CN117035151A

CN117035151A - Unstable water injection working system optimization method and system based on lightGBM algorithm

Info

Publication number: CN117035151A
Application number: CN202310750107.7A
Authority: CN
Inventors: 刘柳
Original assignee: Xian Shiyou University
Current assignee: Xian Shiyou University
Priority date: 2023-06-25
Filing date: 2023-06-25
Publication date: 2023-11-10

Abstract

The application discloses an unstable water injection working system optimization method and system based on a lightGBM algorithm, and aims to predict oil production of an oil well according to data analysis and a machine learning model based on the lightGBM algorithm. The method comprises the steps of collecting data, preprocessing the data, building a model to train and test the data, testing oil production of an oil well and the like. By using the method, the prediction accuracy and reliability can be improved to the greatest extent while the target data and the machine learning model are continuously improved, and the method can guide the production and management of an oil field and improve the oil yield and the exploitation efficiency of the oil well through a faster and more accurate prediction model, so that the maximum exploitation and utilization of an oil reservoir are realized. Meanwhile, the application also discloses a corresponding system, electronic equipment and a computer readable storage medium, which can improve the prediction accuracy and reliability to the greatest extent while continuously improving the target data and the machine learning model, and provide quicker and more accurate guidance for oilfield management and production.

Description

Unstable water injection working system optimization method and system based on lightGBM algorithm

Technical Field

The application relates to the field of petroleum exploration and development, in particular to an unstable water injection working system optimization method and system based on a lightGBM algorithm, which are used for predicting oil production of an oil well.

Background

Most of the oil fields in China enter a high water content exploitation stage, and at present, a conventional water injection mode is often used for exploitation, so that a great deal of researches are carried out on water injection modes, water injection equipment and water injection processes by students for searching the most effective water injection development scheme, and unstable water injection period researches are carried out. Common unstable water injection cycle studies include: laboratory experiments, mining experiments, empirical formula analogy, numerical simulation, etc.

The indoor experimental method is used for realizing unstable period water injection research, namely controlling the water injection period and the water injection quantity through a water injection device, recording the temperature, the pressure, the oil saturation and the like of a test sample through a sensor, processing the data, calculating the core permeability coefficient, the water phase relative permeability and the like, and analyzing and evaluating the indexes such as the production quantity, the water content, the extraction degree and the like of an oil reservoir under unstable water injection according to experimental results.

The mining field experiment method is to select a proper experiment sample at a certain mining field, arrange necessary equipment and instruments, such as a water level meter, a flowmeter and the like, at the mining field by utilizing the data acquired in the field, perform experiments through a water injection experiment device and simulate the situation of actual unstable period water injection. And then processing the data obtained by the field experiment, comparing different parameter changes under different water injection periods, and exploring the influence of the data on the water injection effect.

The empirical formula method is to determine the type of empirical formula, such as exponential type, logarithmic type, power function type, etc., according to the research target, then to use the existing experimental data to interpolate and fit, to determine the parameters in the empirical formula, and to build the empirical formula to predict the water injection quantity of the unstable periodic water injection.

Along with the development of numerical simulation technology, the numerical simulation method is also widely applied to the study of oilfield periodic water injection. The method can simulate various physical and chemical reactions in the water injection process of the oil well and simulate the movement process of well fluid along the oil well pipeline and the reservoir, thereby predicting the exploitation effect of the oil well.

Although the above methods all achieve certain effects, there are still some drawbacks. The laboratory method has the following disadvantages: 1. when the indoor experimental method simulates unstable periodic water injection, certain difference still exists between the experimental method and the actual industrial application, because the industrial application relates to different geological conditions and well site environments, the experimental result is difficult to completely reproduce the actual situation of the site; 2. errors may exist in acquisition and processing of experimental data, and the accuracy and sensitivity of different experimental equipment and sensors also influence the accuracy of experimental results, so that analysis conclusion is influenced; 3. the indoor experiment method can only simulate the actual situation under certain conditions, such as the factors of environmental temperature, geological structure, flow state and the like, and influences the accuracy of the simulation result; 4. the laboratory experiment needs to purchase various experimental equipment and instruments, and the maintenance cost is high, and meanwhile, the environment is possibly polluted.

The mine field experimental method has the following defects: 1. various equipment and instruments, such as a water level meter, a flowmeter and the like, are required to be purchased for carrying out the mine experiment, and the maintenance cost is high; 2. the actual situation of the site, such as large sampling difficulty, sample pollution and other factors, need to be considered in the mine field experiment, and the accuracy and reliability of the experimental result can be indirectly affected; 3. the experimental period of the mine field is long, the time cost is high, and mining strategy planning of a coal mine or a metal mine is required in the experimental process; 4. compared with an indoor experiment, the mine field experiment needs to consider more practical factors, such as the actual environmental influence of well wall stability, other rock physical parameters and the like, and the difficulty and uncertainty of the experiment are increased.

Disadvantages of the empirical formula include: 1. the model established by the empirical formula method is established based on historical data and statistical methods in a deduction mode, so that for a novel study object or when the current environment changes, the obtained model prediction result may be inaccurate, and the model needs to be adjusted or remodelled; 2. the model established by the empirical formula has certain defects and uncertainty, the prediction precision is low, and a high-quality prediction result cannot be provided; 3. the model established by the empirical formula is based on historical data, is sensitive to the current environment change, and is often influenced by interference and noise, so that the robustness of the model is poor; 4. the model established by the empirical formula method is usually based on a simple linear or nonlinear regression model, has low complexity and low interpretability, and is not suitable for processing complex and changeable study objects.

Numerical simulation methods also have certain limitations in application. On the one hand the accuracy of the simulation results is affected. The oil reservoir is complex, and many physical and chemical parameters of the reservoir are difficult to evaluate accurately, such as the porosity of the oil well, the permeability, the interaction of oil and water, etc. Meanwhile, many different factors, such as the advance of a well bore, the water injection quantity, the water injection position and the like, need to be considered in the simulation process, and the analysis and the accuracy of the simulation result are greatly influenced. On the other hand, the numerical simulation method needs to consume higher calculation resources, has longer mining period and overlong calculation time, and can cause the simulation technology to be difficult to copy and can not realize prediction rapidly.

Disclosure of Invention

The application provides an unstable water injection working system optimization method and system based on a lightGBM algorithm, and aims to solve the problems in the prior art.

The technical scheme provided by the application is as follows:

the application provides an unstable water injection working system optimization method based on a lightGBM algorithm, which comprises the following steps:

collecting field data of a target oil field, wherein the field data comprises geological information, injection data and historical production data of the oil field;

preprocessing the field data to obtain first data, wherein the first data comprises training data, test data and characteristic data;

establishing a machine learning model about the relation between an unstable water injection working system and oil increasing amount, wherein the machine learning model is a lightGBM algorithm-based machine learning model, and training and testing the first data through the machine learning model;

and according to training and testing results of the first data, an optimal water injection working system is formulated, and oil production of the oil well is predicted according to the optimal water injection working system.

Further, the predicting the oil production of the oil well according to the optimal water injection working system comprises the following steps:

and carrying out long-term prediction and real-time prediction on the oil yield of the oil well according to the optimal water injection working system.

Further, the collecting the field data of the target oil field specifically comprises the following steps:

and acquiring geological information, production history and natural gas injection data of the target oil field by using Petrel software to obtain field data.

Further, the preprocessing the field data to obtain first data includes:

multiple periodic water injection schemes are formulated based on a target oil field, and numerical simulation is respectively carried out on all the periodic water injection schemes through petrel software to obtain training data;

changing at least one parameter of permeability, porosity and water saturation in the geological information, and adopting the periodic water injection scheme to respectively perform numerical simulation to obtain test data;

extracting target characteristics by extracting the collected field data, and obtaining characteristic data by changing the size of the target characteristic parameters and adopting the periodic water injection scheme to respectively perform numerical simulation.

Further, the target characteristics comprise oil reservoir permeability, water injection quantity, bottom hole flow pressure, oil reservoir pressure and periodic water injection mode.

Further, according to the training and testing results of the first data, the making of the optimal water injection working system specifically includes:

and according to training and testing results of the first data, a plurality of water injection working systems are made to optimize water injection, wherein the water injection working systems comprise five water injection working systems of oil reservoir permeability, water injection quantity, bottom hole flow pressure, oil reservoir pressure and periodic water injection mode, and the optimal water injection working system is determined according to the water injection working systems.

The second aspect of the present application provides an unstable water injection operation system optimization system implemented based on a lightGBM algorithm, configured to implement the above-mentioned unstable water injection operation system optimization method implemented based on the lightGBM algorithm, including:

the data acquisition module is used for collecting field data of a target oil field;

the data processing module is used for preprocessing the field data to obtain first data;

the data training and testing module is used for establishing a machine learning model related to the relation between an unstable water injection working system and an oil increasing amount, wherein the machine learning model is a lightGBM algorithm-based machine learning model, and the first data is trained and tested through the machine learning model;

and the oil production prediction module is used for making an optimal water injection working system according to the training and testing results of the first data and predicting the oil production of the oil well according to the optimal water injection working system.

A third aspect of the present application provides an electronic device, including a processor and a memory, where the memory stores a plurality of instructions, and the processor is configured to read the instructions and execute the unstable water injection operating regime optimization method implemented based on the lightGBM algorithm described above.

A fourth aspect of the present application is a computer readable storage medium storing a plurality of instructions readable by a processor and performing the above-described unstable water injection operating regime optimization method implemented based on the lightGBM algorithm.

Compared with the prior art, the application has the beneficial effects that:

by using the method, the prediction accuracy and reliability can be improved to the greatest extent while the target data and the machine learning model are continuously improved. Compared with the prior prediction method, the method has the following remarkable advantages:

(1) The prediction speed can be increased, and the prediction accuracy can be improved;

(2) The method can adapt to various environments and changes of input data, and the prediction result is relatively stable and accurate;

(3) The automatic preprocessing of the data can be realized, and the cost of manual intervention and preprocessing is reduced.

Drawings

FIG. 1 is a flow chart of an unstable water injection working regime optimization method implemented based on a lightGBM algorithm in an embodiment of the application;

fig. 2 is a schematic diagram of an unstable water injection operation system optimization system implemented based on a lightGBM algorithm in an embodiment of the present application;

fig. 3 is a schematic structural diagram of an electronic device according to an embodiment of the present application;

FIG. 4 is a graph showing the oil production rate with time when the periodic water injection method is adopted in the embodiment of the application.

Detailed Description

In order to make the objects, technical solutions and advantages of the embodiments of the present application more clear, the technical solutions of the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present application. It will be apparent that the embodiments described below are some, but not all, embodiments of the application. The components of the embodiments of the present application generally described and illustrated in the figures herein may be arranged and designed in a wide variety of different configurations.

Accordingly, the following detailed description of the embodiments of the application, taken in conjunction with the accompanying drawings, is intended to represent only selected embodiments of the application, and not to limit the scope of the application as claimed. All other embodiments, which can be made by one of ordinary skill in the art without undue burden on the person of ordinary skill in the art based on the embodiments of the present application, are within the scope of the present application.

Example 1

Referring to fig. 1, the embodiment provides an unstable water injection working system optimization method implemented based on a lightGBM algorithm, which includes the following steps:

step 101, collecting field data of a target oil field, wherein the field data comprises geological information, injection data and historical production data of the oil well.

In this embodiment, the target oil field selects the snare oil field, which is mined in 1 st 1998 and is currently in the first development stage, geological research shows that the oil reservoir is divided into 12 small layers in the longitudinal direction, and seismic data shows that a water body exists in the south of the oil reservoir. There are 5 vertical wells in the field, where the wells PROD1, PROD2, PROD3, PROD4 are located on the central fault block, belonging to the CENTER well group; well production is located in the western world of the field, subordinate to the WEST well group. All wells had a caliper of 0.625 feet and a skin factor of 7.5.

In order to predict future production dynamics of the field, a numerical simulation study needs to be completed, using ECLIPSE reservoir numerical simulation software for simulation and petrel re as a pre-and post-treatment tool. The model data are all made using Field units. It is first necessary to complete a basic data model, then to perform history fitting and scheme prediction.

Model size

And comprehensively considering the existing data, selecting a 3D grid model containing 12 small layers for carrying out the simulation. The model is longitudinally divided and consistent with geological awareness. A 24 x 25 grid system is used on the plane so the overall size of the model is 24 x 25 x 12.

Reservoir fluids

The oil field is an unsaturated oil reservoir, and has no gas cap under initial conditions. However, as development proceeds, the reservoir pressure drops below the bubble point pressure. In this simulation, the fluid properties in the reservoir may be considered to be consistent throughout. PVT analysis indicated that there was only one type of reservoir fluid present, with a bubble point pressure of 1062.2Psia and a corresponding gas-to-oil ratio of 973scf/stb.

From the data in the reservoir that had been drilled, the oil-water interface was determined to be 8200ft, which is consistent with the free water interface. From the log analysis, the pressure reference depth was 7000ft, corresponding to 3035.7Psia. The reservoir has no gas cap in the initial state. The geological engineer analyzes the water body existing in the south of the middle broken block of the oil reservoir, and preliminarily estimates the water body volume to be 10MMstb, so that the supporting capacity of the oil reservoir can be achievedTo 5bbl/day/Psi. The water body can be simulated by Fetkovich water body, and the reference depth of the water body can be defined as the depth of OWC, so that the water body is in balance with an oil reservoir in an initial state. The total compression coefficient of the water body is 1X10 ^-5 (Unit 1/Psi).

Step 102, preprocessing the field data to obtain first data, wherein the first data comprises training data, test data and characteristic data.

In step 102, the acquired target data needs to be preprocessed. The preprocessing comprises the steps of dividing a training set and a testing set, carrying out characteristic derivation and the like, retaining the characteristics with larger influence on oil production, and laying a foundation for model training.

And 103, establishing a machine learning model about the relation between an unstable water injection working system and an oil increasing amount, wherein the machine learning model is a lightGBM algorithm-based machine learning model, and training and testing the first data through the machine learning model.

Based on the collected target data, a machine learning model is established regarding the relationship between the unstable water injection regime and the oil increase. And finally, constructing a regression model or a classification model by using algorithms such as LightGBM and the like so as to analyze the influence of a water injection working system on the oil well production.

The LightGBM algorithm is a machine learning algorithm based on a Gradient Boost Decision Tree (GBDT), advocated by Microsoft in 2017. The core advantage of LightGBM is that it exhibits advantageous speed and accuracy in processing high-dimensional, large-scale data, mainly for classification and regression tasks. Compared to other decision tree based machine learning algorithms, the LightGBM has several features:

1. rapid training speed: the LightGBM algorithm adopts a histogram-based algorithm, can quickly train large-scale high-dimensional data, and can use multithreading parallel processing.

2. High accuracy: the LightGBM algorithm adopts a Leaf-wise growth strategy and a gradient single-side sampling (GOSS) method to improve the accuracy of the model.

3. Low memory occupancy: the LightGBM algorithm employs a compression algorithm to reduce memory usage while also supporting mutually exclusive feature bundling (ExclusiveFeatureBundling, EFB) techniques and random sampling techniques to optimize the storage and computation efficiency of the model.

4. Can be applied to large-scale data: the LightGBM algorithm is adapted to process large-scale data, unlike the traditional GDBT algorithm which selects one large training subset at a time from all data, the LightGBM algorithm is more prone to employ a small training set and each base classifier processes only a portion of the features.

In general, the LightGBM algorithm is a machine learning algorithm with very high running speed and excellent training effect, and a plurality of algorithm models are selected for comparison, so that the LightGBM algorithm is found to be most suitable for the requirement of the machine learning model of the application, and a model is finally built on dataiku according to the algorithm.

And 104, according to training and testing results of the first data, an optimal water injection working system is formulated, and oil production of the oil well is predicted according to the optimal water injection working system.

Optionally, the predicting the oil production of the oil well according to the optimal water injection working system includes:

Optionally, the collecting the field data of the target oil field specifically includes:

Optionally, the preprocessing the field data to obtain first data includes:

and 201, formulating a plurality of periodic water injection schemes based on a target oil field, and respectively performing numerical simulation on all the periodic water injection schemes through petrel software to obtain training data.

In step 201, 24 kinds of periodic water injection schemes are formulated based on the snare oil field, wherein the ratio of the first water injection is 1:1, 1:2, 1:3 and 1:4 respectively, each mode can adopt six reference periods of 3 days, 10 days, 30 days, 90 days, 180 days and one year, so that the scheme of 24 kinds of periodic water injection is totally 4 times 6, the time is set to be 1 month 1 day in 1998 to 1 month 1 year 2008, numerical simulation is respectively carried out on the 24 kinds of periodic water injection schemes through petrel software to obtain an oil production curve, excel data is derived through the oil production curve, the condition that the oil production changes along with time can be clearly seen through the excel data, and the data generated by the 24 kinds of schemes are training data.

And 202, changing at least one parameter of permeability, porosity and water saturation in geological information, and adopting the periodic water injection scheme to respectively perform numerical simulation to obtain test data.

In step 202, in order to ensure accuracy of model prediction, part of geological parameters such as permeability, porosity, water saturation and the like are changed on the basis of the geological model, and then the same 24 kinds of periodic water injection schemes are adopted for the model, numerical simulation is performed in the same manner, and data obtained by the numerical simulation are test data. The geologic model can directly adopt the existing model, and a new model can be built according to the requirement.

And 203, extracting target characteristics by extracting the characteristics of the collected field data, and obtaining characteristic data by changing the size of the target characteristic parameters and adopting the periodic water injection scheme to perform numerical simulation respectively.

In step 203, the collected target data is converted into a form usable for modeling analysis by feature extraction of the target data. The Sklearn is used for carrying out feature engineering treatment, the features related to the water injection working system and the oil increasing amount are extracted, and the finally extracted target features are as follows:

1. geological structure: including the nature of the rock, lithology, faults, formation thickness, etc.

2. Reservoir permeability: the permeability is the resistance of reservoir rock to fluid movement, and the larger the permeability is, the larger the influence on the water injection working system and the oil increasing amount is.

3. And (3) water injection amount: the water injection quantity is the total quantity of manual water injection, and the water injection quantity and the oil increasing quantity are positively correlated.

4. Bottom hole flow pressure: the bottom hole flow pressure refers to the pressure value of a bottom hole measuring point of an oil well, and is one of important factors between a water injection working system and oil increasing quantity.

5. Reservoir pressure: reservoir pressure refers to the pressure in an underground reservoir and is one of the important factors affecting the water injection operating regime and oil enhancement.

Finally, five characteristics of oil reservoir permeability, water injection quantity, bottom hole flow pressure, oil reservoir pressure and periodic water injection mode are selected as target characteristics, oil yield data of different working systems are collected by changing the magnitudes of the five characteristic parameters, prediction and training are carried out in a machine learning mode, and accuracy of a predicted oil yield model is guaranteed.

Optionally, the target features include reservoir permeability, water injection rate, bottom hole pressure, reservoir pressure, and periodic water injection mode.

Optionally, according to the training and testing results of the first data, the making of the optimal water injection working system specifically includes:

Based on the training and predicted data, five water injection working systems are formulated to optimize water injection, and the water injection working systems are respectively as follows:

1. reservoir permeability: other parameters are kept unchanged, and the oil yield is predicted by changing the permeability of the oil reservoir. The higher the permeability is, the better the water injection effect is, and the oil yield of the oil well can be increased.

2. And (3) water injection amount: other parameters are kept unchanged, and the oil yield is predicted by changing the water injection quantity. The data can be used for obtaining that when the water injection quantity is increased initially, the oil yield can be increased, but when the water injection quantity is too high, the oil yield can be increased and the oil yield can be reduced.

3. Bottom hole flow pressure: other parameters are kept unchanged, and the oil production rate is predicted by changing the bottom hole flow pressure. The higher the bottom hole flow pressure is, the higher the oil production of the oil well is.

4. Reservoir pressure: other parameters are kept unchanged, and the oil production rate is predicted by changing the bottom hole flow pressure. From the data, the greater the reservoir pressure, the higher the oil production from the well.

5. Periodic water injection mode: the parameters of the oil reservoir permeability, the water injection amount, the accumulated oil production, the bottom hole flow pressure and the oil reservoir pressure are kept unchanged, four water injection ratios of 1:1, 1:2, 1:3 and 1:4 are respectively used, and each water injection ratio adopts six reference periods of 3 days, 10 days, 30 days, 90 days, 180 days and one year, so that the scheme of 24 periodic water injection is adopted in total, namely 4 times 6, and the oil production rate is predicted by changing the periodic water injection mode. In this embodiment, the time-dependent oil production rate curve when the periodic water injection method is used is shown in fig. 4.

It should be noted that the above factors are not independent and are mutually influenced, and changing one of them may affect other factors at the same time, ultimately affecting the oil production of the well. Therefore, in oil field production and oil extraction planning, various factors need to be comprehensively considered, and an optimal water injection working system is prepared so as to maximally increase the oil yield of the oil well.

Example two

Referring to fig. 2, the embodiment provides an unstable water injection working system optimization system implemented based on a lightGBM algorithm, which is configured to implement the above-mentioned unstable water injection working system optimization method implemented based on the lightGBM algorithm, and includes:

the data acquisition module 1 is used for collecting field data of a target oil field.

And the data processing module 2 is used for preprocessing the field data to obtain first data.

And the data training and testing module 3 is used for establishing a machine learning model about the relation between an unstable water injection working system and an oil increasing amount, wherein the machine learning model is a lightGBM algorithm-based machine learning model, and the first data is trained and tested through the machine learning model.

And the oil production prediction module 4 is used for making an optimal water injection working system according to the training and testing results of the first data and predicting the oil production of the oil well according to the optimal water injection working system.

Referring to fig. 3, in other embodiments, the present application further provides an electronic device, including a processor 5 and a memory 6, where the memory 6 stores a plurality of instructions, and the processor 5 is configured to read the instructions and execute the above-mentioned method for optimizing an unstable water injection working system implemented based on the lightGBM algorithm.

In other embodiments, the application is a computer readable storage medium storing a plurality of instructions readable by a processor and configured to perform the method for optimizing an unstable water injection operating regime based on the lightGBM algorithm described above.

Example III

The embodiment provides a practical application, the method provided in the first embodiment is applied to software, different water injection working systems provided in the above can be selected in the software, and the software manufacturing is specifically divided into the following steps:

1. writing a software requirement specification: according to the requirements and functions of the prediction model, a software requirement specification is written, and the functions, user requirements, application scenes and the like of the software are defined. Because the prediction model is used for realizing unstable periodic water injection, software provides a relevant scheme of the unstable periodic water injection for users.

2. Designing a software framework: according to the requirement specification, a software framework is designed, and the framework comprises the structure, the interface, the function and the like of software.

3. Embedding the predictive model into software: the trained prediction model based on the LightGBM algorithm is embedded into software, and java language is used this time.

4. Designing a software interface: and designing interfaces of the software according to the requirement specification and the software framework, including a user interface, an output interface and the like, so that a user can operate and view the prediction result.

5. Software testing and adjustment: after software development is completed, functional testing and performance testing are performed, problems are found, and repair and adjustment are performed.

6. Release software: release software, including release versions, documents, instructions for use, service support, etc., is provided for users to download and use. The software interface is designed into a clear and easy-to-use graphical interface, so that a user can conveniently input prediction data, check a prediction result, and analyze and output the prediction result. The software interface may be customized and improved for different user needs and application scenarios to provide a better user experience.

The application is based on SNARK oil field, utilize the method of machine learning to set up the model on dataiku, only need to input the ratio of the oil production and benchmark period and some corresponding basic geological parameter to predict the oil production directly, design into a software for users to use finally according to the model set up, the oil field can employ this software to formulate the relevant scheme of unstable water injection directly, and through the software to predict the oil production, the characteristic and effects of this application are as follows:

1. the method can accurately predict the oil production of the oil well, and has important guiding significance for oil field production scheduling.

2. The method can predict according to historical data and real-time production data, and timely find out abnormal conditions of oil well production so as to take corresponding measures.

3. The method can predict oil well characteristics and different geological structures, provide personalized and targeted prediction results, and bring brand new improvement to industry in the aspect of unstable water injection scheme design.

4. The accuracy and efficiency of the production decision are improved, the production cost and resource waste are reduced, and the oil well production benefit and the enterprise profit capability are improved.

5. The method promotes the traditional oil field to be transformed into intelligentized and digitalized, improves the production process, management and technical level of the oil field, and promotes the innovation of the production mode of the oil field.

6. An advanced production prediction tool is provided for oilfield production management personnel, oilfield production real-time monitoring and scheduling management are facilitated, and working efficiency and working quality are improved.

The foregoing description is merely illustrative of the preferred embodiments of the present application, but the scope of the present application is not limited thereto, and any changes or substitutions within the technical scope of the present application should be covered. Therefore, the protection scope of the present application shall be subject to the protection scope of the claims.

Claims

1. The unstable water injection working system optimization method based on the lightGBM algorithm is characterized by comprising the following steps of:

2. The method for optimizing an unstable water injection operating regime based on a lightGBM algorithm according to claim 1, wherein predicting oil production of an oil well according to the optimal water injection operating regime comprises:

3. The optimization method of the unstable water injection working system based on the lightGBM algorithm according to claim 1, wherein the collecting the field data of the target oil field specifically comprises:

4. The method for optimizing an unstable water injection system based on a lightGBM algorithm according to claim 1, wherein the preprocessing the field data to obtain first data comprises:

5. The optimization method of the unstable water injection working system based on the lightGBM algorithm according to claim 4, wherein the optimization method is characterized by comprising the following steps:

the target characteristics comprise oil reservoir permeability, water injection quantity, bottom hole flow pressure, oil reservoir pressure and periodic water injection mode.

6. The method for optimizing an unstable water injection working system based on the lightGBM algorithm according to claim 5, wherein the step of formulating an optimal water injection working system according to the training and testing results of the first data is specifically:

7. An unstable water injection operation system optimization system implemented based on a lightGBM algorithm for implementing the unstable water injection operation system optimization method implemented based on the lightGBM algorithm according to any one of claims 1 to 6, comprising:

8. An electronic device comprising a processor and a memory, wherein the memory stores a plurality of instructions, and the processor is configured to read the instructions and execute the unstable water injection operating regime optimization method implemented based on the lightGBM algorithm according to any one of claims 1 to 6.

9. A computer readable storage medium storing a plurality of instructions readable by a processor and executable by the processor to implement a method of optimizing an unstable water injection regime based on a lightGBM algorithm according to any one of claims 1-6.