Intelligent identification method for well test interpretation model

Abstract

The invention provides an intelligent identification method for a well test interpretation model, and belongs to the technical field of oil and gas field development, and the method comprises the following steps: constructing a well test interpretation model, and establishing a feature element set and an inference rule set corresponding to different flow stages; performing data preprocessing on the data of the well test interpretation model, storing the preprocessed data, and drawing the preprocessed data into a double logarithmic curve graph; according to the size of the initial segmentation window and the primitive discrimination threshold value, feature primitive extraction and curve tracking are carried out on the double logarithmic curve graph, and an initial recognition result is determined; and according to the morphological change of the double logarithmic curve graph and the primitive feature point information, screening and verifying the correctness of the initial identification result by using a DTS method, and completing the identification of the well test interpretation model. According to the method, the problems of poor well test interpretation efficiency and quality and the like caused by inaccurate identification of the current well test interpretation model are solved, the steps are simple, the model inference result is accurate, and the accuracy of the method is proved through software fitting result comparison.

Description

Technical Field

[0001] The invention relates to the technical field of oil and gas field development, and in particular to an intelligent identification method for a well test interpretation model. Background Art

[0002] Well test analysis is an effective method to describe the characteristics of oil and gas reservoirs. In well test analysis, the correct interpretation of production / pressure data can obtain reliable information about well type, inner boundary type, reservoir type and outer boundary type. In the process of well test interpretation, early, middle and late flow stages may appear, which are reflected in the orderly arrangement of time in the well test interpretation diagnostic diagram. Therefore, the identification of well test interpretation models can be regarded as a model classification problem consisting of flow stages corresponding to different time series. Since there are many types of well test interpretation models and a single model will produce multiple curve shapes due to the sensitivity of parameters, it increases the difficulty of interpretation for well test interpretation staff.

[0003] At present, some scholars are able to perform intelligent identification of well test interpretation models through two methods: feature point extraction and neural network. However, both methods still have certain limitations. For the feature point extraction method, the data processing process is relatively demanding and the steps are cumbersome. With the development of computer technology, the neural network method is used more and more widely, but this method requires a large number of samples to be collected for training, and the output results are relatively small. At present, there are two common problems in the intelligent identification methods of various well test interpretation models: 1) It is impossible to accurately handle the non-uniqueness of the interpretation results; 2) The accuracy of identification cannot be guaranteed.

[0004] In general, the current methods for identifying well test interpretation models have certain limitations. Therefore, there is an urgent need for an intelligent identification method suitable for well test interpretation models. Summary of the invention

[0005] In order to overcome the shortcomings of the prior art, the purpose of the present invention is to provide an intelligent identification method for a well test interpretation model to solve the problems of low efficiency and poor quality of well test interpretation caused by inaccurate identification of the current well test interpretation model. The steps are simple and the model inference results are accurate. The accuracy of the present invention is proved by comparing the software fitting results. In addition, the method also completes the division of the flow stage, laying a theoretical foundation for the convenient development of subsequent well test interpretation work.

[0006] To achieve the above object, the present invention provides the following solutions:

[0007] An intelligent identification method for a well test interpretation model comprises the following steps:

[0008] Construct well test interpretation model and establish feature primitives and inference rule sets corresponding to different flow stages;

[0009] Preprocessing the data of the well test interpretation model, saving the preprocessed data, and plotting the preprocessed data into a double logarithmic curve graph;

[0010] According to the initial segmentation window size and the primitive discrimination threshold, feature primitives are extracted and curve traced on the double logarithmic curve graph to determine the initial recognition result;

[0011] According to the morphological changes of the double logarithmic curve graph and the primitive feature point information, the DTS method is used to screen and verify the correctness of the initial recognition result to complete the recognition of the well test interpretation model.

[0012] Preferably, the construction of the well test interpretation model and the establishment of characteristic primitive sets and inference rule sets corresponding to different flow stages include:

[0013] Constructing the well test interpretation model according to the well type, inner boundary type, reservoir type, and outer boundary type;

[0014] Based on the pressure and pressure derivative data of the well test interpretation model, a double logarithmic curve graph is drawn, and different flow stages in the well test interpretation model are identified according to characteristic points and characteristic lines on the double logarithmic curve graph; the pressure and pressure derivative curves of the well test interpretation model are analyzed and synthesized to determine a characteristic primitive set;

[0015] The inference rule set is established based on the characteristic parameters of the well test interpretation model and the experience of actual interpretation.

[0016] Preferably, data preprocessing of the well test interpretation model data includes:

[0017] Acquire the pressure data of the well test interpretation model, perform wavelet transformation on the pressure data, use Daubechies 4 (db4) as an orthogonal wavelet transformation function to perform wavelet decomposition of the pressure data, use Sqtwolog as a threshold calculation rule to quantify the pressure data after wavelet decomposition, and finally perform wavelet reconstruction on the quantized pressure data to generate denoised pressure data;

[0018] Polishing the denoised pressure data and calculating pressure derivative data using a differential method;

[0019] Regularization is performed on the time, denoised pressure data, and pressure derivative data.

[0020] Preferably, the pressure derivative data is calculated using a differential method, including:

[0021] Get p i In t i The derivative at time satisfies |log(t)-log(ti )|≤d, determine the difference between the two points and then calculate. The derivative formula is expressed as:

[0022]

[0023] in, is the pressure derivative value at the calculation point; t i is the time corresponding to the calculation point; t1 is the time corresponding to calculation point 1; t2 is the time corresponding to calculation point 2; is the pressure corresponding to the calculation point; is the pressure corresponding to calculation point 1; is the pressure corresponding to calculation point 2; d

[0024] is the polishing coefficient.

[0025] Preferably, the time, denoised pressure data and pressure derivative data are normalized, and the normalized expression is:

[0026] p Ni ←[log(p i )-[log(min(p,p′))]×100 / [log(max(p,p′))-log(min(p,p′))]

[0027] p′ Ni ←[log(p′ i )-log(min(p,p′))]×100 / [log(max(p,p′))-log(min(p,p′))]

[0028] t Ni ←[log(t i )-log(min(t))}×100 / [log(max(t))-log(min(t))]

[0029] Among them, p Ni is the normalized pressure value; p' Ni is the normalized pressure derivative value; t Ni is the normalized time value; p i is the pressure value at the calculation point; p' i is the pressure derivative value at the calculation point; t i The time value of the calculation point.

[0030] Preferably, the extracting characteristic primitives and tracing the curves of the double logarithmic curve graph according to the initial segmentation window size and the primitive discrimination threshold to determine the initial recognition result includes:

[0031] Setting the initial segmentation window size and primitive discrimination threshold as recognition conditions, and performing the curve tracing and primitive division on the double logarithmic curve graph through a finite state machine;

[0032] Based on the inference rule set, all intervals of the double logarithmic curve graph are divided into feature primitives, and a primitive sequence set is obtained by extracting the feature primitives;

[0033] By matching and comparing the primitive sequence set with the characteristic primitive set of the well test interpretation model, the matched well test interpretation model is used as an initial recognition result.

[0034] Preferably, the expression of the finite state machine is:

[0035] M=(Q,Σ,δ,q0,F)

[0036] Among them, M is a finite state machine; Q is a non-empty finite set of states; Σ is the input alphabet; δ is the state transition function; q0 is the starting state; and F is the ending state.

[0037] Preferably, the initial identification results are screened and verified by a DTS method to complete the identification of the well test interpretation model, including:

[0038] According to the morphological changes of the double logarithmic curve graph, the primitive feature point information of the well test interpretation model is obtained, and the correctness of the initial recognition result is screened and tested using the DTS method to obtain the final output model and perform parameter fitting to complete the recognition of the well test interpretation model.

[0039] According to the specific embodiments provided by the present invention, the present invention discloses the following technical effects:

[0040] The present invention provides an intelligent identification method for a well test interpretation model. Based on the well test interpretation theory, a well test interpretation model library and a model inference rule set are established. In combination with the syntactic pattern recognition method, a feature primitive set for well test interpretation model identification is set, and structural information for distinguishing different feature primitives is established. The actual well test interpretation data is subjected to wavelet transform and polishing preprocessing steps, primitive extraction and curve tracing are performed through a finite state machine to obtain a primitive sequence set of the model, a preliminary inference model is obtained by matching with the model library, and the output model is screened and verified in combination with the DTS technology to obtain the final model inference result. The present invention solves the problems of low efficiency and poor quality of well test interpretation caused by inaccurate identification of well test interpretation models. The steps are simple, the model inference results are accurate, and the accuracy of the present invention is proved by comparing the software fitting results. In addition, the method also completes the division of flow stages, laying a theoretical foundation for the convenient development of subsequent well test interpretation work. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention, and for ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0042] Figure 1 A flow chart of an intelligent identification method of a well test interpretation model of the present invention;

[0043] Figure 2 An initial double logarithmic curve graph of pressure data and pressure derivative data provided in the first embodiment of the present invention;

[0044] Figure 3 A characteristic primitive diagram provided in the first embodiment of the present invention;

[0045] Figure 4 A schematic diagram of a primitive extraction and curve tracing process provided in the first embodiment of the present invention;

[0046] Figure 5 A schematic diagram of the final model inference process of the DTS method provided in the first embodiment of the present invention;

[0047] Figure 6 This is a comparison chart of actual data and fitted data provided in Example 1 of the present invention. DETAILED DESCRIPTION

[0048] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments 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.

[0049] In order to make the purpose, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0050] like Figure 1 As shown, the present invention provides an intelligent identification method for a well test interpretation model, comprising the following steps:

[0051] Step 100: construct a well test interpretation model and establish feature primitive sets and inference rule sets corresponding to different flow stages;

[0052] Step 200: preprocessing the data of the well test interpretation model, saving the preprocessed data, and plotting the preprocessed data into a double logarithmic curve graph;

[0053] Step 300: extracting characteristic primitives and tracing curves on the double logarithmic curve graph according to the initial segmentation window size and primitive discrimination threshold, and determining an initial recognition result;

[0054] Step 400: Based on the morphological changes of the double logarithmic curve graph and the primitive feature point information, the DTS method is used to screen and verify the correctness of the initial recognition result to complete the recognition of the well test interpretation model.

[0055] Wherein, in step 100, the construction of the well test interpretation model and the establishment of the characteristic primitive set and the inference rule set corresponding to different flow stages include:

[0056] S101, constructing the well test interpretation model according to the well type, inner boundary type, reservoir type, and outer boundary type;

[0057] S102, based on the pressure and pressure derivative data of the well test interpretation model, refer to Figure 3 As shown, a double logarithmic curve graph is drawn, and different flow stages in the well test interpretation model are identified according to characteristic points and characteristic lines on the double logarithmic curve graph; the pressure and pressure derivative curves of the well test interpretation model are analyzed and synthesized to determine a characteristic primitive set;

[0058] S103: Establish the inference rule set according to the characteristic parameters of the well test interpretation model and the actual interpretation experience.

[0059] The establishment of the inference rule set is based on the relationship between the well test interpretation model and the flow stage, mainly including:

[0060] (1) When making model inferences, we first need to identify the type of well, such as conventional well, water injection well, gas well, or mining well. Based on the known well type, we can select an appropriate well test interpretation model.

[0061] (2) Each primitive is usually associated with a specific flow stage. Different primitives can be assigned priorities based on the characteristics of different flow stages and their importance to model reasoning.

[0062] (3) Due to the complexity of the curve shape of the well test interpretation model, the three flow stages of early, middle and late flow may not all exist. Figure 3 The morphological sequence diagram shown uses the state sequence diagram to map the possibilities of different flow stages, and there are various combinations of flow stages when outputting the explanation model.

[0063] (4) There are multiple explanatory models corresponding to one primitive sequence.

[0064] Secondly, in step 200, data preprocessing is performed on the data of the well test interpretation model, including:

[0065] S201, obtaining pressure data of the well test interpretation model, performing wavelet transform on the pressure data, using Daubechies 4 (db4) as an orthogonal wavelet transform function to perform wavelet decomposition of the pressure data, using Sqtwolog as a threshold calculation rule to quantify the pressure data after wavelet decomposition, and finally performing wavelet reconstruction on the quantized pressure data to generate denoised pressure data;

[0066] S202, polishing the denoised pressure data, and calculating pressure derivative data using a differential method;

[0067] The pressure derivative data calculated by the difference method above include:

[0068] Get p i In t i The derivative at time satisfies |log(t)-log(t i )|≤d, determine the difference between the two points and then calculate. The derivative formula is expressed as:

[0069]

[0070] in, is the pressure derivative value at the calculation point; t i is the time corresponding to the calculation point; t1 is the time corresponding to calculation point 1; t2 is the time corresponding to calculation point 2; is the pressure corresponding to the calculation point; is the pressure corresponding to calculation point 1; is the pressure corresponding to calculation point 2; d

[0071] is the polishing coefficient.

[0072] S203, regularizing the time, denoised pressure data and pressure derivative data.

[0073] The normalized expressions of the above variables are:

[0074] p Ni ←[log(p i )-log(min(p,p′))]×100 / [log(max(p,p′))-log(min(p,p′))]

[0075] p Ni ←[log(p′ i )-log(min(p,p′))]×100 / [log(max(p,p′))-log(min(p,p′))]

[0076] t Ni ←[log(t i )-log(min(t))]×100 / [log(max(t))-log(min9t))]

[0077] Among them, p Ni is the normalized pressure value; p' Ni is the normalized pressure derivative value; t Ni is the normalized time value; p i is the pressure value at the calculation point; p' i is the pressure derivative value at the calculation point; t i The time value of the calculation point.

[0078] In step 300, the feature primitives are extracted and the curve is traced on the double logarithmic curve graph according to the initial segmentation window size and the primitive discrimination threshold to determine the initial recognition result, including:

[0079] S301, setting the initial segmentation window size and primitive discrimination threshold as recognition conditions, and performing the curve tracing and primitive division on the double logarithmic curve graph through a finite state machine;

[0080] The expression of the above finite state machine is:

[0081] M=(Q,∑,δ,q0,F)

[0082] Among them, M is a finite state machine; Q is a non-empty finite set of states; Σ is the input alphabet; δ is the state transition function; q0 is the starting state; and F is the ending state.

[0083] S302, based on the inference rule set, dividing all intervals of the double logarithmic curve graph into feature primitives, and obtaining a primitive sequence set by extracting the feature primitives;

[0084] S303, matching and comparing the primitive sequence set with the characteristic primitive set of the well test interpretation model, and taking the matched well test interpretation model as an initial recognition result.

[0085] In step 400, the initial identification results are screened and verified by the DTS method to complete the identification of the well test interpretation model, including:

[0086] S401. According to the morphological changes of the double logarithmic curve graph, the primitive feature point information of the well test interpretation model is obtained, and the correctness of the initial recognition result is screened and verified by the DTS method to obtain the final output model and perform parameter fitting to complete the recognition of the well test interpretation model.

[0087] The above-mentioned DTS (Direct Type-curve Synthesis) method refers to extracting key characteristic points and characteristic lines from the pressure derivative double logarithmic curve by analyzing and processing the well test data, and verifying the effectiveness of these characteristic points and characteristic lines, so as to more accurately construct and optimize the well test interpretation model.

[0088] Embodiment 1

[0089] The above identification method is applied to the interpretation of the well test data of a certain oil well provided in this embodiment, specifically:

[0090] The basic data of the well include a well radius of 0.07 m, a formation thickness of 4.5 m, a porosity of 0.158, and a formation volume coefficient of 1.175 m. 3 / stm 3 , crude oil viscosity is 5.84cp, comprehensive compression coefficient is 0.002MPa -1 , the daily crude oil production is 4.28m 3 / D. The double logarithmic curve of the original pressure and pressure derivative of the pressure recovery test of the well is as follows Figure 2 Then, the data of the well test interpretation model is preprocessed, including denoising, polishing and regularization, and by setting the initial segmentation interval of 5 to 10, a finite state machine is used for curve tracing and primitive extraction. The specific implementation process is as follows Figure 4 As shown, the basic element sequence of the model can be obtained as RM1+R1+U+H+R1+H.

[0091] According to the obtained primitive sequence, it is matched with the characteristic primitive sequence of each model in the well test interpretation model library, and three preliminary inference models can be obtained: Result I vertical well-well reservoir + skin-homogeneous-one parallel fault, Result II vertical well-well reservoir + skin-radial composite-infinite, and Result III vertical well-well reservoir + skin-linear composite-infinite. Subsequently, the DTS method is used to screen and test the accuracy of the model. The final inference process of the entire model is as follows: Figure 5 As shown, it includes: first, obtaining the pressure derivative value of the radial flow stage corresponding to the first horizontal primitive as 0.048MPa, and calculating the permeability of the radial flow stage as 125.22mD, and converting the pressure and pressure derivative data of the double logarithmic curve graph into dimensionless form; secondly, obtaining the dimensionless pressure derivative value of the radial flow stage of the second horizontal primitive as 1.5, thereby excluding the ResultⅠ vertical well-well reservoir + skin-homogeneous-one parallel fault model and the ResultⅢ vertical well-well reservoir + skin-linear composite-infinite model. The dimensionless pressure derivative values ​​of the second horizontal primitive radial flow stage of the above two models should be 1.0 and 0.5~1.0 respectively, thereby determining that the well is the ResultⅡ vertical well-well reservoir + skin-radial composite-infinite model.

[0092] Finally, in this embodiment, according to the intelligent recognition inference model results, the above-determined model is fitted with parameters to obtain the interpretation parameters of the well, such as Figure 6 As shown in the figure, it can be intuitively found that the model interpretation parameter fitting results are reliable and the model selection is correct. The specific fitting parameters include: the wellbore storage coefficient is 0.275m 3 / MPa, the skin coefficient is -4.05, the permeability is 120mD, the inner zone radius is 56.59m, the mobility ratio is 3.01, and the diffusion coefficient ratio is 1.88.

[0093] Therefore, the above-mentioned intelligent identification method of a well test interpretation model is adopted to solve the problems of low efficiency and poor quality of well test interpretation caused by inaccurate identification of the current well test interpretation model. The steps are simple and the model inference results are accurate. The accuracy of the present invention is proved by comparing the software fitting results. In addition, the method also completes the division of the flow stage, laying a theoretical foundation for the convenient implementation of subsequent well test interpretation work.

[0094] The principles and implementation methods of the present invention are described in this article using specific examples. The description of the above embodiments is only used to help understand the method and core idea of ​​the present invention. At the same time, for those skilled in the art, according to the idea of ​​the present invention, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting the present invention.

Claims

1. An intelligent identification method for a well test interpretation model, characterized in that: The following steps are involved: Construct well test interpretation model and establish feature primitives and inference rule sets corresponding to different flow stages; Preprocessing the data of the well test interpretation model, saving the preprocessed data, and plotting the preprocessed data into a double logarithmic curve graph; According to the initial segmentation window size and the primitive discrimination threshold, feature primitives are extracted and curve traced on the double logarithmic curve graph to determine the initial recognition result; According to the morphological changes of the double logarithmic curve graph and the primitive feature point information, the DTS method is used to screen and verify the correctness of the initial recognition result to complete the recognition of the well test interpretation model.

2. The intelligent identification method of a well test interpretation model according to claim 1, characterized in that: The construction of the well test interpretation model and the establishment of the characteristic primitive sets and inference rule sets corresponding to different flow stages include: Constructing the well test interpretation model according to the well type, inner boundary type, reservoir type, and outer boundary type; Based on the pressure and pressure derivative data of the well test interpretation model, a double logarithmic curve graph is drawn, and different flow stages in the well test interpretation model are identified according to characteristic points and characteristic lines on the double logarithmic curve graph; the pressure and pressure derivative curves of the well test interpretation model are analyzed and synthesized to determine a characteristic primitive set; The inference rule set is established based on the characteristic parameters of the well test interpretation model and the experience of actual interpretation.

3. The intelligent identification method of a well test interpretation model according to claim 1, characterized in that: The data of the well test interpretation model is preprocessed, including: Acquire the pressure data of the well test interpretation model, perform wavelet transform on the pressure data, use Daubechies4 (db4) as an orthogonal wavelet transform function to perform wavelet decomposition of the pressure data, use Sqtwolog as a threshold calculation rule to quantify the pressure data after wavelet decomposition, and finally perform wavelet reconstruction on the quantized pressure data to generate denoised pressure data; Polishing the denoised pressure data and calculating pressure derivative data using a differential method; Regularization is performed on the time, denoised pressure data, and pressure derivative data.

4. The intelligent identification method of a well test interpretation model according to claim 3 is characterized in that: The pressure derivative data is calculated using the difference method, including: Get p i In t i The derivative at time satisfies |log(t)-log(t i )|≤d, determine the difference between the two points and then calculate. The derivative formula is expressed as: in, is the pressure derivative value at the calculation point; t i is the time corresponding to the calculation point; t1 is the time corresponding to calculation point 1; t2 is the time corresponding to calculation point 2; is the pressure corresponding to the calculation point; is the pressure corresponding to calculation point 1; is the pressure corresponding to the calculation point 2; d is the polishing coefficient.

5. The intelligent identification method of a well test interpretation model according to claim 3, characterized in that: The time, denoised pressure data and pressure derivative data are normalized, and the normalized expression is: p Ni ←[log(p i )-log(min(p,p’))]×100 / [log(max(p,p’))-log(min(p,p))] p′ Ni ←[log(p’ i )-log(min(p,p’))]×100 / [log(max(p’p’))-log(min(p,p’))] t Ni ←[log(t i )-log(min(t))]×100 / [log(max(t))-log(min(t))] Among them, p Ni is the normalized pressure value; p′ Ni is the normalized pressure derivative value; t Ni is the normalized time value; p i is the pressure value at the calculation point; p′ i is the pressure derivative value at the calculation point; t i The time value of the calculation point.

6. The intelligent identification method of a well test interpretation model according to claim 1, characterized in that: The step of extracting characteristic primitives and tracing curves on the double logarithmic curve graph according to the initial segmentation window size and primitive discrimination threshold to determine the initial recognition result includes: Setting the initial segmentation window size and primitive discrimination threshold as recognition conditions, and performing the curve tracing and primitive division on the double logarithmic curve graph through a finite state machine; Based on the inference rule set, all intervals of the double logarithmic curve graph are divided into feature primitives, and a primitive sequence set is obtained by extracting the feature primitives; By matching and comparing the primitive sequence set with the characteristic primitive set of the well test interpretation model, the matched well test interpretation model is used as an initial recognition result.

7. The intelligent identification method of a well test interpretation model according to claim 6, characterized in that: The expression of the finite state machine is: M=(Q,∑,δ,q0,F) Among them, M is a finite state machine; Q is a non-empty finite set of states; Σ is the input alphabet; δ is the state transition function; q0 is the starting state; and F is the ending state.

8. The intelligent identification method of a well test interpretation model according to claim 1, characterized in that: The initial identification results are screened and verified by the DTS method to complete the identification of the well test interpretation model, including: According to the morphological changes of the double logarithmic curve graph, the primitive feature point information of the well test interpretation model is obtained, and the correctness of the initial recognition result is screened and tested using the DTS method to obtain the final output model and perform parameter fitting to complete the recognition of the well test interpretation model.

Images