A multi-objective optimization method for intelligent well completion based on multi-objective genetic algorithm

Abstract

The present application relates to the technical field of well completion data processing, and particularly relates to a multi-objective optimization method for intelligent well completion based on a multi-objective genetic algorithm. The method comprises the following steps: obtaining a complete set of well completion parameters and a set of objective functions for intelligent well completion; performing dimension reduction on the complete set of well completion parameters to obtain a set of key well completion parameters; constructing a dynamic penalty factor generation rule; constructing a multi-level agent modeling system based on the set of key well completion parameters and the set of objective functions to obtain a hierarchical agent model; constructing an objective function evaluation system for multi-objective optimization to obtain fitness evaluation criteria; and performing multi-population parallel genetic evolution based on the hierarchical agent model and the fitness evaluation criteria to obtain a set of multi-objective optimization candidate solutions. The present application realizes an efficient, stable and globally optimized intelligent well completion parameter optimal configuration method by fusing dimension reduction, multi-objective collaborative evaluation, parallel evolution and a dynamic constraint mechanism.

Description

Technical Field

[0001] This invention relates to the field of well completion data processing technology, and in particular to an intelligent well completion multi-objective optimization method based on a multi-objective genetic algorithm. Background Technology

[0002] In the field of intelligent oil and gas well completion design, the selection and configuration of completion parameters are gradually moving towards multi-objective collaboration and intelligent optimization. Traditional well completion schemes are mostly based on engineering experience and single-objective evaluation for parameter configuration, making it difficult to balance multiple performance indicators such as production capacity, cost, risk, and system stability. Moreover, parameter selection is limited by subjective human judgment, which can easily lead to insufficient scheme optimization or imbalanced resource allocation. Existing intelligent well completion multi-objective optimization methods have the following key technical bottlenecks:

[0003] First, the failure to effectively integrate dimensional reduction and hierarchical modeling mechanisms resulted in bloated model structures or distorted responses.

[0004] Secondly, conventional genetic algorithms perform poorly in maintaining population diversity and handling constraints, lack population cooperation and adaptive evolution strategies, and are prone to getting trapped in local optima.

[0005] Third, traditional constraint mechanisms mostly rely on static penalty functions, which cannot dynamically respond to the degree of constraint violation during the optimization process, making it difficult to guarantee the feasibility and stability of the optimization results. Summary of the Invention

[0006] Therefore, it is necessary for the present invention to provide an intelligent well completion multi-objective optimization method based on a multi-objective genetic algorithm to solve at least one of the above-mentioned technical problems.

[0007] To achieve the above objectives, a smart well completion multi-objective optimization method based on a multi-objective genetic algorithm includes the following steps:

[0008] Step S1: Obtain the complete set of completion parameters and the set of objective functions for intelligent well completion; reduce the dimensionality of the complete set of completion parameters to obtain the set of key completion parameters; construct the dynamic penalty factor generation rule;

[0009] Step S2: Construct a multi-level surrogate modeling system based on the set of key completion parameters and the set of objective functions to obtain a hierarchical surrogate model; construct a multi-objective optimization objective function evaluation system to obtain a fitness evaluation standard;

[0010] Step S3: Perform multi-population parallel genetic evolution based on the hierarchical surrogate model and fitness evaluation criteria to obtain a multi-objective optimization candidate solution set;

[0011] Step S4: Use the dynamic penalty factor generation rule to constrain the candidate solution set of multi-objective optimization to obtain a feasible solution set;

[0012] Step S5: Perform Pareto boundary analysis and multi-objective sorting on the feasible solution set to obtain the optimal intelligent well completion parameter configuration scheme.

[0013] This invention significantly reduces the redundancy and computational complexity of the well completion parameter space by introducing a strategy that integrates dimensional reduction and hierarchical modeling. This allows for the effective extraction of key parameters while maintaining their physical meaning, effectively mitigating model overfitting and response instability caused by high-dimensional data. By constructing a hierarchical proxy modeling system, the nonlinear relationships between different levels of objectives and parameters in complex well completion systems are expressed in a structured manner, improving the accuracy of objective responses and modeling efficiency. Furthermore, by adopting a multi-objective collaborative evaluation standard, it ensures a balanced consideration of multiple objectives such as production capacity, cost, and risk control during the optimization process, enhancing the systematicness and rationality of performance evaluation. The multi-population parallel evolution mechanism and adaptive solution interaction strategy significantly enhance the global search capability and population diversity control capability of the population, effectively overcoming the local convergence problem caused by a single evolution path. The dynamic penalty factor mechanism adjusts the constraint processing intensity in real time according to the change of violation deviation at different stages, realizing dynamic identification and flexible intervention of constraint violation, and improving the feasibility judgment and stability control capability of the solution. Finally, the feasible solution set is optimized through multi-objective non-dominated analysis and multi-level ranking strategy, realizing the trade-off coordination among various objectives and the rapid extraction of the optimal configuration scheme. Overall, it significantly improves the scientificity, synergy and reliability of intelligent well completion parameter design in practical applications. Attached Figure Description

[0014] Other features, objects, and advantages of the invention will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings:

[0015] Figure 1 This is a flowchart illustrating the steps of an intelligent well completion multi-objective optimization method based on a multi-objective genetic algorithm according to the present invention.

[0016] Figure 2 for Figure 1 A detailed flowchart of step S1;

[0017] Figure 3 This is a diagram illustrating the parallel genetic evolution process of multiple populations in an embodiment of the present invention. Detailed Implementation

[0018] The technical method of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0019] Furthermore, the accompanying drawings are merely illustrative of the invention and are not necessarily drawn to scale. The same reference numerals in the drawings denote the same or similar parts, and therefore repeated descriptions of them will be omitted. Some block diagrams shown in the drawings are functional entities and do not necessarily correspond to physically or logically independent entities. These functional entities can be implemented in software, in one or more hardware modules or integrated circuits, or in different network and / or processor methods and / or microcontroller methods.

[0020] It should be understood that although the terms "first," "second," etc., may be used herein to describe various units, these units should not be limited by these terms. These terms are used merely to distinguish one unit from another. For example, without departing from the scope of the exemplary embodiments, a first unit may be referred to as a second unit, and similarly, a second unit may be referred to as a first unit. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0021] To achieve the above objectives, please refer to Figures 1 to 3 This invention provides an intelligent well completion multi-objective optimization method based on a multi-objective genetic algorithm, the method comprising the following steps:

[0022] Step S1: Obtain the complete set of completion parameters and the set of objective functions for intelligent well completion; reduce the dimensionality of the complete set of completion parameters to obtain the set of key completion parameters; construct the dynamic penalty factor generation rule;

[0023] Step S2: Construct a multi-level surrogate modeling system based on the set of key completion parameters and the set of objective functions to obtain a hierarchical surrogate model; construct a multi-objective optimization objective function evaluation system to obtain a fitness evaluation standard;

[0024] Step S3: Perform multi-population parallel genetic evolution based on the hierarchical surrogate model and fitness evaluation criteria to obtain a multi-objective optimization candidate solution set;

[0025] Step S4: Use the dynamic penalty factor generation rule to constrain the candidate solution set of multi-objective optimization to obtain a feasible solution set;

[0026] Step S5: Perform Pareto boundary analysis and multi-objective sorting on the feasible solution set to obtain the optimal intelligent well completion parameter configuration scheme.

[0027] Preferably, step S1 includes the following steps:

[0028] Step S11: Collect multi-source intelligent well completion related data to obtain the complete set of well completion parameters and the set of objective functions;

[0029] Step S12: Perform correlation analysis on the entire set of completion parameters based on the objective function set to obtain the initial feature importance ranking;

[0030] Step S13: Perform principal component analysis on the initial feature importance ranking to obtain the key principal component set;

[0031] Step S14: Perform feature space mapping transformation on the complete set of completion parameters based on the key principal component set to obtain a dimensionless parameter representation;

[0032] Step S15: Perform hierarchical clustering on the dimensionality-reduced parameter representation to obtain the parameter hierarchical grouping structure;

[0033] Step S16: Select a high-weight subset based on the parameter hierarchical grouping structure and objective function set to form a key completion parameter set;

[0034] Step S17: Construct an initial family of constraint functions based on the fitness evaluation criteria to obtain the constraint function definition set;

[0035] Step S18: Perform function transformation and weight adjustment on the constraint function definition set to obtain the dynamic penalty factor generation rule.

[0036] In this embodiment of the invention, multiple types of data acquisition devices deployed on the surface of oil and gas well sites, including wellbore pressure sensors, casing temperature sensors, flow meters, production testing instruments, and formation property analysis equipment, are used to collect no less than 20 original well completion-related parameter data, including wellbore pressure, casing temperature, formation porosity, permeability, completion fluid type, injection rate, packer location, and perforation density. Based on the historical intelligent well completion project operation results database and economic evaluation model, a set of objective functions is constructed, specifying four types of indicators: operation cost, production growth rate, reservoir damage control rate, and safety stability. The acquired complete set of well completion parameters and the set of objective functions are calculated using a bivariate matrix method based on Pearson correlation coefficient. The linear correlation coefficient between each well completion parameter and each objective indicator is calculated sequentially, and the absolute values ​​are weighted and averaged to obtain a comprehensive correlation score for the parameters. The parameters are then ranked from highest to lowest score to form an initial feature importance ranking. Using eigenvalues ​​greater than 1, the top 10 completion parameters were selected as principal component inputs. A covariance matrix was constructed, and eigenvalue decomposition was performed to obtain the top three principal component combinations with a cumulative contribution rate of no less than 85%, forming a key principal component set. This key principal component set was applied to all parameters in the original complete set of completion parameters through a linear combination mapping. Specifically, a 3×20-dimensional mapping weight matrix was constructed to perform a linear transformation, resulting in a 3-dimensional reduced parameter representation. An iterative clustering method based on Euclidean distance was used for the reduced parameter representation, with an initial cluster size of 3 and a maximum iteration count of 100. After each iteration, the within-cluster variance was calculated, and the convergence condition was checked. Finally, the reduced data was divided into three stable hierarchical groups, resulting in a parameter hierarchical grouping structure. Based on the ranking of the average relevance scores of the completion parameters to the objective function set in each group, the top two completion parameters in each group were selected, forming a key completion parameter set containing six parameters. Referring to the scoring limits and evaluation methods of the objective function in the fitness assessment criteria, and combining the engineering limit conditions summarized from failed cases in historical well completion operations, six initial constraint functions were constructed, including a maximum completion fluid pressure not exceeding 40 MPa, a casing temperature not exceeding 180℃, and a minimum distance of 1 meter between the packer and the upper boundary of the oil layer, forming a constraint function definition set. An exponential decay function was applied to each constraint function for intensity adjustment, constructing a penalty expression for each constraint. Based on the rule that the penalty intensity increases linearly with the number of iterations, the function penalty weight was set to an initial value of 0.1, with a maximum dynamic range of 0.8. A dynamic penalty factor generation rule was formed by gradually increasing the weight.

[0037] This invention introduces multi-source data acquisition and quantification processing methods to ensure the comprehensiveness of well completion parameter information and the engineering adaptability of the objective function system, providing a high-quality input foundation for subsequent optimization. Through correlation analysis between parameters and the objective function, the system identifies key factors that significantly affect the performance response of multiple objectives, reducing the dimensionality burden of the feature space. Principal component extraction and mapping transformation are employed to improve the efficiency of information expression and correlation compression capabilities among the original parameters, significantly reducing computational resource consumption and enhancing the physical consistency of data expression. Combined with hierarchical clustering, the dimensionality-reduced parameters are grouped according to structural features, effectively preserving the structural coupling relationships between parameters during the optimization process, enhancing the hierarchical clarity and interpretability of subsequent modeling. A key subset selection strategy further strengthens the ability to extract objective-oriented parameters, achieving accurate mapping of parameter sensitivity differences among multiple objectives. Simultaneously, a constraint function definition system is constructed, enabling the optimization process to have comprehensive constraint response capabilities. Combined with function morphology reconstruction and dynamic weight adjustment mechanisms, a penalty control system that can automatically adjust with the evolutionary process is built, significantly improving the ability to distinguish infeasible solutions and dynamic adaptability, providing a rigorous constraint mechanism and refined parameter support for high-quality optimization throughout the entire process.

[0038] Preferably, step S2, which involves constructing a multi-level proxy modeling system based on the set of key completion parameters and the set of objective functions, includes:

[0039] The key completion parameter set is stratified and classified by feature to obtain completion layer structure data;

[0040] Based on the well completion layer structure data and the objective function set, sub-models for each layer are constructed to obtain a set of surrogate model substructures.

[0041] Model fusion is performed on the set of substructures of the proxy model to obtain a unified hierarchical proxy model.

[0042] In this embodiment of the invention, the key completion parameter set is input into the feature hierarchical classification module. Parameter hierarchies are performed using a dual-factor partitioning method based on attribute scale and effective area. The attribute scale dimension includes three categories: "continuous," "discrete," and "categorical." The effective area dimension is divided into three categories: "wellbore segment," "formation segment," and "surface segment." After locating each parameter in two-dimensional coordinates, they are grouped according to clustering boundary conditions to obtain the well completion hierarchical structure data. The wellbore segment includes parameters such as casing inner diameter and packer position; the formation segment includes formation porosity and permeability; and the surface segment includes injection rate and completion fluid density. The number of parameters in each segment is limited to 2 to 4. Based on the above hierarchical structure data and objective function set, a series of equally spaced parameter combination samples are constructed in the value space of each parameter subset using finite difference approximation. By consulting field experimental reports, operating curves, and engineering experience calculation formulas, the corresponding objective function response values ​​are extracted to form sample input and output pairs. An input-output mapping structure is then established for each layer, forming a surrogate model substructure set. After obtaining multiple substructures, a fusion operation is performed. First, the outputs of each substructure are uniformly normalized using a min-max linear normalization method to map the response values ​​to the [0,1] interval. Then, a three-dimensional response matrix (dimensions of sample number × number of sub-models × number of objective functions) is constructed to summarize the normalized outputs of each substructure. Next, a cross-layer mapping operation is performed with the objective function number as the index dimension, and the response values ​​of each objective function in different substructures are weighted and averaged. The weights are set inversely proportional to the mean absolute error of the corresponding substructure on the previous validation set. Finally, the weighted fusion results of all objective functions are arranged and combined according to the sample number to form a unified response vector set, and an input-output pair relationship between the key completion parameter set and the unified response vector is established, thereby obtaining a unified hierarchical surrogate model.

[0043] This invention introduces a feature-based hierarchical classification mechanism to structure and organize well completion parameters with significant differences in their application areas and attribute features. This makes the modeling process more physically logical and interpretable, effectively improving the ability to express the correlation between parameters. The separate establishment of sub-models at each level makes the mapping relationship between local features and the objective function more refined, avoiding the decline in generalization ability caused by parameter interference in global modeling, and improving the accuracy of response fitting and the ability to predict boundary behavior. After normalization and unification, the results of the multi-substructure are fused to fully integrate the response advantages provided by information from different levels, enhancing the robustness and response consistency of the overall fitting of the objective function, and reducing the impact of abnormal inputs on output stability. The resulting unified hierarchical proxy structure has hierarchical adaptability and overall convergence, and can provide more accurate and balanced evaluation references for different schemes in the genetic evolution stage, significantly improving the screening efficiency and convergence speed of global solutions in multi-objective optimization.

[0044] Preferably, model fusion of the proxy model substructure set includes:

[0045] Perform output consistency analysis on the set of substructures of the proxy model to obtain normalized model response data;

[0046] A cross-layer fusion mapping structure is constructed based on the model response normalized data to obtain the structure mapping matrix;

[0047] Inter-layer information aggregation is performed on the structure mapping matrix to obtain the aggregated output representation;

[0048] By combining aggregated output representations with a set of objective functions to perform joint regression modeling, a unified hierarchical proxy model is obtained.

[0049] In this embodiment of the invention, based on multiple well completion layered substructure proxy models, the output prediction results of each substructure model are first read sequentially on a unified test sample set (the number of samples is not less than 100, the sample dimension is equal to the number of key well completion parameters, and the number of key well completion parameters does not exceed 20) to form an original response result matrix. This matrix is ​​then normalized column-wise using standard deviation normalization, and a maximum-minimum value interval linear stretching method is used to normalize all model output results to the closed interval [0,1], obtaining normalized model response data. Based on the well completion parameter hierarchy information corresponding to each substructure model, a cross-layer fusion mapping structure is established. Specifically, the model output results under the same level are linearly combined row-wise, and the model results across levels are layered. The weighted fusion method sets weights based on the normalized weight values ​​of the corresponding level objectives in the objective function set (the normalized weight precision is set to three decimal places, and the sum of the weights equals 1), thereby constructing a structural mapping matrix. This structural mapping matrix is ​​then subjected to matrix row and column summation operations in hierarchical order to perform inter-level information aggregation, extracting the aggregated output representations of each level to form a fusion feature set. Using this fusion feature set as input and the multi-objective response values ​​from the objective function set as output, a joint regression expression is established using a multiple linear regression method based on the least squares error criterion. The input-output correspondence is clear, each objective function corresponds to a set of regression coefficients, and the regression residual is controlled within 5%. Finally, a unified hierarchical surrogate model is generated.

[0050] This invention effectively eliminates response bias caused by inconsistent target response scales among substructures through output consistency analysis and normalization, improving the comparability of outputs at each layer under a unified evaluation system. The construction of a cross-layer fusion mapping structure enables effective transmission and integration of information between different layers, enhancing the collaborative expressive ability of each substructure in predicting the objective function. The inter-layer aggregation operation of the structure mapping matrix further strengthens the complementarity of local response features between different layers, enabling the overall output to maintain detailed differences while possessing stronger global consistency. The final aggregated output representation is associated with the set of objective functions through joint regression, giving the unified hierarchical proxy model a comprehensive mapping capability for the influence of multi-source and multi-scale parameters. This significantly improves the stability, response continuity, and prediction accuracy of multi-objective evaluation under complex constraints, providing more reliable performance support for subsequent optimal solution selection and evolution.

[0051] Preferably, the objective function evaluation system for multi-objective optimization in step S2 includes:

[0052] Based on the set of objective functions, a hierarchical structure of objective functions is constructed to obtain the objective function weight mapping data;

[0053] The objective function weight mapping data is normalized to obtain a standardized objective weight set;

[0054] A family of fitness evaluation functions for multi-objective optimization is constructed based on the standardized objective weight set and objective function set, and a fitness evaluation standard is obtained.

[0055] In this embodiment of the invention, based on a set of objective functions, the engineering significance of each objective function in intelligent well completion operations is functionally decomposed. These functions are grouped and categorized according to three dimensions: "production efficiency," "operational cost," and "risk control," constructing a three-layer nested hierarchical structure of objective functions to form a multi-objective hierarchical index table. The number of objective functions under each dimension shall not exceed five, and the naming of the objective functions shall remain consistent with the original data. Initial weight values ​​are assigned to each objective function according to the hierarchical structure. The initial allocation method is based on an engineering experience scoring table, using an integer table of 1 to 9. The scoring results within each layer are used to construct a pairwise comparison matrix and verify consistency with the maximum eigenvalue through AHP (Analytic Hierarchy Process), generating objective function weight mapping data. This weight mapping... The emission data is normalized by dividing each function weight by the sum of all function weights to ensure that the sum of all objective function weights is 1, and that the decimal precision is uniformly retained to three decimal places. After constructing the standardized objective weight set, an evaluation sub-function is defined for each objective function based on the standardized weights and the objective function set. The evaluation sub-function adopts a weighted linear scoring method, specifically, the objective response value is multiplied by the corresponding standardized weight coefficient to obtain the individual score value of each objective function. All scores are weighted and summed to obtain the total evaluation score of a single well completion parameter combination, which constitutes the fitness evaluation function set for multi-objective optimization. This evaluation function set is used as a unified metric for measuring the quality of each solution in population evolution, forming the fitness evaluation standard.

[0056] This invention, by constructing a hierarchical structure of the objective function, clarifies the subordinate relationships and importance of each optimization objective, avoids interference from multi-objective conflicts in the optimization results, and enhances the organizational logic of the overall evaluation. Weight normalization ensures that each objective is evaluated on a unified numerical scale, effectively preventing evaluation bias caused by differences in the original index dimensions and improving the fairness of fitness calculations for each objective. The establishment of fitness evaluation criteria provides quantifiable evaluation criteria for each optimization scheme during iteration, enhancing not only the ability to distinguish between superior and inferior schemes during population evolution but also ensuring the dynamic response capability of the convergence path to different objective trade-offs, thereby improving the overall efficiency, stability, and decision rationality of multi-objective optimization.

[0057] Preferably, step S3 includes the following steps:

[0058] Step S31: Construct an initial multi-population genetic coding system based on the hierarchical surrogate model and fitness evaluation criteria to obtain the population initialization parameter set;

[0059] Step S32: Perform selection, crossover, and mutation operations in parallel for each population in the population initialization parameter set, and judge their quality by combining the fitness evaluation criteria to obtain the fitness data of the first generation of multi-objective populations;

[0060] Step S33: Adjust the fitness evaluation criteria based on the fitness data of the first-generation multi-objective population to obtain the updated fitness function;

[0061] Step S34: Perform the solution exchange operation for each population in parallel evolution according to the preset migration frequency based on the updated fitness function to obtain population migration update data;

[0062] Step S35: Perform multiple rounds of parallel evolutionary iterations based on the updated fitness penalty function and population migration update data to obtain a multi-objective optimization candidate solution set.

[0063] In this embodiment of the invention, based on the hierarchical surrogate model and fitness evaluation criteria, each parameter in the key completion parameter set is encoded using integer encoding. The encoding precision is set to three decimal places, and the encoding range is strictly limited to the upper and lower limits of the parameter's physical parameters. A multi-population genetic encoding structure is constructed, wherein the number of parallel populations is fixed at 4, and each population contains no less than 50 groups of individuals. The population initialization process is generated by uniform distribution sampling to form a population initialization parameter set. Five rounds of individual screening, crossover and recombination, and perturbation transformation operations are performed on each parallel population. The specific process includes: first, performing a roulette wheel selection on all individuals in the population, retaining no less than 60% of the individuals; then, performing a crossover and recombination operation on the selected individuals in pairs, with the crossover method being single-point crossover and the crossover probability set to 0.9; next, applying a perturbation transformation operation to the crossover individuals, with the perturbation method being Gaussian perturbation, the perturbation mean being 0, the perturbation standard deviation being set to 5% of the initial parameter range, and the transformation probability being set to 0.1. After completion, the fitness evaluation criteria are used to calculate the value of each transformed individual. The total evaluation score is categorized and summarized by population ID to form the first generation of multi-objective population fitness data. Based on the fitness data, the weights of each objective function in the original evaluation criteria are fine-tuned using a proportional offset adjustment, with an offset range not exceeding ±0.03. This adjustment is based on the variance data of the objective function scores in the previous generation; objective functions with larger variances have their weights increased, while those with smaller variances have their weights decreased, generating an updated fitness function. According to the updated fitness function and the unified population migration mechanism, a solution interaction operation is performed on the four parallel populations every three generations. The interaction method is the exchange of optimal solutions between populations, with five sets of solutions exchanged each time. The results after the exchange are merged with the original population and re-sorted, retaining the top 50 sets to form the population migration update data. Based on the updated fitness function and the population migration update data, a parallel evolutionary iteration process of no less than 20 rounds is executed. Each round repeats the above selection, crossover, perturbation, evaluation, and migration operations, ultimately converging to obtain a multi-objective optimization candidate solution set where the total objective function score reaches a stable threshold (score change rate less than 1% for five consecutive generations).

[0064] This invention enhances the coverage of the solution space and reduces the risk of getting trapped in local optima by setting up multiple populations for parallel evolution, thereby improving global search capabilities. Introducing fitness evaluation criteria during genetic operations such as selection, crossover, and mutation helps to increase solution diversity while preserving superior individuals, enhancing the evolutionary vitality of the population structure. The dynamic adjustment mechanism of the fitness evaluation criteria allows the optimization direction to be adaptively corrected based on the performance of each generation of the population, enhancing the responsiveness of the search strategy to changes in the objective function. The inter-population solution transfer mechanism under a preset migration frequency enables the co-evolution of solutions, effectively promoting the sharing of excellent features across different evolutionary paths and improving the convergence speed of solutions. Introducing an updated fitness penalty function in multiple rounds of evolutionary iteration further strengthens the screening control of solutions that do not meet the constraints, thereby improving the overall quality of the candidate solution set in terms of multi-objective consideration and constraint satisfaction, providing a solid foundation for ultimately selecting a highly practical intelligent well completion parameter configuration.

[0065] Preferably, step S32 includes the following steps:

[0066] Step S321: Perform individual selection operation on the population initialization parameter set to obtain the population selection result set;

[0067] Step S322: Perform genetic crossover operation based on the population selection result set to obtain the crossover and recombination individual set;

[0068] Step S323: Perform mutation perturbation operation on the crossover and recombination individual set to obtain the genetically mutated individual set;

[0069] Step S324: Based on the genetic variation individual set and fitness assessment criteria, judge the quality of individuals to obtain individual fitness score data;

[0070] Step S325: Based on the individual fitness score data and the dynamic penalty factor generation rule, perform constraint penalty correction to obtain the corrected fitness dataset;

[0071] Step S326: Group and summarize the corrected fitness dataset according to the population label to obtain the first generation multi-objective population fitness data.

[0072] In this embodiment of the invention, for each population initialization parameter set, a roulette-style probability screening operation is performed. The roulette weight is based on the individual's score under the initial fitness assessment criteria. The retention ratio is set to 60% of the original population's number of individuals, and the remaining individuals are discarded and do not participate in subsequent operations, resulting in a population selection result set. Individuals in each selection result set are grouped into pairs in sequence, and a single-point crossover operation is performed. Specifically, for each pair of individuals, the parameter at the same position is selected as the crossover point. The crossover position is uniformly drawn from the 3rd to the 7th position, generating a new crossover and recombination individual set. During the crossover operation, the crossover probability is fixed at 0.9 to maintain the number of individuals after crossover consistent with the original population. A perturbation operation is performed on each crossover and recombination individual set. The perturbation method is Gaussian perturbation, with the perturbation mean set to 0 and the standard deviation set to 5% of the original parameter definition interval. With a perturbation probability of 0.1, parameter values ​​exceeding the original physical boundary after perturbation are truncated to within the boundary range, resulting in a genetically mutated individual set. Using fitness evaluation criteria, weighted scores for each objective function are calculated and summed for each genetically mutated individual to form the individual's total score data, which in turn constitutes the individual fitness score data. Based on the dynamic penalty factor generation rule, constraint condition checks are performed on each genetically mutated individual to obtain its corresponding constraint violation deviation value. This deviation value is then matched with a preset penalty mapping relationship to generate a penalty correction value. Finally, the original fitness score of each individual is subtracted from the corresponding penalty correction value to form a corrected fitness dataset. The corrected fitness dataset is grouped by population number, and the scores of all individuals within each group are categorized and summarized to form the first-generation multi-objective population fitness data.

[0073] This invention effectively preserves individuals with high fitness in the current population through individual selection operations, enhancing the continuity of superior genes and providing a stable foundation for subsequent population evolution. Genetic crossover operations promote the recombination of different characteristics within the population, expanding the search space and accelerating the diversity expansion of the population within the solution domain. Mutation perturbation operations further increase the diversity of solutions, strengthen the population's ability to explore unknown regions, and enhance the overall innovation of optimization. The introduction of fitness evaluation criteria in the individual merit judgment stage, assigning a clear evaluation value to each individual, helps to construct a unified evaluation scale and accurately identify potential optimal solutions. A dynamic penalty factor correction mechanism suppresses individuals that do not meet the constraints, effectively guiding the optimization process to avoid infeasible regions, thereby improving the overall constraint satisfaction of the population. Finally, the results are summarized and grouped based on population labels, giving the fitness evaluation results clear structural division characteristics and providing stable, high-resolution population fitness information for subsequent collaborative parallel evolution of multiple populations.

[0074] Preferably, step S4 includes the following steps:

[0075] Step S41: Evaluate the default degree of the candidate solution set for multi-objective optimization based on the dynamic penalty factor generation rule to obtain penalty score data;

[0076] Step S42: Perform threshold judgment on the penalty scoring data, remove solutions that exceed the limit, and obtain the feasible solution screening results;

[0077] Step S43: Extract a subset of candidate solutions based on the feasible solution screening results to obtain a feasible solution set.

[0078] In this embodiment of the invention, for a multi-objective optimization candidate solution set, the key completion parameter combinations corresponding to each solution are extracted, and all constraint condition parameters are extracted based on the constraint function definition set. Constraint condition checks are performed on each solution, specifically including boundary constraint determination (whether the physical upper and lower limits of the parameters are exceeded), process boundary checks (whether the safety boundaries of the completion process are met), and logical mutual exclusion checks (whether the process dependency logic between parameters is violated). The result of each type of constraint check is quantified as a binary value of 0 or 1. All constraint check results are then input into the dynamic penalty factor generation rule, and a dynamic weight is assigned to each default item. The dynamic weight is adjusted based on the current score volatility of the associated targets in the objective function set. The score volatility calculation formula is: σ i / μ i , where σ i Let μ be the standard deviation of the score of objective function i in the current population. i The average value is used to form a comprehensive weight value for the default deviation of each solution. This value is combined with the original fitness score to construct a penalty scoring expression. The expression is used to calculate the final penalty score for each candidate solution, resulting in penalty scoring data. A threshold judgment operation is performed on the penalty scoring data, setting the maximum allowable default score to 10% of the total score of the objective function. Solutions exceeding this proportion are considered out-of-limit solutions and are removed from the penalty scoring data. The remaining solutions that do not exceed the limit are marked as feasible solutions, forming the feasible solution screening results. Based on the feasible solution screening results, all solutions that pass the screening are extracted from the original multi-objective optimization candidate solution set to form the final feasible solution set.

[0079] This invention improves the accuracy of solution feasibility determination by introducing a dynamic penalty factor generation rule for default assessment, enabling the quantification of the deviation degree of each candidate solution under constraints. This mechanism implements differentiated scoring based on the default magnitude of solutions, strengthening the ability to distinguish between mild and severe default schemes and enhancing the flexibility of penalty intensity control. Setting threshold judgment criteria ensures that the screening process has clear judgment criteria, avoiding the risk of misjudgment caused by fuzzy decision-making and improving the stability and repeatability of feasible solution screening. The elimination of out-of-limit solutions directly reduces the consumption of subsequent computational resources by invalid solutions, improving overall optimization efficiency. Based on the screening results, a subset is extracted, and the final feasible solution set strictly conforms to physical constraints and engineering safety boundaries while ensuring multi-objective response performance, laying a solid foundation for subsequent optimization ranking and optimal solution extraction.

[0080] Preferably, step S41 includes the following steps:

[0081] Step S411: Extract constraints from the candidate solution set for multi-objective optimization to obtain constraint feature data of the solution set;

[0082] Step S412: Perform constraint deviation analysis based on the solution set constraint feature data and the objective function set to obtain solution set default deviation data;

[0083] Step S413: Perform dynamic penalty intensity mapping on the default deviation data of the solution set according to the dynamic penalty factor generation rule to obtain the penalty factor data of the solution set;

[0084] Step S414: Construct a default scoring expression based on the solution set penalty factor data and fitness evaluation criteria to obtain the solution set penalty scoring function;

[0085] Step S415: Calculate the solution set penalty scoring function for the candidate solution set of multi-objective optimization to obtain the penalty scoring data.

[0086] In this embodiment of the invention, for a multi-objective optimization candidate solution set, key completion parameter combinations are extracted for each solution, and parameter-level constraint extraction is performed according to the constraint function definition set. The constraint types include three categories: physical limits, process safety thresholds, and logical dependency constraints. Physical limits are determined by comparing upper and lower limits to see if they exceed the limits; process safety thresholds are confirmed by comparing with field experience threshold ranges to see if they exceed the limits; and logical dependencies are retrieved by using a defined table of mutual exclusion or inclusion rules to find conflicting relationships. All results are summarized to form solution set constraint feature data, in a two-dimensional matrix format. Each row corresponds to a candidate solution, and each column represents the test result value of a certain constraint. After binarization, a default matrix is ​​formed. The solution set constraint feature data is then subjected to joint analysis with the objective function set. Statistical analysis is used to calculate the deviation of each solution under each constraint dimension. The deviation is represented by a normalized deviation, expressed by the formula: (v...) i -bi ) / r i , where v i b represents the constraint value of the current solution. i To constrain lower or upper boundary values, r i The standard deviation of this constraint in historical data yields the default deviation data of the solution set, stored as a multi-dimensional deviation vector corresponding to the candidate solution number. Based on the dynamic penalty factor generation rule, a mapping operation is performed on each item of the default deviation data in the solution set. The mapping method is linear enhancement mapping, which multiplies each deviation by the score volatility coefficient of the objective function to which the current constraint belongs. The score volatility is calculated as σ. i / μ i , where σ i With μ i Let $j$ be the standard deviation and mean of the objective function $j$ in the current candidate solution set, respectively. The results are summarized to form the solution set penalty factor data; a larger value indicates a stronger penalty. Based on the solution set penalty factor data and the fitness evaluation criteria, a default scoring expression is defined, with the structure: F = F0 - ∑(w) k ×p k ), where F is the final score, F0 is the original fitness assessment score, and w k p is the standardized weight of the k-th objective function. k The constraint penalty factor value associated with the k-th objective function is expressed in the structure to ensure that the score is negatively correlated with the constraint penalty strength, thus forming the solution set penalty scoring function. The penalty scoring function is called for each set of solutions in the multi-objective optimization candidate solution set to calculate and output the penalty correction score value in sequence, and the result forms the penalty scoring data.

[0087] This invention ensures that the constraint attributes of each candidate solution are fully quantified before evaluation by extracting constraint conditions, thereby improving the clarity of the data structure and the relevance of subsequent analysis. The constraint deviation analysis process clarifies the degree of deviation of each solution on different physical constraint indicators, providing a precise basis for the subsequent penalty mechanism. The penalty factor data formed by associating deviation data with dynamic penalty factor generation rules makes the penalty intensity proportional to the severity of the breach, enhancing the difference and adaptability of breach identification. The introduction of fitness evaluation criteria to construct a breach scoring expression makes the scoring function goal-oriented, ensuring that the response to constraint violation and the overall optimization goal form a unified evaluation logic. The final output penalty scoring data has both numerical continuity and comparability, and can effectively distinguish between boundary feasible solutions and obviously breached solutions during the screening stage, thereby improving the accuracy, stability and computational efficiency of the entire feasibility evaluation system.

[0088] Preferably, step S5 includes the following steps:

[0089] Step S51: Perform objective function response extraction on the feasible solution set to obtain a multi-objective response dataset;

[0090] Step S52: Perform non-dominated solution identification operation based on the multi-objective response dataset to obtain a Pareto solution candidate set;

[0091] Step S53: Calculate the crowding distance of the Pareto solution candidate set to obtain Pareto boundary level data;

[0092] Step S54: Perform weighted sorting based on Pareto boundary level data and standardized target weight set to obtain multi-objective optimization sequence;

[0093] Step S55: Extract the first and second solutions from the multi-objective optimization sequence to obtain the optimal intelligent well completion parameter configuration scheme.

[0094] In this embodiment of the invention, based on the feasible solution set, key completion parameter combinations are read one by one and input into the hierarchical surrogate model for objective function response calculation to obtain the corresponding multi-objective response values. A two-dimensional response matrix is ​​constructed using the feasible solution number as an index, where each row represents a solution and each column corresponds to an objective function value, forming a multi-objective response dataset. The dataset is then compared pairwise according to the Pareto optimality criterion. For any two solutions i and j, if solution i is not inferior to solution j in all objective function values ​​and is superior to solution j in at least one objective, then i is determined to dominate j. All solutions not dominated by any other solution are marked as non-dominated solutions, and the set output is used as a Pareto solution candidate set. Based on the multi-objective response values ​​in the Pareto solution candidate set, congestion distance is calculated for each solution. The calculation method is as follows: for each objective function, the congestion distance is calculated separately according to the numerical values. The solutions are sorted in ascending order, and their positions within the current objective function are recorded. Then, the linear differences between adjacent distances of each solution on each objective function are accumulated to obtain the total crowding distance value. A larger value indicates that the solution is located in a sparse region. The calculation results form Pareto boundary level data. Using a standardized objective weight set, a weighted sorting operation is performed on each solution in the Pareto boundary level data. The sorting score is a single-value index obtained by multiplying each objective response value by its corresponding weight and summing the results. The crowding distance value is used as a secondary criterion; when sorting scores are consistent, the solution with the higher crowding is selected first, ultimately forming a multi-objective optimization sequence. The solution at the top of the sorting sequence is directly extracted, and its corresponding key completion parameter combination is the optimal intelligent completion parameter configuration scheme. This scheme simultaneously satisfies the constraints and has the best comprehensive performance under the multi-objective evaluation system.

[0095] The objective function response extraction process in this invention ensures a comprehensive quantification of the performance of each solution, providing a stable input basis for subsequent superiority / inferiority identification. The non-dominated solution identification mechanism can filter out the solution set that cannot be strictly surpassed by other solutions on any objective in the context of conflicting multiple objectives, effectively preserving balance and diversity. The calculation of crowding distance further characterizes the distribution density of solutions in the objective space, strengthens the differentiation of differences in the solution set structure, and makes the subsequent ranking more discriminative. The introduction of a standardized objective weight set for weighted ranking effectively integrates the priority information among multiple objectives, improving the rationality and pertinence of the ranking sequence in application decision-making. Finally, the top-ranked solution is extracted as the optimal configuration scheme, which not only ensures its comprehensive advantages in multiple objective dimensions, but also significantly improves the decision-making accuracy, feasibility, and global optimality of intelligent well completion operation parameter configuration.

[0096] Therefore, the embodiments should be considered as exemplary and non-limiting in all respects, and the scope of the invention is not limited by the foregoing description. Thus, all changes falling within the meaning and scope of the equivalents of the application are intended to be included within the scope of the invention.

[0097] The above description is merely a specific embodiment of the present invention, enabling those skilled in the art to understand or implement the invention. Various modifications to these embodiments will be readily 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 invention. Therefore, the present invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features of the invention herein.

Claims

1. A smart well completion multi-objective optimization method based on a multi-objective genetic algorithm, characterized in that, Includes the following steps: Step S1: Obtain the complete set of completion parameters and the set of objective functions for intelligent well completion; reduce the dimensionality of the complete set of completion parameters to obtain the set of key completion parameters; construct the dynamic penalty factor generation rule; Step S2: Construct a multi-level proxy modeling system based on the set of key completion parameters and the set of objective functions to obtain a hierarchical proxy model; construct a multi-objective optimization objective function evaluation system to obtain a fitness evaluation standard; wherein, the construction of the multi-level proxy modeling system based on the set of key completion parameters and the set of objective functions in step S2 includes: The key completion parameter set is stratified and classified by feature to obtain completion layer structure data; Based on the well completion layer structure data and the objective function set, sub-models for each layer are constructed to obtain a set of surrogate model substructures. Model fusion is performed on the set of proxy model substructures to obtain a unified hierarchical proxy model; the model fusion of the set of proxy model substructures includes: Perform output consistency analysis on the set of substructures of the proxy model to obtain normalized model response data; A cross-layer fusion mapping structure is constructed based on the model response normalized data to obtain the structure mapping matrix; Inter-layer information aggregation is performed on the structure mapping matrix to obtain the aggregated output representation; By combining the aggregated output representation with the set of objective functions to perform joint regression modeling, a unified hierarchical proxy model is obtained; Step S3: Perform multi-population parallel genetic evolution based on the hierarchical surrogate model and fitness evaluation criteria to obtain a multi-objective optimization candidate solution set; Step S4: Use the dynamic penalty factor generation rule to constrain the candidate solution set of multi-objective optimization to obtain a feasible solution set; Step S5: Perform Pareto boundary analysis and multi-objective sorting on the feasible solution set to obtain the optimal intelligent well completion parameter configuration scheme.

2. The intelligent well completion multi-objective optimization method based on multi-objective genetic algorithm according to claim 1, characterized in that, Step S1 includes the following steps: Step S11: Collect multi-source intelligent well completion related data to obtain the complete set of well completion parameters and the set of objective functions; Step S12: Perform correlation analysis on the entire set of completion parameters based on the objective function set to obtain the initial feature importance ranking; Step S13: Perform principal component analysis on the initial feature importance ranking to obtain the key principal component set; Step S14: Perform feature space mapping transformation on the complete set of completion parameters based on the key principal component set to obtain a dimensionless parameter representation; Step S15: Perform hierarchical clustering on the dimensionality-reduced parameter representation to obtain the parameter hierarchical grouping structure; Step S16: Select a high-weight subset based on the parameter hierarchical grouping structure and objective function set to form a key completion parameter set; Step S17: Construct an initial family of constraint functions based on the fitness evaluation criteria to obtain the constraint function definition set; Step S18: Perform function transformation and weight adjustment on the constraint function definition set to obtain the dynamic penalty factor generation rule.

3. The intelligent well completion multi-objective optimization method based on multi-objective genetic algorithm according to claim 1, characterized in that, Step S2 involves constructing an evaluation system for the objective function of multi-objective optimization, including: Based on the set of objective functions, a hierarchical structure of objective functions is constructed to obtain the objective function weight mapping data; The objective function weight mapping data is normalized to obtain a standardized objective weight set; A family of fitness evaluation functions for multi-objective optimization is constructed based on the standardized objective weight set and objective function set, and a fitness evaluation standard is obtained.

4. The intelligent well completion multi-objective optimization method based on multi-objective genetic algorithm according to claim 1, characterized in that, Step S3 includes the following steps: Step S31: Construct an initial multi-population genetic coding system based on the hierarchical surrogate model and fitness evaluation criteria to obtain the population initialization parameter set; Step S32: Perform selection, crossover, and mutation operations in parallel for each population in the population initialization parameter set, and judge their quality by combining the fitness evaluation criteria to obtain the fitness data of the first generation of multi-objective populations; Step S33: Adjust the fitness evaluation criteria based on the fitness data of the first-generation multi-objective population to obtain the updated fitness function; Step S34: Perform the solution exchange operation for each population in parallel evolution according to the preset migration frequency based on the updated fitness function to obtain population migration update data; Step S35: Perform multiple rounds of parallel evolutionary iterations based on the updated fitness penalty function and population migration update data to obtain a multi-objective optimization candidate solution set.

5. The intelligent well completion multi-objective optimization method based on multi-objective genetic algorithm according to claim 4, characterized in that, Step S32 includes the following steps: Step S321: Perform individual selection operation on the population initialization parameter set to obtain the population selection result set; Step S322: Perform genetic crossover operation based on the population selection result set to obtain the crossover and recombination individual set; Step S323: Perform mutation perturbation operation on the crossover and recombination individual set to obtain the genetically mutated individual set; Step S324: Based on the genetic variation individual set and fitness assessment criteria, judge the quality of individuals to obtain individual fitness score data; Step S325: Based on the individual fitness score data and the dynamic penalty factor generation rule, perform constraint penalty correction to obtain the corrected fitness dataset; Step S326: Group and summarize the corrected fitness dataset according to the population label to obtain the first generation multi-objective population fitness data.

6. The intelligent well completion multi-objective optimization method based on multi-objective genetic algorithm according to claim 1, characterized in that, Step S4 includes the following steps: Step S41: Evaluate the default degree of the candidate solution set for multi-objective optimization based on the dynamic penalty factor generation rule to obtain penalty score data; Step S42: Perform threshold judgment on the penalty scoring data, remove solutions that exceed the limit, and obtain the feasible solution screening results; Step S43: Extract a subset of candidate solutions based on the feasible solution screening results to obtain a feasible solution set.

7. The intelligent well completion multi-objective optimization method based on multi-objective genetic algorithm according to claim 6, characterized in that, Step S41 includes the following steps: Step S411: Extract constraints from the candidate solution set for multi-objective optimization to obtain constraint feature data of the solution set; Step S412: Perform constraint deviation analysis based on the solution set constraint feature data and the objective function set to obtain solution set default deviation data; Step S413: Perform dynamic penalty intensity mapping on the default deviation data of the solution set according to the dynamic penalty factor generation rule to obtain the penalty factor data of the solution set; Step S414: Construct a default scoring expression based on the solution set penalty factor data and fitness evaluation criteria to obtain the solution set penalty scoring function; Step S415: Calculate the solution set penalty scoring function for the candidate solution set of multi-objective optimization to obtain the penalty scoring data.

8. The intelligent well completion multi-objective optimization method based on multi-objective genetic algorithm according to claim 1, characterized in that, Step S5 includes the following steps: Step S51: Perform objective function response extraction on the feasible solution set to obtain a multi-objective response dataset; Step S52: Perform non-dominated solution identification operation based on the multi-objective response dataset to obtain a Pareto solution candidate set; Step S53: Calculate the crowding distance of the Pareto solution candidate set to obtain Pareto boundary level data; Step S54: Perform weighted sorting based on Pareto boundary level data and standardized target weight set to obtain multi-objective optimization sequence; Step S55: Extract the first and second solutions from the multi-objective optimization sequence to obtain the optimal intelligent well completion parameter configuration scheme.

Images