Drilling accident complex alarm method based on small-layer research and related equipment

Abstract

The invention relates to the field of drilling measurement, in particular to a drilling accident complex alarm method and related equipment based on small-layer research, and the drilling accident complex alarm method based on small-layer research comprises the steps: constructing a label library, target geological information and target accident complex condition data; historical well data information is acquired, and an accident complex alarm model is constructed; acquiring field data, determining a risk interval according to the tag library, acquiring field real-time data and tag library data, predicting the complex probability of various accidents occurring in the current well, and generating alarm information; and carrying out credibility evaluation on the alarm information, if the credibility reaches a preset threshold value, giving an alarm, and if the credibility does not reach the threshold value, not giving an alarm. The method has the advantages that risk prompting can be automatically and comprehensively carried out on batch wells before drilling, real-time alarming can be carried out on complex accidents during drilling, the occurrence rate of the complex accidents is reduced, the complex time efficiency of drilling accidents is reduced, and the purpose of reducing the drilling cost is achieved.

Description

Technical Field

[0001] The invention relates to the field of drilling measurement, and in particular to a complex drilling accident alarm method based on small layer research and related equipment. Background Art

[0002] During the drilling operation, accidents such as overflow, leakage, stuck drill, broken drill tools, etc. often occur. Once an accident occurs, drilling needs to be stopped to deal with the accident. Firstly, it takes a long time, consuming about 5% of the total drilling time. Secondly, the tools and materials consumed in dealing with the accident cause great losses.

[0003] Therefore, it is necessary to be able to provide early risk warnings for complex accidents. On the one hand, this can reduce the incidence of complex accidents, and on the other hand, it can improve the efficiency of handling complex accidents and reduce the cost of handling complex accidents. Summary of the invention

[0004] In order to provide risk warning for complex accidents, reduce the occurrence rate of complex accidents, improve the efficiency of complex accident handling, and reduce the cost and timeliness of complex accident handling, the present invention provides a drilling accident complex alarm method and related equipment based on small layer research.

[0005] The present invention provides a complex alarm method for drilling accidents based on small layer research, which adopts the following technical solutions:

[0006] A complex alarm method for drilling accidents based on small layer research, including

[0007] Build a tag library, target geological information, target accident complex situation data, and create a database;

[0008] Obtain historical well data information and build complex accident alarm models;

[0009] Obtain field data and determine the risk interval based on the tag library.

[0010] Acquire the real-time data on site and the tag library data, predict the probability of various accidents occurring in the current well, and generate alarm information;

[0011] Performing credibility assessment on the alarm information,

[0012] If the credibility reaches a preset threshold, an alarm is triggered.

[0013] If the credibility does not reach the threshold, no alarm is given.

[0014] In a specific possible implementation scheme, the tag library includes, as primary keys, north coordinates, east coordinates and altitude data;

[0015] Supported depth domain data measurement step as data interval;

[0016] Determine data categories and parameter items, and complete the data in the data categories and parameter items that cannot be directly measured by obtaining indirect data through calculation.

[0017] In a specific possible implementation manner, the data classification and the parameter items adopt a unified dimension.

[0018] In a specific possible implementation scheme, when constructing the accident complex alarm model, the historical well data is analyzed, including: noise processing, unified dimension, data reduction, data synthesis and data dimension reduction.

[0019] In a specific possible implementation scheme, the accident complex alarm model includes taking the historical well data as samples and dividing the time domain of the historical well data into time periods;

[0020] Features are extracted from the historical well data to mark the differences between the features of normal operation, accident complex symptoms, and accident complex occurrence.

[0021] In a specific implementation scheme, according to the characteristics, the screening algorithm is

[0022] The historical well data is put into the algorithm, machine learning is performed, the model is fitted, and the optimal algorithm is determined according to the degree of model fitting.

[0023] In a specific feasible implementation scheme, based on the historical well data, the accuracy and recall rate of the accident complex alarm model are verified by performing calculations according to the confusion matrix, and the weights of the parameters in the historical well data are adjusted to obtain the optimal model.

[0024] The present invention also provides a complex drilling accident alarm device based on small layer research, which adopts the following technical solution:

[0025] A complex alarm device for drilling accidents based on small layer research, comprising: one or more processors; a memory;

[0026] And one or more computer programs, wherein the one or more computer programs are stored in the memory, and the one or more computer programs include instructions, which, when executed by the alarm device, enable the alarm device to execute the complex drilling accident alarm method based on small layer research as described in any one of claims 1-7.

[0027] The present invention also provides an electronic device, which adopts the following technical solution:

[0028] An electronic device, comprising: one or more processors; a memory;

[0029] And one or more computer programs, wherein the one or more computer programs are stored in the memory, and the one or more computer programs include instructions, which, when executed by the electronic device, enable the electronic device to execute the above-mentioned complex drilling accident alarm method based on small layer research.

[0030] The present invention also provides a non-transitory computer-readable storage medium storing computer instructions, which adopts the following technical solution:

[0031] A non-transitory computer-readable storage medium storing computer instructions, wherein the computer instructions are used to enable a computer to execute the above-mentioned complex drilling accident alarm method based on small layer research.

[0032] In summary, the present invention includes at least one of the following beneficial technical effects:

[0033] 1. Use the complex alarm model for engineering accidents trained by supervised machine learning, combined with the geological risk set of the small layer label library and real-time data of actual drilling, to carry out risk estimation and alarm in the drilling process, reduce the incidence of complex accidents, improve the efficiency of complex accident handling, and reduce the cost and timeliness of complex accident handling.

[0034] 2. By adjusting algorithms and models, the accuracy of risk estimation can be improved, the incidence of complex accidents can be further reduced, and the efficiency of complex accident handling can be further improved, thereby reducing the cost and timeliness of complex accident handling. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 It is the overall flow chart of this embodiment.

[0036] Figure 2 This is a flowchart of building a small layer tag library.

[0037] Figure 3 It is a flow chart for constructing a complex accident alarm model. DETAILED DESCRIPTION

[0038] The following is combined with Figure 1-3 The present invention is described in further detail.

[0039] Reference Figure 1 The complex alarm method for drilling accidents based on small layer research includes the following steps:

[0040] S100, builds a small-layer label library and a complex accident alarm model.

[0041] Reference Figure 2 ,Constructing a small layer tag library includes the following steps:

[0042] A100, build a small layer label library template. Determine the north coordinate, east coordinate and altitude of the well as the primary key. Determine the depth domain data measurement step supported by the logging instrument as the data interval. Determine geology, engineering, accident complexity, etc. as data classification and parameter items. Create a database by entering relevant data.

[0043] A200, regional small layer label library data processing. Obtain real-time data during the actual drilling process, unify the step size of real-time data, that is, unify the data interval of the data. According to the API standard, unify the dimension of real-time data. Some data that can be obtained by indirect measurement are indirect data obtained based on computer software modeling and calculation. Indirect data is the indirect measurement data, or can replace the indirect measurement data for reference and calculation.

[0044] Reference Figure 3 ,Constructing a complex accident alarm model includes the following steps:

[0045] B100, obtain geological, engineering and logging time domain data of historical wells as modeling sample data.

[0046] B200, based on the business expert theory and actual drilling experience, divides the logging time domain into time periods according to the operating status such as normal operation, complex accident signs, complex accident occurrence, and complex accident handling. In other words, the sample data is manually divided into time periods according to the operating status. The expert label library is constructed based on the sample data of the divided time periods. The construction method is the same as the above-mentioned small layer label library, which will not be repeated here.

[0047] B300, obtains parameters such as outlet flow, total pool volume, and mechanical drilling speed in the sample data, and extracts features such as skewness, kurtosis value, range, and correlation.

[0048] Based on data operations such as noise data processing, unified dimension, data reduction, data synthesis, and data dimension reduction, the sample data is identified.

[0049] B400, based on the acquired features, performs machine learning according to different algorithms, such as cluster analysis, neural network, XGBOOST gradient forest, etc., repeatedly fits the model, screens the algorithm according to the degree of fit between the algorithm results and the model, and selects the optimal algorithm.

[0050] Algorithm screening includes the following steps:

[0051] a100, obtain relevant data in the accident complex as raw data.

[0052] a200, according to the average value algorithm of the first-order difference, processes the original data to form feature data between 0 and 1. Based on algorithms such as XGBOOST random forest algorithm, neural network algorithm, SVM algorithm, etc., a variety of prediction models are generated.

[0053] a300, based on the complex data of historical accidents, calls the prediction model for prediction and tests the performance of the prediction model. By comparing the accuracy of the prediction model's prediction results and the time required to calculate the results, the selected algorithm is determined.

[0054] According to the sample data, based on the calculation of the confusion matrix, the model accuracy and recall rate are verified, and the optimal model is obtained by adjusting the weights of the parameters of the sample data multiple times.

[0055] Model validation includes the following steps:

[0056] b100, adjust the weights of the original data to an initial value, and use the screened algorithm to train the model.

[0057] b200, repeatedly adjust the weights of the parameters to find the best sensitivity parameters. The data that can maximize the prediction accuracy by changing the weights in the parameters is the best sensitivity parameter.

[0058] Reference Figure 1 , S200, obtain the actual drilling data and determine the risk interval based on the small layer label library. Specifically, it is divided into two aspects. The first is geological risk analysis. According to the ECD, pore pressure, leakage pressure and other data in the small layer label library, the geological risks such as leakage, overflow, and collapse in each well section are analyzed. The second is engineering risk analysis. According to the accident complexity marked in the expert label library and the stratigraphic relationship between the actual drilling well and the target well, the engineering risks such as stuck drill and blocked drill in each well section of the actual drilling well are identified.

[0059] Reference Figure 1 ,S300, analyzes the real-time data of actual drilling based on the small layer label library and the accident complex alarm model.,Determine the risk interval based on the small label library.

[0060] According to the real-time data of the drilling process and the small layer label library, the probability of various accident complexes occurring at present is identified. According to the layer and the number of accident complexes, the risk is estimated. Specifically, by calculating the average of different accident complexes occurring in the same layer, the corresponding threshold is set to estimate the risk of different accident complexes.

[0061] Reference Figure 1, S400, evaluate the credibility of the alarm information. If the credibility of the alarm information exceeds the preset threshold, an alarm is triggered; if it does not exceed the preset threshold, no alarm is triggered. Specifically, the current actual well depth is obtained, and the layer is matched through the small layer label library. The probability value obtained by the real-time data in the prediction model is obtained.

[0062] Based on the probability of the accident complexity and the risk range, the credibility level is divided. If the probability of occurrence P ≥ 75 and the accident complexity is within the risk range, the credibility is level 3 and the alarm process is executed.

[0063] If P<75, the accident complexity is within the risk range, the credibility is level 2, and the alarm process is executed.

[0064] If P<75, and the accident is complex and not within the risk range, the credibility is level 1 and the alarm process is not executed.

[0065] Example:

[0066] Obtain the geological, public testing and accident-related real drilling data of historical wells in the block where the target well belongs, and build a small layer label library.

[0067] The well depth, drill bit position, total pool volume, total pump stroke, inlet flow, outlet flow, total pool volume, standpipe pressure, casing pressure, etc. are obtained as raw data, and the raw data are processed according to the average value algorithm of the first-order difference to form feature data between 0 and 1. Three prediction models are generated based on the XGBOOST random forest algorithm, neural network algorithm, and SVM algorithm. Based on the complex data of historical accidents, the prediction model is called for prediction, and the performance of the prediction model is tested.

[0068] The accuracy of the XGBOOST random forest algorithm reached over 95%, and it could calculate the result within 2 seconds; the accuracy of the neural network algorithm was about 85%, and it could calculate the result within 30 minutes; the accuracy of the SVM vector machine algorithm was about 90%, and it could calculate the result within 20 minutes. After comparing the performance and prediction results of the three models, the XGBOOST random forest algorithm was finally selected.

[0069] The weights of well depth, drill bit position, total pool volume, total pump stroke, inlet flow, outlet flow, total pool volume, standpipe pressure, casing pressure and other data were adjusted to 11.11% and input into the XGBOOST random forest algorithm for model training. The optimal sensitivity parameter is the total pool volume. When the weight of the total pool volume is increased to 80%, the prediction accuracy of the model can reach more than 95%.

[0070] Obtain data sent by automatic parameter acquisition instruments such as drilling parameters, logging instruments, LWD, RSS, MPD, etc., and conduct complex analysis and prediction of accidents.

[0071] For the same layer, if the average number of lost circulation events X is less than 1, it is considered low risk; if 1≤X<2, it is considered medium risk; and if X≥2, it is considered high risk.

[0072] For the same layer, if the average number of overflow or drill bit breakage Y is less than 0.5, it is a low risk; if 0.5≤Y<1, it is a medium risk; if Y≥1, it is a high risk.

[0073] At the same layer, if the average number of blockages Z is less than 2, it is a low risk; if 2≤Z<4, it is a medium risk; if Z≥4, it is a high risk.

[0074] For the same layer, if the average number of stuck drill bits W is less than 3, it is considered low risk. If 3≤W<5, it is considered medium risk. If W≥5, it is considered high risk.

[0075] According to the probability of accident occurrence and risk range, the prediction results are divided into credibility levels, and an alarm is automatically issued based on the division results.

[0076] The present invention also discloses a complex drilling accident alarm device based on small layer research, comprising: one or more processors; a memory; and one or more computer programs, wherein the one or more computer programs are stored in the memory, and the one or more computer programs include instructions, which, when executed by the alarm device, enable the alarm device to execute the above-mentioned complex drilling accident alarm method based on small layer research.

[0077] The present invention also discloses an electronic device, comprising: one or more processors; a memory; and one or more computer programs, wherein the one or more computer programs are stored in the memory, and the one or more computer programs include instructions, which, when executed by the electronic device, enable the electronic device to execute the above-mentioned complex drilling accident alarm method based on small layer research.

[0078] The present invention also discloses a non-transient computer-readable storage medium storing computer instructions, wherein the computer instructions are used to enable a computer to execute the above-mentioned complex drilling accident alarm method based on small layer research.

[0079] The above are all preferred embodiments of the present invention, and are not intended to limit the protection scope of the present invention. Therefore, any equivalent changes made based on the structure, shape, and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A complex drilling accident alarm method based on small layer research, characterized by: include Build a tag library, target geological information, target accident complex situation data, and create a database; Obtain historical well data information and build complex accident alarm models; Obtain field data and determine the risk interval based on the tag library. Acquire the real-time data on site and the tag library data, predict the probability of various accidents occurring in the current well, and generate alarm information; Performing credibility assessment on the alarm information, If the credibility reaches a preset threshold, an alarm is triggered. If the credibility does not reach the threshold, no alarm is given.

2. The complex drilling accident alarm method based on small layer research according to claim 1 is characterized by: The tag library includes, north coordinate, east coordinate and altitude data as primary keys; Supported depth domain data measurement step as data interval; Determine data categories and parameter items, and complete the data in the data categories and parameter items that cannot be directly measured by obtaining indirect data through calculation.

3. The complex drilling accident alarm method based on small layer research according to claim 2 is characterized by: The data classification and the parameter items adopt a unified dimension.

4. The complex alarm method for drilling accidents based on small layer research according to claim 1 is characterized by: When constructing the complex accident alarm model, the historical well data is analyzed, including: noise processing, unified dimension, data reduction, data synthesis and data dimension reduction.

5. The complex alarm method for drilling accidents based on small layer research according to claim 1 is characterized by: The complex accident alarm model includes taking the historical well data as samples and dividing the time domain of the historical well data into time periods; Features are extracted from the historical well data to mark the differences between the features of normal operation, accident complex symptoms, and accident complex occurrence.

6. The complex drilling accident alarm method based on small layer research according to claim 5 is characterized by: Based on the characteristics, the algorithm is selected. The historical well data is put into the algorithm, machine learning is performed, the model is fitted, and the optimal algorithm is determined according to the degree of model fitting.

7. The complex alarm method for drilling accidents based on small layer research according to claim 5 is characterized by: Based on the historical well data, the accuracy and recall rate of the accident complex alarm model are verified by calculating according to the confusion matrix, and the weights of the parameters in the historical well data are adjusted to obtain the optimal model.

8. A complex alarm device for drilling accidents based on small layer research, characterized by: include: one or more processors; Memory; And one or more computer programs, wherein the one or more computer programs are stored in the memory, and the one or more computer programs include instructions, which, when executed by the alarm device, enable the alarm device to execute the complex drilling accident alarm method based on small layer research as described in any one of claims 1-7.

9. An electronic device, characterized in that: include: one or more processors; Memory; And one or more computer programs, wherein the one or more computer programs are stored in the memory, and the one or more computer programs include instructions, which, when executed by the electronic device, enable the electronic device to execute the complex drilling accident alarm method based on small layer research as described in any one of claims 1-7.

10. A non-transitory computer-readable storage medium storing computer instructions, characterized in that: The computer instructions are used to enable a computer to execute the complex drilling accident alarm method based on small layer research described in any one of claims 1-7.

Images