Effective data efficient cleaning method for well drilling acceleration

Abstract

The invention discloses an effective data efficient cleaning method for well drilling acceleration. The method comprises the steps that well drilling big data and well drilling related parameters derived from the well drilling big data are collected; the drilling big data comprises well logging, well logging, daily report, special report, well history and a work supervision system; the well drilling related parameters comprise a drill bit, a tool, a drilling tool combination, mechanical parameters and hydraulic parameters; determining drilling acceleration effective data from the collected drilling big data; effective data of well drilling acceleration can be efficiently cleaned. According to the method, efficient cleaning of effective drilling acceleration data can be achieved, the integrity, consistency, accuracy and reliability of the data are improved, and the application prospect is wide.

Description

Technical Field

[0001] The invention relates to the technical field of oil drilling big data, and in particular to a method for efficiently cleaning effective data for accelerating drilling. Background Art

[0002] Drilling big data is extracted and organized from multiple business systems (logging, mud recording, daily reports, special reports, well history and industrial supervision systems, etc.). In the process of extraction and organization, there are human factors and export reasons, which will lead to errors or conflicts between some data. This type of data is called "dirty data". Data cleaning is the process of identifying and correcting dirty data, and its purpose is to improve the accuracy, timeliness, consistency and completeness of data.

[0003] Data cleaning is generally only targeted at specific applications, so there is no unified method and steps, but corresponding solutions can be given according to the type of dirty data. The types of dirty data are: missing data, similar duplicate data, and outlier data. For missing data, direct deletion is generally used. The advantages of this method are simple operation and easy implementation. The disadvantage is that the packet loss rate is likely to be too high, resulting in hidden information. For similar duplicate data, the basic method of merging / clearing is generally used. The advantages of this type of method are simple and easy to operate, but the disadvantages are high packet loss rate, serious data distortion, and no use value. For outlier data (outliers, high leverage values, and strong influence values), direct deletion or regression processing is generally used. Direct deletion is likely to cause distortion of the data set, and due to the uncertainty of the data, there are problems such as low accuracy of regression processing and difficulty in determining the regression function.

[0004] Therefore, how to reduce the packet loss rate in the process of speeding up effective data cleaning while ensuring the original data characteristics of the data set based on the premise of making full use of drilling big data plays an important role in the establishment of digital oil fields. Summary of the invention

[0005] In view of this, the present invention provides a method for efficiently cleaning effective data for accelerating drilling, so as to solve the above technical problems.

[0006] The present invention discloses a method for efficiently cleaning effective data for drilling speed-up, which comprises:

[0007] Step 1: Collect drilling big data and drilling related parameters derived from the drilling big data; the drilling big data includes well logging, mud logging, daily reports, special reports, well history and work supervision system; the drilling related parameters include drill bits, tools, drilling tool assemblies, mechanical parameters, and hydraulic parameters;

[0008] Step 2: Determine effective data for drilling speed increase from the drilling big data collected in step 1;

[0009] Step 3: Efficient cleaning of effective data for drilling speed-up.

[0010] Furthermore, the step 2 comprises:

[0011] Step 21: extracting samples and drilling related parameters from the drilling big data to form an original data matrix;

[0012] Step 22: Based on the original data matrix, calculate the Pearson coefficient matrix and the entropy weight of each parameter respectively;

[0013] Step 23: Calculate the entropy weight of all drilling related parameters in step 21, which is the degree of influence on the drilling speed, that is, the contribution rate. Finally, select the parameters with a cumulative contribution rate of more than 10% as the effective data parameters for speeding up.

[0014] Furthermore, the step 21 comprises:

[0015] Assume there are n samples and any one of p drilling related parameters, forming a sample matrix x of size n*p:

[0016]

[0017] First calculate the mean of each parameter in formula (1):

[0018]

[0019] Where: is the mean of any parameter; x ij is any parameter value in the sample matrix;

[0020] Then calculate the standard deviation:

[0021]

[0022] Where: p is the number of parameters; S j is the standard deviation of any parameter j;

[0023] Data Standardization:

[0024]

[0025] Therefore, the matrix of the original sample matrix after standardization is:

[0026]

[0027] Where: X ij is the standardized data of any parameter j; X is the standardized matrix.

[0028] Calculate the covariance matrix based on the standardized matrix:

[0029]

[0030] Where: p is the number of parameters; is the mean of the i-th row in the covariance matrix; X ki is any value in the i-th row; is the mean of the jth column in the covariance matrix; X kj is any value in the jth column;

[0031] Written in matrix form:

[0032]

[0033] Where: r ij is the covariance of a parameter after standardization; R is the covariance matrix of the data set.

[0034] Further, the step 22 comprises:

[0035] Based on the covariance matrix, the Pearson coefficient matrix and the entropy weights of each parameter are calculated respectively:

[0036]

[0037]

[0038] h j =1-e j (10)

[0039]

[0040] Where: R XY is the Pearson coefficient matrix, where the correlation between the same parameters is 1; S X , S Y are the sample standard deviations of any parameter X and any parameter Y respectively; j is information entropy; R ij is the Pearson coefficient of any parameter; h j is the information utility value; w j is the weight.

[0041] Furthermore, the step 3 comprises:

[0042] Step 31: Based on the principle of normal distribution concentration, a preset packet loss rate is selected for a single parameter as a basis for single parameter cleaning: if the packet loss rate of the effective data of drilling speed-up is less than the preset packet loss rate, the effective data of drilling speed-up is deleted; otherwise, cubic spline interpolation is used to achieve efficient single parameter cleaning;

[0043] Step 32: Based on the principle of normal distribution concentration, for the data group, the preset cumulative weight is selected as the basis for data group cleaning: if the effective data of drilling speed-up is less than the preset cumulative weight, the effective data of drilling speed-up is deleted; otherwise, the entropy weight-filtering method is used to achieve efficient cleaning of the data group.

[0044] Furthermore, the basic form of the cubic spline interpolation is:

[0045]

[0046] Where: α j (x) and β j (x) is the interpolation basis function; m j is the first derivative of the cubic spline interpolation basis function of a certain parameter and a certain node; y j is the function value at a given node.

[0047] Furthermore, the specific operation steps of the cubic spline interpolation are:

[0048] Assumption: The second-order derivative of the i-th point is M i , then for any parameter [x i-1 ,x i ] node, the interval length is h i-1 , whose second-order derivative is a linear function:

[0049] S″ i (x) = M i (13)

[0050] Then in [x i-1 ,x i ] to perform linear interpolation:

[0051]

[0052] h i =x i -x i-1 (15)

[0053]

[0054] Performing the quadratic integration yields the function:

[0055]

[0056] Get the original function:

[0057]

[0058] Derivative each segment:

[0059]

[0060]

[0061] According to the continuity condition: S′(x i -0) = S′(x i +0) combined with the boundary conditions, we can get:

[0062]

[0063] μ i M i-1 +2M i +λ i M i+1 =g i (twenty two)

[0064] Written in matrix form:

[0065]

[0066] in: That is, the weighted value in a given interval; g i For the interval [x i-1 ,x i ] on Newton interpolation results.

[0067] Furthermore, the entropy weight-filtering method is used to achieve efficient cleaning of the data set, including:

[0068] Perform a wavelet transform on a data array:

[0069]

[0070] Where: a is the scale factor, and a>0, to achieve the basic wavelet Perform scaling transformation; τ is the translation factor, which realizes the translation transformation of the basic wavelet on the time axis; W f (τ,a) is the denoised data after wavelet transform.

[0071] Furthermore, the method of using the entropy weight-filtering method to achieve efficient cleaning of the data set also includes:

[0072] Introduce entropy weight based on wavelet transform:

[0073]

[0074] h j =1-e j (26)

[0075]

[0076] Where: W ij is the denoised data of any parameter; ej is information entropy; h j is the information utility value; w j is the weight.

[0077] Due to the adoption of the above technical solution, the present invention has the following advantages:

[0078] Based on drilling big data, the present invention selects data with a cumulative contribution rate of more than 10% as effective data for speeding up through the entropy weight-Pearson method. On this basis, the single parameter and the data group are cleaned respectively. For the single parameter, the packet loss rate of 30% is used as the processing basis for abnormal points (outliers, strong influences and high leverage points). The packet loss rate is less than 30% and the processing method of direct deletion is adopted. The packet loss rate is greater than 30% and the cubic spline interpolation is used for interpolation supplement. For the data group, the cumulative contribution rate of the abnormal point parameter is 30% as the processing basis. The cumulative contribution rate is less than 30% and the processing method of direct deletion is adopted. If it is greater than 30%, all abnormal points are processed by the entropy weight-filtering method. In practical applications, the present invention can realize the efficient cleaning of effective data for drilling speed-up under the premise of meeting the low packet loss rate (<10%), improve the integrity, consistency, accuracy and reliability of the data, and has broad application prospects. BRIEF DESCRIPTION OF THE DRAWINGS

[0079] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the embodiments of the present invention. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0080] Figure 1 This is a flow chart of a method for efficiently cleaning effective data for drilling speed-up according to an embodiment of the present invention;

[0081] Figure 2 It is an entropy weight-Pearson correlation analysis diagram of an embodiment of the present invention;

[0082] Figure 3 is a normal distribution diagram of an embodiment of the present invention;

[0083] Figure 4(a) to Figure 4(f) This is a single parameter high efficiency cleaning diagram of an embodiment of the present invention;

[0084] FIG. 5( a ) and FIG. 5( b ) are comparison diagrams of the parameter groups before and after efficient cleaning according to an embodiment of the present invention, respectively. DETAILED DESCRIPTION

[0085] The present invention is further described in conjunction with the accompanying drawings and embodiments. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by ordinary technicians in this field should fall within the scope of protection of the embodiments of the present invention.

[0086] See also Figures 1 to 3 , the technical solution of this embodiment is: based on the field drilling big data, firstly, the entropy weight-Pearson method is used to clarify the effective data and its influence degree, and then based on the principle of normal distribution concentration, the packet loss rate of 30% is selected as the basis for single parameter cleaning, and the packet loss rate is less than 30%, and the deletion process is adopted; the packet loss rate is greater than 30%, and the cubic spline interpolation is adopted to achieve efficient single parameter cleaning, and finally the cumulative weight of 30% is selected as the basis for data group cleaning, and the cumulative weight is less than 30%, and the deletion process is adopted; the cumulative weight is greater than 30%, and the entropy weight-filtering method is adopted to achieve efficient data group cleaning. Specifically, the following steps are included:

[0087] Step 1: Collect drilling big data;

[0088] In this embodiment, the drilling big data (drill bits, tools, drill bit assemblies, mechanical parameters, hydraulic parameters, etc.) derived from well logging, mud logging, daily reports, special reports, well histories, and work supervision systems are integrated into the same workbook to form a data set.

[0089] Step 2: Determine effective data for drilling speed increase;

[0090] The big data that affects the drilling speed has the characteristics of large data set dimensions, data encoding, mutual influence between parameters, and unclear parameter correlation. The entropy weight-Pearson method is selected to clarify the effective data for speed improvement and its influence degree. The specific operation is as follows: first, the original data matrix is ​​standardized and the covariance matrix and Pearson coefficient of the standard matrix are calculated. Secondly, the entropy weight is introduced to calculate the entropy weight value of the variable (the degree of influence on the drilling speed, that is, the contribution rate). Finally, the parameters with a cumulative contribution rate of more than 10% are selected as the effective data parameters for speed improvement.

[0091] Assuming there are n samples, p of which are drilling-related parameters derived from logging, mud logging, daily reports, special reports, well histories, and industrial monitoring systems, a sample matrix x of size n*p can be constructed:

[0092]

[0093] First, calculate the mean value of each parameter according to formula (1):

[0094]

[0095] Where: is the mean of variable j; x ijis any parameter value in the sample matrix.

[0096] Then calculate the standard deviation of each data:

[0097]

[0098] Where: p is the number of variables; S j is the standard deviation of parameter j.

[0099] Normalize the data based on the calculated standard deviation:

[0100]

[0101] Therefore, the original sample matrix becomes after standardization:

[0102]

[0103] Where: X ij is the standardized data of variable j; X is the standardized matrix.

[0104] Based on the standardized matrix, the covariance and covariance matrix corresponding to each parameter are obtained.

[0105]

[0106] Where: p is the number of parameters; is the mean of the i-th row in the covariance matrix; X ki is any value in the i-th row; is the mean of the jth column in the covariance matrix; X kj is any value in the jth column;

[0107] Written in matrix form:

[0108]

[0109] Where: r ij is the covariance of the standardized variables; R is the covariance matrix of the data set.

[0110] The Pearson coefficient matrix (i.e., correlation coefficient matrix) and entropy weight are shown below:

[0111]

[0112]

[0113] h j =1-e j (10)

[0114]

[0115] Where: R XY is the Pearson coefficient matrix, where the correlation between the same parameters is 1; S X , S Y are the standard deviations corresponding to different parameters X and Y respectively; e j is information entropy; R ij is the Pearson coefficient of a parameter; h j is the information utility value; w j is the weight.

[0116] It should be noted that the correlation between different parameters and ROP is given by heat maps (e.g. Figure 2 The contribution rates of different parameters are given in tabular form (as shown in Table 1).

[0117] The effective data for drilling speed-up in this embodiment include 9 parameters: drill bit, rotary guide, BHA, drilling pressure, torque, displacement, rotation speed, penetration density, and vertical pressure.

[0118] Table 1 Summary of correlation coefficients and weights

[0119]

[0120]

[0121] Step 3: Efficient cleaning of effective data to speed up drilling.

[0122] Based on the principle of normal distribution concentration, a packet loss rate of 30% is selected as the basis for single parameter cleaning. If the packet loss rate is less than 30%, deletion is used. If the packet loss rate is greater than 30%, cubic spline interpolation is used to achieve efficient single parameter cleaning. For the data group, a cumulative weight of 30% is selected as the basis for data group cleaning. If the cumulative weight is less than 30%, deletion is used. If the cumulative weight is greater than 30%, entropy weight-filtering method is used to achieve efficient data group cleaning. The specific steps of step three are as follows:

[0123] The normal distribution principle is: The 3σ principle of normal distribution is: the probability of the value being distributed in (μ-σ,μ+σ) is 0.6; the probability of the value being distributed in (μ-2σ,μ+2σ) is 0.956; the probability of the value being distributed in (μ-3σ,μ+3σ) is 0.998. It can be considered that the values ​​of Y are almost all concentrated in the interval (μ-3σ,μ+3σ), and the possibility of exceeding this range is less than 0.2%. Therefore, the packet loss rate (cumulative weight) is taken as 30% as the selection standard. The normal distribution diagram is as follows: Figure 2 shown.

[0124] For a single parameter, if the packet loss rate is less than 30%, deletion is used. If the packet loss rate is greater than 30%, cubic spline interpolation is used to achieve efficient cleaning of the single parameter. The following are the specific steps of cubic spline interpolation:

[0125] The basic form of cubic spline interpolation is:

[0126]

[0127] Among them, α j (x) and β j (x) is the interpolation basis function; m j is the first derivative of the cubic spline interpolation basis function of a certain parameter and a certain node; y j is the function value at a given node.

[0128] It should be noted that the specific operation steps of cubic spline interpolation are:

[0129] Assumption: The second-order derivative of the i-th point is M i , then for any parameter [x i-1 ,x i ] node, the interval length is h i-1 , whose second-order derivative is a linear function:

[0130] S″ i (x) = M i (13)

[0131] Then in [x i-1 ,x i ] to perform linear interpolation:

[0132]

[0133] h i =x i -x i-1 (15)

[0134]

[0135] Performing the quadratic integration yields the function:

[0136]

[0137] Get the original function:

[0138]

[0139] Derivative each segment:

[0140]

[0141]

[0142] According to the continuity condition: S′(x i -0) = S′(x i +0) combined with the boundary conditions, we can get:

[0143]

[0144] μ i M i-1 +2M i +λ i M i+1 =g i (twenty two)

[0145] Written in matrix form:

[0146]

[0147] in: That is, the weighted value of the step length in a given interval; g i For the interval [x i-1 ,x i ] on Newton interpolation results.

[0148] The above are the specific steps of cubic spline interpolation. It should be noted that cubic spline interpolation is the fitting of adjacent points by cubic functions, which has high accuracy and can meet general engineering needs. Cubic spline interpolation is efficient in cleaning single parameter results. Figure 4(a) to Figure 4(f) As shown; among them, the packet loss rate of Figure 4(a) and Figure 4(b) is 40% ↘ 3.4%; the packet loss rate of Figure 4(c) and Figure 4(d) is 42% ↘ 2.5%; the packet loss rate of Figure 4(e) and Figure 4(f) is 83% ↘ 1.6%.

[0149] For the data group, the cumulative weight of 30% is selected as the basis for data group cleaning. If the cumulative weight is less than 30%, deletion is adopted. If the cumulative weight is greater than 30%, the entropy weight-filtering method is used to achieve efficient data group cleaning.

[0150] The principle of the entropy weight-filtering method is to introduce entropy weight on the basis of wavelet filtering, and use weights to clarify the degree of influence between different parameters. The specific steps are as follows:

[0151] Perform a wavelet transform on a data array:

[0152]

[0153] Where: a is the scale factor, and a>0, to achieve the basic wavelet Perform scaling transformation; τ is the translation factor, which realizes the translation transformation of the basic wavelet on the time axis; W f (τ,a) is the denoised data after wavelet transform.

[0154] Introduce entropy weight based on wavelet transform:

[0155]

[0156] h j =1-e j (26)

[0157]

[0158] Where: e j is information entropy; W ij is the denoised data of any parameter; e j is information entropy; h j is the information utility value; w j is the weight.

[0159] It should be noted that the drilling parameters affect each other. To process the parameters with “dirty data”, we cannot only process the single parameter, but should judge it by the weight of the parameter’s influence on ROP. The cleaning results of the parameter group are shown in Figure 5(a) and Figure 5(b).

[0160] In summary, the above method has a good application effect in the efficient cleaning of effective data for drilling speed-up, and plays an important role in improving the integrity, consistency, accuracy and reliability of data.

[0161] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the relevant field should understand that the specific implementation methods of the present invention can still be modified or replaced by equivalents. Any modification or equivalent replacement that does not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A method for efficiently cleaning effective data for drilling speed-up, characterized in that: include: Step 1: Collect drilling big data and drilling related parameters derived from the drilling big data; the drilling big data includes well logging, mud logging, daily reports, special reports, well history and work supervision system; the drilling related parameters include drill bits, tools, drilling tool assemblies, mechanical parameters, and hydraulic parameters; Step 2: Determine effective data for drilling speed increase from the drilling big data collected in step 1; Step 3: Efficient cleaning of effective data for drilling speed-up.

2. The method according to claim 1, characterized in that: The step 2 comprises: Step 21: extracting samples and drilling related parameters from the drilling big data to form an original data matrix; Step 22: Based on the original data matrix, calculate the Pearson coefficient matrix and the entropy weight of each parameter respectively; Step 23: Calculate the entropy weight of all drilling related parameters in step 21, which is the degree of influence on the drilling speed, that is, the contribution rate. Finally, select the parameters with a cumulative contribution rate of more than 10% as the effective data parameters for speeding up.

3. The method according to claim 2, characterized in that The step 21 comprises: Assume there are n samples and any one of p drilling related parameters, forming a sample matrix x of size n*p: First calculate the mean of each parameter in formula (1): Where: is the mean of any parameter; x ij is any parameter value in the sample matrix; Then calculate the standard deviation: Where: p is the number of parameters; S j is the standard deviation of any parameter j; Data Standardization: Therefore, the matrix of the original sample matrix after standardization is: Where: X ij is the standardized data of any parameter j; X is the standardized matrix. The covariance matrix is ​​calculated based on the standardized matrix: Where: p is the number of parameters; is the mean of the i-th row in the covariance matrix; X ki is any value in the i-th row; is the mean of the jth column in the covariance matrix; X kj is any value in the jth column; Written in matrix form: Where: r ij is the covariance of a parameter after standardization; R is the covariance matrix of the data set.

4. The method according to claim 3, characterized in that The step 22 comprises: Based on the covariance matrix, the Pearson coefficient matrix and the entropy weights of each parameter are calculated respectively: h j =1-e j (10) Where: R XY is the Pearson coefficient matrix, where the correlation between the same parameters is 1; S X , S Y are the sample standard deviations of any parameter X and any parameter Y respectively; j is information entropy; R ij is the Pearson coefficient of any parameter; h j is the information utility value; w j is the weight.

5. The method according to claim 1, characterized in that The step 3 comprises: Step 31: Based on the principle of normal distribution concentration, a preset packet loss rate is selected for a single parameter as a basis for single parameter cleaning: if the packet loss rate of the effective data of drilling speed-up is less than the preset packet loss rate, the effective data of drilling speed-up is deleted; otherwise, cubic spline interpolation is used to achieve efficient single parameter cleaning; Step 32: Based on the principle of normal distribution concentration, for the data group, the preset cumulative weight is selected as the basis for data group cleaning: if the effective data of drilling speed-up is less than the preset cumulative weight, the effective data of drilling speed-up is deleted; otherwise, the entropy weight-filtering method is used to achieve efficient cleaning of the data group.

6. The method according to claim 5, characterized in that The basic form of the cubic spline interpolation is: Where: α j (x) and β j (x) is the interpolation basis function; m j is the first derivative of the cubic spline interpolation basis function of a certain parameter and a certain node; y j is the function value at a given node.

7. The method according to claim 5, characterized in that The specific operation steps of the cubic spline interpolation are: Assumption: The second-order derivative of the i-th point is M i , then for any parameter [x i-1 ,x i ] node, the interval length is h i-1 , whose second-order derivative is a linear function: S i ″(x)=M i (13) Then in [x i-1 ,x i ] to perform linear interpolation: h i =x i -x i-1 (15) Performing the quadratic integration yields the function: Get the original function: Derivative each segment: According to the continuity condition: S′(x i -0) = S′(x i +0) combined with the boundary conditions, we can get: m i M i-1 +2M i +λ i M i+1 =g i (22) Written in matrix form: in: That is, the weighted value in a given interval; g i For the interval [x i-1 ,x i ] on Newton interpolation results.

8. The method according to claim 5, characterized in that The entropy weight-filtering method is used to achieve efficient cleaning of data sets, including: Perform a wavelet transform on a data array: Where: a is the scale factor, and a>0, to achieve the basic wavelet Perform scaling transformation; τ is the translation factor, which realizes the translation transformation of the basic wavelet on the time axis; W f (τ,a) is the denoised data after wavelet transform.

9. The method according to claim 8, characterized in that The method of using entropy weight-filtering method to achieve efficient cleaning of data sets also includes: Introduce entropy weight based on wavelet transform: h j =1-e j (26) Where: W ij is the denoised data of any parameter; e j is information entropy; h j is the information utility value; w j is the weight.

Images