and the Grey Wolf Optimization (GWO) algorithm of the prior art using Test Function 2; Figure 4 It is a schematic diagram of the result comparison between the scheme (GWO_FA) of Experimental Group 1 of the present invention and the Grey Wolf Optimization (GWO) algorithm of the prior art using Test Function 3; Figure 5 It is a schematic diagram of the result comparison between the scheme (GWO_FA) of Experimental Group 1 of the present invention and the Grey Wolf Optimization (GWO) algorithm of the prior art using Test Function 4; Figure 6 It is a schematic diagram of the result comparison between the scheme (GWO_FA) of Experimental Group 1 of the present invention and the Grey Wolf Optimization (GWO) algorithm of the prior art using Test Function 5; Figure 7 Schematic diagram of the result comparison between the solution (GWO_FA) of Experimental Group 1 of the present invention and the Grey Wolf Optimization algorithm (GWO) of the prior art using Test Function 6; Figure 8 Schematic diagram of the result comparison between the solution (GWO_FA) of Experimental Group 1 of the present invention and the Grey Wolf Optimization algorithm (GWO) of the prior art using Test Function 7; Figure 9 Schematic diagram of the result comparison between the solution (GWO_FA) of Experimental Group 1 of the present invention and the Grey Wolf Optimization algorithm (GWO) of the prior art using Test Function 8; Figure 10 Schematic diagram of the result comparison between the solution (GWO_FA_FWA) of Experimental Group 2 of the present invention, the solution (GWO_FA) of Experimental Group 1, and the Grey Wolf Optimization algorithm (GWO) of the prior art using Test Function 1; Figure 11 Schematic diagram of the result comparison between the solution (GWO_FA_FWA) of Experimental Group 2 of the present invention, the solution (GWO_FA) of Experimental Group 1, and the Grey Wolf Optimization algorithm (GWO) of the prior art using Test Function 2; Figure 12 Schematic diagram of the result comparison between the solution (GWO_FA_FWA) of Experimental Group 2 of the present invention, the solution (GWO_FA) of Experimental Group 1, and the Grey Wolf Optimization algorithm (GWO) of the prior art using Test Function 3; Figure 13 Schematic diagram of the result comparison between the solution (GWO_FA_FWA) of Experimental Group 2 of the present invention, the solution (GWO_FA) of Experimental Group 1, and the Grey Wolf Optimization algorithm (GWO) of the prior art using Test Function 4; Figure 14 Schematic diagram of the result comparison between the solution (GWO_FA_FWA) of Experimental Group 2 of the present invention, the solution (GWO_FA) of Experimental Group 1, and the Grey Wolf Optimization algorithm (GWO) of the prior art using Test Function 5; Figure 15 Schematic diagram of the result comparison between the solution (GWO_FA_FWA) of Experimental Group 2 of the present invention, the solution (GWO_FA) of Experimental Group 1, and the Grey Wolf Optimization algorithm (GWO) of the prior art using Test Function 6; Figure 16 Schematic diagram of the result comparison between the solution (GWO_FA_FWA) of Experimental Group 2 of the present invention, the solution (GWO_FA) of Experimental Group 1, and the Grey Wolf Optimization algorithm (GWO) of the prior art using Test Function 7; Figure 17 Schematic diagram of the result comparison between the solution (GWO_FA_FWA) of Experimental Group 2 of the present invention, the solution (GWO_FA) of Experimental Group 1, and the Grey Wolf Optimization algorithm (GWO) of the prior art using Test Function 8; Figure 18 Schematic diagram of the comparison results of the solution (GWO_FA_FWA_DA) of Experimental Group 3, the solution (GWO_FA_FWA) of Experimental Group 2, the solution (GWO_FA) of Experimental Group 1, and the Grey Wolf Algorithm (GWO) of the prior art using Test Function 1; Figure 19 Schematic diagram of the comparison results of the solution (GWO_FA_FWA_DA) of Experimental Group 3, the solution (GWO_FA_FWA) of Experimental Group 2, the solution (GWO_FA) of Experimental Group 1, and the Grey Wolf Algorithm (GWO) of the prior art using Test Function 2; Figure 20 Schematic diagram of the comparison results of the solution (GWO_FA_FWA_DA) of Experimental Group 3, the solution (GWO_FA_FWA) of Experimental Group 2, the solution (GWO_FA) of Experimental Group 1, and the Grey Wolf Algorithm (GWO) of the prior art using Test Function 3; Figure 21 Schematic diagram of the comparison results of the solution (GWO_FA_FWA_DA) of Experimental Group 3, the solution (GWO_FA_FWA) of Experimental Group 2, the solution (GWO_FA) of Experimental Group 1, and the Grey Wolf Algorithm (GWO) of the prior art using Test Function 4; Figure 22 Schematic diagram of the comparison results of the solution (GWO_FA_FWA_DA) of Experimental Group 3, the solution (GWO_FA_FWA) of Experimental Group 2, the solution (GWO_FA) of Experimental Group 1, and the Grey Wolf Algorithm (GWO) of the prior art using Test Function 5; Figure 23 Schematic diagram of the comparison results of the solution (GWO_FA_FWA_DA) of Experimental Group 3, the solution (GWO_FA_FWA) of Experimental Group 2, the solution (GWO_FA) of Experimental Group 1, and the Grey Wolf Algorithm (GWO) of the prior art using Test Function 6; Figure 24 Schematic diagram of the comparison results of the solution (GWO_FA_FWA_DA) of Experimental Group 3, the solution (GWO_FA_FWA) of Experimental Group 2, the solution (GWO_FA) of Experimental Group 1, and the Grey Wolf Algorithm (GWO) of the prior art using Test Function 7; Figure 25 Schematic diagram of the comparison results of the solution (GWO_FA_FWA_DA) of Experimental Group 3, the solution (GWO_FA_FWA) of Experimental Group 2, the solution (GWO_FA) of Experimental Group 1, and the Grey Wolf Algorithm (GWO) of the prior art using Test Function 8; Figure 26Schematic diagram of the comparison results of the solution of experimental group 1 (GWO_FA), the solution of experimental group 2 (GWO_FA_FWA), the solution of experimental group 3 (GWO_FA_FWA_DA) of the present invention and the solution of the prior art using test function 1; Figure 27 Schematic diagram of the comparison results of the solution of experimental group 1 (GWO_FA), the solution of experimental group 2 (GWO_FA_FWA), the solution of experimental group 3 (GWO_FA_FWA_DA) of the present invention and the solution of the prior art using test function 2; Figure 28 Schematic diagram of the comparison results of the solution of experimental group 1 (GWO_FA), the solution of experimental group 2 (GWO_FA_FWA), the solution of experimental group 3 (GWO_FA_FWA_DA) of the present invention and the solution of the prior art using test function 3; Figure 29 Schematic diagram of the comparison results of the solution of experimental group 1 (GWO_FA), the solution of experimental group 2 (GWO_FA_FWA), the solution of experimental group 3 (GWO_FA_FWA_DA) of the present invention and the solution of the prior art using test function 4; Figure 30 Schematic diagram of the comparison results of the solution of experimental group 1 (GWO_FA), the solution of experimental group 2 (GWO_FA_FWA), the solution of experimental group 3 (GWO_FA_FWA_DA) of the present invention and the solution of the prior art using test function 5; Figure 31 Schematic diagram of the comparison results of the solution of experimental group 1 (GWO_FA), the solution of experimental group 2 (GWO_FA_FWA), the solution of experimental group 3 (GWO_FA_FWA_DA) of the present invention and the solution of the prior art using test function 6; Figure 32 Schematic diagram of the comparison results of the solution of experimental group 1 (GWO_FA), the solution of experimental group 2 (GWO_FA_FWA), the solution of experimental group 3 (GWO_FA_FWA_DA) of the present invention and the solution of the prior art using test function 7; Figure 33 Schematic diagram of the comparison results of the solution of experimental group 1 (GWO_FA), the solution of experimental group 2 (GWO_FA_FWA), the solution of experimental group 3 (GWO_FA_FWA_DA) of the present invention and the solution of the prior art using test function 8; Figure 34 Schematic diagram of the architecture of this solution; Figure 35 Schematic diagram of the lithology identification confusion matrix results of this solution; Figure 36This is a graph showing the curve changes of the accuracy rate and loss value during the training process of the model of this solution. Detailed implementation manner

[0023] To make the objectives, technical solutions and advantages of the present invention more clear and understandable, the present invention will be further described in detail below in combination with the implementation manners and the accompanying drawings. Herein, the illustrative implementation manners of the present invention and their descriptions are used to explain the present invention, but do not limit the present invention.

[0024] Herein, it should also be noted that in order to avoid obscuring the present invention due to unnecessary details, only the structures and / or processing steps closely related to the solution according to the present invention are shown in the drawings, while other details less related to the present invention are omitted.

[0025] Introduction to the prior art: The grey wolf algorithm shows significant advantages in high-dimensional complex spaces. Especially in identifying key factors and constructing structured models with clear mapping relationships, it has been widely applied in many fields such as path planning, clustering analysis, feature selection, and power dispatching. In recent years, the research on the grey wolf algorithm has been continuously deepened, and many scholars have extended its theoretical framework and application scenarios. In addition, some researchers have combined optimization methods such as particle swarm optimization (PSO) and differential evolution (DE) to enhance the diversity of the grey wolf algorithm and its ability to jump out of local optima. In recent years, with the development of artificial intelligence and deep learning, the grey wolf algorithm has also been applied to neural network training, image processing, and complex engineering optimization problems, and has shown strong adaptability and application value.

[0026] Although the grey wolf algorithm has achieved success in many fields, it still faces many challenges in practical applications. First of all, in the process of multi-dimensional feature selection, the grey wolf algorithm cannot record the historical data of each individual iteration, thus affecting the acquisition of the global optimal solution. In addition, in task scheduling optimization, the traditional combination strategy has uncertainty, making it difficult for the algorithm to balance precise solution and probabilistic solution during calculation. At the same time, the requirements of homogeneous and heterogeneous forms are not satisfied simultaneously during the data processing process, and these factors together lead to algorithm optimization deviation and affect its solution accuracy. Secondly, in solving complex optimization problems, there is still room for improvement in the convergence speed and search efficiency of the grey wolf algorithm. Especially when facing high-dimensional and uncertain problems, it is easy to fall into local optima and affect the final optimization effect. In addition, the current research still lacks a systematic optimization method for how to improve the balance between the global exploration ability and local development ability of the grey wolf algorithm.

[0027] In response to the above problems, this solution proposes an improved optimization strategy for the Grey Wolf Algorithm. First, a holistic strategy framework is constructed using the decision hierarchy and execution hierarchy of the Grey Wolf Algorithm, and the absorption parameter and mate list mechanism of the Firefly Algorithm are introduced to record historical information, thereby dynamically adjusting parameters during the search process to ensure the consistency between path exploration and the target location, while avoiding falling into local optima. Second, to meet the requirements of discrete multi-processor task scheduling, the Fireworks Algorithm is integrated to enhance the stability and flexibility of the algorithm during execution. In addition, by calculating the complexity of parameter configuration, the crossover and mutation operations are optimized to improve the effectiveness of the algorithm and reduce the performance degradation caused by premature convergence. Finally, this paper combines the optimization processes of the Firefly Algorithm and the Dragonfly Algorithm, and through the synergistic effect of the two meta-heuristic algorithms, the exploration and exploitation capabilities of the Grey Wolf Algorithm are enhanced. A comparative optimization method is introduced to further improve the quality of the solution, accelerate the convergence speed, and improve the search efficiency. Generally speaking, this optimization strategy can effectively make up for the deficiencies of the traditional Grey Wolf Algorithm and provide a better solution for solving high-dimensional complex problems.

[0028] Thus, it can be seen that there is a bottleneck problem in how to break through the Grey Wolf Algorithm in the optimization and solution process. Although the current optimization methods have improved the computational performance of the algorithm to a certain extent, it is still difficult to find the optimal solution in high-dimensional complex problems. Especially in key fields such as task scheduling, feature selection, and resource allocation, the solution accuracy and adaptability of the existing algorithms still need to be improved. Therefore, the optimization strategy proposed in this solution not only focuses on the balance between the search ability and exploitation ability of the Grey Wolf Algorithm, but also combines the advantages of multiple meta-heuristic algorithms to achieve a more intelligent and efficient solution process. In addition, the improved algorithm of this solution can adapt to more complex optimization environments, including non-linear, multi-constrained, and dynamically changing optimization problems, providing theoretical support and practical basis for further expanding the application scenarios of the Grey Wolf Algorithm.

[0029] As Figure 1 shown, the present invention proposes a model preprocessing method for a lithology identification model, and the steps of this method include: Step S100, obtain the hyperparameter ranges of multiple hyperparameters in the pre-constructed lithology identification model, and construct a first preset number of individuals based on the parameter ranges of the multiple hyperparameters to complete the initialization of the population; And update the initialized population for a preset number of rounds; Each update round includes global optimization and local optimization; Step S210, in the step of global optimization, calculate the fitness of the individuals in the current population, screen a second preset number of individuals as target individuals based on the fitness of the individuals, and update the position of the individual based on the distance between the individual and the target individual to obtain the updated position of the individual; Step S220. In the local optimization step, calculate the individual influence parameter group of each individual after global optimization. The individual influence parameter group includes at least one individual influence parameter. Based on the individual influence parameter group, re-optimize the position of each individual in the globally optimized population to complete the local optimization. Step S300. After a preset number of rounds of updates, select the final individual through fitness screening. The lithology identification model uses the hyperparameters of this individual for model training, and performs lithology identification after the training is completed.

[0030] The model of this solution uses a BiLSTM model.

[0031] Adopting the above solution, this solution can pre-select the hyperparameters of the lithology identification model, and each update round includes two position optimization processes, namely global optimization and local optimization. In the process of global optimization, determine the target individual through fitness, and can continuously iteratively track the position of the target. Through the observation and analysis of the target, the purpose of accurately pursuing the target is achieved; in the process of local optimization, introduce the individual influence parameters of the individual, consider various aspects of the influence of the individual, and realize the optimization process of the state by appropriately adjusting these parameters, improve the behavior efficiency of the population, and increase the convergence speed.

[0032] In some embodiments of the present invention, the steps of global optimization further include: Update the position of the individual based on the distance between the individual and the target individual to obtain the updated position of the individual as the first updated position; Calculate the Gaussian distribution, calculate the second updated position of the individuals in the population based on the first updated position of the individuals in the population and the Gaussian distribution, and use the second updated position of the individual as the final position of the individual in the global optimization step.

[0033] In some embodiments of the present invention, in the step of updating the position of the individual based on the distance between the individual and the target individual to obtain the updated position of the individual, calculate the step size of each individual relative to a target individual based on the position of the individual and the position of the target individual, and calculate the updated position of the individual based on the step size of the individual relative to each target individual.

[0034] In some embodiments of the present invention, in the step of calculating the step size of each individual relative to a target individual based on the position of the individual and the position of the target individual, and calculating the updated position of the individual based on the step size of the individual relative to each target individual, calculate the step size of the individual relative to a target individual based on the following formula: ; where represents the step size of individual 1 relative to the target individual of, Indicates the target individual 's position, represents the coefficient vector of individual 1; represents the distance between individual 1 and the target individual ;

[0035] Based on the following formula, calculate the updated position of an individual based on the step size of the individual relative to each target individual: ; where, represents the updated position of the individual, represents the superimposed value of the step size of this individual relative to each target individual, represents the number of target individuals.

[0036] In the specific implementation process, the following formula can be used to calculate the updated position of the individual: t represents the current iteration round.

[0037] The calculation method of, can also be: ; ; .

[0038] Specifically, is used to simulate the exploration behavior of grey wolves when searching for prey. In this paper, randomness is introduced to force the individual to deviate from the current target position, so as to conduct global search and avoid the algorithm falling into the local optimal solution prematurely; represents the attraction, and each individual has its unique attraction, represents at and the distance between i and j at positions. The observation coefficient parameter is very important in determining the convergence speed and behavior of the FA algorithm, because it affects the change of attraction and theoretically affects its value ; but in most applications, it is usually from 0.01 to 100, = 1 and ∈[0, 1].

[0039] In the specific implementation process, traditional algorithms lack detailed descriptions in terms of the execution or state change of individuals, resulting in poor performance in distinguishing the effectiveness of individuals. Therefore, this paper introduces the fireworks algorithm to address this problem and make up for the deficiencies of traditional algorithms in expressing natural laws, which can comprehensively express the individual mutation, individual strength, the change range of individual iteration positions, the moving state, as well as the mapping rules and selection strategies of the algorithm.

[0040] First, according to the initialization of the population, indicators such as the intensity of state change, the threshold of state change, and the vector value of displacement are proposed, and a better fitness value generates a benign iterative calculation of the optimal number of individuals; secondly, according to the fitness value, the effectiveness and reliability of the population state change are determined; then, a best range affected by the state change is selected to accommodate individuals with relatively poor strength as much as possible, avoid the "premature aging" phenomenon of individuals, and give full play to the maximum ability of individuals with poor strength; finally, the intensity of state change and the threshold of state change are calculated to achieve the vector operation of individual displacement, as shown in the following formula: represents the value of the position vector of individual i in the k-th dimension. represents after i changes occur in the population state in the k-th dimension. represents 0 and a random number between.

[0041] This paper uses Gaussian distribution to express the pre-formed Gaussian mutation of individuals in any dimension of the population; in addition, the modulo operation rule is used to map individuals that cross the boundary back to the best range of the population to further improve the diversity of the population. The formula is as follows: In each iteration of the grey wolf algorithm, a part of the grey wolf individuals are selected for Gaussian mutation operation to achieve Gaussian mutation, and the formula is as follows: .

[0042] represents the value of the position vector of the Gaussian mutation individual generated by population i in the k-th dimension. n follows a Gaussian distribution with a mean of 1 and a variance of 1. and represent the upper and lower bounds of individual i in the k-th dimension respectively.

[0043] In some embodiments of the present invention, in the step of calculating the second updated position of an individual in the population based on the first updated position of the individual in the population and the Gaussian distribution, and taking the second updated position of the individual as the final position of the individual in the step of global optimization, To accurately express the behavior of the algorithm after determining the target position, this solution expresses the process of approaching the target by tracking the movement state of the positioning target. From the perspective of the group, gradually discover and use the parameters of individuals to optimize the static and dynamic behaviors of the group, analyze from the coverage range of the algorithm and the movement path of the group, and describe the implementation process of tracking the target at the current stage. Therefore, use From 2 linearly decreasing to 0 linearly decreasing is represented by the following formula: ; Where, represents the second updated position, represents the first updated position, represents the preset maximum number of iterations, , , and γ are all preset constants, D represents the number of dimensions of the position vector, represents the value of a dimension in the position vector of the first updated position, represents the corresponding dimension value in the position vector of the target individual, represents the preset Gaussian mutation constant, represents the position vector of the target individual; represents the position vector of the first updated position; represents the Gaussian distribution.

[0044] In some embodiments of the present invention, in the step of calculating the second updated position of an individual in the population based on the first updated position of the individual in the population and the Gaussian distribution, and taking the second updated position of the individual as the final position of the individual in the step of global optimization, calculate the second updated position of an individual based on each target individual respectively, and calculate the average value of multiple second updated positions as the final position of the individual.

[0045] This solution introduces the concept of group cohesion, avoids conflicts between adjacent individuals by separating the group from other individuals; at the same time, considers the speed of each individual from the perspective of group behavior, and uses the vector calculation of the group to represent the cohesion of the population with the central attraction, thereby calculating the attraction of the objective function and the interference ability to the target. Therefore, set the influence ranges of the group attraction value, dynamic inertia weight value, group cohesion value, target attraction value, and assumed individual interference value of each individual as the domain, and optimize the state by appropriately adjusting these parameters to improve the behavior efficiency of the population.

[0046] In some embodiments of the present invention, the individual influence parameters include a group attraction value, a dynamic inertia weight value, a group cohesion value, a target attraction value, and a hypothetical individual interference value. In the step of calculating the individual influence parameter group of each individual after global optimization, the group attraction value, the dynamic inertia weight value, the group cohesion value, the target attraction value, and the hypothetical individual interference value are calculated according to the following formulas: ; ; ; ; ; wherein S, A, C, F, and E respectively represent the group attraction value, the dynamic inertia weight value, the group cohesion value, the target attraction value, and the hypothetical individual interference value for an individual; X represents the current position of the calculated individual, represents the current position of any individual j other than the calculated individual in the current population; N represents the number of individuals other than the calculated individual in the current population; represents the current velocity of individual j; represents the attraction of the target individual to the calculated individual; represents the interference value of the hypothetical individual to the calculated individual.

[0047] In some embodiments of the present invention, in the step of re-optimizing the position of each individual in the global optimization group based on the individual influence parameter group, the movement amount of the individual's re-movement is calculated based on the individual influence parameter group, and the target position of the individual is calculated based on the movement amount of the individual's re-movement.

[0048] In some embodiments of the present invention, in the step of calculating the movement amount of the individual's re-movement based on the individual influence parameter group, the movement amount of the individual's re-movement is calculated according to the following formula: ; wherein, represents the movement amount of the individual's re-movement; s, a, c, f, and e represent the calculation weights corresponding to the group attraction value, the dynamic inertia weight value, the group cohesion value, the target attraction value, and the hypothetical individual interference value; represents the distance between the current position of the individual and the final position of the previous round, represents the corresponding calculation weight.

[0049] Experimental Example First, this project improves the gray wolf algorithm by introducing the description method of the attraction to the target itself in the firefly algorithm. The purpose is to force individuals to deviate from the current target position, avoid the algorithm falling into the local optimal solution prematurely, and thus enter the global search stage. For this experimental process, data generation and simulation are completed, and the experimental environment is Windows 11 and MATLAB R2023b.

[0050] This scheme uses a variety of test functions, as shown in Table 1 below: Table 1 Experimental group 1 The scheme of experimental group 1 is as follows: Obtain the hyperparameter ranges of multiple hyperparameters in the pre-constructed lithology recognition model, construct the first preset number of individuals based on the parameter ranges of multiple hyperparameters, complete the initialization of the population, and perform preset rounds of updates on the initialized population; Each update round includes global optimization and local optimization; In the steps of global optimization, calculate the fitness of individuals in the current population, screen the second preset number of individuals as target individuals based on the fitness of individuals, and update the position of the individual based on the distance between the individual and the target individual to obtain the updated position of the individual; Specifically, update the position of the individual based on the distance between the individual and the target individual to obtain the updated position of the individual as the first updated position, calculate the step size of each individual relative to a target individual based on the position of the individual and the position of the target individual, and calculate the updated position of the individual based on the step size of the individual relative to each target individual. The step size of an individual relative to a target individual is calculated using the following formula: ; Where represents the step size of individual 1 relative to the target individual , represents the position of the target individual , represents the coefficient vector of individual 1; represents the distance between individual 1 and the target individual .

[0051] The updated position of the individual is calculated based on the step size of the individual relative to each target individual using the following formula: ; Where represents the updated position of the individual, represents the superimposed value of the step sizes of the individual relative to each target individual, Indicates the number of target individuals; After the update in the preset round, the final individuals are screened by fitness, and the hyperparameters of the individuals are applied by the lithology identification model for model training, and lithology identification is performed after the training is completed.

[0052] Experimental group 2 In the global optimization step of the solution of experimental group 2 in the solution of experimental group 1, the calculation of the Gaussian distribution is added, and the second updated position of the individuals in the population is calculated based on the first updated position of the individuals in the population and the Gaussian distribution, and the second updated position of the individuals is used as the final position of the individuals in the global optimization step; The following formula is used to calculate the second updated position: ; Among them, Indicates the second updated position, Indicates the first updated position, Indicates the preset maximum number of iterations, , , and γ are all preset constants, D represents the number of dimensions of the position vector, Indicates the value of a dimension in the position vector of the first updated position, Indicates the corresponding value of the dimension in the position vector of the target individual, Indicates the preset Gaussian mutation constant, Indicates the position vector of the target individual; Indicates the position vector of the first updated position; Indicates the Gaussian distribution.

[0053] Experimental group 3 Based on the solution of experimental group 2, in the local optimization step, the individual influence parameter group of each individual after global optimization is calculated, and the individual influence parameter group includes at least one individual influence parameter, and the position of each individual in the globally optimized population is re-optimized based on the individual influence parameter group to complete the local optimization; The individual influence parameters include the group attraction value, the dynamic inertia weight value, the group cohesion value, the target attraction value and the assumed individual interference value. In the step of calculating the individual influence parameter group of each individual after global optimization, the group attraction value, the dynamic inertia weight value, the group cohesion value, the target attraction value and the assumed individual interference value are calculated according to the following formula: ; ; ; ; ; Among them, S, A, C, F, and E respectively represent the group attraction value, dynamic inertia weight value, group cohesion value, target attraction value, and assumed individual interference value for an individual; X represents the current position of the calculated individual, represents the current position of any individual j other than the calculated individual in the current population; N represents the number of individuals other than the calculated individual in the current population; represents the current velocity of individual j; represents the attraction of the target individual to the calculated individual; represents the interference value of the assumed individual to the calculated individual; Calculate the movement amount of the individual's next movement based on the individual influence parameter group, and calculate the target position of the individual based on the movement amount of the individual's next movement; calculate the movement amount of the individual's next movement based on the following formula: ; Among them, represents the movement amount of the individual's next movement; s, a, c, f, and e represent the calculation weights corresponding to the group attraction value, dynamic inertia weight value, group cohesion value, target attraction value, and assumed individual interference value; represents the distance between the current position of the individual and the final position of the previous round, represents the corresponding calculation weight.

[0054] Secondly, a random function is used to generate data, the population size is set to 50, the upper and lower bounds of the variables are set to [10, 10] and [-10, -10] respectively, and the fitness of the algorithm is simulated and calculated. The results are as follows Figures 2 to 9 shown. By introducing the attraction parameter in the firefly algorithm, the exploration behavior of the improved grey wolf algorithm when searching for prey is improved, and then the calculation strategy of the distance variable is improved (GWO_FA); the abscissa of the image is the number of iterations, increasing from left to right. The ordinate is the objective function value, and the position variable gradually converges to the optimal value as the number of iterations increases.

[0055] Overall, it can be seen that the improved algorithm GWO_FA of this scheme not only has a faster descent speed in the initial stage, but also reaches a lower fitness value within fewer iterations, indicating that its optimization efficiency and effect are better than the original GWO algorithm.

[0056] The , and of the test function all represent the input values of each test function.

[0057] Figure 2 and Figure 3All are the performances of the algorithm on unimodal functions, but Figure 2 the search space of the corresponding function is larger than Figure 3 the corresponding function. During the iteration process, the fitness values of the algorithm gradually decrease and finally tend to the minimum value. Therefore, the algorithm performs well on unimodal functions with different search spaces.

[0058] Figure 4 、 Figure 5 are the performances of the algorithm on unimodal and multimodal functions respectively. The GWO_FA algorithm shows a very fast convergence speed in Figure 4 and almost reaches a near-optimal fitness value within the first 5 iterations. In Figure 5 , the GWO_FA algorithm also shows fast convergence characteristics. However, compared with unimodal functions, it requires slightly more iterations to converge to the final stable value and shows certain volatility. This is because multimodal functions have multiple local optimal solutions, and the algorithm needs more iterations to avoid falling into local optima when exploring the solution space. Figure 5 The final fitness value stabilizes at around -7.89, indicating that the GWO_FA algorithm can also effectively find the global optimal solution on multimodal functions.

[0059] Figure 6 、 Figure 7 are the performances of the algorithm on different multimodal functions. Figure 6 In the function, it is required to find the global minimum among multiple local minima, which requires the algorithm to have good global search ability. For the Figure 7 function, the algorithm also needs to handle the complexity of exponential and cosine functions. From the results in Figure 6 、 Figure 7 , we can also see that compared with the GWO algorithm, the GWO_FA algorithm has better global search ability and performs better in handling the complexity of exponential and cosine functions.

[0060] Figure 8 、 Figure 9 are the performances of the algorithm on different multimodal functions. Figure 8 The function is a univariate function, while Figure 9 is a multivariate function. In these two figures, the GWO_FA algorithm shows fast convergence characteristics, and the fitness value drops rapidly in the first few iterations.

[0061] Secondly, for the improved GWO_FA algorithm mentioned above, the fireworks algorithm is further introduced to perform Gaussian mutation on individuals to expand the scope of state change and avoid the "premature aging" phenomenon of individuals. Simulations are carried out for this process. During the experiment, the population size, the upper and lower bounds of variables remain unchanged. The probability of Gaussian mutation is set to 0.1, and the initial standard deviation of the Gaussian distribution is set to 0.5. Random functions are used to generate data, and the position variables of individuals are simulated. The results are as follows Figures 10 to 17 As shown: In the simulation results, the circular curve is the unimproved Grey Wolf Optimization (GWO); the asterisk curve is the Gaussian mutation process with the introduction of the fireworks algorithm, which aims to perform Gaussian mutation operations on individuals in any dimension of the Grey Wolf Optimization algorithm to further improve the population diversity (GWO_FA_FWA).

[0062] It can be seen from the figure that GWO_FA_FWA generally outperforms GWO in terms of convergence speed, especially for multi-modal functions. Moreover, the algorithm GWO_FA_FWA has a very fast convergence speed in the initial stage of iteration, quickly approaching the optimal solution, and performs well on different types of objective functions. Although the fitness value of GWO_FA gradually decreases during the iteration process, its final fitness value is higher than that of GWO_FA_FWA. This indicates that the improved function GWO_FA did not find the optimal value in this experiment and was unable to effectively explore the solution space during the search process, resulting in the problem of algorithm instability.

[0063] In summary, the improved algorithm GWO_FA_FWA is more effective in dealing with complex problems and has good generality and adaptability.

[0064] In Figure 10 、 Figure 11 , the GWO_FA_FWA algorithm quickly reduces the fitness value in the initial iteration and maintains the lowest fitness value throughout the process, showing the optimal convergence performance.

[0065] According to Figure 12 、 Figure 13 , the GWO_FA_FWA algorithm has a faster convergence speed and better effect on unimodal and multi-modal functions compared to the previous GWO_FA.

[0066] Analyzing Figure 14 the curves, it can be seen that all three algorithms quickly reduce the fitness value in the initial iteration. However, the GWO_FA algorithm finally stabilizes at a relatively high fitness value. Compared with the GWO_FA_FWA algorithm, GWO_FA lacks an effective global search mechanism, which makes it difficult for it to escape from the local optimal solution. Therefore, the algorithm has poor stability and low final convergence performance. The GWO_FA_FWA algorithm, due to the introduction of the Gaussian mutation strategy, has enhanced global search ability. Therefore, in Figure 14 、 15A stable state closer to the global optimal solution is achieved within fewer iteration times in both cases.

[0067] In Figure 16 , 17 the GWO_FA algorithm still stabilizes at a relatively high fitness value, which further verifies the conclusion obtained above, that is, when dealing with functions involved in optimization problems with multiple local minima, a large search space or a high dimension, the GWO_FA algorithm has low efficiency in global search.

[0068] Finally, the dragonfly algorithm is introduced to consider the individual movement mode from the perspective of group behavior, and the attributes of individuals are adjusted through parameters such as (s, a, c, f, and e), thereby improving the efficiency of population behavior. During the experiment, the population size, the upper and lower bounds of variables remain unchanged, the Gaussian mutation probability and the initial standard deviation of the Gaussian distribution remain unchanged, and random functions are used to generate data for simulating the position variables of individuals. The results are as follows Figures 18 to 25 shown: In the simulation results, the circular curve is the unimproved Grey Wolf Optimization (GWO); combined with Figure 1 the simulation results, the asterisk curve is the algorithm (GWO_FA) obtained by introducing the parameter of attraction β in the Firefly Algorithm into the calculation strategy of the distance variable for the exploration behavior of grey wolves when searching for prey in the Grey Wolf Optimization algorithm; the triangle curve is the improved algorithm (GWO_FA_FWA _DA) after Gaussian mutation of individuals, introducing the dragonfly algorithm, considering the influence of group factors on individuals, and iterating the individual positions again.

[0069] Among them, the GWO algorithm found a better solution in the early stage, but there was no significant improvement in the subsequent iteration process. Compared with the GWO algorithm, the curve of the GWO_FA algorithm has a faster descent speed in the initial stage and tends to be stable after about 10 iterations, indicating that the improved algorithm can find a better solution faster. The curve of the GWO_FA_FWA _DA algorithm has the fastest convergence speed among all the above algorithms, reaching the lowest fitness value almost within the first 3 iterations and remaining the lowest throughout the iteration process. This experimental result shows its powerful optimization ability.

[0070] It can be seen that after 30 iterations, compared with other algorithms, the GWO_FA_FWA _DA algorithm obtains a better average fitness value, is closer to the global optimal value, and its performance is also better than other algorithms before improvement.

[0071] In Figure 18 , the GWO_FA_FWA_DA algorithm has a faster convergence speed, indicating that the algorithm achieves a better balance between global search and local search. While in Figure 19 although the GWO_FA_FWA_DA algorithm still has a fast convergence speed, compared withFigure 18 Compared with [the other algorithms], more iteration times are required to reach the steady state. This indicates that the algorithm has the ability to handle non - linear and non - convex characteristics.

[0072] In Figure 20 and Figure 21 the GWO_FA_FWA_DA algorithm still demonstrates fast convergence characteristics. The convergence speeds of GWO and GWO_FA are relatively slow, and there are slight fluctuations during the iteration process, but they also tend to lower fitness values eventually.

[0073] Figure 22 During the iteration process of the three algorithms in [a certain situation], there are slight fluctuations, but finally both the GWO_FA_FWA_DA algorithm and the GWO algorithm converge to relatively low fitness values, while the GWO_FA algorithm converges to a higher fitness value. Figure 23 During the convergence process in [a certain situation], the GWO_FA_FWA_DA algorithm maintains the relatively lowest fitness value, which proves that the GWO_FA_FWA_DA algorithm performs better when dealing with complex multimodal problems.

[0074] In Figure 24 and Figure 25 at the beginning, the GWO and GWO_FA algorithms are stable at a relatively high fitness value for a short period, and finally jump out of the local optimal solution during the iteration. While for the GWO_FA_FWA_DA algorithm, it converges to the lowest fitness value at the beginning, which indicates that the GWO_FA_FWA_DA algorithm has good convergence characteristics and is not easily trapped in the local optimal solution.

[0075] The grey wolf algorithm (GWO), the improved grey wolf algorithm (GWO_FA) in the experimental process, the particle swarm optimization algorithm (PSO), the fruit fly algorithm (FOA) and the final algorithm obtained in this paper (GWO_FA_FWA_DA) are selected for comparison. Random functions are used to generate data during the simulation process to simulate the convergence of the above algorithms on different functions. The results are as follows Figure 4 shown: The dotted line is the particle swarm optimization algorithm (PSO); the square curve is the fruit fly algorithm (FOA); and combined with Figures 26 to 33 the simulation results, the circular curve is the unimproved grey wolf algorithm (GWO); the star - shaped curve is the algorithm (GWO_FA) obtained by introducing the attraction β parameter in the firefly algorithm and improving the calculation strategy of the distance variable for the exploration behavior of grey wolves when searching for prey in the grey wolf algorithm; the triangular curve is the improved algorithm (GWO_FA_FWA _DA) obtained by performing Gaussian mutation on individuals, introducing the dragonfly algorithm, considering the influence of the group factor on individuals, and iterating the individual positions again.

[0076] As can be seen from the figure, the GWO_FA_FWA _DA algorithm usually reaches a lower fitness value within fewer iterations, and in many cases, it is close to the final fitness value in the early iterations; in the later iterations, the fitness value still further decreases. Moreover, the final algorithm in this paper reaches the lowest fitness value for different types of test functions, demonstrating good adaptability. However, GWO_FA has shown significant instability in previous simulations, and such problems also occurred in this simulation, further verifying the foregoing description.

[0077] Therefore, it can be concluded that compared with the original GWO algorithm and other algorithms, the final GWO_FA_FWA _DA algorithm has a fast convergence speed and is more stable in the process of finding the optimal solution. Whether it is a single-peak or multi-peak function, it can effectively find the optimal solution by continuously optimizing the quality of the solution.

[0078] In Figure 26 , Figure 27 the GWO_FA_FWA_DA algorithm drops rapidly in the first few iterations. The fitness values of the PSO algorithm are relatively high in both figures, and its optimization effect is not as good as that of other algorithms. The GWO and GWO_FA algorithms perform similarly in both figures and finally converge to lower fitness values. The FOA algorithm maintains the lowest fitness value throughout the iteration process, showing the best optimization effect.

[0079] Figure 28 The overall trend is similar to the above image. However, in Figure 29 the FOA has a slow convergence speed and finally stabilizes at around -4, performing the worst in the function test. This indicates that the multi-peak problem is likely to cause the FOA to get lost in local optimal solutions.

[0080] In Figure 30 the FOA algorithm has the slowest convergence speed, but it still maintains good optimization ability in the later stage of exploration and finally reaches the global optimal solution. The GWO_FA_FWA_DA algorithm proposed in this paper has a relatively good overall convergence effect, but like other algorithms, it falls into a local optimal solution close to the global optimal solution. In Figure 31 the GWO_FA_FWA_DA algorithm performs the best overall, rapidly decreasing in the initial stage and finally reaching the lowest fitness value and remaining stable. Followed by GWO_FA and GWO, the overall convergence speed is slightly lower than that of the GWO_FA_FWA_DA algorithm, but they finally converge to the global optimal solution.

[0081] In Figure 32 the GWO_FA algorithm finally converges to a higher fitness value, further proving the conclusion that the algorithm has low stability. Combining Figure 33As a result, the FOA algorithm has strong global search ability in the initial stage and can quickly find the area close to the optimal solution, which are around -0.8 and 101 respectively. However, in the later stage, the optimization accuracy is insufficient and it lacks efficient local search ability, and finally neither converges to the optimal solution.

[0082] In this paper, the lithology identification method based on GWO_FA_FWA_DA-BiLSTM is applied to the publicly available well logging dataset, which is specifically used for machine learning applications of underground reservoir lithology identification and classification. The dataset contains 7 attributes: _CAL (Caliper) well diameter; _GR (Gamma Ray) gamma ray; _SP (Spontaneous Potential) spontaneous potential logging; _LLD (Laterolog Deep) deep laterolog resistivity; _LLS (Laterolog Shallow) shallow laterolog resistivity; _AC (Acoustic) acoustic logging; _DEN (Density) density logging; _PEF (Photoelectric Factor) photoelectric absorption factor; Lith_Section formation lithology.

[0083] A total of 2385 data points, that is, 2385 well logging data, are used in this experiment. Among them, there are 1179 data points of mudstone, 383 data points of siltstone, 444 data points of silty mudstone, 344 data points of muddy siltstone, and 35 data points of oil shale. The prediction lithology is sampled and divided according to the training set: test set = 7:3. The experimental environment is Windows 11 and MATLAB R2023b.

[0084] The specific method process is shown in Figure 34 。

[0085] The experiment is run 5 times. Since the search range is relatively wide in the initial stage of the experiment, it is easy to cause the algorithm to lose its direction in high-dimensional search. Based on the initial experimental results, the range is gradually narrowed. Finally, the parameter settings are as follows: the number of gray wolf individuals is 20, the number of dragonfly individuals is 15, and the maximum number of iterations is 40 and 50 times respectively. The optimization ranges of the number of hidden layer neurons, the maximum number of training times, and the learning rate of the BiLSTM model by the optimization algorithm in this paper are [50, 150], [30, 80], and [0.005, 0.01] respectively. The intensity of Gaussian noise is also adjusted in the experiment: reduce sigma (the standard deviation value of Gaussian noise is 0.2) to avoid large-scale fluctuations. On this basis, the penalty function penalty is added to punish the solutions that deviate far from the current optimal solution and improve the convergence stability.

[0086] In the experiment, siltstone mudstone, mudstone, oil shale, siltstone, and argillaceous siltstone in the sample data are defined as 1, 2, 3, 4, and 5 respectively. Select CAL, GR, SP, LLD, LLS, DEN, PEF, and Lith_Section as input values, and the lithology identification result as the output value. Update the hyperparameter values of the BiLSTM model through GWO_FA_FWA _DA to obtain the lithology identification confusion matrix diagram based on GWO_FA_FWA _DA - BiLSTM as Figure 35 shown. It can be seen from Figure 35 that for the identification of siltstone mudstone by the model, 33 data points are correctly identified, with a 24.8% probability of being identified as mudstone, a 25.7% probability of being identified as siltstone, and a 19.3% probability of being identified as argillaceous siltstone; for the identification of mudstone, 302 data points are correctly identified and 49 data points are incorrectly identified, with 2.0%, 2.6%, and 9.4% probabilities of being identified as siltstone mudstone, siltstone, and argillaceous siltstone respectively; for the identification of oil shale, 3 data points are correctly identified and 7 data points are incorrectly identified; for the identification of siltstone, 91 data points are correctly identified and 21 data points are incorrectly identified, with 8.9%, 4.5%, 0.9%, and 4.5% probabilities of being identified as siltstone mudstone, mudstone, oil shale, and argillaceous siltstone respectively; for the identification of argillaceous siltstone, 52 data points are correctly identified and 79 data points are incorrectly identified. The results show that the model has good identification effects on mudstone and siltstone.

[0087] Figure 36 The upper part in shows the change trend of the accuracy rate on the training set. The solid line below shows the change trend of the loss value on the validation set. Since the algorithm is in the exploration stage in the initial stage, there are large fluctuations in the target value and the error value is relatively high. As the number of iterations increases, the accuracy curve finally stabilizes at about 75%. On the contrary, the error value gradually decreases, and finally the loss curve also tends to be stable, indicating that the optimization target gradually converges.

[0088] Specifically, when the algorithms of the prior art process data with a hierarchical structure and certain natural inspiration, due to the mismatch between the parameter changes and the iterative process, the algorithms tend to emphasize local search and ignore global search, resulting in the emergence of local optimum problems, thereby reducing the robustness of the entire algorithm, directly affecting the parallelism and efficiency of algorithm execution, and delaying the convergence speed of the algorithm. To address the above problems, this solution reasonably controls the exploration ability by setting balance parameters, and uses the position vector control coefficient in the process of controlling the movement mode through historical records. The purpose is to control the iterative process by changing the coefficient and absorb new parameters during the iterative process to balance the exploration ability of the algorithm. Secondly, the discretization mapping method of the continuous space is realized by setting the adaptive dimension to control the Gaussian mutation process of the population, accelerate the algorithm convergence speed, shorten the time to find the optimal task scheduling sequence with fitness as the calculation standard, and reduce the time complexity of the algorithm. And finally, the algorithm accuracy is improved through the robust process set backpropagation learning (OBL) algorithm. Finally, through four simulation experiments, it is analyzed that the algorithm proposed in this paper is significantly superior to the traditional algorithms in terms of fitness, convergence speed and performance. Based on multiple experiments, the publicly available logging data set is used, and this algorithm is used for experiments. The recognition rate is significantly better than that of the traditional BiLSTM model. It can be seen that the method designed in this solution has good effects compared with the traditional recognition methods.

[0089] For processing high-dimensional data with a hierarchical structure and certain natural inspiration, based on the characteristics of the grey wolf algorithm, this solution considers aspects such as distance, dimension, behavior process and the execution efficiency of the population, improves from aspects such as global search, target change mode and multi-dimensional attributes of individuals, compares with the traditional algorithms, and is verified through three types of simulation methods. Finally, it is tested with actual data sets, indicating that the algorithm proposed in this paper is superior to several traditional algorithms in terms of fitness, convergence speed and performance.

[0090] The embodiment of the present invention also provides a model preprocessing system for a lithology recognition model. The system includes a computer device, the computer device includes a processor and a memory, the memory stores computer instructions, and the processor is used to execute the computer instructions stored in the memory. When the computer instructions are executed by the processor, the system implements the steps implemented by the method described above.

[0091] An embodiment of the present invention further provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps implemented by the model preprocessing method of the foregoing lithology identification model are realized. The computer-readable storage medium may be a tangible storage medium, such as a random access memory (RAM), memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, register, floppy disk, hard disk, removable storage disk, CD-ROM, or any other form of storage medium well known in the technical field.

[0092] Those of ordinary skill in the art should understand that the various exemplary components, systems, and methods described in conjunction with the embodiments disclosed herein can be implemented in hardware, software, or a combination of both. Specifically, whether to implement in hardware or software depends on the specific application and design constraints of the technical solution. Professional technicians can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of the present invention. When implemented in hardware, it can be, for example, an electronic circuit, an application-specific integrated circuit (ASIC), appropriate firmware, a plug-in, a functional card, and so on. When implemented in software, the elements of the present invention are programs or code segments used to execute the required tasks. The program or code segment can be stored in a machine-readable medium or transmitted through a data signal carried in a carrier wave on a transmission medium or a communication link.

[0093] It should be clear that the present invention is not limited to the specific configurations and processes described above and illustrated in the figures. For the sake of brevity, detailed descriptions of known methods are omitted here. In the above embodiments, several specific steps are described and illustrated as examples. However, the method process of the present invention is not limited to the specific steps described and illustrated. Those skilled in the art can make various changes, modifications, and additions, or change the order between steps after understanding the spirit of the present invention.

[0094] In the present invention, the features described and / or illustrated for one embodiment can be used in the same or similar manner in one or more other embodiments, and / or combined with the features of other embodiments or replace the features of other embodiments.

[0095] The above are only the preferred embodiments of the present invention and are not used to limit the present invention. For those skilled in the art, various changes and modifications can be made to the embodiments of the present invention. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A method for preprocessing a lithology identification model, characterized in that The steps of the method include: Obtain the hyperparameter ranges of multiple hyperparameters in a pre-built lithology identification model, construct an initial population of a first preset number of individuals based on the parameter ranges of the multiple hyperparameters, and perform a preset number of rounds of updates on the initial population; Each update round includes global optimization and local optimization; In the steps of global optimization, calculate the fitness of individuals in the current population, screen out a second preset number of individuals as target individuals based on the fitness of the individuals, update the position of the individual based on the distance between the individual and the target individual, and obtain the updated position of the individual; In the steps of local optimization, calculate the individual influence parameter group of each individual after global optimization, where the individual influence parameter group includes at least one individual influence parameter, and re-optimize the position of each individual in the globally optimized population based on the individual influence parameter group to complete local optimization; After the preset number of rounds of updates, screen the final individuals through fitness, and the lithology identification model uses the hyperparameters of the individuals for model training, and performs lithology identification after training is completed.

2. The model preprocessing method of the lithology identification model according to claim 1, characterized in that, The steps of the global optimization further include: Taking the update of the position of the individual based on the distance between the individual and the target individual and obtaining the updated position of the individual as the first updated position; Calculate the Gaussian distribution, calculate the second updated position of the individuals in the population based on the first updated position of the individuals in the population and the Gaussian distribution, and take the second updated position of the individual as the final position of the individual in the steps of global optimization.

3. The model preprocessing method of the lithology identification model according to claim 2, characterized in that In the step of updating the position of the individual based on the distance between the individual and the target individual and obtaining the updated position of the individual, calculate the step size of each individual relative to a target individual based on the position of the individual and the position of the target individual, and calculate the updated position of the individual based on the step size of the individual relative to each target individual.

4. The model preprocessing method of the lithology identification model according to claim 3, characterized in that, In the step of calculating the step size of each individual relative to a target individual based on the position of the individual and the position of the target individual and calculating the updated position of the individual based on the step size of the individual relative to each target individual, calculate the step size of the individual relative to a target individual based on the following formula: ； Among them, represents the step size of individual 1 relative to the target individual ; represents the position of the target individual ; represents the coefficient vector of individual 1; represents the distance between individual 1 and the target individual ; Calculate the updated position of the individual based on the step size of the individual relative to each target individual based on the following formula: Among them, represents the updated position of the individual, represents the superimposed value of the step sizes of the individual relative to each target individual, represents the number of target individuals.

5. The model preprocessing method of the lithology identification model according to claim 2, characterized in that, In the step of calculating the second updated position of the individuals in the population based on the first updated position of the individuals in the population and the Gaussian distribution and taking the second updated position of the individual as the final position of the individual in the steps of global optimization, calculate the second updated position using the following formula: ； Among them, represents the second update position, represents the first update position, represents the preset maximum number of iterations, 、 、 and γ are both preset constants, D represents the number of dimensions of the position vector, represents the value of a dimension in the position vector of the first update position, represents the value of the corresponding dimension in the position vector of the target individual, represents the preset Gaussian mutation constant, represents the position vector of the target individual; represents the position vector of the first update position; represents the Gaussian distribution.

6. The model preprocessing method of the lithology identification model according to claim 2, characterized in that, In the step of calculating the second updated position of the individuals in the population based on the first updated position of the individuals in the population and the Gaussian distribution and taking the second updated position of the individual as the final position of the individual in the steps of global optimization, calculate the second updated position of an individual based on each target individual respectively, and calculate the average value of the multiple second updated positions as the final position of the individual.

7. The model preprocessing method of the lithology identification model according to any one of claims 1 to 6, characterized in that The individual influence parameters include a group attraction value, a dynamic inertia weight value, a group cohesion value, a target attraction value, and an assumed individual interference value. In the step of calculating the individual influence parameter group of each individual after global optimization, the group attraction value, the dynamic inertia weight value, the group cohesion value, the target attraction value, and the assumed individual interference value are calculated according to the following formula: ； ； ； ； ； Among them, S, A, C, F, and E respectively represent the group attraction value, dynamic inertia weight value, group cohesion value, target attraction value, and hypothetical individual interference value for an individual; X represents the current position of the calculated individual, represents the current position of any individual j other than the calculated individual in the current population; N represents the number of individuals other than the calculated individual in the current population; represents the current velocity of individual j; represents the attraction of the target individual to the calculated individual; represents the interference value of the hypothetical individual to the calculated individual.

8. The method for preprocessing the lithology identification model according to claim 7, characterized in that, In the step of re-optimizing the position of each individual in the globally optimized group based on the individual influence parameter group, the movement amount of the individual's re-movement is calculated based on the individual influence parameter group, and the target position of the individual is calculated based on the movement amount of the individual's re-movement.

9. The method for preprocessing a lithology identification model according to claim 7, wherein In the step of calculating the movement amount of the individual's re-movement based on the individual influence parameter group, the movement amount of the individual's re-movement is calculated according to the following formula: ； Among them, represents the amount of movement for the individual to move again; s, a, c, f, and e represent the calculation weights for the corresponding group attraction value, dynamic inertia weight value, group cohesion value, target attraction value, and assumed individual interference value; represents the distance between the current position of the individual and the final position of the previous round, represents the corresponding calculation weight.

10. A model preprocessing system for a lithology identification model, characterized in that, The system includes a computer device, the computer device includes a processor and a memory, computer instructions are stored in the memory, the processor is configured to execute the computer instructions stored in the memory, and when the computer instructions are executed by the processor, the system implements the steps implemented by the method according to any one of claims 1 to 9.

Images