Analog reservoir recommendation method based on dynamic and static data sets

Abstract

The present invention discloses an analog reservoir recommendation method based on dynamic and static data sets. The specific steps are: extracting dynamic data and static data from a professional oil well database and performing normalization cleaning and dynamic data normalization in sequence to obtain a temporary data table; specifying a target oil reservoir and constructing an oil reservoir static similarity model based on the temporary data table using a K-means clustering algorithm to obtain multiple oil reservoirs with similar geological characteristics; selecting discrimination parameters and constructing a discrimination model from the multiple oil reservoirs with similar geological characteristics using an expert fuzzy system to obtain multiple similar oil reservoirs at development stages; from the multiple similar oil reservoirs at development stages, determining the main controlling factors and weights affecting production using a random forest algorithm to obtain a dynamic curve similarity calculation model; the oil reservoir with the highest curve similarity determination coefficient, that is, the oil reservoir with the most similar development effect to the target oil reservoir, is used to complete the recommendation of the analog reservoir. The problem in the prior art that it is difficult for humans to judge the similarity between oil reservoirs at different development stages is solved.

Description

Technical Field

[0001] The invention belongs to the technical field of oilfield analogy methods, and in particular relates to an analogy reservoir recommendation method based on dynamic and static data sets. Background Art

[0002] Due to the diverse geological, production, and mining parameters of oil and gas fields, and the varying importance of reservoir characteristics on production under different geological conditions, defining and determining reservoir similarity has become a challenge. Reservoir analogy is a key technical tool for SEC reserve assessment, recoverable reserve calibration, development plan compilation, and reservoir development performance prediction. However, given the diverse nature of reservoir properties, development stages, and development outcomes, accurately and quickly identifying similar reservoirs is particularly challenging.

[0003] In the prior art, patent publication number CN107133879A discloses a method for screening similar oil fields. Based on grey theory and similarity theory, analog parameters and weights are quantitatively determined according to the essential association between analog parameters and analog targets. According to the reservoir type of the studied object, parameter threshold normal distribution is introduced to process analog parameters, calculate similarity, and comprehensively evaluate and screen similar oil fields. However, only the static data association relationship of the oil fields is considered, and the analog parameters are relatively simple. Patent publication number CN113240321A discloses an oil field analogy evaluation method and system based on deep learning. It proposes to use one of the static characteristic parameters of the oil field to be evaluated or the oil field development and production parameters as the analogy target function, and compare and screen the parameters one by one. However, the development and production parameters only involve the characteristics at a single time point, and the similarity of the reservoirs is not considered from a time series perspective.

[0004] Therefore, there is an urgent need for a method that is easy to use, highly accurate, and suitable for recommending similar reservoirs at all development stages. Summary of the Invention

[0005] The purpose of the present invention is to provide an analog reservoir recommendation method based on dynamic and static data sets, which solves the problem in the prior art that it is difficult for humans to judge the similarity between reservoirs at different development stages.

[0006] The technical solution of the present invention is an analog reservoir recommendation method based on dynamic and static data sets, which is specifically implemented according to the following steps: Step 1, extracting dynamic data and static data from a professional oil well database and performing normalization cleaning and dynamic data normalization in sequence to obtain a temporary data table;

[0007] Step 2: Specify the target reservoir and build a static similarity model of the reservoir based on the temporary data table using the K-means clustering algorithm to obtain multiple reservoirs with similar geological characteristics;

[0008] Step 3: Select discrimination parameters and construct a discrimination model from multiple reservoirs with similar geological characteristics through an expert fuzzy system to obtain similar reservoirs at multiple development stages;

[0009] Step 4: From similar reservoirs at multiple development stages, the main controlling factors and weights affecting production are determined using the random forest algorithm to obtain a dynamic curve similarity calculation model. The reservoir with the highest curve similarity determination coefficient is the reservoir with the most similar development effect to the target reservoir, completing the recommendation of analogous reservoirs.

[0010] The present invention is also characterized in that: the dynamic data in step 1 include decline rate, water content, oil production and recovery degree of recoverable reserves; the static data in step 1 include development layer, permeability, effective thickness, fracture development, edge and bottom water development, formation coefficient, mobility and porosity-to-roar ratio.

[0011] The normalization and cleaning in step 1 includes filling missing values, removing outliers, and converting text categorical variables.

[0012] In step 1, the dynamic data is normalized to transform the dynamic data into a dimensionless expression between [0, 1].

[0013] Step 2 is implemented as follows:

[0014] Step 2.1, specify the target reservoir from the entire reservoir and construct a reservoir static similarity model using a one-dimensional or multi-dimensional K-means clustering algorithm based on the temporary data table and several static parameters;

[0015] Step 2.2: Extract other reservoirs in the same cluster as the target reservoir in the reservoir static similarity model, which is the reservoir group with similar geological characteristics.

[0016] Step 3 is implemented as follows:

[0017] Step 3.1, selecting a discrimination parameter from the temporary data table, wherein the discrimination parameter includes one or more of the three parameters: water cut, oil production, and degree of recovery of recoverable reserves;

[0018] Step 3.2: From multiple reservoirs with similar geological characteristics, a discrimination model is constructed using an expert fuzzy system to extract reservoirs with similar development stages.

[0019] Step 3.2 is implemented as follows:

[0020] Step 3.2.1: Determine the oil production stage based on the change rate of oil production in the reservoir in the last two months. The calculation formula for the change rate is:

[0021]

[0022] Among them, ρ cis the oil production change rate, m2 is the oil production of the latest month, and m1 is the oil production of the previous month; when ρ c Greater than 5% is considered as high yield. c Less than 5% is a decreasing trend, and a change rate of 5% is a stable yield;

[0023] Step 3.2.2: Determine the extent of recoverable reserves. The calculation formula is:

[0024]

[0025] Among them, ρ k is the recoverable reserves rate, V y is the geological reserves of crude oil, V k is recoverable reserves; when ρ k When the recoverable reserve ratio is less than 20%, it is a low recoverable reserve ratio. k When the ratio is less than 60%, it is medium recoverable reserves rate. When 60%≤ρ k When <80%, it is a high recoverable reserve rate. When 80%≤ρ k When is the ultra-high recoverable reserves rate;

[0026] Step 3.2.3: Based on the comprehensive water content of the reservoir f w Divide, when f w When the moisture content is less than 20%, it is low moisture content. When 20%≤f w When the moisture content is less than 60%, it is medium moisture content. When 60%≤f w When the moisture content is less than 80%, it is high moisture content. When 80%≤f w When the moisture content is very high;

[0027] Step 3.2.4: Based on the oil production change rate ρ in step 3.2.1 c , the recoverable reserve rate ρ in step 3.2.2 k and the comprehensive reservoir water content f in step 3.2.3 w Label naming is performed, a discriminant model is constructed, and multiple reservoirs with similar geological characteristics are selected, that is, similar reservoirs in multiple development stages.

[0028] Step 4 is implemented as follows:

[0029] Step 4.1: Based on similar reservoirs at multiple development stages, determine the main controlling factors and their weights that affect production changes using the random forest algorithm;

[0030] Step 4.2: Draw a curve based on one or more parameters of the main control factors affecting yield changes, and determine the similarity of the curves. The curve similarity coefficient R 2 The calculation formula is

[0031]

[0032] Among them, y is the data of the specified parameters of the target reservoir, The parameter data for other comparative reservoirs are specified, i is the sequence number of the i-th value in the parameter data, n is the number of parameter data values, is the mean value of y;

[0033] Curve similarity coefficient R 2 is the weighted average of the similarity and influence weight of each main control factor curve, which is

[0034]

[0035] Among them, R 2 is the curve similarity coefficient, R i is the similarity coefficient of each main control factor curve, V i is the weight of the impact of the main controlling factors on output, and the sum of the weights of all main controlling factors is equal to 1;

[0036] Step 4.3: Based on the reservoir similarity R 2 The reservoir with the highest curve similarity determination coefficient among similar reservoirs at multiple development stages is the reservoir with the most similar development effect to the target reservoir, thus completing the recommendation of analogous reservoirs.

[0037] The beneficial effects of the present invention are as follows: the present invention automatically extracts static parameters of oil reservoirs and dynamic parameters based on time series, applies K-means clustering algorithm, fuzzy expert system and random forest algorithm to construct an integrated model, and comprehensively recommends other oil reservoirs similar to the target oil reservoir from multiple dimensions and multiple levels; the present invention overcomes the shortcomings of single comparison index and insufficient consideration of dynamic development rules, has the characteristics of convenient application and high accuracy, is suitable for recommending similar oil reservoirs at various development stages, solves the problem of human difficulty in judging the degree of similarity between oil reservoirs, and provides technical support for in-depth research on the relationship between multiple oil reservoirs. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 is a flow chart of the analog reservoir recommendation method based on dynamic and static data sets of the present invention;

[0039] Figure 2 is an operational flow chart of the recommended method for analog reservoirs of the present invention;

[0040] Figure 3 This is a cluster analysis diagram of Example 1 of the analog reservoir recommendation method of the present invention;

[0041] Figure 4 It is a similar reservoir in the development stage as in Example 1 of the analog reservoir recommendation method of the present invention;

[0042] Figure 5 is a similarity curve between the recommended reservoir and the target reservoir in Example 1 of the analog reservoir recommendation method of the present invention;

[0043] Figure 6 This is a cluster analysis diagram of Example 2 of the analog reservoir recommendation method of the present invention;

[0044] Figure 7 It is a similar reservoir in the development stage as in Example 2 of the analog reservoir recommendation method of the present invention;

[0045] Figure 8 is a similarity curve of water drive reserve production between the recommended reservoir and the target reservoir in Example 2 of the analog reservoir recommendation method of the present invention;

[0046] Figure 9 is a similarity curve between the recoverability of the recommended reservoir and the target reservoir in Example 2 of the analog reservoir recommendation method of the present invention;

[0047] Figure 10 This is a cluster analysis diagram of Example 3 of the analog reservoir recommendation method of the present invention;

[0048] Figure 11 It is a similar reservoir in the development stage as in Example 3 of the analog reservoir recommendation method of the present invention;

[0049] Figure 12 is a similarity curve of water drive reserve production between the recommended reservoir and the target reservoir in Example 3 of the analog reservoir recommendation method of the present invention;

[0050] Figure 13 is a similarity curve of monthly comprehensive decline rate of old wells in the recommended reservoir and the target reservoir in Example 3 of the analog reservoir recommendation method of the present invention;

[0051] Figure 14 This is a cluster analysis diagram of Example 4 of the analog reservoir recommendation method of the present invention;

[0052] Figure 15 It is a similar reservoir in the development stage of Example 4 of the analog reservoir recommendation method of the present invention;

[0053] Figure 16 is a similarity curve of water drive reserve production between the recommended reservoir and the target reservoir in Example 4 of the analog reservoir recommendation method of the present invention;

[0054] Figure 17 It is a similarity curve of the monthly comprehensive decline rate of old wells in the recommended reservoir and the target reservoir in Example 4 of the analog reservoir recommendation method of the present invention. DETAILED DESCRIPTION

[0055] The present invention will be described in detail below with reference to the accompanying drawings and specific embodiments.

[0056] The present invention is based on the analog reservoir recommendation method of dynamic and static data sets, such as Figure 1 、 Figure 2As shown, please follow the steps below:

[0057] Step 1, data preparation stage: extract dynamic data and static data from the oil well professional database and perform standardization and cleaning and dynamic data normalization in sequence to obtain a temporary data table;

[0058] The dynamic data in step 1 include decline rate, water cut, oil production and recovery degree of recoverable reserves; the static data in step 1 include development layer, permeability, effective thickness, fracture development, edge and bottom water development, formation coefficient, mobility and porosity-to-roar ratio;

[0059] The normalization and cleaning in step 1 includes filling missing values, removing outliers, and converting text categorical variables.

[0060] In step 1, the dynamic data is normalized to transform the dynamic data into a dimensionless expression between [0, 1].

[0061] Step 2, Recommendation of Reservoirs with Similar Geological Characteristics: Specify the target reservoir and, based on the temporary data table, construct a static similarity model of the reservoir using the K-means clustering algorithm to obtain multiple reservoirs with similar geological characteristics;

[0062] Step 2.1, specify the target reservoir from the entire reservoir and construct a reservoir static similarity model using a one-dimensional or multi-dimensional K-means clustering algorithm based on the temporary data table and several static parameters;

[0063] Step 2.2, extracting other reservoirs in the same cluster as the target reservoir in the reservoir static similarity model, i.e., a reservoir group with similar geological characteristics;

[0064] Step 3, Recommendation of Similar Reservoirs at Development Stages: Select discrimination parameters and construct a discrimination model from multiple reservoirs with similar geological characteristics through an expert fuzzy system to obtain multiple reservoirs with similar development stages;

[0065] Step 3.1, selecting a discrimination parameter from the temporary data table, wherein the discrimination parameter includes one or more of the three parameters: water cut, oil production, and degree of recovery of recoverable reserves;

[0066] Step 3.2: From multiple reservoirs with similar geological characteristics, a discriminant model is constructed using an expert fuzzy system to extract reservoirs with similar development stages;

[0067] Step 3.2.1: Determine the oil production stage based on the change rate of oil production in the reservoir in the last two months. The calculation formula for the change rate is:

[0068] ρ c =(m2-m1) / m2×100% (1)

[0069] Among them, ρ cis the oil production change rate, m2 is the oil production of the latest month, and m1 is the oil production of the previous month; when ρ c Greater than 5% is considered as high yield. c Less than 5% is a decreasing trend, and a change rate of 5% is a stable yield;

[0070] Step 3.2.2: Determine the extent of recoverable reserves. The calculation formula is:

[0071] ρ k =(V y -V k ) / V y ×100% (2)

[0072] Among them, ρ k is the recoverable reserves rate, V y is the geological reserves of crude oil, V k is recoverable reserves; when ρ k When the recoverable reserve ratio is less than 20%, it is a low recoverable reserve ratio. k When the ratio is less than 60%, it is medium recoverable reserves rate. When 60%≤ρ k When <80%, it is a high recoverable reserve rate. When 80%≤ρ k When is the ultra-high recoverable reserves rate;

[0073] Step 3.2.3: Based on the comprehensive water content of the reservoir f w Divide, when f w When the moisture content is less than 20%, it is low moisture content. When 20%≤f w When the moisture content is less than 60%, it is medium moisture content. When 60%≤f w When the moisture content is less than 80%, it is high moisture content. When 80%≤f w When the moisture content is very high;

[0074] Step 3.2.4: Based on the oil production change rate ρ in step 3.2.1 c , the recoverable reserve rate ρ in step 3.2.2 k and the comprehensive reservoir water content f in step 3.2.3 w The label naming rule for reservoir development stages is "comprehensive water cut + technically recoverable reserves recovery degree + oil production," for example: "high water cut, declining production stage." A discriminant model is constructed to select reservoirs with similar geological characteristics from multiple reservoirs at similar development stages.

[0075] Step 4: Recommendation of reservoirs with similar development effects: From similar reservoirs at multiple development stages, the main controlling factors and weights affecting production are determined using the random forest algorithm, and a dynamic curve similarity calculation model is obtained. The reservoir with the highest curve similarity determination coefficient is the reservoir with the development effect most similar to the target reservoir, completing the recommendation of analogous reservoirs.

[0076] Step 4.1: Based on similar reservoirs at multiple development stages, determine the main controlling factors and their weights that affect production changes using the random forest algorithm;

[0077] Step 4.2: Draw a curve based on one or more parameters of the main control factors affecting yield changes, and determine the similarity of the curves. The curve similarity coefficient R 2 The calculation formula is

[0078]

[0079] Among them, y is the data of the specified parameters of the target reservoir, The parameter data for other comparative reservoirs are specified, i is the sequence number of the i-th value in the parameter data, n is the number of parameter data values, is the mean value of y;

[0080] Curve similarity coefficient R 2 is the weighted average of the similarity and influence weight of each main control factor curve, which is

[0081]

[0082] Among them, R 2 is the curve similarity coefficient, R i is the similarity coefficient of each main control factor curve, V i is the weight of the impact of the main controlling factors on output, and the sum of the weights of all main controlling factors is equal to 1;

[0083] Step 4.3, according to the curve similarity coefficient R 2 The reservoir with the highest curve similarity determination coefficient among similar reservoirs at multiple development stages is the reservoir with the most similar development effect to the target reservoir, thus completing the recommendation of analogous reservoirs.

[0084] Example 1

[0085] This embodiment provides an analog reservoir recommendation method based on dynamic and static data sets, which is specifically implemented according to the following steps:

[0086] Step 1: Extract dynamic data and static data from the oil well professional database and perform normalization cleaning and dynamic data normalization in sequence to obtain a temporary data table;

[0087] The data extractor automatically extracts relevant data from the reservoir static database and the oil, gas and water well production database. The data cleaner fills missing values, removes outliers, and converts text classification variables. The data processor performs dimensionless transformation of dynamic data to [0,1], and then all processed data are saved in the data storage.

[0088] Step 2: Specify the target reservoir and build a static similarity model of the reservoir based on the temporary data table using the K-means clustering algorithm to obtain multiple reservoirs with similar geological characteristics;

[0089] like Figure 3 As shown, the analog object, reservoir range and static parameters (one or more) are specified through the parameter selector 1, the cluster analyzer performs cluster analysis on the relevant reservoir parameters in the data storage, and the recommender displays the best cluster and clustering results. The other reservoirs in the cluster where the target reservoir is located are the reservoirs with similar geological characteristics.

[0090] Step 3: Select discrimination parameters and construct a discrimination model from multiple reservoirs with similar geological characteristics through an expert fuzzy system to obtain similar reservoirs at multiple development stages;

[0091] like Figure 4 As shown, one or more parameters including annual oil production, technically recoverable reserves recovery rate, and comprehensive water content are specified through parameter selector 2, and all reservoir-related parameters of the cluster where the target reservoir is located in the data storage are identified through the expert fuzzy system. Reservoirs with the same name are similar reservoirs in the development stage.

[0092] Step 4: From similar reservoirs at multiple development stages, the main controlling factors and weights affecting production are determined using the random forest algorithm to obtain a dynamic curve similarity calculation model. The reservoir with the highest curve similarity determination coefficient is the reservoir with the most similar development effect to the target reservoir, completing the recommendation of analogous reservoirs.

[0093] like Figure 5 As shown, the normalized time series data of the reservoir are subjected to random forest learning through the main controlling factor analyzer, and the production parameters are used as the input set and the other parameters are used as the input set to perform feature importance analysis on the changes in the production data, and determine the main controlling factors (one or more). The main controlling factor curve is drawn by the graph plotter, and the similarity of each main controlling factor and the reservoir similarity are calculated separately by the similarity calculator. The reservoir with the highest score is the reservoir with similar development effect.

[0094] Example 2

[0095] This embodiment provides an analog reservoir recommendation method based on dynamic and static data sets, which is specifically implemented according to the following steps:

[0096] Step 1: Extract dynamic data and static data from the oil well professional database and perform normalization cleaning and dynamic data normalization in sequence to obtain a temporary data table;

[0097] Step 2: Specify the target reservoir and build a static similarity model of the reservoir based on the temporary data table using the K-means clustering algorithm to obtain multiple reservoirs with similar geological characteristics;

[0098] like Figure 6 As shown in Table 1, taking the analogous reservoir of recommended reservoir 46 as an example, the fracture development status, development degree, formation coefficient, mobility, and average roar radius are selected as clustering parameters, and the optimal clustering clusters are determined to be 4 clusters.

[0099] Table 1

[0100]

[0101]

[0102] Step 3: Select discrimination parameters and construct a discrimination model from multiple reservoirs with similar geological characteristics through an expert fuzzy system to obtain similar reservoirs at multiple development stages;

[0103] like Figure 7 As shown, taking the cluster where reservoir 46 is located as input, the fuzzy expert system is used to judge the similarity of development stages, and the reservoirs with similar development stages are obtained.

[0104] Step 4: From similar reservoirs at multiple development stages, the main controlling factors and weights affecting production are determined using the random forest algorithm to obtain a dynamic curve similarity calculation model. The reservoir with the highest curve similarity determination coefficient is the reservoir with the most similar development effect to the target reservoir, completing the recommendation of analogous reservoirs.

[0105] like Figure 8 、 Figure 9 As shown, the reservoir 46, which is located in the "high water content and ultra-high recovery reduction stage", is used as input to calculate the curve similarity, and the reservoir 43 with similar development effect is determined.

[0106] Example 3

[0107] Step 1: Extract dynamic data and static data from the oil well professional database and perform normalization cleaning and dynamic data normalization in sequence to obtain a temporary data table;

[0108] Step 2: Specify the target reservoir and build a static similarity model of the reservoir based on the temporary data table using the K-means clustering algorithm to obtain multiple reservoirs with similar geological characteristics;

[0109] like Figure 10 As shown in Table 2, taking the analogous reservoir of recommended reservoir 28 as an example, matrix permeability_air permeability, fracture development status_development degree, and edge and bottom water development status_development degree are selected as clustering parameters. As shown in Table 2, 6 is selected as the best cluster, and reservoirs with similar geological characteristics are obtained.

[0110] Table 2

[0111]

[0112] Step 3: Select discrimination parameters and construct a discrimination model from multiple reservoirs with similar geological characteristics through an expert fuzzy system to obtain similar reservoirs at multiple development stages;

[0113] like Figure 11 As shown, taking the cluster where reservoir 28 is located as input, the fuzzy expert system is used to judge the similarity of development stages, and the reservoirs with similar development stages are obtained.

[0114] Step 4: From similar reservoirs at multiple development stages, the main controlling factors and weights affecting production are determined using the random forest algorithm to obtain a dynamic curve similarity calculation model. The reservoir with the highest curve similarity determination coefficient is the reservoir with the most similar development effect to the target reservoir, completing the recommendation of analogous reservoirs.

[0115] like Figure 12 、 Figure 13 As shown, the reservoir 28, which is located in the "medium water content and high production reduction stage", is used as input to calculate the curve similarity, and the reservoir 47 with similar development effect is determined.

[0116] Example 4

[0117] This embodiment provides an analog reservoir recommendation method based on dynamic and static data sets, which is specifically implemented according to the following steps:

[0118] Step 1: Extract dynamic data and static data from the oil well professional database and perform normalization cleaning and dynamic data normalization in sequence to obtain a temporary data table;

[0119] Step 2: Specify the target reservoir and build a static similarity model of the reservoir based on the temporary data table using the K-means clustering algorithm to obtain multiple reservoirs with similar geological characteristics;

[0120] like Figure 14 As shown in Figure 3, taking the analogous reservoir of recommended reservoir 20 as an example, matrix permeability_air permeability, fracture development status_development degree, fracture development status_fracture direction, edge and bottom water development status_development degree, edge and bottom water development status_contact type, formation coefficient, mobility, average roar channel radius, and porosity-roar ratio are selected as clustering parameters, and 10 is selected as the best cluster, as shown in Table 3, which shows the reservoirs with similar geological characteristics.

[0121] Table 3

[0122]

[0123]

[0124] Step 3: Select discrimination parameters and construct a discrimination model from multiple reservoirs with similar geological characteristics through an expert fuzzy system to obtain similar reservoirs at multiple development stages;

[0125] like Figure 15 As shown, taking the cluster where reservoir 20 is located as input, the fuzzy expert system is used to judge the similarity of development stages, and the reservoirs with similar development stages are obtained.

[0126] Step 4: From similar reservoirs at multiple development stages, the main controlling factors and weights affecting production are determined using the random forest algorithm to obtain a dynamic curve similarity calculation model. The reservoir with the highest curve similarity determination coefficient is the reservoir with the most similar development effect to the target reservoir, completing the recommendation of analogous reservoirs.

[0127] like Figure 16 、 Figure 17 As shown, the reservoir 20, which is located in the "ultra-high water content and ultra-high recovery reduction stage", is used as input to calculate the curve similarity, and it is determined that the reservoir 14 has a similar development effect.

[0128] The advantage of the present invention is that it comprehensively considers the dynamic and static parameters of the reservoir, and based on machine learning methods such as K-means clustering algorithm, fuzzy expert system, random forest algorithm, etc., establishes a multi-dimensional, multi-level, progressive integrated analog reservoir recommendation method, which has the advantages of comprehensive consideration of indicators and high similarity of recommendation results. Data extraction, cleaning, and analysis are fully automated, and users can freely choose the operation steps according to the application purpose, which greatly reduces the difficulty of manual judgment of similar reservoirs and effectively supports business applications such as recoverable reserves calibration, development plan preparation, and reservoir development effect evaluation.

Claims

1. The analog reservoir recommendation method based on dynamic and static data sets is characterized by: The specific implementation is as follows: Step 1, extract dynamic data and static data from the oil well professional database and perform standardization cleaning and dynamic data normalization in sequence to obtain a temporary data table; Step 2: specify the target reservoir and construct a reservoir static similarity model based on the temporary data table using a K-means clustering algorithm to obtain multiple reservoirs with similar geological characteristics; Step 3: Selecting discrimination parameters and constructing a discrimination model from the multiple reservoirs with similar geological characteristics through an expert fuzzy system to obtain multiple reservoirs with similar development stages; Step 4: From multiple reservoirs with similar development stages, determine the main controlling factors and weights affecting production through the random forest algorithm, and obtain a dynamic curve similarity calculation model. The reservoir with the highest curve similarity determination coefficient is the reservoir with the most similar development effect to the target reservoir, completing the recommendation of the analog reservoir.

2. The analog reservoir recommendation method based on dynamic and static data sets according to claim 1, characterized in that: The dynamic data in step 1 include decline rate, water cut, oil production and recovery rate of recoverable reserves; the static data in step 1 include development layer, permeability, effective thickness, fracture development, edge and bottom water development, formation coefficient, mobility and porosity-to-roar ratio.

3. The analog reservoir recommendation method based on dynamic and static data sets according to claim 1, characterized in that: The normalization cleaning described in step 1 includes filling missing values, removing outliers, and converting text categorical variables.

4. The analog reservoir recommendation method based on dynamic and static data sets according to claim 1, characterized in that: The normalization of the dynamic data in step 1 is to transform the dynamic data into a dimensionless expression between [0, 1].

5. The analog reservoir recommendation method based on dynamic and static data sets according to claim 1, characterized in that: The step 2 is specifically implemented according to the following steps: Step 2.1, specifying a target reservoir from the entire reservoir and constructing a reservoir static similarity model using a one-dimensional or multi-dimensional K-means clustering algorithm based on the temporary data table and a number of static parameters; Step 2.2: Extract other reservoirs in the same cluster as the target reservoir in the reservoir static similarity model, which is the reservoir group with similar geological characteristics.

6. The analog reservoir recommendation method based on dynamic and static data sets according to claim 1, characterized in that: The step 3 is specifically implemented according to the following steps: Step 3.1, selecting a discrimination parameter from the temporary data table, wherein the discrimination parameter includes one or more of the three parameters: water cut, oil production, and degree of recovery of recoverable reserves; Step 3.2: From the multiple reservoirs with similar geological characteristics, a discrimination model is constructed through an expert fuzzy system to extract multiple reservoirs with similar development stages.

7. The analog reservoir recommendation method based on dynamic and static data sets according to claim 6, characterized in that: The step 3.2 is specifically implemented according to the following steps: Step 3.2.1: Determine the oil production stage based on the change rate of oil production in the reservoir in the last two months. The calculation formula for the change rate is: ρ c =(m2-m1) / m2×100% (1) Among them, ρ c is the oil production change rate, m2 is the oil production of the latest month, and m1 is the oil production of the previous month; when ρ c Greater than 5% is considered as high yield. c Less than 5% is a decreasing trend, and a change rate of 5% is a stable yield; Step 3.2.2: Determine the extent of recoverable reserves. The calculation formula is: ρ k =(V y -V k ) / V y ×100% (2) Among them, ρ k is the recoverable reserves rate, V y is the geological reserves of crude oil, V k is recoverable reserves; when ρ k When the recoverable reserve ratio is less than 20%, it is a low recoverable reserve ratio. k When the ratio is less than 60%, it is medium recoverable reserves rate. When 60%≤ρ k When <80%, it is a high recoverable reserve rate. When 80%≤ρ k When is the ultra-high recoverable reserves rate; Step 3.2.3: Based on the comprehensive water content of the reservoir f w Divide, when f w When the moisture content is less than 20%, it is low moisture content. When 20%≤f w When the moisture content is less than 60%, it is medium moisture content. When 60%≤f w When the moisture content is less than 80%, it is high moisture content. When 80%≤f w When the moisture content is very high; Step 3.2.4: According to the oil production change rate ρ in step 3.2.1 c , the recoverable reserve rate ρ in step 3.2.2 k and the comprehensive reservoir water content f in step 3.2.3 w Label naming is performed, a discrimination model is constructed, and oil reservoirs with similar geological characteristics are selected from multiple reservoirs, that is, oil reservoirs with similar development stages.

8. The analog reservoir recommendation method based on dynamic and static data sets according to claim 1, characterized in that: The step 4 is specifically implemented according to the following steps: Step 4.1: Based on similar reservoirs at multiple development stages, determine the main controlling factors and their weights that affect production changes using the random forest algorithm; Step 4.2: Draw a curve based on one or more parameters of the main control factors affecting yield changes, and determine the similarity of the curves. The curve similarity coefficient R 2 The calculation formula is Among them, y is the data of the specified parameters of the target reservoir, The parameter data for other comparative reservoirs are specified, i is the sequence number of the i-th value in the parameter data, n is the number of parameter data values, is the mean value of y; Curve similarity coefficient R 2 is the weighted average of the similarity and influence weight of each main control factor curve, which is Among them, R 2 is the curve similarity coefficient, R i is the similarity coefficient of each main control factor curve, V i is the weight of the impact of the main controlling factors on output, and the sum of the weights of all main controlling factors is equal to 1; Step 4.3: Based on the curve similarity coefficient R 2 The reservoir with the highest curve similarity determination coefficient among similar reservoirs at multiple development stages is the reservoir with the most similar development effect to the target reservoir, thus completing the recommendation of analogous reservoirs.

Images