Oil field production allocation method and device based on deep learning algorithm

Abstract

The invention provides an oil field production allocation method and device based on a deep learning algorithm, and relates to the technical field of oil field exploitation, and the method comprises the steps: obtaining the number of blocks in an oil field, each oil well parameter included in each block, and the production parameter of each block, and the oil well parameters comprise an oil well number and a plurality of oil well exploitation parameter data; screening the plurality of oil well exploitation parameters to obtain at least one piece of sensitive parameter data; according to all the sensitive parameter data and a prediction model based on a time sequence, the yield of each oil well is obtained through prediction; constructing a yield objective function according to the yield of each oil well, and constructing a profit objective function and a fund investment objective function according to the production parameters of each block; and according to the yield objective function, the profit objective function, the capital investment objective function and a preset constraint function, constructing an oil and gas production allocation multi-objective optimization model, and solving the oil and gas production allocation multi-objective optimization model to obtain an oil field production allocation result.

Description

Technical Field

[0001] The present invention relates to the technical field of oilfield exploitation, and in particular to an oilfield production allocation method and device based on a deep learning algorithm. Background Art

[0002] With the increasing demand for oil in my country and the continuous growth of oil imports, the world energy supply and demand relationship directly affects my country's modernization construction. How to use limited oil resources to ensure the sustained and rapid growth of the national economy is a key issue that the oil industry urgently needs to solve. After years of exploitation, most of my country's oil fields are in the middle and late stages of exploitation. Insufficient new reserves, high difficulty in stabilizing production, shortage of funds and difficulty in raising funds, and serious aging of equipment are urgent problems to be solved. In view of the severe situation at home and abroad, it is necessary to have a correct understanding of the current status of my country's oil reserves and production allocation, reasonably predict oil production from the overall perspective, carry out scientific production allocation, and improve the effectiveness and scientificity of decision-making. This will provide certain guidance for the overall energy development direction of the country and promote the sustainable development of the oil industry. Oil and gas development is an international system engineering with the remarkable characteristics of high technology, high investment and high risk. In the work of oil and gas production allocation, due to the complexity of the actual implementation, it involves a wide range of factors and is affected by many factors. Moreover, oil and gas production allocation is closely related to the allocation of human, material and property resources in oil fields.

[0003] Domestic and foreign scholars have conducted a lot of exploration and research on oilfield production allocation, forming a series of theories and methods. There has also been a blowout in computer science. However, the current artificial intelligence methods cannot perform fast calculations when production does not conform to the actual situation. Therefore, there is an urgent need for a reasonable allocation of comprehensive resources and a way to optimize oilfield production allocation. Summary of the invention

[0004] The purpose of the present invention is to provide an oilfield production allocation method and device based on a deep learning algorithm to improve the above problems. In order to achieve the above purpose, the technical solution adopted by the present invention is as follows:

[0005] On the one hand, the present application provides an oilfield production allocation method based on a deep learning algorithm, comprising:

[0006] Obtaining the number of blocks in the oil field, the parameters of each oil well included in each block, and the production parameters of each block, wherein the oil well parameters include the oil well number and multiple oil well production parameter data;

[0007] Screening and processing a plurality of oil well production parameters to obtain at least one sensitive parameter data;

[0008] Based on all the sensitive parameter data and the time-series based prediction model, the production of each oil well is predicted;

[0009] Based on the production of each oil well, a production objective function is constructed, and based on the production start parameters of each block, a profit objective function and a capital investment objective function are constructed;

[0010] Based on the production objective function, the profit objective function, the capital investment objective function and the preset constraint function, a multi-objective optimization model for oil and gas production allocation is constructed, and the multi-objective optimization model for oil and gas production allocation is solved to obtain the oilfield production allocation result. The oilfield production allocation result is the result of the maximum total production of all blocks, the minimum capital investment of all blocks and the sum of the block profits of all blocks.

[0011] Furthermore, screening and processing of the production parameters of multiple oil wells are carried out to obtain at least one sensitive parameter data, including:

[0012] Taking the well number as the index value, the production parameters of the oil wells form an original data set;

[0013] Processing the data set for outliers and missing values to obtain an initial data set;

[0014] Performing dimensionless quantization processing on the initial data set to obtain a data set;

[0015] According to the grey relational analysis method, data screening is carried out on the parameters in the data set to obtain at least one sensitive parameter data.

[0016] Furthermore, according to the grey relational analysis method, data screening is carried out on the parameters in the data set to obtain at least one sensitive parameter data, including:

[0017] According to the grey relational analysis algorithm, pairwise analysis is carried out on the parameter data in the data set for the curves changing with time;

[0018] The parameter data corresponding to the similar curve shapes are used as the sensitive parameter data for production prediction.

[0019] Furthermore, based on all the sensitive parameter data and the time-series based prediction model, the production of each oil well is predicted, including:

[0020] According to the sliding window method, the sensitive parameter data are divided into a training set and a validation set;

[0021] According to the training set, the preset long short-term memory network model is trained to obtain a production prediction model;

[0022] According to the production prediction model, the validation set is predicted to obtain the prediction result of the validation set;

[0023] According to the prediction result of the validation set and the validation set, the prediction error is calculated;

[0024] If the prediction error is greater than the threshold, restart the division of the sensitive parameter data according to the sliding window method until the prediction error is less than the threshold.

[0025] Furthermore, construct a multi-objective optimization model for oil and gas production allocation according to the production target function, profit target function, capital investment target function and the preset constraint function, and solve the multi-objective optimization model for oil and gas production allocation to obtain the oilfield production allocation result, including:

[0026] Population initialization: Randomly generate an initial population with a size of N, and then use the deep learning model Transformer to evaluate the fitness of each individual in the population to obtain the fitness value of each individual;

[0027] Fast non-dominated sorting: Use the fast non-dominated sorting algorithm to sort each population, and divide each population individual into different non-dominated levels to distinguish the advantages and disadvantages of each solution;

[0028] Crowding degree assignment: Evaluate the distribution density of individuals in the solution space through crowding degree calculation to select individuals that maintain population diversity and avoid falling into local optimal solutions;

[0029] Population crossover and mutation: Generate new individuals using Transformer based on the fitness value, non-dominated level and crowding degree of individuals;

[0030] Repeat starting the fast non-dominated sorting until the number of iterations reaches the preset target to obtain the oilfield production allocation result.

[0031] On the other hand, the present application also provides an oilfield production allocation device based on deep learning and multi-objective optimization, including:

[0032] A data acquisition module for acquiring the number of blocks in the oilfield, each well parameter included in each block and the production start parameters of each block, where the well parameters include well numbers and multiple well production parameter data;

[0033] A data screening module for screening and processing multiple well production parameters to obtain at least one sensitive parameter data;

[0034] A production prediction module for predicting the production of each well according to all the sensitive parameter data and a time series-based prediction model;

[0035] A function construction module for constructing a production target function according to the production of each well, and constructing a profit target function and a capital investment target function according to the production start parameters of each block;

[0036] The production allocation model module is used to construct a multi-objective optimization model for oil and gas production allocation according to the production target function, profit target function, capital investment target function and preset constraint functions, and solve the multi-objective optimization model for oil and gas production allocation to obtain the oilfield production allocation result, where the oilfield production allocation result is the result of maximizing the total production of all blocks, minimizing the capital investment of all blocks and summing the block profits of all blocks.

[0037] Furthermore, the data screening module includes:

[0038] The data classification module is used to form the original data set of oil well production parameters with the well number as the index value;

[0039] The data cleaning module is used to process the data set for outliers and missing values to obtain the initial data set;

[0040] The data dimensionless module is used to perform dimensionless processing on the initial data set to obtain the data set;

[0041] The data extraction module is used to screen the parameters in the data set according to the grey relational analysis method to obtain at least one sensitive parameter data.

[0042] Furthermore, the data extraction module includes:

[0043] The curve calculation module is used to analyze the curves of the parameter data in the data set changing with time pairwise according to the grey relational analysis algorithm;

[0044] The curve judgment module is used to take the parameter data corresponding to the similar curve shapes as the sensitive parameter data for production prediction.

[0045] Furthermore, the production prediction module includes:

[0046] The data division module is used to divide the sensitive parameter data into a training set and a validation set according to the sliding window method;

[0047] The training module is used to train the preset long short-term memory network model according to the training set to obtain the production prediction model;

[0048] The validation module is used to predict the validation set according to the production prediction model to obtain the prediction result of the validation set;

[0049] The error calculation module is used to calculate the prediction error according to the prediction result of the validation set and the validation set;

[0050] The error judgment module is used to, if the prediction error is greater than the threshold, restart dividing the sensitive parameter data according to the sliding window method until the prediction error is less than the threshold.

[0051] Furthermore, the production prediction module includes:

[0052] An initialization module for population initialization: randomly generate an initial population of size N, and then use the deep learning model Transformer to evaluate the fitness of each individual in the population to obtain the fitness value of each individual;

[0053] A non-dominated sorting module for fast non-dominated sorting: use the fast non-dominated sorting algorithm to sort each population, and divide each population individual into different non-dominated levels to distinguish the advantages and disadvantages of each solution;

[0054] A crowding degree allocation module for crowding degree allocation: evaluate the distribution density of individuals in the solution space through crowding degree calculation to select individuals that maintain population diversity and avoid falling into local optimal solutions;

[0055] A crossover and mutation module for population crossover and mutation: generate new individuals using Transformer based on the fitness value, non-dominated level, and crowding degree of individuals;

[0056] A loop module for repeating the start of fast non-dominated sorting until the number of iterations reaches the preset target to obtain the oilfield production allocation result.

[0057] The beneficial effects of the present invention are as follows:

[0058] By combining the actual situation of the oilfield, the present invention takes single wells and blocks as the research objects respectively, constructs model training samples by filling missing values, processing outliers, feature standardization, and feature selection for the data required in single wells, and constructs a prediction model based on deep learning methods to accurately predict the production of single wells and blocks. At the same time, in the present invention, a multi-objective optimization model for oil and gas production allocation is established to assist in decision-making for production allocation. The three items of maximizing profit, maximizing production, and minimizing capital investment are used as the objective functions, and conditions such as production, cost, and reserve-production balance are used as constraints. A multi-objective solution algorithm is applied to solve the multi-objective model problem. Using the multi-objective model can comprehensively consider from the overall perspective of the company, and at the same time fully combine the characteristics of the local geology, exploration technology, and human and material resources, etc., adopt a scientific and effective production allocation plan, so as to reasonably allocate resources, maximize the oil and gas exploitation benefits, and provide an effective production allocation plan for the oilfield company. Description of the Drawings

[0059] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the drawings required in the embodiments. It should be understood that the following drawings only show some embodiments of the present invention, and therefore should not be regarded as limiting the scope. For those of ordinary skill in the art, other related drawings can be obtained based on these drawings without creative efforts.

[0060] Figure 1Schematic diagram of the oilfield production allocation method based on deep learning and multi-objective optimization in the embodiments of the present invention;

[0061] Figure 2 Schematic diagram of the structure of the oilfield production allocation device based on deep learning and multi-objective optimization in the embodiments of the present invention. Detailed implementation manners

[0062] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. Usually, the components of the embodiments of the present invention described and illustrated herein can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the drawings is not intended to limit the scope of the claimed invention, but merely represents selected embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0063] As Figure 1 shown, this embodiment provides an oilfield production allocation method based on deep learning and multi-objective optimization, which mainly includes steps S1, S2, S3, S4, and S5, including:

[0064] S1. Obtain the number of blocks in the oilfield, each well parameter included in each block, and the production start parameters of each block. The well parameters include well numbers and multiple well production parameter data.

[0065] It should be noted that in this application, the well production parameters include: daily liquid production of the well, daily oil production of the well, daily water injection of the water well, formation pressure, stroke, pumping frequency, oil pressure, casing pressure, gas-oil ratio, dynamic liquid level depth, static liquid level depth, production start time, water injection start time, oil production before operation, increased oil production after operation, cumulative oil production fitting, formation pressure change, injection-production ratio change, and formation pressure change. The production start parameters include: block water cut fitting, input cost of each block, natural production of the block, and production of the block by measures.

[0066] S2. Screen and process multiple well production parameters to obtain at least one sensitive parameter data.

[0067] S3. Predict the production of each well based on all the sensitive parameter data and a time-series prediction model.

[0068] S4. Construct a production objective function based on the production of each well, and construct a profit objective function and a capital investment objective function based on the production start parameters of each block.

[0069] It should be noted that in the application, three objectives are constructed to maximize profit, maximize production, and minimize capital investment, where:

[0070] 1. The objective pursued by the production objective function is to maximize the total annual production of the block, and the total production of the block is equal to the sum of the production allocated to each oil well.

[0071] 2. The objective pursued by the capital investment objective function is to minimize the total capital investment of the block, and the total investment of each block is equal to the sum of the investment allocated to each block.

[0072] 3. The objective pursued by the profit objective function is to maximize the total profit of the block, and in the model, the objective is to maximize the sum of the profits of each block.

[0073] Moreover, in this application, for the three objectives of maximizing total production, maximizing total profit, and minimizing total investment, the multi-objective block production allocation operation decision variables are determined. The production target Q i and the investment target I i allocated to the i-th block are selected as variables, that is:

[0074]

[0075]

[0076] At the same time, it should also be noted that constraint conditions need to be set in the process of solving the multi-objective optimization model for subsequent oil and gas production allocation. Therefore, in this application, the constraint conditions regarding total production; secondly, the constraint conditions regarding the production of each oilfield block; and finally, the constraint conditions regarding the total cost. The total production needs to be equal to or exceed the production tasks of all oilfields; the overall investment of the oilfield cannot be higher than the planned cost expenditure. It includes the following items:

[0077] 1) The sum of the total costs allocated to each oil well cannot exceed the limit of the total cost planned by the oilfield company.

[0078] 2) The sum of the production of each block, that is, the total production, cannot be lower than the lower limit of the total production capacity and cannot be higher than the upper limit of the total production capacity.

[0079] 3) The production task of the i-th block cannot be lower than the lower limit of the production capacity of this block and cannot be higher than the upper limit of the production capacity of this block.

[0080] 4) Reserve-production balance constraint. The reserve-production balance constraint refers to the limitation on the reserve-production balance coefficient. For the sustainable development of the oilfield, the reserve-production balance coefficient should be no less than 1 every year. The ratio of the newly recoverable reserves in a block to the total oil production in the block in the current year is greater than or equal to 1. The reserve-production balance coefficient is the ratio of the recoverable reserves in a block in the i-th year of the oilfield development plan period to the oil production in the block in the i-th year of the oilfield development plan period.

[0081] S5. Construct a multi-objective optimization model for oil and gas production allocation according to the production target function, profit target function, capital investment target function, and preset constraint functions, and solve the multi-objective optimization model for oil and gas production allocation to obtain the oilfield production allocation result. The oilfield production allocation result is the result with the maximum total production of all blocks, the minimum capital investment of all blocks, and the sum of the block profits of all blocks.

[0082] Specifically, in some specific embodiments, step S2 further includes the following steps to implement the screening of sensitive data. That is, it includes step S21, step S22, step S23, and step S24.

[0083] S21. Construct an original data set with the well numbers as index values for the oil well production parameters;

[0084] S22. Process the data set for outliers and missing values to obtain an initial data set;

[0085] S23. Perform dimensionless processing on the initial data set to obtain a data set;

[0086] S24. Screen the parameters in the data set according to the grey relational analysis method to obtain at least one sensitive parameter data.

[0087] Furthermore, it is also very important how to screen sensitive parameter data through the grey relational analysis method. Specifically, in this application, step S24 includes step S241 and step S242.

[0088] S241. Analyze the curves of the parameter data in the data set changing with time pairwise according to the grey relational analysis algorithm;

[0089] S242. Use the parameter data corresponding to the similar curve shapes as the sensitive parameter data for production prediction.

[0090] In this application, the correlation degree between parameters is judged according to the similarity degree of the curves of each parameter changing with time. If the curve shapes are similar, it is considered that the parameter correlation degree is large, that is, it is selected as the sensitive parameter for production prediction.

[0091] Furthermore, in this application, step S3 further specifically includes step S31, step S32, step S33, step S34, and step S35.

[0092] S31. Divide the sensitive parameter data into a training set and a validation set according to the sliding window method;

[0093] S32. Train a preset long short-term memory network model according to the training set to obtain a production prediction model;

[0094] S33. Predict the validation set according to the production prediction model to obtain the prediction result of the validation set;

[0095] S34. Calculate the prediction error according to the prediction result of the validation set and the validation set;

[0096] S35. If the prediction error is greater than the threshold, restart dividing the sensitive parameter data according to the sliding window method until the prediction error is less than the threshold.

[0097] It should be noted that in this application, the sliding window size is selected by judging the error performance on different window sizes and data sets. The production of each single well is accurately predicted and then accumulated to obtain the production of the block as the later production constraint of the block.

[0098] Finally, how to obtain the oilfield production allocation result is also extremely important. Among them, step S5 includes step S51:

[0099] S51. Population initialization: Randomly generate an initial population of size N, and then use the deep learning model Transformer to evaluate the fitness of each individual in the population to obtain the fitness value of each individual;

[0100] S52. Fast non-dominated sorting: Use the fast non-dominated sorting algorithm to sort each population, and divide each population individual into different non-dominated levels to distinguish the advantages and disadvantages of each solution;

[0101] S53. Crowding degree assignment: Evaluate the distribution density of individuals in the solution space through crowding degree calculation to select individuals that maintain population diversity and avoid falling into local optimal solutions;

[0102] S54. Population crossover and mutation: Generate new individuals using Transformer based on the fitness value, non-dominated level and crowding degree of individuals;

[0103] S55. Repeat starting fast non-dominated sorting until the number of iterations reaches the preset target to obtain the oilfield production allocation result.

[0104] Embodiment 2:

[0105] As Figure 2 shown, this embodiment provides an oilfield production allocation device based on deep learning and multi-objective optimization. The device includes:

[0106] A data acquisition module for acquiring the number of blocks in the oilfield, each well parameter included in each block, and the production start parameters of each block, where the well parameters include well numbers and multiple well production parameter data;

[0107] A data screening module for screening and processing multiple well production parameters to obtain at least one sensitive parameter data;

[0108] A production prediction module for predicting the production of each well based on all the sensitive parameter data and a time-series-based prediction model;

[0109] A function construction module for constructing a production objective function based on the production of each well, and constructing a profit objective function and a capital investment objective function based on the production start parameters of each block;

[0110] A production allocation model module for constructing a multi-objective optimization model for oil and gas production allocation based on the production objective function, the profit objective function, the capital investment objective function, and a preset constraint function, and solving the multi-objective optimization model for oil and gas production allocation to obtain an oilfield production allocation result, where the oilfield production allocation result is the result of maximizing the total production of all blocks, minimizing the capital investment amount of all blocks, and the sum of the block profits of all blocks.

[0111] In some specific embodiments, the data screening module includes:

[0112] A data classification module for constructing an original data set of well production parameters with the well number as the index value;

[0113] A data cleaning module for processing outliers and missing values in the data set to obtain an initial data set;

[0114] A data dimensionless quantification module for performing dimensionless quantification processing on the initial data set to obtain a data set;

[0115] A data extraction module for screening the parameters in the data set according to the grey relational analysis method to obtain at least one sensitive parameter data.

[0116] In some specific embodiments, the data extraction module includes:

[0117] A curve calculation module for pairwise analyzing the curve of the parameter data in the data set changing with time according to the grey relational analysis algorithm;

[0118] A curve judgment module for using the parameter data corresponding to similar curve shapes as the sensitive parameter data for production prediction.

[0119] In some specific embodiments, the production prediction module includes:

[0120] A data partitioning module for partitioning the sensitive parameter data into a training set and a validation set according to the sliding window method;

[0121] A training module, configured to train a preset long short-term memory network model according to a training set to obtain a production prediction model;

[0122] A verification module, configured to predict a verification set according to the production prediction model to obtain a prediction result of the verification set;

[0123] An error calculation module, configured to calculate a prediction error according to the prediction result of the verification set and the verification set;

[0124] An error judgment module, configured to, if the prediction error is greater than a threshold, restart partitioning the sensitive parameter data according to the sliding window method until the prediction error is less than the threshold.

[0125] In some specific embodiments, the production allocation model module includes:

[0126] An initialization module, configured to perform population initialization: randomly generate an initial population with a size of N, and then use the deep learning model Transformer to evaluate the fitness of each individual in the population to obtain the fitness value of each individual;

[0127] A non-dominated sorting module, configured to perform fast non-dominated sorting: use the fast non-dominated sorting algorithm to sort each population, and divide each population individual into different non-dominated levels to distinguish the advantages and disadvantages of each solution;

[0128] A crowding degree allocation module, configured to perform crowding degree allocation: evaluate the distribution density of individuals in the solution space through crowding degree calculation to select individuals that maintain population diversity and avoid falling into local optimal solutions;

[0129] A crossover and mutation module, configured to perform population crossover and mutation: generate new individuals using Transformer based on the fitness value, non-dominated level, and crowding degree of individuals;

[0130] A loop module, configured to repeatedly start fast non-dominated sorting until the number of iterations reaches a preset target to obtain the oilfield production allocation result.

[0131] It can be understood that the present invention is described through some embodiments. Those skilled in the art know that without departing from the spirit and scope of the present invention, various changes or equivalent replacements can be made to these features and embodiments. Additionally, under the teaching of the present invention, these features and embodiments can be modified to adapt to specific situations and materials without departing from the spirit and scope of the present invention. Therefore, the present invention is not limited by the specific embodiments disclosed herein, and all embodiments falling within the scope of the claims of this application belong to the scope protected by the present invention.

Claims

1. An oilfield production allocation method based on deep learning and multi-objective optimization, characterized in that, it includes: Obtain the number of blocks in the oilfield, each well parameter included in each block, and the production start parameters of each block. The well parameters include well numbers and multiple well production parameter data; Perform screening processing on multiple well production parameters to obtain at least one sensitive parameter data; Based on all the sensitive parameter data and a time-series-based prediction model, predict the production of each well; Construct a production target function according to the production of each well, and construct a profit target function and a capital investment target function according to the production start parameters of each block; Construct an oil and gas production allocation multi-objective optimization model according to the production target function, the profit target function, the capital investment target function and a preset constraint function, and solve the oil and gas production allocation multi-objective optimization model to obtain the oilfield production allocation result. The oilfield production allocation result is the result of the maximum total production of all blocks, the minimum capital investment of all blocks, and the sum of the block profits of all blocks.

2. The oilfield production allocation method based on deep learning and multi-objective optimization according to claim 1, characterized in that, Performing screening processing on multiple well production parameters to obtain at least one sensitive parameter data includes: Taking the well number as the index value to form an original data set of well production parameters; Perform outlier and missing value processing on the data set to obtain an initial data set; Perform dimensionless processing on the initial data set to obtain a data set; According to the grey relational analysis method, perform data screening on the parameters in the data set to obtain at least one sensitive parameter data.

3. The oilfield production allocation method based on deep learning and multi-objective optimization according to claim 2, characterized in that, According to the grey relational analysis method, performing data screening on the parameters in the data set to obtain at least one sensitive parameter data includes: Analyze the curves of the parameter data in the data set changing with time pairwise according to the grey relational analysis algorithm; Take the parameter data corresponding to the similar curve shapes as the sensitive parameter data for production prediction.

4. The oilfield production allocation method based on deep learning and multi-objective optimization according to claim 1, characterized in that, Based on all the sensitive parameter data and a time-series-based prediction model, predicting the production of each well includes: Divide the sensitive parameter data into a training set and a validation set according to the sliding window method; Train a preset long short-term memory network model according to the training set to obtain a production prediction model; Predict the validation set according to the production prediction model to obtain the prediction result of the validation set; Calculate the prediction error according to the prediction result of the validation set and the validation set; If the prediction error is greater than the threshold, restart dividing the sensitive parameter data according to the sliding window method until the prediction error is less than the threshold.

5. The oilfield production allocation method based on deep learning and multi-objective optimization according to claim 1, characterized in that, Construct an oil and gas production allocation multi-objective optimization model according to the production target function, the profit target function, the capital investment target function and a preset constraint function, and solve the oil and gas production allocation multi-objective optimization model to obtain the oilfield production allocation result, including: Population initialization: Randomly generate an initial population of size N. Subsequently, use the deep learning model Transformer to evaluate the fitness of each individual in the population to obtain the fitness value of each individual; Fast non-dominated sorting: Use the fast non-dominated sorting algorithm to sort each population, and divide each population individual into different non-dominated levels to distinguish the advantages and disadvantages of each solution; Crowding degree assignment: Evaluate the distribution density of individuals in the solution space through crowding degree calculation to select individuals that maintain population diversity and avoid falling into local optimal solutions; Population crossover and mutation: Generate new individuals using Transformer based on the fitness value, non-dominated level, and crowding degree of individuals; Repeat the fast non-dominated sorting until the number of iterations reaches the preset target to obtain the oilfield production allocation result.

6. An oilfield production allocation device based on deep learning and multi-objective optimization, Characterized in that, Comprising: A data acquisition module for acquiring the number of blocks in the oilfield, each well parameter included in each block, and the production start parameters of each block. The well parameters include well numbers and multiple well production parameter data; A data screening module for screening and processing multiple well production parameters to obtain at least one sensitive parameter data; A production prediction module for predicting the production of each well based on all sensitive parameter data and a time-series-based prediction model; A function construction module for constructing a production target function based on the production of each well, and constructing a profit target function and a capital investment target function based on the production start parameters of each block; A production allocation model module for constructing a multi-objective optimization model for oil and gas production allocation according to the production target function, the profit target function, the capital investment target function, and a preset constraint function, and solving the multi-objective optimization model for oil and gas production allocation to obtain the oilfield production allocation result. The oilfield production allocation result is the result of the maximum total production of all blocks, the minimum capital investment of all blocks, and the sum of the block profits of all blocks.

7. The oilfield production allocation method based on deep learning and multi-objective optimization according to claim 6, Characterized in that, The data screening module includes: A data classification module for constructing an original data set with well production parameters using the well number as the index value; A data cleaning module for processing outliers and missing values in the data set to obtain an initial data set; A data dimensionless quantification module for performing dimensionless quantification processing on the initial data set to obtain a data set; A data extraction module for screening at least one sensitive parameter data from the parameters in the data set according to the grey relational analysis method.

8. The oilfield production allocation method based on deep learning and multi-objective optimization according to claim 7, Characterized in that, The data extraction module includes: A curve calculation module for pairwise analyzing the parameter data curves in the data set that change over time according to the grey relational analysis algorithm; A curve judgment module for using the parameter data corresponding to similar curve shapes as sensitive parameter data for production prediction.

9. The oilfield production allocation method based on deep learning and multi-objective optimization according to claim 6, Characterized in that, The production prediction module includes: A data division module, which is used to divide the sensitive parameter data into a training set and a validation set according to the sliding window method; A training module, which is used to train a preset long short-term memory network model according to the training set to obtain a production prediction model; A validation module, which is used to predict the validation set according to the production prediction model to obtain the prediction result of the validation set; An error calculation module, which is used to calculate the prediction error according to the prediction result of the validation set and the validation set; An error judgment module, which is used to restart the division of the sensitive parameter data according to the sliding window method if the prediction error is greater than the threshold until the prediction error is less than the threshold.

10. The oilfield production allocation method based on deep learning and multi-objective optimization according to claim 6, characterized in that, The production prediction module includes: An initialization module, which is used for population initialization: randomly generate an initial population with a size of N, and then use the deep learning model Transformer to evaluate the fitness of each individual in the population to obtain the fitness value of each individual; A non-dominated sorting module, which is used for fast non-dominated sorting: use the fast non-dominated sorting algorithm to sort each population, and divide each population individual into different non-dominated levels to distinguish the advantages and disadvantages of each solution; A crowding degree allocation module, which is used for crowding degree allocation: evaluate the distribution density of individuals in the solution space through crowding degree calculation to select individuals that maintain population diversity and avoid falling into local optimal solutions; A crossover and mutation module, which is used for population crossover and mutation: generate new individuals using Transformer based on the fitness value, non-dominated level and crowding degree of individuals; A loop module, which is used to repeat the start of fast non-dominated sorting until the number of iterations reaches the preset target to obtain the oilfield production allocation result.

Images