A karst reservoir leakage risk assessment method based on random forest-comprehensive fuzzy evaluation

Abstract

The present invention relates to the technical field of karst reservoir leakage risk assessment, and provides a karst reservoir leakage risk assessment method based on random forest-comprehensive fuzzy evaluation, comprising the following steps: S10, collecting influencing factors of karst reservoir leakage, including river valley morphology, leakage type, stratum lithology, karst development degree, fracture development degree, karst hydrodynamic condition type, reservoir capacity, underground watershed water level, fracture development degree; S20, analyzing the influencing factors of karst reservoir leakage, which are divided into qualitative description features, qualitative variable features and quantitative features; S30, quantifying and dividing non-quantitative influencing factors into grades, and the influencing factors are divided into 3-5 grades, and specific divisions are carried out according to different factors. The method of the present invention can ensure the safe operation of the reservoir, and through the combination of random forest model and comprehensive fuzzy evaluation model, the information of various influencing factors is fully utilized to improve the accuracy and reliability of karst reservoir leakage risk assessment.

Description

Technical Field

[0001] The invention relates to the technical field of karst reservoir leakage risk assessment, and in particular to a karst reservoir leakage risk assessment method based on random forest-comprehensive fuzzy evaluation. Background Art

[0002] Karst reservoir is a special type of reservoir, and its geological conditions and hydrological characteristics make it have a certain leakage risk. At present, the research on the leakage risk assessment of karst reservoirs mainly focuses on qualitative description and empirical judgment.

[0003] Traditional leakage risk assessment methods mainly rely on expert experience and statistical data, lacking systematic quantitative analysis. In addition, due to the complex and diverse geological conditions and hydrological characteristics of karst reservoirs, existing assessment methods are difficult to meet the needs of different types of karst reservoirs. Summary of the invention

[0004] In view of the shortcomings of the prior art, the present invention provides a karst reservoir leakage risk assessment method based on random forest-integrated fuzzy evaluation, which solves the problem that the prior art relies on long-term statistical data and expert experience.

[0005] To achieve the above objectives, the present invention is implemented through the following technical solutions: a karst reservoir leakage risk assessment method based on random forest-comprehensive fuzzy evaluation, comprising the following steps:

[0006] S10. Collect the influencing factors of karst reservoir leakage, including river valley morphology, leakage type, stratum lithology, karst development degree, fracture development degree, karst hydrodynamic condition type, reservoir capacity, underground watershed water level, fracture development degree;

[0007] S20. Analyze the influencing factors of karst reservoir leakage, which are divided into qualitative description characteristics, qualitative variable characteristics and quantitative characteristics;

[0008] S30. Quantitatively classify the non-quantitative influencing factors into levels 3-5, and make specific classifications according to different factors;

[0009] S40. Quantify the weights of factors affecting karst reservoirs using random forest model;

[0010] S50. Establish a karst reservoir leakage evaluation index system;

[0011] S60. Based on the weights of influencing factors determined by the random forest model and combined with the comprehensive fuzzy model, the leakage risk assessment of karst reservoirs is realized.

[0012] Preferably, the qualitative description feature in step S20 is: qualitative evaluation of influencing factors, including river valley morphology, seepage channels and karst hydrodynamic condition types;

[0013] The characteristics of qualitative variables are as follows: the influencing factors have relatively clear change characteristics, including bedrock permeability characteristics, fracture development degree, anti-seepage treatment and bedrock permeability characteristics;

[0014] Quantitative characteristics are: specific data quantifying the characteristics of influencing factors, including reservoir capacity, seepage head, underground watershed water level, bedrock karst development degree and stratigraphic lithology characteristics.

[0015] Preferably, in the step S30, when the non-quantitative influencing factors are graded, the greater the potential possibility of leakage, the higher the grade and the larger the quantitative value, and vice versa, the lower the grade and the smaller the quantitative value.

[0016] Preferably, in step 4 of S40, when using the random forest model to quantify the influencing factors of karst reservoirs, 70% of the samples are used as training samples and 30% of the samples are used as test samples; the random forest algorithm parameters are adjusted, and the tree depth and number of iterations are cyclically searched; it is determined whether the value of the loss function MAE on the validation set has reached the minimum, and if not, the previous step is repeated; and finally, the contribution rate of the influencing factors is output.

[0017] Preferably, in step S50, evaluation criteria are established through standards, literature and expert consultation.

[0018] Preferably, in step S60, the weights of the influencing factors are determined according to the random forest model, and the leakage risk of the karst reservoir is evaluated in combination with the comprehensive fuzzy model.

[0019] The fuzzy comprehensive evaluation method is mainly carried out in the following steps:

[0020] S1. Determine the comprehensive evaluation factor set: The first-level factor set is U = [U1, U2, ..., U n ], the second-level factor set is U i =[U i1 , U i2 , ..., U im ].

[0021] S2. Establish a comprehensive evaluation comment set: Combined with the specifications and experts' suggestions, the results of karst reservoir leakage are divided into 4 levels, and the comment set is V = [V1, V2, V3, V4] = [no leakage, small leakage risk, large leakage risk, large leakage risk], and the corresponding scores are 90, 70, 50 and 20 points, respectively, and the corresponding comprehensive evaluation score ranges are 100-80, 80-60, 60-40 and 40-0 points;

[0022] S3. Model building: The weights of different indicators in the secondary evaluation index are W j =(W j1 , W j2 , ..., W jn ), the membership matrix is:

[0023]

[0024] The single-level evaluation model is:

[0025] B j =W j ·R j

[0026] B j is the result of the single factor evaluation of the first-level evaluation index, so the membership matrix of the first-level evaluation index can also be obtained. The weight vector of the first-level index is W = (W1, W2, ..., W n ), the membership matrix is:

[0027]

[0028] Then, the comprehensive evaluation model is:

[0029] B=W·R

[0030] B is the comprehensive evaluation result.

[0031] The present invention provides a karst reservoir leakage risk assessment method based on random forest-comprehensive fuzzy evaluation. It has the following beneficial effects:

[0032] The present invention can help reservoir managers to timely discover potential leakage risks, take corresponding prevention and control measures, and effectively ensure the safe operation of the reservoir. Through the combination of random forest model and comprehensive fuzzy evaluation, it can make full use of the information of various influencing factors, improve the accuracy and reliability of karst reservoir leakage risk assessment, and can be applied to various types of karst reservoirs, with good versatility and applicability. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 It is a schematic diagram of the karst reservoir leakage evaluation index system of the present invention;

[0034] Figure 2 This is a schematic diagram of the weights and ranking of the factors affecting the leakage of karst reservoirs according to the present invention;

[0035] Figure 3 It is a schematic diagram of fuzzy comprehensive evaluation of the present invention. DETAILED DESCRIPTION

[0036] The technical solution of the present invention will be described clearly and completely below in conjunction with the accompanying drawings of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0037] Example:

[0038] The embodiment of the present invention provides a karst reservoir leakage risk assessment method based on random forest-comprehensive fuzzy evaluation, comprising the following steps:

[0039] S10. Collect the influencing factors of karst reservoir leakage, including river valley morphology, leakage type, stratum lithology, karst development degree, fracture development degree, karst hydrodynamic condition type, reservoir capacity, underground watershed water level, fracture development degree;

[0040] S20. Analyze the influencing factors of karst reservoir leakage, which are divided into qualitative description characteristics, qualitative variable characteristics and quantitative characteristics;

[0041] S30. Quantitatively classify the non-quantitative influencing factors into levels 3-5, and make specific classifications according to different factors;

[0042] S40. Determine the weights of factors affecting karst reservoirs using the random forest model;

[0043] S50. Establish a karst reservoir leakage evaluation index system;

[0044] S60. Based on the weights of influencing factors determined by the random forest model and combined with the comprehensive fuzzy model, the leakage risk assessment of karst reservoirs is realized.

[0045] By combining the random forest model with the comprehensive fuzzy model, the information of various influencing factors is fully utilized to improve the accuracy and reliability of the karst reservoir leakage risk assessment.

[0046] The qualitative description features in step S20 are: qualitative evaluation of influencing factors, including valley morphology, seepage channels, and types of karst hydrodynamic conditions;

[0047] The characteristics of qualitative variables are as follows: the influencing factors have relatively clear change characteristics, including bedrock permeability characteristics, fracture development degree, anti-seepage treatment and bedrock permeability characteristics;

[0048] Quantitative characteristics are: specific data quantifying the characteristics of influencing factors, including leakage head, underground watershed water level, bedrock karst development degree and stratigraphic lithology characteristics.

[0049] In step S30, when the non-quantitative influencing factors are graded, the greater the potential possibility of leakage, the higher the grade and the larger the quantitative value, and vice versa, the lower the grade and the smaller the quantitative value.

[0050] In step 4 of S40, when using the random forest model to quantify the influencing factors of karst reservoirs, 70% of the samples are used as training samples and 30% of the samples are used as test samples; the random forest algorithm parameters are adjusted, and the tree depth and number of iterations are searched cyclically; it is determined whether the value of the loss function MAE on the validation set has reached the minimum, and if not, the previous step is repeated; finally, the contribution rate of the influencing factors is output.

[0051] In step S50, the evaluation criteria are established according to the specifications, literature and expert consultation. Figure 1 .

[0052] In step S60, the weights of the influencing factors are determined according to the random forest model, combined with the comprehensive fuzzy model, and then the leakage risk of the karst reservoir is evaluated.

[0053] Through multi-level quantification and qualitative description of different types of different features, after being introduced into the random forest model, it can help reservoir managers to timely discover potential leakage risks and take corresponding prevention and control measures through machine learning prediction. It can be applied to different types of karst reservoirs, has good versatility and applicability, and can be adjusted and applied according to actual conditions.

[0054] The fuzzy comprehensive evaluation method is mainly carried out in the following steps:

[0055] S1. Determine the comprehensive evaluation factor set: The first-level factor set is U = [U1, U2, ..., U n ], the second-level factor set is U i =[U i1 , U i2 , ..., U im ];

[0056] S2. Establish a comprehensive evaluation comment set: Combined with the specifications and experts' suggestions, the results of karst reservoir leakage are divided into 4 levels, and the comment set is V = [V1, V2, V3, V4] = [no leakage, small leakage risk, large leakage risk, large leakage risk], and the corresponding scores are 90, 70, 50 and 20 points, respectively, and the corresponding comprehensive evaluation score ranges are 100-80, 80-60, 60-40 and 40-0 points;

[0057] S3. Model building: The weights of different indicators in the secondary evaluation index are W j =(W j1 , W j2 , ..., W jn ), the membership matrix is:

[0058]

[0059] The single-level evaluation model is:

[0060] B j =W j ·R j

[0061] B j is the result of the single factor evaluation of the first-level evaluation index, so the membership matrix of the first-level evaluation index can also be obtained. The weight vector of the first-level index is W = (W1, W2, ..., W n ), the membership matrix is:

[0062]

[0063] Then, the comprehensive evaluation model is:

[0064] B=W·R

[0065] B is the comprehensive evaluation result.

[0066] The classification and quantitative standard table of factors affecting karst reservoir leakage is as follows:

[0067]

[0068]

[0069] According to the specifications, literature and expert consultation, the evaluation criteria are established as shown in the following table:

[0070]

[0071] Note: The percentage in F2 represents the proportion of carbonate rock content.

[0072] The random forest model is used for training and fitting, and the results are as follows:

[0073] project MSE RMSE MAE MAPE <![CDATA[R 2 ]]> Training set 1.346 1.16 0.653 31.3 0.967 Test Set 11.028 3.321 1.721 64.604 0.77

[0074] Model evaluation results:

[0075] In the step S40, the weights and rankings of the factors affecting karst reservoir leakage are obtained. Figure 2 shown.

[0076] According to the results of the random forest model, the leakage risk of karst reservoirs was assessed by combining the comprehensive fuzzy evaluation model:

[0077] On the basis of determining the weight of each factor and the fuzzy comprehensive evaluation method, several experts in karst hydrogeology were invited to grade the factors affecting the leakage risk of a karst reservoir according to the evaluation criteria in the evaluation standard table. The membership degree of each indicator was calculated using the assignment type, and the results are shown in the table below.

[0078]

[0079] The fuzzy comprehensive evaluation of geological characteristics, karst characteristics, hydrogeology, topography and engineering measures was calculated and the results are shown in the attached Figure 3 shown.

[0080] The result of the target layer (leakage risk of a karst reservoir) is calculated as [0.518, 0.371, 0.064, 0.052]. Then, according to the maximum membership principle, the corresponding comment set scores are 90, 70, 50 and 20 points respectively, and the fuzzy comprehensive score is calculated as follows:

[0081] 0.518×20+0.371×50+0.064×70+0.052×90=38.07;

[0082] The evaluation results show that a certain karst reservoir has a high risk of leakage and further human intervention is needed to reduce the risk of leakage.

[0083] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A karst reservoir leakage risk assessment method based on random forest-comprehensive fuzzy evaluation, characterized in that: The following steps are involved: S10. Collect the influencing factors of karst reservoir leakage, including river valley morphology, leakage type, stratum lithology, karst development degree, fracture development degree, karst hydrodynamic condition type, reservoir capacity, underground watershed water level, fracture development degree; S20, analyzing the influencing factors of karst reservoir leakage, which are divided into qualitative description characteristics, qualitative variable characteristics and quantitative characteristics; the qualitative description characteristics in the step S20 are: qualitative evaluation of the influencing factors, including river valley morphology, leakage channel and karst hydrodynamic condition type; The characteristics of qualitative variables are as follows: the influencing factors have relatively clear change characteristics, including bedrock permeability characteristics, fracture development degree, anti-seepage treatment and bedrock permeability characteristics; Quantitative characteristics are: specific data quantifying the characteristics of influencing factors, including reservoir capacity, seepage head, underground watershed level, bedrock karst development degree and stratum lithology characteristics; S30, quantifying and classifying the non-quantitative influencing factors into 3-5 levels, and performing specific classification according to different factors; in the step S30, when the non-quantitative influencing factors are classified into levels, the greater the potential possibility of leakage, the higher the level, the greater the quantitative value, and vice versa, the lower the level, the smaller the quantitative value; S40, using the random forest model to quantify the weights of the influencing factors of karst reservoirs; in the step S40, when using the random forest model to quantify the influencing factors of karst reservoirs, 70% of the samples are used as training samples and 30% of the samples are used as test samples; adjust the parameters of the random forest algorithm, and cyclically search the tree depth and the number of iterations; determine whether the value of the loss function MAE on the validation set has reached the minimum, and if not, repeat the previous step; finally, output the contribution rate of the influencing factors; in the step S40, when using the random forest model to quantify the influencing factors of karst reservoirs, 70% of the samples are used as training samples and 30% of the samples are used as test samples; adjust the parameters of the random forest algorithm, and cyclically search the tree depth and the number of iterations; determine whether the value of the loss function MAE on the validation set has reached the minimum, and if not, repeat the previous step; finally, output the contribution rate of the influencing factors; S50. Establish a karst reservoir leakage evaluation index system; S60. Based on the weights of influencing factors determined by the random forest model and combined with the comprehensive fuzzy model, the leakage risk assessment of karst reservoirs is realized; In the step S60, the weights of the influencing factors are determined according to the random forest model, and the leakage risk of the karst reservoir is evaluated in combination with the comprehensive fuzzy model; The fuzzy comprehensive evaluation method is mainly carried out in the following steps: S1. Determine the comprehensive evaluation factor set: The first-level factor set is U=[ U 1, U 2,……, U n ], the second-level factor set is U i =[ U i1 , U i2 ,……, U im ]; S2. Establish a comprehensive evaluation comment set: Combined with the specifications and experts' suggestions, the results of karst reservoir leakage are divided into 4 levels. The comment set is V =[ V 1, V 2, V 3. V 4]=[no leakage, small leakage risk, large leakage risk, large leakage risk], the corresponding scores are 90, 70, 50 and 20 points, and the corresponding comprehensive evaluation score ranges are 100-80, 80-60, 60-40 and 40-0 points; S3. Model building: The weights of different indicators in the secondary evaluation index are W j =( W j1 , W j2 ,……, W jn ), the membership matrix is: The single-level evaluation model is: B j = W j · R j B j is the result of the single factor evaluation of the first-level evaluation index, so the membership matrix of the first-level evaluation index can also be obtained; the weight vector of the first-level index is W =( W 1, W 2,……, W n ), the membership matrix is: Then, the comprehensive evaluation model is: B = W · R B It is the comprehensive evaluation result.

2. The method for assessing the leakage risk of karst reservoirs based on random forest-comprehensive fuzzy evaluation according to claim 1 is characterized in that: In the step S50, evaluation criteria are established through standards, literature and expert consultation.

Images