A grading evaluation method for offshore oilfield development effect

Abstract

The present invention relates to a hierarchical evaluation method for offshore oilfield development effect, comprising: S1, using a SHAP model to perform positive and negative judgment and normalization on each indicator of a neural network model with preset multiple input indicators, and fitting the impact of the indicators output by the SHAP model; S2, calculating the proportion of samples under each indicator, calculating the entropy value of a preset number of features formed by sample distribution, so as to obtain the weight of each indicator; S3, based on the weight of each indicator, using a grey comprehensive evaluation method to determine the indicator weight and the grey correlation coefficient of the indicator itself to perform an overall comprehensive evaluation; and S4, using a fuzzy comprehensive evaluation method to determine the hierarchical evaluation result of the offshore oilfield.

Description

Technical Field

[0001] The present invention relates to the technical field of oil and gas production equipment, and in particular to a method for grading and evaluating the development effects of different types of offshore oil fields, and belongs to the field of deep mining of big data value in offshore oil field development. Background Art

[0002] The overall offshore oil fields are in the medium-high water content stage, and it is difficult to further control the decline rate and water content increase rate, and the situation of sustained stable production is severe. Looking at the producing oil fields in the whole sea area, the development effects of different types of oil fields and different development stages are different. Therefore, it is of great significance to deeply understand the development laws and main controlling factors of different types of oil fields and different development stages, and systematically propose a set of grading evaluation indicators, evaluation standards and evaluation systems for the development effects of different types of oil fields, so as to continuously improve the level of refined reservoir management, accurately formulate technical policies and adjust the direction of potential tapping, and further improve the development effects of different types of oil fields.

[0003] However, the existing oilfield development level classification system relies on the calculation and evaluation of single indicators, which is relatively mature, but the overall classification evaluation is still in the semi-quantitative stage and has not yet achieved quantitative comprehensive evaluation. It ignores the differences in importance between indicators and the internal influencing factors of each indicator and treats them equally. Therefore, it is necessary to systematically propose a set of comprehensive evaluation methods to quantitatively judge the development effect and development status of oilfields and lay the foundation for the comprehensive evaluation system of offshore oilfield development effect. Summary of the invention

[0004] The present invention provides a method for grading and evaluating the development effect of offshore oil fields. The Shapley Additive ExPlanations (SHAP) method is used to interpret the artificial neural network model, and the SHAP method is used to study the response relationship between input indicators and key development indicators, and a new set of empirical formulas is fitted through the explained quantitative relationship. This method avoids the problems of individual experience differences among experts and lack of individual knowledge in questionnaires.

[0005] To achieve the above object, the present invention adopts the following technical solutions:

[0006] In a first aspect, the present application provides a method for grading and evaluating the development effect of an offshore oil field, comprising:

[0007] S1, using the SHAP model to make positive and negative judgments and normalize each indicator of the neural network model with preset multiple input indicators, and fitting the impact of the indicators output by the SHAP model;

[0008] S2, calculate the proportion of samples under each indicator, calculate the entropy value of the preset number of features formed by the sample distribution, and thus obtain the weight of each indicator;

[0009] S3, based on the weight of each indicator, the indicator weight determined by the grey comprehensive evaluation method and the grey correlation coefficient of the indicator itself are used for overall comprehensive evaluation; and

[0010] S4, the fuzzy comprehensive evaluation method is used to determine the classification evaluation results of offshore oil fields.

[0011] In one implementation, the S1 includes:

[0012] S11, using the entropy weight method to make positive and negative judgments on each indicator, and performing normalized calculations on the indicators based on the results of the positive and negative judgments;

[0013] S12, fitting the functional relationship between each indicator and the output, and assigning corresponding weights to each indicator according to the order of the indicator sequence, using linear superposition fitting based on weights for all indicators.

[0014] In one implementation, the normalized calculation method includes:

[0015] The normalization formula of the positive indicator is as follows:

[0016]

[0017] The normalization formula for negative indicators is:

[0018]

[0019] In one implementation, the S2 specifically includes:

[0020] There are n samples, each with m-dimensional features. The formula for calculating the proportion of each value in the i-th dimension is as follows:

[0021]

[0022] After calculating the proportion, the entropy value of each feature can be calculated. The specific calculation formula of the entropy value is as follows:

[0023]

[0024] k=1 / ln(n)

[0025] e j That is the entropy value of the jth indicator;

[0026] Entropy redundancy d is used to determine the weight j Assign weights:

[0027] d j =1-e j

[0028]

[0029] The final W j is the weight of the jth feature.

[0030] In one implementation, the S3 includes:

[0031] S31, selecting the optimal value of the indicator based on the preset strategy;

[0032] S32, generating an optimal value sequence according to the selection result of the optimal value of the indicator, and calculating the grey correlation degree between the indicator and the optimal value sequence; and

[0033] S33, calculating the quantitative score of the overall comprehensive evaluation according to the determined weight and grey correlation degree.

[0034] In one implementation, in S31, for the positive indicator and the negative indicator, the maximum value and the minimum value in the feature data are selected as the optimal value of the indicator through the positive and negative characteristics.

[0035] In one implementation, in S33, the calculation formula is:

[0036]

[0037] where w j is the weight determined by the entropy weight method, r j is the grey relational coefficient, and m is the dimension of the input feature.

[0038] In one implementation, S4 specifically includes:

[0039] S41, determining the factor set and weights;

[0040] S42, determining a comment set for offshore oilfield classification evaluation, wherein the comment set includes a set number of level classification results;

[0041] S43, determining the membership function of each level of classification results, thereby obtaining a membership matrix as an evaluation matrix for fuzzy evaluation calculation;

[0042] S44, calculating a fuzzy evaluation result based on the evaluation matrix calculated by the fuzzy evaluation.

[0043] In one implementation, the negative indicators include the moisture content, and the moisture content is used to determine the level classification results in the comment set.

[0044] The technical solution of the present invention proposes a hierarchical evaluation method for oilfield development effect based on entropy weight method, gray comprehensive evaluation method, fuzzy comprehensive evaluation method and SHAP network interpretation model. First, the entropy weight method is used to judge the positive and negative direction of the indicators, and then the SHAP network model is used to perform linear fitting on the output indicator impact. Finally, the gray comprehensive rating method and fuzzy comprehensive evaluation method are used to evaluate and score the indicators. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 is a SHAP flow chart in one embodiment of the present invention;

[0046] Figure 2 It is the SHAP schematic diagram;

[0047] Figure 3 is a SHAP feature correlation summary diagram in one embodiment of the present invention;

[0048] Figure 4 is a dependency graph of input features and output features in one embodiment of the present invention;

[0049] Figure 5 This is a comparison chart between the SHAP score ranking of each oil field in a calculation example and the actual oil field ranking. DETAILED DESCRIPTION

[0050] In order to make the purpose, technical solution and advantages of the embodiment of the present invention clearer, the technical solution of the embodiment of the present invention will be clearly and completely described below in conjunction with the drawings of the embodiment of the present invention. Obviously, the described embodiment is a part of the embodiment of the present invention, not all of the embodiments. Based on the described embodiment of the present invention, all other embodiments obtained by ordinary technicians in this field belong to the scope of protection of the present invention.

[0051] In view of the defects and problems of the prior art, the present application provides a method for grading and evaluating the development effect of an offshore oil field, comprising:

[0052] S1, using the SHAP model to make positive and negative judgments and normalize each indicator of the neural network model with preset multiple input indicators, and fitting the impact of the indicators output by the SHAP model;

[0053] S2, calculate the proportion of samples under each indicator, calculate the entropy value of the preset number of features formed by the sample distribution, and thus obtain the weight of each indicator;

[0054] S3, based on the weight of each indicator, the indicator weight determined by the grey comprehensive evaluation method and the grey correlation coefficient of the indicator itself are used for overall comprehensive evaluation; and

[0055] S4, the fuzzy comprehensive evaluation method is used to determine the classification evaluation results of offshore oil fields.

[0056] The above method is described below in a more detailed embodiment in conjunction with more drawings.

[0057] The present invention proposes a method for grading and evaluating the development effects of different types of offshore oil fields, and the specific steps are as follows:

[0058] S1: When using the entropy weight method to normalize indicators, it is necessary to make positive and negative judgments on the indicators. The SHAP network interpretation model can be used to make judgments on positive and negative indicators, and perform linear fitting on the impact of the indicators output by the SHAP model.

[0059] Step S1 includes:

[0060] S11: Data normalization. The entropy weight method operation first needs to normalize the data. Different from the standard normalization of other data, the entropy weight method needs to first make a positive or negative judgment on the indicator when normalizing the indicator. That is, the larger the positive indicator, the better the effect, and the smaller the negative indicator, the better the effect. The calculation method is also different when normalizing positive and negative indicators.

[0061] The normalization formula for positive indicators is as shown in formula (1), and the normalization formula for negative indicators is as shown in formula (2):

[0062]

[0063] For positive indicators, the closer the data is to the maximum value, the larger the value is after normalization. For negative indicators, the closer the data is to the maximum value, the smaller the value is after normalization.

[0064] S12: Construction of SHAP network interpretation model. The SHAP model is based on Shapely Value, a contribution calculation method based on game theory. It is used to calculate the contribution of participating indicators in the process of obtaining the final result. When calculating, it is necessary to take into account each order of joining the participating indicators, and assign weights to the contributions according to the order of joining. The global contribution of the participating indicator is finally obtained, which is the weighted average of all contributions. The size of the weight setting is related to the order of participation of the participating indicators, that is, the weight obtained by participating first is higher.

[0065] The SHAP model can explain the deep learning model from both global and local perspectives. The local explanation process usually explains how the input indicators in a single or multiple predictions affect the generation of prediction results, while the global explanation explains the input and output of the entire model. The basic process of the SHAP model is as follows: Figure 1 and Figure 2 As shown. Figure 3As shown, each row represents a feature, each point in the row represents a sample, and the contribution of the sample is evaluated by the color of the sample point (see Figure 4 ), the closer to red in the figure, the greater the contribution. In addition to calculating the global contribution, the SHAP model also supports drawing the relationship between the target and the feature. This method can intuitively reflect the relationship between the feature and the output, such as linear and nonlinear relationships, positive and negative correlations, and monotonic and complex correlations.

[0066] Through the correlation between this indicator and the output provided by SHAP, the function formula f can be fitted i (x), and set the weight w according to the correlation coefficient of the formula i Finally, several feature combinations with strong correlation are selected and fitted into a new empirical formula (Formula 3), which realizes the interpretation of the deep learning model, provides theoretical support for improving the effect of oilfield development, and can apply the empirical formula to similar scenarios to improve work efficiency. When it is needed, the prediction formula can be efficiently obtained by combining specific data.

[0067]

[0068] S2: Use the entropy weight method to calculate the proportion of each sample of each indicator in the data, and then calculate the entropy value of each feature to obtain the weight of each indicator.

[0069] The specific content of S2 is:

[0070] Before calculating the entropy value, it is necessary to calculate the proportion of each sample j of each indicator i in the data to determine the importance of the sample in the feature. Assuming there are n samples, each sample has m-dimensional features, the formula for calculating the proportion of each value in the i-th dimension in the i-th dimension is as follows:

[0071]

[0072] After calculating the proportion, the entropy value of each feature can be calculated. The specific calculation formula of the entropy value is as follows:

[0073] e j That is, the entropy value of the jth indicator. Because the weight judgment is inversely proportional to the entropy value, the entropy redundancy d is used to determine the weight. j Assign weights:

[0074]

[0075] The final W jThat is the weight of the jth feature, which is used in the subsequent grey comprehensive evaluation and fuzzy comprehensive evaluation algorithms.

[0076] S3: First, the positive and negative values ​​of the features are selected from the data to determine the maximum and minimum values ​​of each feature as the optimization index. Next, the grey correlation between each feature and its corresponding maximum and minimum values ​​is calculated. Finally, the index weight determined by the entropy weight method is used to conduct a comprehensive evaluation with the grey correlation of each indicator to obtain the overall evaluation result.

[0077] Step S3 includes:

[0078] S31: Selection of optimal indicator values. First, it is necessary to clarify the optimal indicator among all indicators, that is, the value of each development indicator under the optimal development effect. There are two methods to determine the optimal value of the indicator. One is to follow the unified standard that has reached a consensus, such as the numerical range specified in the "Oilfield Development Level Classification" standard. The second is to follow the law of indicator change, that is, the positive and negative indicators mentioned in the entropy weight method, and select the maximum and minimum values ​​in the feature data as the optimal value of the indicator through the positive and negative characteristics. The selection of the optimal value of the indicator will have a greater impact on the final comprehensive evaluation results.

[0079] S32: Grey correlation analysis. In the grey comprehensive evaluation process, the parent sequence is no longer the sequence of indicators to be predicted, but a new sequence composed of the optimal values ​​of each indicator. The grey correlation between each indicator and the sequence of the optimal values ​​of the indicators is calculated and finally used as the evaluation matrix in the grey comprehensive evaluation process.

[0080] S33: Comprehensive evaluation and scoring. The overall comprehensive evaluation is carried out by combining the indicator weights determined by the entropy weight method with the grey correlation coefficient of the indicator itself. The final score of the grey comprehensive evaluation is:

[0081]

[0082] where w j is the weight determined by the entropy weight method, r j is the grey relational coefficient, and m is the dimension of the input feature.

[0083] S4: First, the factor set and weights are determined using the entropy weight method. Then the comment set is determined, and the corresponding membership function is determined for each level, and then the membership degree is calculated. Finally, the obtained fuzzy evaluation matrix is ​​combined with the weights calculated by the entropy value method to obtain the membership degree of a certain evaluation.

[0084] Step S4 includes:

[0085] S41: First, the factor set and weights need to be determined, which is done using the entropy weight method in S1 and S2.

[0086] S42: Next, you need to determine the comment set, which is the category for scoring, similar to {"excellent", "good", "medium", "poor"}, and each level should have a corresponding range. For example, the moisture content is divided into three stages {"good", "relatively good", "average"}, and the corresponding value for each stage is {≤60%, 69-90%, >90%}.

[0087] S43: Finally, each level needs to have its corresponding membership function. Taking moisture content as an example, the rectangular membership function is selected for division:

[0088] The membership function of moisture content to “good” is:

[0089]

[0090] The membership function of moisture content to “better” is:

[0091]

[0092] The membership function of moisture content to “difference” is:

[0093]

[0094] For each piece of input feature data, the membership degree of the three comments needs to be calculated separately, and finally the fuzzy evaluation matrix R is obtained:

[0095]

[0096] where r ij The membership of the i-th indicator of the input evaluation index to the indicator comment i. The membership of the whole to a comment can be obtained by combining the fuzzy evaluation matrix with the weight calculated by the entropy method, and the comment with the largest membership is selected as the overall comment according to the maximum membership principle.

[0097] Here is a direct example of the above method:

[0098] In an offshore oil field, when constructing the best indicator sequence, the SHAP network can be used to explain the model analysis results. By judging the positive and negative directions of the indicators, the maximum value of the positive indicator and the minimum value of the negative indicator are selected as the optimal values. The comprehensive evaluation indicators are shown in Table 1:

[0099]

[0100] Table 1

[0101] After determining the positive and negative directions of the indicators, grey correlation analysis can be performed to obtain the grey correlation coefficient between each indicator and the optimal state of the indicator in the oil field production process. Combined with the weights of each indicator calculated by the entropy weight method, the grey comprehensive evaluation score of the oil field can be calculated. The following are the calculation results of the grey evaluation coefficient of the oil field and the weight of the entropy weight method:

[0102]

[0103]

[0104] Table 2

[0105] The larger the grey evaluation coefficient is, the closer the indicator is to the optimal development effect, while the larger the entropy weight method is, the more stable the overall fluctuation of the indicator is. For extremely unstable indicators, the entropy weight method will assign negative weights, indicating that the overall development effect has been affected.

[0106] The score calculated by the grey comprehensive evaluation is a quantitative evaluation and cannot directly and qualitatively evaluate the effect of oilfield development. In order to qualitatively evaluate the development effect of oilfields, other oilfields are selected for grey comprehensive evaluation and scoring, and then compared with the target oilfield. After comparison, the target oilfield has the highest overall comprehensive evaluation score of 0.9026, so it is classified as level 1.

[0107]

[0108] Table 3

[0109] The determination of the membership function refers to the standard of "Oilfield Development Effect Level Classification", and the membership function is determined by clearly defining the relevant standardized indicators in the standard. First, according to the standard of "Oilfield Development Effect Level Classification", combined with the indicator library constructed in this paper, the evaluation indicators are selected, and the evaluation level is divided into 3 levels according to the national standard. The specific indicator selection and the details of each indicator division are shown in the following table.

[0110]

[0111]

[0112] Table 4

[0113] Since all indicators are divided into three stages, combined with the basic form of the membership function, the present invention intends to use a triangular membership function as the membership function of all indicators, where the comprehensive decline rate, natural decline rate, and water content index are negative indicators from the stage definition, that is, the smaller the value, the better the oil field development effect. Therefore, the relevant calculation method of negative indicators will be used in the design of the membership function and data calculation. The hierarchical membership matrix of each indicator of the oil field can be obtained by bringing the oil field production data into each membership function. The development level of the corresponding indicator of the oil field can be judged by the size of the membership of each indicator, providing a theoretical basis for the subsequent adjustment of the oil field. Here, the oil field production data is the final data of the oil field production.

[0114] Since all indicators are divided into three stages, and combined with the basic form of the membership function, the present invention adopts a triangular membership function as the membership function of all indicators. Among them, the comprehensive decline rate, natural decline rate and water content indicators are negative indicators, that is, the smaller the value, the better the oil field development effect. Therefore, the relevant calculation method of negative indicators will be adopted in the design of the membership function and data calculation. Substituting the oil field production data into each membership function, the hierarchical membership matrix of each indicator of the oil field can be obtained. According to the size of the membership of each indicator, the development level of the corresponding indicator of the oil field can be judged, providing a theoretical basis for the subsequent adjustment of the oil field. The final data of the oil field production is used for calculation here.

[0115]

[0116] Table 5

[0117] The entropy weight method is used to determine the weight of each indicator. The overall development grade membership of the oil field can be obtained through matrix multiplication. The oil field development effect grade is determined according to the maximum membership principle. The overall membership of the oil field is finally calculated as follows. Based on the maximum membership principle, the final fuzzy comprehensive evaluation result of the target oil field is level one.

[0118]

[0119] Table 6

[0120] like Figure 5 As shown in FIG. 1 , the scoring ranking of each oil field calculated by combining the SHAP network interpretation model is basically consistent with the real oil field ranking, which shows the effectiveness of the method provided by the present invention.

[0121] In the several embodiments provided by the present invention, it should be understood that the disclosed method can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the above units is only a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0122] The above are only preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for grading and evaluating the development effect of an offshore oil field, characterized in that: include: S1, using the SHAP model to make positive and negative judgments and normalize each indicator of the neural network model with preset multiple input indicators, and fitting the impact of the indicators output by the SHAP model; S2, calculate the proportion of samples under each indicator, calculate the entropy value of the preset number of features formed by the sample distribution, and thus obtain the weight of each indicator; S3, based on the weight of each indicator, the indicator weight determined by the grey comprehensive evaluation method and the grey correlation coefficient of the indicator itself are used for overall comprehensive evaluation; as well as S4, the fuzzy comprehensive evaluation method is used to determine the classification evaluation results of offshore oil fields.

2. The offshore oilfield development effect classification evaluation method according to claim 1, characterized in that: The S1 includes: S11, using the entropy weight method to make positive and negative judgments on each indicator, and performing normalized calculations on the indicators based on the results of the positive and negative judgments; S12, fitting the functional relationship between each indicator and the output, and assigning corresponding weights to each indicator according to the order of the indicator sequence, using linear superposition fitting based on weights for all indicators.

3. The offshore oilfield development effect classification evaluation method according to claim 2, characterized in that: The normalized calculation method includes: The normalization formula of the positive indicator is as follows: The normalization formula for negative indicators is:

4. The offshore oilfield development effect classification evaluation method according to claim 2, characterized in that: The S2 specifically includes: There are n samples, each with m-dimensional features. The formula for calculating the proportion of each value in the i-th dimension is as follows: After calculating the proportion, the entropy value of each feature can be calculated. The specific calculation formula of the entropy value is as follows: k=1 / ln(n) e j That is the entropy value of the jth indicator; Entropy redundancy d is used to determine the weight j Assign weights: d j =1-e j The final W j is the weight of the jth feature.

5. The offshore oilfield development effect classification evaluation method according to claim 4, characterized in that: The S3 includes: S31, selecting the optimal value of the indicator based on the preset strategy; S32, generating an optimal value sequence according to the selection result of the optimal value of the indicator, and calculating the grey correlation degree between the indicator and the optimal value sequence; and S33, calculating the quantitative score of the overall comprehensive evaluation according to the determined weight and grey correlation degree.

6. The offshore oilfield development effect classification evaluation method according to claim 5, characterized in that: In S31, for the positive index and the negative index, the maximum value and the minimum value in the feature data are selected as the optimal value of the index through the positive and negative characteristics.

7. The offshore oilfield development effect classification evaluation method according to claim 5, characterized in that: In S33, the calculation formula is: where w j is the weight determined by the entropy weight method, r j is the grey relational coefficient, and m is the dimension of the input feature.

8. The offshore oilfield development effect classification evaluation method according to claim 4, characterized in that: S4 specifically includes: S41, determining the factor set and weights; S42, determining a comment set for offshore oilfield classification evaluation, wherein the comment set includes a set number of level classification results; S43, determining the membership function of each level classification result, thereby obtaining a membership matrix as an evaluation matrix for fuzzy evaluation calculation; S44, calculating a fuzzy evaluation result based on the evaluation matrix calculated by the fuzzy evaluation.

9. The offshore oilfield development effect classification evaluation method according to claim 8, characterized in that: Negative indicators include moisture content, and the moisture content is used to determine the level classification results in the comment set.

Images