Method for optimizing pump change displacement of producing well based on machine learning

Abstract

The invention provides a method for optimizing pump replacement displacement of a producing well based on machine learning. The method mainly solves the problem that the pump replacement displacement is difficult to accurately determine by an existing artificial qualitative analysis method. The method comprises the following steps: establishing sample data of block historical pump changing wells; main influence factors of daily oil increase in the initial stage of pump replacement are determined; a machine learning algorithm is selected and designed, and a pump change initial daily oil increase machine learning prediction model is established; based on the established pump change initial daily oil increase machine learning prediction model, the displacement ratio corresponding to the maximum initial daily oil increase after pump change under the condition that the water content is not increased is determined and serves as the optimal displacement ratio; and determining the optimal post-pump-change displacement by combining the pre-pump-change displacement and the optimal displacement ratio. According to the method for optimizing the pump change displacement of the oil production well based on machine learning, the daily increase oil quantitative prediction model at the initial stage of pump change is established, then the daily increase oil and the moisture content value under different displacement after pump change are predicted, the optimal displacement of pump change of the oil well is determined, and therefore the pump change effect is optimal.

Description

Technical Field

[0001] The present invention relates to the technical field of oilfield development, and in particular to a method for optimizing the displacement of oil well pump replacement based on machine learning. Background Art

[0002] As an important means of increasing oilfield production, the choice of displacement after pump replacement is crucial to the effect of pump replacement. At present, the displacement of pump replacement is mainly determined by qualitative analysis of factors such as daily liquid production, flow pressure, and dynamic liquid level before oil well pump replacement. There is a lack of quantitative optimization methods. Therefore, a method for optimizing the displacement of oil well pump replacement based on machine learning has been established, which is of great significance. Summary of the invention

[0003] The technical problem to be solved by the present invention is to overcome the problem that the existing manual qualitative analysis methods in the background technology are difficult to accurately determine the displacement of pump replacement, and to provide an oil well pump replacement displacement optimization method based on machine learning. The oil well pump replacement displacement optimization method based on machine learning establishes a quantitative prediction model for the daily oil increase in the initial stage of pump replacement, and then predicts the daily oil increase and water content values ​​under different displacements after pump replacement, determines the optimal displacement of the oil well pump replacement, and thus achieves the best pump replacement effect.

[0004] The present invention solves the problem through the following technical solution: The method for optimizing the displacement of oil well pump replacement based on machine learning comprises the following steps:

[0005] Establish sample data of historical pump replacement wells in the block;

[0006] Based on the established sample data of historical pump replacement wells in the block, determine the main factors affecting the daily oil production increase in the initial stage of pump replacement;

[0007] Based on the historical daily oil increase in the initial stage of pump replacement wells and its main influencing factors, a machine learning algorithm is designed to establish a machine learning prediction model for the daily oil increase in the initial stage of pump replacement;

[0008] Based on the establishment of a machine learning prediction model for the initial daily oil increase after pump replacement, the displacement ratio corresponding to the maximum initial daily oil increase after pump replacement under the condition that the water content does not increase is determined as the optimal displacement ratio;

[0009] The optimal displacement after pump replacement is determined by combining the displacement before pump replacement and the optimal displacement ratio.

[0010] Preferably, the establishment of the block historical pump replacement well sample data includes:

[0011] (1) The design response target data is the daily oil increase in the initial stage of pump replacement;

[0012] (2) Based on the factors that affect oil well production, 19 analysis indicators in the following seven aspects are designed:

[0013] 1) Permeability: average permeability;

[0014] 2) Formation energy: formation pressure, flow pressure, subsidence, and daily fluid production before replacement;

[0015] 3) Production thickness: sandstone thickness, effective thickness, converted thickness;

[0016] 4) Oil content: daily oil production, water content, water absorption ratio of connected water wells, daily oil production and water content of surrounding oil wells before pump replacement;

[0017] 5) Oil leakage area: injection and production well spacing, number of connected directions;

[0018] 6) Interlayer interference: permeability variation coefficient, number of small production layers;

[0019] 7) Process: the ratio of the displacement before pump replacement, the displacement after pump replacement and the displacement before pump replacement (referred to as displacement ratio).

[0020] Preferably, the determination of the main influencing factors of the daily oil increase in the initial stage of pump replacement includes:

[0021] The Spearman rank correlation analysis algorithm was used to calculate the correlation between the initial daily oil production increase of historical pump replacement wells and various influencing factors.

[0022]

[0023] In the formula, r is the correlation degree; d i is the order difference of two data; n is the number of data items of historical pump replacement wells.

[0024] The calculated correlations were sorted, and factors with correlations greater than 0.3 were identified as the main factors affecting the daily oil increase in the initial stage of pump replacement.

[0025] Preferably, the establishment of the machine learning prediction model for daily oil increase in the initial stage of pump replacement includes:

[0026] The values ​​of each main control factor are normalized according to the maximum and minimum value method;

[0027]

[0028] x 归一 is the normalized value of a certain main control factor; x is the actual value of a certain main control factor; x max is the maximum value of a main control factor in the sample data; x min is the minimum value of a main control factor in the sample data

[0029] Design a machine learning algorithm to conduct machine learning between the initial daily oil increase of historical pump replacement wells and the normalized main control factor values;

[0030] The ratio of the number of historical data for training and testing is set, and the daily oil increase prediction model for the initial stage of pump replacement is determined based on the principle of the highest test compliance rate.

[0031] Preferably, the machine learning algorithm includes any one of decision tree, random forest, support vector machine, multi-layer perceptron neural network, integrated boosting tree regression, gradient descent regression, bagging and stacking integrated regression algorithms;

[0032] The ratio of the number of historical data items for training and testing is set to 8:2.

[0033] Preferably, the method for determining the displacement ratio when the initial daily oil increase is maximum under the condition that the water cut of the oil well does not increase after the pump is replaced includes:

[0034] For the wells that are going to change pumps, calculate the values ​​of the main factors that affect the daily oil increase in the initial stage of pump change; set the displacement ratio, and set n values ​​according to the set range with a step length of 0.1;

[0035] The values ​​of different displacement ratios and other calculated main factors of daily oil increase in the initial stage of pump replacement are normalized and input into the optimally determined daily oil increase prediction model for pump replacement. There are n groups in total, and n daily oil increase values ​​in the initial stage of pump replacement can be output;

[0036] According to n displacement ratios, the daily liquid production of the oil well after the pump is replaced can be calculated; furthermore, the water content after the pump is replaced can be calculated;

[0037] Finally, the n initial daily oil increase values ​​and water content values ​​after the pump replacement are obtained, and the displacement ratio with a water content not higher than that before the pump replacement and the largest daily oil increase is selected as the optimal displacement ratio.

[0038] Preferably, the daily liquid production of the oil well after the pump is replaced is calculated as follows:

[0039] q 后 =q 前 *k (3)

[0040] q 后 is the daily fluid production after the pump is replaced; q 前 is the daily liquid production before pump replacement; k is the different displacement ratios set.

[0041] The water content calculation formula after the pump is changed is:

[0042]

[0043] fwWater content after pump replacement, q o前 To change the previous Nissan oil, q o增 Initial daily oil increase after pump replacement

[0044] Preferably, the method for determining the optimal displacement after pump replacement is to multiply the displacement before pump replacement by the optimal displacement ratio to obtain the optimal displacement after pump replacement.

[0045] Compared with the above-mentioned background technology, the present invention has the following beneficial effects: the method for optimizing the displacement of oil well pump replacement based on machine learning comprehensively considers the main factors affecting the daily increase in oil in the initial stage of pump replacement, establishes a prediction model for the daily increase in oil in the initial stage of pump replacement based on machine learning, and determines the optimal displacement of pump replacement by predicting the daily increase in oil and water content after pump replacement under different displacements. It solves the problem that manual qualitative analysis is difficult to accurately determine the displacement of pump replacement, realizes quantitative optimization of pump replacement displacement, and provides a new way to further improve the effect of pump replacement in oil wells. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 This is a flow chart of a method for optimizing the displacement of oil well pump replacement based on machine learning in the present invention;

[0047] Figure 2 It is a schematic diagram of the influence of various factors on the daily oil increase in the initial stage of pump replacement in the embodiment of the present invention;

[0048] Figure 3 It is a schematic diagram of the test compliance rates of various machine learning algorithms in an embodiment of the present invention. Specific implementation method:

[0049] The present invention will be further described below in conjunction with the accompanying drawings:

[0050] As attached Figure 1 As shown, the method for optimizing the displacement of oil well pump replacement based on machine learning includes the following steps:

[0051] Step 1: A method for establishing sample data of historical pump replacement wells in a block, including:

[0052] Collect and organize the daily oil increase data in the initial stage of pump replacement, as well as 19 analytical index data in the following 7 aspects:

[0053] 1.1 Permeability: average permeability;

[0054] 1.2 Formation energy: formation pressure, flow pressure, subsidence, and daily fluid production before replacement;

[0055] 1.3 Production thickness: sandstone thickness, effective thickness, converted thickness;

[0056] 1.4 Oil content: daily oil production, water content, water absorption ratio of connected water wells, daily oil production and water content of surrounding oil wells before pump replacement;

[0057] 1.5 Oil leakage area: injection and production well spacing, number of connected directions;

[0058] 1.6 Interlayer interference: permeability variation coefficient, number of small production layers;

[0059] 1.7 Process: the ratio of the displacement before pump replacement, the displacement after pump replacement and the displacement before pump replacement (referred to as displacement ratio).

[0060] Step 2: Based on the established sample data of historical pump replacement wells in the block, determine the main factors affecting the daily oil production increase in the initial stage of pump replacement; including:

[0061] The Spearman rank correlation analysis algorithm was used to calculate the correlation between the initial daily oil production increase of historical pump replacement wells and various influencing factors.

[0062]

[0063] In the formula, r is the degree of correlation;

[0064] d i is the order difference of two data;

[0065] n is the number of data items of historical pump replacement wells;

[0066] The calculated correlations were sorted, and factors with correlations greater than 0.3 were identified as the main factors affecting the daily oil increase in the initial stage of pump replacement.

[0067] Step 3: Based on the historical daily oil increase in the initial stage of the pump replacement well, a machine learning algorithm is selected to establish a method for a machine learning prediction model for the daily oil increase in the initial stage of the pump replacement; including:

[0068] The values ​​of each main control factor are normalized according to the maximum and minimum value method;

[0069]

[0070] In the formula, x 归一 It is the normalized value of a certain main control factor;

[0071] x is the actual value of a certain main control factor;

[0072] x max is the maximum value of a main controlling factor in the sample data;

[0073] x min It is the minimum value of a main controlling factor in the sample data.

[0074] Design a machine learning algorithm to conduct machine learning between the initial daily oil increase of historical pump replacement wells and the normalized main control factor values; the machine learning algorithm includes decision tree, random forest, support vector machine, multi-layer perceptron neural network, integrated boosting tree regression, gradient descent regression, bagging and stacking integrated regression algorithms;

[0075] The ratio of the number of historical data for training and testing is set to 8:2, and the daily oil increase prediction model for the initial stage of pump replacement is determined based on the principle of the highest test compliance rate.

[0076] Step 4: Based on the machine learning prediction model for the initial daily oil increase after pump replacement, determine the displacement ratio corresponding to the maximum initial daily oil increase after pump replacement under the condition that the water content does not increase, as the optimal displacement ratio, including:

[0077] For the wells that are going to change pumps, calculate the values ​​of the main factors that affect the daily oil increase in the initial stage of pump change; set the displacement ratio, set n values ​​in a certain range with a step length of 0.1;

[0078] Input the values ​​of different displacement ratios and other calculated main factors of initial daily oil increase in pump replacement into the optimally determined initial daily oil increase prediction model for pump replacement, a total of n groups, and n initial daily oil increase values ​​for pump replacement can be output;

[0079] According to n displacement ratios, the daily liquid production of the oil well after the pump is replaced can be calculated; furthermore, the water content after the pump is replaced can be calculated;

[0080] Finally, the n initial daily oil increase values ​​and water content values ​​after the pump replacement are obtained, and the displacement ratio with a water content not higher than that before the pump replacement and the largest daily oil increase is selected as the optimal displacement ratio.

[0081] The calculation formula for the daily liquid production of the oil well after the pump is replaced is:

[0082] q 后 =q 前 *k (3)

[0083] In the formula, q 后 This is the daily fluid produced after the pump is replaced;

[0084] q 前 This is the daily fluid production before the pump is replaced;

[0085] k is the different displacement ratios set.

[0086] The water content calculation formula after the pump is changed is:

[0087]

[0088] Where, fw is the water content after the pump is replaced;

[0089] q o前 To change the former Nissan oil;

[0090] q o增 This is the initial daily oil increase after replacing the pump.

[0091] Step 5: Determine the optimal displacement after pump replacement by combining the displacement before pump replacement and the optimal displacement ratio.

[0092] The optimal displacement after pump change is obtained by multiplying the displacement before pump change by the optimal displacement ratio.

[0093] Embodiment 1:

[0094] Taking the pump replacement of Well S748 in a development zone of Daqing Oilfield as an example, the implementation process of the method of the present invention is explained.

[0095] The present invention is described below based on the present embodiment, but it is worth noting that the present invention is not limited to these embodiments. In the detailed description of the present invention below, some specific details are described in detail. However, for the part not described in detail, those skilled in the art can also fully understand the present invention.

[0096] like Figure 1 As shown, a method for optimizing the displacement of oil well pump replacement based on machine learning mainly includes the following steps:

[0097] Step S101, collect and organize the initial daily oil increase data of 410 historical pump replacement wells in the block where the S748 well is located, as well as 19 analytical index data in 7 major aspects:

[0098] 1.1 Permeability: average permeability;

[0099] 1.2 Formation energy: formation pressure, flow pressure, subsidence, and daily fluid production before replacement;

[0100] 1.3 Production thickness: sandstone thickness, effective thickness, converted thickness;

[0101] 1.4 Oil content: daily oil production, water content, water absorption ratio of connected water wells, daily oil production and water content of surrounding oil wells before pump replacement;

[0102] 1.5 Oil leakage area: injection and production well spacing, number of connected directions;

[0103] 1.6 Interlayer interference: permeability variation coefficient, number of small production layers;

[0104] 1.7 Process: the ratio of the displacement before pump replacement, the displacement after pump replacement and the displacement before pump replacement (referred to as displacement ratio).

[0105] Step S102: Analyze and determine the main influencing factors of daily oil increase in the initial stage of pump replacement.

[0106] The correlation between the initial daily oil production increase of historical pump replacement wells and various influencing factors was calculated by using spearman rank correlation. Figure 2 It is the influence degree of various factors on the daily oil increase in the initial stage of pump replacement.

[0107] The calculated correlations were sorted, and factors with correlations greater than 0.3 were identified as the main factors affecting the daily oil increase in the initial stage of pump replacement.

[0108] According to the order of correlation, six factors with correlation greater than 0.3 were identified as the main influencing factors of daily oil production in the initial stage of pump replacement, namely: daily oil production before replacement, displacement ratio, daily liquid production before replacement, submergence before replacement, average daily oil production of surrounding oil wells before replacement, and production converted thickness.

[0109] Step S103: Establish a machine learning prediction model for daily oil increase in the initial stage of pump replacement.

[0110] 3.1 Normalize the six main control factors;

[0111]

[0112] In the formula, x 归一 It is the normalized value of a certain main control factor;

[0113] x is the actual value of a certain main control factor;

[0114] x max is the maximum value of a main controlling factor in the sample data;

[0115] x min It is the minimum value of a main controlling factor in the sample data.

[0116] 3.2 Select decision tree, random forest, support vector machine, multi-layer perceptron neural network, integrated boosting tree regression, gradient descent regression, bagging and stacking integrated regression algorithms, conduct machine learning between the initial daily oil increase of historical pump replacement wells and the normalized main controlling factors, and set the ratio of training and testing historical data to 8:2;

[0117] 3.3 According to the prediction results of the machine learning algorithm, the compliance rate of various algorithms is calculated (when the absolute value of the initial daily oil increase is greater than 3 tons, the relative error is within 10% and it is marked as compliant; when the absolute value of the initial daily oil increase is less than or equal to 3 tons, the absolute error is within 0.3 tons and it is marked as compliant). Based on the principle of the highest test compliance rate, a prediction model for the initial daily oil increase of pump replacement based on the multi-layer perception neural network algorithm is preferably established. The model is applied to test the 82 historical pump replacement wells reserved. The predicted daily oil increase results of 66 wells are consistent with the actual daily oil increase, and the compliance rate reaches 80.5%. Figure 3 shown.

[0118] Step S104: Based on the established machine learning prediction model for the initial daily oil increase after pump replacement, determine the displacement ratio corresponding to the maximum initial daily oil increase after pump replacement under the condition that the water content does not increase, as the optimal displacement ratio.

[0119] 4.1 Calculate the five main influencing factors of the pump replacement well S748, except the pump replacement displacement ratio. The five main influencing factors of the initial daily oil increase of the S748 well pump replacement are shown in Table 1.

[0120]

[0121] 4.2 In the range of 1.1 to 3.0, set the displacement ratio in steps of 0.1.

[0122] 4.3 The daily oil increase in the initial stage of pump replacement was predicted for the 20 groups of factor data of Well S748. The results of the daily oil increase and water cut in the initial stage of pump replacement predicted by the data of different displacement ratios of Well S748 are shown in Table 2.

[0123] Table 2

[0124]

[0125] 4.4 It can be seen from Table 2 that under the condition that the water content does not increase, the displacement ratio corresponding to the maximum daily oil increase is 1.8, which is the optimal displacement ratio.

[0126] Step S105, calculating the optimal pump replacement displacement.

[0127] S748 displacement before pump replacement is 43.8m 3 / d, optimal displacement of pump change = displacement before change * optimal displacement ratio = 43.8m 3 / d*1.8=78.84m 3 / d.

[0128] The above examples specifically illustrate the entire process of the method for optimizing the displacement of oil well pump replacement based on machine learning in the present invention. The analysis results can be directly used for optimizing the displacement of pump replacement wells to further improve the pump replacement effect.

[0129] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for optimizing the displacement of oil well pump replacement based on machine learning, characterized in that: The following steps are involved: Establish sample data of historical pump replacement wells in the block; Based on the established sample data of historical pump replacement wells in the block, determine the main factors affecting the daily oil production increase in the initial stage of pump replacement; Based on the historical daily oil increase in the initial stage of pump replacement wells and its main influencing factors, a machine learning algorithm was designed to establish a machine learning prediction model for the daily oil increase in the initial stage of pump replacement; Based on the establishment of a machine learning prediction model for the initial daily oil increase after pump replacement, the displacement ratio corresponding to the maximum initial daily oil increase after pump replacement under the condition that the water content does not increase is determined as the optimal displacement ratio; The optimal displacement after pump replacement is determined by combining the displacement before pump replacement and the optimal displacement ratio.

2. The method for optimizing the displacement of oil well pump replacement based on machine learning according to claim 1, characterized in that: The establishment of the block historical pump replacement well sample data includes: 2.1 The design response target data is the daily oil increase in the initial stage of pump replacement; 2.2 Based on the factors that affect oil well production, 19 analysis index data in 7 major aspects are designed.

3. The method for optimizing the displacement of oil well pump replacement based on machine learning according to claim 2 is characterized in that: starting from the factors affecting the oil well production, the 19 analysis index data in 7 aspects are designed as follows: 3.1 Permeability: average permeability; 3.2 Formation energy: formation pressure before replacement, flow pressure, subsidence, and daily fluid production; 3.3 Production thickness: sandstone thickness, effective thickness, converted thickness; 3.4 Oil content: daily oil production, water content, water absorption ratio of connected water wells, daily oil production and water content of surrounding oil production wells before pump replacement; 3.5 Oil leakage area: injection and production well spacing, number of connected directions; 3.6 Interlayer interference: coefficient of variation of permeability, number of small production layers; 3.7 Process aspects: the ratio of the displacement before pump replacement, the displacement after pump replacement and the displacement before pump replacement.

4. The method for optimizing displacement of oil well pump replacement based on machine learning according to claim 1, characterized in that: The determination of the main influencing factors of the daily oil increase in the initial stage of pump replacement includes: The Spearman rank correlation analysis algorithm was used to calculate the correlation between the initial daily oil production increase of the historical pump replacement wells and various influencing factors; the calculation expression of the correlation degree is: In the formula, r is the degree of correlation; d i is the order difference of two data; n is the number of data items of historical pump replacement wells; The calculated correlations were sorted, and factors with correlations greater than 0.3 were identified as the main factors affecting the daily oil increase in the initial stage of pump replacement.

5. The method for optimizing displacement of oil well pump replacement based on machine learning according to claim 1, characterized in that: The establishment of the machine learning prediction model for daily oil increase in the initial stage of pump replacement includes: The values ​​of each main control factor are normalized according to the maximum and minimum value method; the normalization formula is: Where: x 归一 It is the normalized value of a certain main control factor; x is the actual value of a certain main control factor; x max is the maximum value of a main controlling factor in the sample data; x min is the minimum value of a main control factor in the sample data; Design a machine learning algorithm to conduct machine learning between the initial daily oil increase of historical pump replacement wells and the normalized main control factor values; The ratio of the number of historical data for training and testing is set, and the daily oil increase prediction model for the initial stage of pump replacement is determined based on the principle of the highest test compliance rate.

6. The method for optimizing displacement of oil well pump replacement based on machine learning according to claim 5, characterized in that: The machine learning algorithms include decision tree, random forest, support vector machine, multi-layer perceptron neural network, integrated boosting tree regression, gradient descent regression, bagging and stacking integrated regression algorithms; The ratio of the number of historical data items for training and testing is set to 8:

2.

7. The method for optimizing displacement of oil well pump replacement based on machine learning according to claim 1 or 5, characterized in that: The method for determining the displacement ratio when the initial daily oil increase is maximum under the condition that the water content of the oil well does not increase after the pump is replaced comprises the following steps: 7.1 For the wells that are going to change pumps, calculate the values ​​of the main factors that affect the daily oil increase in the initial stage of pump change; set the displacement ratio, according to the setting range of the displacement ratio, with 0.1 as the step, set n values; the setting range of the displacement ratio is 1 to 3; 7.2 After normalizing the values ​​of the main factors of different displacement ratios and other calculated daily oil increase in the initial stage of pump replacement, input them into the optimally determined daily oil increase prediction model for pump replacement, a total of n groups, and n initial daily oil increase values ​​for pump replacement can be output; 7.3 According to n displacement ratios, calculate the daily liquid production of the oil well after the pump is replaced; and then calculate the water content after the pump is replaced; 7.4 Finally, the n initial daily oil increase values ​​and water content values ​​after the pump replacement are obtained, and the displacement ratio with a water content not higher than that before the pump replacement and the largest daily oil increase is selected as the optimal displacement ratio.

8. The method for optimizing displacement of oil well pump replacement based on machine learning according to claim 7, characterized in that: The calculation formula for the daily liquid production of the oil well after the pump is replaced is: q 后 =q 前 *k (3) Where: q 后 This is the daily fluid produced after the pump is replaced; q 前 This is the daily fluid production before the pump is replaced; k is the different displacement ratios set.

9. The method for optimizing displacement of oil well pump replacement based on machine learning according to claim 7 or 8, characterized in that: The water content calculation formula after the pump is changed is: Where: fw is the water content after the pump is replaced; q o前 To change the former Nissan oil; q o增 This is the initial daily oil increase after replacing the pump.

10. The method for optimizing displacement of oil well pump replacement based on machine learning according to claim 1 or 7, characterized in that: The method for determining the optimal displacement after pump replacement is to multiply the displacement before pump replacement by the optimal displacement ratio to obtain the optimal displacement after pump replacement.

Images