Mud logging anomaly identification and early warning training system and method of using the same

Abstract

The application discloses a mud logging abnormality identification and early warning training system and a use method thereof, relates to the technical field of mud logging professional technical training systems and methods, and comprises a mud logging real-time database, a mud logging abnormality prediction knowledge base, a well control learning unit and a warning learning unit. The warning learning unit comprises an abnormal curve simulation module, an extraction module, a labeling module, a student prediction filling module, an answer prediction generation module, an abnormality prediction comparison module and a prediction result analysis module. Based on the system, the whole process of online scene simulation of mud logging site abnormality reporting can be completed, so as to realize the training of mud logging abnormality identification and reporting, well control knowledge and the like in combination with real-time mud logging abnormality data, and the learning results of the trainees can be examined and evaluated, so that the training efficiency is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of mud logging professional technical training system and method, in particular to a mud logging abnormality identification and early warning training system and a method thereof. BACKGROUND

[0002] Abnormality prediction is an important part of mud logging work, which is related to the control of drilling safety and drilling cost, and also has the requirement of 100% abnormality prediction rate in mud logging quality assessment. However, due to the lack of experience of field staff and uneven business level, accurate judgment of abnormality of mud logging parameters cannot be made in time, which easily leads to the reduction of abnormality prediction rate and accuracy. The existing mud logging prediction training is mostly limited to PPT teaching material mode, which cannot combine with the dynamic change characteristics of mud logging parameter curve, and it is difficult to give the trainees a direct impression, which easily causes the problems of single training mode and limited training content, and it is difficult to achieve the expected training effect. The comprehensiveness and timeliness of the existing mud logging abnormality prediction training need to be improved, the training content received by each employee is not comprehensive and systematic enough, and it is difficult to effectively train all employees in time in the face of new mud logging abnormalities generated under new technologies and new processes.

[0003] For example, the patent document with the title of logging technology training simulation client system, the publication number of CN103854530A, which was published on June 11, 2014, records a logging technology training simulation client system, which comprises a teaching demonstration module, an interactive practice module, a skill examination module, a user information module and a database module; the teaching demonstration module is used for displaying the name of an object when the object is passed by a mouse in a virtual digital well site and popping up the corresponding course content when double-clicking an object in the digital well site scene, and the corresponding teaching demonstration screen appears when the user selects any course; the interactive contact module is used for the user to select the tool or accessory to be used in the tool box according to the corresponding business logic and place it in the position where the tool is to be used or the position where the accessory is to be installed, and automatically execute the corresponding action when placed correctly, and give an error prompt when placed incorrectly; the skill examination module is used for giving the user a corresponding score after the user practices according to a certain examination content. For another example, the patent document with the title of comprehensive logging simulation system, the publication number of CN104680878A, which was published on June 3, 2015, records a comprehensive logging simulation system, which comprises a teacher console, a sensor detection system, a sand table model, a power control system and a cabinet, the teacher console comprises a main control machine and a graphic machine, the cabinet comprises a chromatograph, a CO2 analyzer, a hydrogen generator, a collection computer, a chromatograph computer and a client group, the sensor detection system comprises a plurality of sensors, the plurality of sensors are collected in respective junction boxes and connected with the main control machine through a sensor signal bus cable, and various virtual simulation program modules are arranged on the main control machine; a virtual presentation module is arranged on the graphic machine. The prior art has the above-mentioned defects.

[0004] According to the demand investigation and business analysis, it is found that it is necessary to train the logging operation personnel for logging prediction, and a more efficient, fast and flexible training method is needed: a logging abnormality prediction knowledge base is established, a logging prediction training system is developed, the logging prediction knowledge base is continuously updated, the training and practice of employees are realized, and the abnormality prediction ability of employees is improved. SUMMARY

[0005] The purpose of the present application is to provide a logging abnormality identification and early warning training system and a use method thereof, which can complete online scene simulation of the whole process of logging site abnormality reporting, so as to realize training of logging abnormality identification and reporting, well control knowledge and the like in combination with real-time logging abnormality data online, thereby improving the training efficiency, and also realizing examination and evaluation of the learning results of the trainees.

[0006] The present application is realized by the following technical scheme:

[0007] The mud logging abnormality identification and early warning training system comprises a mud logging real-time database, a mud logging abnormality prediction knowledge base, a well control learning unit and an early warning learning unit,

[0008] The early warning learning unit comprises an abnormal curve simulation module, an extraction module, a labeling module, a student prediction filling module, an answer prediction generation module, an abnormality prediction comparison module and a prediction result analysis module.

[0009] The abnormal curve simulation module is connected with the mud logging abnormality prediction knowledge base, and the function of the abnormal curve simulation module is to simulate and generate corresponding mud logging curves according to different types of mud logging abnormality cases.

[0010] The well control learning unit comprises a well control knowledge simulation module and a result scoring module.

[0011] The mud logging real-time database is used to obtain mud logging data in a mud logging operation site.

[0012] The mud logging abnormality prediction database is connected with the mud logging real-time database, and the mud logging abnormality prediction database is used to standardize mud logging data into various mud logging abnormality parameters and store different types of mud logging abnormality cases.

[0013] The extraction module is connected with the mud logging abnormality prediction knowledge base and the abnormal curve simulation module, and the function of the extraction module is to randomly extract mud logging abnormality cases from the mud logging abnormality prediction knowledge base and play back mud logging curves corresponding to the mud logging abnormality cases.

[0014] The function of the labeling module is to label abnormal data points in the mud logging curves to generate labeling information.

[0015] The student prediction filling module is used to generate a student prediction filling sheet.

[0016] The answer prediction generation module is connected with the extraction module, and the answer prediction generation module is used to automatically generate an answer prediction sheet in combination with mud logging parameters.

[0017] The abnormality prediction comparison module is connected with the student prediction filling module and the answer prediction generation module, and the abnormality prediction comparison module is used to compare the student prediction filling sheet with the answer prediction sheet and output a student prediction result score.

[0018] The prediction result analysis module is connected with the abnormality prediction comparison module, and the prediction result analysis module is used to generate prediction result analysis and modification suggestions for the student prediction result.

[0019] The well control knowledge simulation module is used to realize a well control knowledge simulation examination.

[0020] The result scoring module is connected with the well control knowledge simulation module, and the result scoring module is used to complete result scoring.

[0021] Further, the generated annotation information in the annotation module includes annotation curve type, annotation time and annotation content.

[0022] Further, the answer forecast generated by the answer forecast generation module includes hash marks, date, well depth, abnormal type, abnormal start time, forecast time, forecast parameter, parameter change, analysis result forecast and suggested treatment measures in the student forecast fill-in form generated by the student forecast fill-in module.

[0023] Further, the logging abnormality prediction database can input new logging data.

[0024] Further, the logging abnormality prediction knowledge base includes each exploration area in the logging abnormality prediction case.

[0025] A use method of a logging abnormality identification and early warning training system, based on the foregoing logging abnormality identification and early warning training system, comprising the following steps:

[0026] S1, for different logging abnormality events, collect logging data in logging operation site;

[0027] S2, according to the parameter change characteristics in different types of logging abnormality events, standardize the logging data into each logging abnormality parameter;

[0028] S3, establish a logging abnormality prediction database to store logging abnormality parameters of different types of logging abnormality cases, and simulate corresponding logging curves for each logging abnormality parameter;

[0029] S4, according to the logging abnormality cases extracted by the student, automatically generate an answer forecast form, and compare the student forecast fill-in form with the answer forecast form, output the student forecast result score, forecast result analysis and modification suggestion.

[0030] Further, in step S4, the student forecast result score is obtained by scoring the student forecast result according to the established comprehensive evaluation model, different content on the forecast form can be made into different scoring model, and the comprehensive evaluation model can be made by combining each scoring model, the specific method is as follows:

[0031] S41, for pure numerical type answer content, an error scoring strategy is used to select an error range between the student's answer and the standard answer, the smaller the error range, the higher the single item score, wherein the pure numerical type answer content includes well depth, abnormal start time, prediction time, parameter change condition;

[0032] S42, for text answer content, the scoring of the recommended treatment measure answer item is to first train the text of the standard answer library using the Word2Vector model, establish the mapping relationship between the text and the numerical vector, and when scoring, convert the standard answer and the student's answer through the trained Word2Vector model between the word vector and the numerical vector, calculate the cosine distance of the converted numerical vector, and score the student's answer through the distance metric value;

[0033] S43, on the basis of calculating the scores of numerical type and text type answers, comprehensive scoring is performed.

[0034] Further, in step S41, the scoring rules for well depth, abnormal start time, prediction time, and parameter change condition are as follows:

[0035] a). The specific scoring calculation rule for the well depth item satisfies the relationship of formula (1):

[0036] (1),

[0037] In formula (1), S D represents the score of the student in the well depth answer item, which is dimensionless; D stu represents the answer filled by the student in the well depth answer item, m ; D sta represents the standard answer in the well depth answer item, m ; D rang is the error value allowed in the well depth answer item, m ;

[0038] b). The specific scoring calculation rule for the abnormal start time item satisfies the following formula (2) relationship:

[0039] First, convert the time format to seconds, and the calculation formula is as follows:

[0040] BT = H *3600+ M *60+ S (2),

[0041] In formula (2), BT represents the number of seconds of the abnormal start time, s; H is the time value in the time format, dimensionless; M is the minute value in the time format, dimensionless; S is the second value in the time format, dimensionless;

[0042] Then, on the basis of converting the time format into seconds, the score of the abnormal start time item is calculated, and the specific scoring rule satisfies the following formula (3):

[0043] (3),

[0044] In formula (3), S BT represents the score of the student in the abnormal start time item, dimensionless; BT stu represents the answer filled by the student in the abnormal start time answer item, s ; BT sta represents the standard answer in the abnormal start time answer item, s; BT rang is the allowable error value in the abnormal start time answer item, s;

[0045] c). The specific scoring rule for the forecast time item satisfies the following relationship formula (4):

[0046] First, the time format is converted into seconds, and the calculation formula is as follows:

[0047] WT = H *3600+ M *60+ S (4),

[0048] In formula (4), WT represents the second of the forecast time, s ; H is the time value in the time format; M is the minute value in the time format; S is the second value in the time format;

[0049] Then, on the basis of converting the time format into seconds, the score of the forecast time item is calculated, and the specific scoring rule satisfies the following formula (5):

[0050] (5),

[0051] In formula (5), S WT represents the score of the student in the forecast time item, dimensionless; WT stuan answer filled by the student in the prediction time answer item, s ; WT sta s, a standard answer in the prediction time answer item, WT rang an error value allowed in the prediction time answer item, s ;

[0052] d). The parameter change item specific scoring rule satisfies the following formula (6) relationship:

[0053] (6),

[0054] In formula (6), an answer filled by the student in the parameter change item, i an index of the parameter, taking a value in the range [1, M], M representing the total number of change parameters considered in a single prediction; P stu an answer filled by the student in the parameter change item, P sta s, a standard answer in the parameter change item, P rang an error value allowed in the parameter change item.

[0055] Further, in step S42, the method for obtaining the text answer content is:

[0056] i). First, the standard answer in the answer library is segmented, and when the text description type is segmented, irrelevant words and punctuation marks that do not consider the semantic information of the text are not considered;

[0057] ii). Then all the words in the answer library are encoded by one-hot encoder, and the length of the encoding is the total number of words in the answer library. On the basis of encoding, word vector mapping containing semantic information is performed, and word embedding word vector after word2vector model training is adopted;

[0058] iii). The scoring rule of the suggestion processing measure answer item satisfies the following formula (7) relationship:

[0059] (7),

[0060] In formula (7), S W a score of the suggestion processing measure answer item, an answer filled by the student in the suggestion processing measure answer item, i the value of the word embedding encoding of the answer filled by the student in the suggestion processing measure answer item at the i-th position, dimensionless; The numerical value of the word embedding encoding of the standard answer in the recommended processing measure answer item i , dimensionless.

[0061] Further, in step S43, on the basis of the numerical type and text type answer score calculation, comprehensive scoring is performed, and the scoring method satisfies the following formula (8) and (9) relationship:

[0062] (8);

[0063] (9);

[0064] In formula (8) and formula (9), S is the score of the student, dimensionless; w1 is the score weight of the depth item answer, dimensionless; w2 is the score weight of the abnormal start time item answer, dimensionless; w3 is the score weight of the abnormal early warning time item answer, dimensionless; w4 is the score weight of the parameter change item answer, dimensionless; and w5 is the score weight of the recommended processing measure item answer, dimensionless.

[0065] Compared with the prior art, the present application has the following advantages and beneficial effects:

[0066] First, in the present application, the establishment of the abnormal prediction knowledge base first faces the standardized conversion of the text abnormal report document and related data and the rule making, wherein the diversification of the data source and the case selection into the library are adjusted by various parameters, the latest one-to-one comparison between the actual and the document is carried out for many times, and the parameter adjustment and setting are carried out in combination with the artificial operation thinking, to form the knowledge base case basis. An independent mud logging abnormal case recognition, learning model and process are formed in the establishment process. A complete result evaluation, measurement analysis model and related multiple reinforcement learning strategy mechanism are constructed.

[0067] Second, in the present application, the mud logging curve simulation is an independent parameter model formed according to a large number of similar case analyses, to realize the virtualization of a similar abnormal case according to the set type under a specific working condition. At present, there are some mud logging abnormal recognition technologies, but there is a lack of simulation case learning technology. The technology will enrich the existing knowledge base and flexibly construct the case parameters and trend states, and is more variable, diverse and random. Through long-time learning, the sensitivity of personnel to the recognition of mud logging abnormality after the occurrence can be greatly improved.

[0068] Three, in the application, the mud logging data playback is to gradually display real or simulated case data in the form of real-time refreshing. The data playback of the scheme can realize the adjustment of different rates at any time and greatly enhance the controllability of the case, realizes the manual intervention in the playback process to pre-pause and start the speed adjustment. Technically, the svg vector mode is adopted in the curve display, which can realize higher definition of curve shape and vector compatibility. In the aspect of real-time data extraction, the mobile scale form is adopted to extract and display the data in real time through the floating layer. In the aspect of interface design, multiple floating layers can coexist and share, and each layer has operation ability to realize data communication and control.

[0069] Four, in the application, based on the mud logging abnormality identification and early warning training system, the scheme has the following advantages: first, the online training of mud logging abnormality identification and report, well control knowledge and other contents can be realized, and the training efficiency is improved; second, through the development of the system, the prediction and prediction level of the field mud logging personnel is further improved, and the safety awareness of well control of mud logging is improved; third, new cases can be continuously supplemented, which avoids the uneven training of employees, and the training is more comprehensive and systematic. BRIEF DESCRIPTION OF DRAWINGS

[0070] Figure 1 It is a structure diagram of the mud logging abnormality identification and early warning training system in the application.

[0071] Figure 2 It is the overall function framework of the mud logging abnormality identification and early warning training system in the application.

[0072] Figure 3 It is a common type selection diagram.

[0073] Figure 4 It is an abnormal curve marking a.

[0074] Figure 5 It is a student prediction filling diagram.

[0075] Figure 6 It is an answer viewing screenshot.

[0076] Figure 7 It is an abnormal prediction comparison diagram.

[0077] Figure 8 It is a result analysis screenshot.

[0078] Figure 9 It is a word2vector model structure diagram. DETAILED DESCRIPTION

[0079] The application will be further described in detail below in combination with examples, but the implementation manner of the application is not limited thereto.

[0080] Example 1

[0081] In order to carry out more efficient, fast and flexible mud logging prediction training for mud logging personnel, the mud logging abnormality identification and early warning training system is provided, the training system establishes the abnormality prediction case mud logging curve playback function; the output function of the student identifying abnormality, marking, picture grabbing, early warning, filling out the prediction form and the like is established; the output function of the student completely describing the oil, gas and water leakage display is established; the student output information is compared and analyzed with the standardized abnormality prediction knowledge base to judge the student training result.

[0082] More specifically, referring to Figure 1 , the mud logging abnormality identification and early warning training system comprises a mud logging real-time database, a mud logging abnormality prediction knowledge base, a well control learning unit and an early warning learning unit.

[0083] The early warning learning unit comprises an abnormal curve simulation module, an extraction module, a marking module, a student prediction filling module, an answer prediction generation module, an abnormality prediction comparison module and a prediction result analysis module; the well control learning unit comprises a well control knowledge simulation module and a result scoring module.

[0084] The overall function framework of the mud logging abnormality identification and early warning training system is shown in Figure 2 .

[0085] In the system, the mud logging real-time database is used to acquire mud logging data of a mud logging operation site.

[0086] The mud logging abnormality prediction database and the mud logging real-time database are configured to standardize the mud logging data into various mud logging abnormality parameters and store the mud logging abnormality cases of different types.

[0087] It should be noted that the mud logging abnormality prediction database should have the functions of continuous improvement and continuous expansion, and the cases should cover gas logging abnormality, gas invasion, overflow, drill tool stabbing, well leakage, water eye plugging, hydrogen sulfide abnormality, sticking and the like. The operation site of the prediction case includes various exploration well areas and other geographic locations, and has good universality.

[0088] The abnormality prediction case categories include gas logging abnormality, gas invasion, overflow, drill tool stabbing, well leakage, water eye plugging, hydrogen sulfide abnormality, sticking and the like. The collected case conditions are as shown in Table 1.

[0089] Table 1: Abnormality category statistics table of collected cases

[0090]

[0091] In this embodiment, the abnormal curve simulation module is connected with the mud logging abnormality prediction knowledge base, and the abnormal curve simulation module is used to simulate and generate corresponding mud logging curves according to different types of mud logging abnormality cases.

[0092] In this embodiment, the extraction module is connected with the logging abnormality prediction knowledge base and the abnormal curve simulation module respectively, and is used for randomly extracting logging abnormality cases from the logging abnormality prediction knowledge base and playing back logging curves corresponding to the logging abnormality cases. It is referred to Figure 3 As shown in the figure, Figure 3 The logging abnormality cases obtained according to the selected abnormal type are extracted.

[0093] In this embodiment, the labeling module is configured to label the abnormal data points in the logging curve to generate labeling information. The labeling information includes labeling curve type, labeling time and labeling content. As Figure 4 The labeling result of the abnormal curve is shown in the figure.

[0094] In this embodiment, the student prediction filling module is used to generate a student prediction filling sheet. As Figure 5 The page of student prediction filling is shown in the figure.

[0095] In this embodiment, the answer prediction generation module is connected with the extraction module, and the answer prediction generation module is used to automatically generate an answer prediction sheet in combination with logging parameters.

[0096] In this embodiment, the student prediction filling sheet and the answer prediction sheet both include well number, date, well depth, abnormal type, abnormal start time, prediction time, prediction parameter, parameter change, analysis result prediction and suggested treatment measures. As Figure 6 The generated prediction answer is shown in the figure.

[0097] In this embodiment, the abnormal prediction comparison module is connected with the student prediction filling module and the answer prediction generation module respectively, and the abnormal prediction comparison module is used to compare the student prediction filling sheet with the answer prediction sheet, and output a student prediction result score. As Figure 7 The abnormal prediction comparison result is shown in the figure.

[0098] In this embodiment, the prediction result analysis module is connected with the abnormal prediction comparison module, and the prediction result analysis module is used to generate prediction result analysis and modification suggestions for the student prediction result. As Figure 8 The result analysis result is shown in the figure.

[0099] In this embodiment, the well control knowledge simulation module is used to realize a well control knowledge simulation test.

[0100] In this embodiment, the result scoring module is connected with the well control knowledge simulation module, and the result scoring module is used to complete result scoring.

[0101] In summary, the logging abnormality identification and early warning training system in this embodiment can mainly realize the following functions:

[0102] (1) Curve playback, curve playback of the time database;

[0103] (2) Curve annotation, annotate curve abnormal points;

[0104] (3) Curve management, add, delete curves, and increase data lines for observation;

[0105] (4) Automatic image capture, form a curve graph of the data in a specific time period and store it;

[0106] (5) Oil, gas and water leakage display output function, automatically convert parameters into standard output format;

[0107] (6) Filling of logging engineering anomaly report form to form a standardized prediction form;

[0108] (7) Learning evaluation, scoring the answers filled by the students;

[0109] (8) Simulated curve, simulate abnormal curves according to abnormal prediction cases.

[0110] Based on the above logging abnormality identification and early warning training system, the present application also provides a use method of the logging abnormality identification and early warning training system, and the training method comprises the following steps:

[0111] S1, for different logging abnormal events, collect logging data in the logging operation site.

[0112] S2, according to the parameter change characteristics in different types of logging abnormal events, the logging data is standardized to form various logging abnormal parameters. The logging abnormal prediction parameter reference table 2 is formed after the standardization processing, and table 2 shows part of the logging abnormal prediction parameter characteristics after the standardization processing.

[0113] Table 2: Logging abnormal prediction parameter characteristics table (part)

[0114]

[0115] S3, establish a logging abnormal prediction database to store logging abnormal parameters of different types of logging abnormal cases, and simulate corresponding logging curves for each logging abnormal parameter.

[0116] S4, according to the logging abnormal cases extracted by the students, automatically generate answer prediction forms, and compare the student prediction forms with the answer prediction forms, output the student prediction result score, prediction result analysis and modification suggestions.

[0117] The student prediction result score is obtained by scoring the student prediction result according to the established comprehensive evaluation model, different content on the prediction sheet can develop different scoring models, and the comprehensive evaluation model can be developed in combination with each scoring model. The specific implementation steps of the student prediction result score are as follows:

[0118] S41, for pure numerical answer content, an error scoring strategy is adopted to select an error range between the student answer and the standard answer, and the smaller the error range, the higher the single item score. Among them, the pure numerical answer content includes well depth, abnormal start time, prediction time, parameter change, etc.

[0119] a). The specific scoring calculation rule of well depth item is as follows:

[0120] (1),

[0121] In formula (1), S D represents the score of the student in the well depth answer item, dimensionless; D stu represents the answer filled by the student in the well depth answer item, m ; D sta represents the standard answer in the well depth answer item, m ; D rang is the error value allowed in the well depth answer item, m .

[0122] b). The specific scoring calculation rule of abnormal start time item is as follows:

[0123] First, convert the time format to seconds, and the calculation formula is as follows:

[0124] BT = H *3600+ M *60+ S (2),

[0125] In formula (2), BT represents the seconds of the abnormal start time, s ; H is the time format hour value, dimensionless; M is the time format minute value, dimensionless; S is the time format second value, dimensionless.

[0126] Then, on the basis of converting the time format to seconds, the scoring of the abnormal start time item is calculated, and the specific scoring rule is as follows:

[0127] (3),

[0128] In formula (3), S BT represents the score of the student in the abnormal start time item, dimensionless; BT stu represents the answer filled by the student in the abnormal start time answer item, s ; BT sta represents the standard answer in the abnormal start time answer item, s; BT rang is the allowable error value in the abnormal start time answer item, s.

[0129] c). The specific scoring rules for the forecast time item are as follows:

[0130] First, convert the time format to seconds, and the calculation formula is as follows:

[0131] WT = H *3600+ M *60+ S (4),

[0132] In formula (4), WT represents the number of seconds of the forecast time, s ; H is the hour value in the time format; M is the minute value in the time format; S is the second value in the time format.

[0133] Then, on the basis of converting the time format to seconds, the scoring of the forecast time item is calculated, and the specific scoring rules are as follows:

[0134] (5),

[0135] In formula (5), S WT represents the score of the student in the forecast time item, dimensionless; WT stu represents the answer filled by the student in the forecast time answer item, s ; WT sta represents the standard answer in the forecast time answer item, s; WT rang is the allowable error value in the forecast time answer item, s .

[0136] d). The specific scoring rules for the parameter change situation item are as follows:

[0137] (6),

[0138] In equation (6), This represents the student's score in the parameter variation section, and is dimensionless. i This indicates the parameter number, with a value range of [1, M], where M represents the total number of varying parameters considered in a single forecast. P stu This indicates the answers filled in by the trainees in the parameter change section; the specific units depend on the parameter type. P sta This indicates the standard answer in the answer section for parameter variation items; the specific unit depends on the parameter type. P rang This represents the allowable error value in the answer section for parameter variation; the specific unit depends on the parameter type.

[0139] S42. For text answer content, such as the scoring of suggested treatment measures, the standard answer library text is first trained using a Word2Vector model to establish a mapping relationship between text and numerical vectors. When scoring, the standard answer and the student's answer are converted from word vectors to numerical vectors through the trained Word2Vector model. The cosine distance of the converted numerical vectors is calculated, and the distance metric is used to score the student's answer.

[0140] The specific calculation process is as follows:

[0141] i) First, segment the answers to the processing measures in the standard answer database into words. For example, the specific word segmentation steps are as follows:

[0142] [“Stop pump, observe, plug leak”]->[“Stop pump”, “Observe”, “Plug leak”].

[0143] When segmenting text description types here, irrelevant words such as modal particles and stop words, as well as punctuation marks, are not considered.

[0144] ii) Then, all words in the answer database are encoded using a one-hot encoder, meaning each word has a unique code. The specific encoding process is as follows:

[0145] "Stop pump" -> [1,0,0,0,0,0,0,...,0];

[0146] “Observation”->[0,1,0,0,0,0,0,...,0];

[0147] "Plug the leak" -> [0,0,1,0,0,0,0,...,0].

[0148] The length of the encoding is the total number of words in the answer database. Based on the encoding, word vector mapping containing semantic information can be performed. The model used in this patent is the word2vector model, and the model structure diagram is as follows. Figure 9 As shown. In the model, the input vector x k ( k =1,2,3,…, V () is the one-hot vector encoding of a single word; V This indicates the total number of words in the answer bank; h i The numerical value of the coding layer, N This is the length of the output word vector; y j ( j =1,2,3,…, V ) represents the word embeddings after the model has been trained.

[0149] iii) The scoring rules for the suggested solutions are as follows:

[0150] (7);

[0151] In equation (7), S W The score for the suggested treatment measures in the answer is dimensionless. The word embedding code represents the answer that the trainee filled in in the suggested treatment measures answer item. i The numerical value at each position is dimensionless; The word embedding code in the answer item indicating the suggested treatment measures is the first word of the standard answer. i The numerical value at each position is dimensionless.

[0152] S43. A comprehensive score will be awarded based on the calculated scores for both numerical and text-based answers. The scoring details are as follows:

[0153] (8);

[0154] (9);

[0155] In equations (8) and (9), S represents the student's score, which is dimensionless; w1 represents the weight of the answer score for the depth item, which is dimensionless; w2 represents the weight of the answer score for the abnormal start time item, which is dimensionless; w3 represents the weight of the answer score for the abnormal warning time item, which is dimensionless; w4 represents the weight of the answer score for the parameter change item, which is dimensionless; and w5 represents the weight of the answer score for the suggested handling measures item. The specific values ​​of the above weights are given by experts and are dimensionless.

[0156] The above is only the preferred embodiment of the present application, and does not limit the present application in any form. Any simple modification or equivalent change of the above embodiment according to the technical essence of the present application falls within the protection scope of the present application.

Claims

1. A mud logging abnormality identification and early warning training system, characterized in that: The well control learning unit comprises a well control knowledge simulation module and a result scoring module. The real-time logging database is used to acquire logging data of a logging operation site. The logging abnormality prediction database is connected with the real-time logging database, and is used to standardize the logging data into logging abnormality parameters and divide the logging abnormality parameters into different types of logging abnormality cases for storage. The labeling module is used to label abnormal data points in the logging curve to generate labeling information. The student prediction filling module is used to generate a student prediction filling sheet. The answer prediction generation module is connected with the extraction module, and is used to automatically generate an answer prediction sheet in combination with logging parameters. The abnormality prediction comparison module is connected with the student prediction filling module and the answer prediction generation module, and is used to compare the student prediction filling sheet with the answer prediction sheet and output a student prediction result score. The prediction result analysis module is connected with the abnormality prediction comparison module, and is used to generate prediction result analysis and modification suggestions for the student prediction result. The well control knowledge simulation module is used to realize well control knowledge simulation examination. The result scoring module is connected with the well control knowledge simulation module, and is used to complete result scoring. The labeling information generated in the labeling module comprises labeling curve types, labeling times and labeling contents. The answer prediction sheet generated by the answer prediction generation module and the student prediction filling sheet generated by the student prediction filling module both comprise hash marks, dates, well depths, abnormality types, abnormality start times, prediction times, prediction parameters, parameter change conditions, analysis result predictions and suggested treatment measures. The logging abnormality prediction database can input new logging data. The logging abnormality prediction knowledge base comprises different exploration well areas.

2. The mud logging anomaly identification and early warning training system of claim 1, wherein: The logging abnormality prediction knowledge base comprises gas logging abnormality, gas invasion, overflow, drill tool sticking, well leakage, water eye plugging, hydrogen sulfide abnormality and sticking.

3. The mud logging anomaly identification and early warning training system of claim 1, wherein: The logging abnormality identification and early warning training system according to claim 1 comprises the following steps:

4. The mud logging anomaly identification and early warning training system of claim 1, wherein: S1, logging data of a logging operation site is collected for different logging abnormality events; 5. The mud logging anomaly identification and early warning training system of claim 1, wherein: S2, logging data is standardized into logging abnormality parameters according to parameter change characteristics in different types of logging abnormality events; 6. A method for using a training system for identifying and warning of logging anomalies, characterized in that, ​ ​ ​ S3, establish a mud logging anomaly prediction database to store mud logging anomaly parameters of different types of mud logging anomaly cases, and simulate corresponding mud logging curves for each mud logging anomaly parameter; S4, according to the mud logging anomaly cases extracted by the students, automatically generate an answer prediction sheet, and compare the student prediction sheet with the answer prediction sheet to output the student prediction result score, prediction result analysis and modification suggestions.

7. The use of a mud logging anomaly identification and early warning training system according to claim 6, characterized in that, In step S4, the student prediction result score is obtained by scoring the student prediction result according to the established comprehensive evaluation model. Different content on the prediction sheet can have different scoring models. The comprehensive evaluation model can be established by combining the scoring models. The specific method is as follows: S41, for pure numerical answer content, an error scoring strategy is used to select an error range between the student answer and the standard answer. The smaller the error range, the higher the single score. Pure numerical answer content includes well depth, abnormal start time, prediction time, and parameter change; S42, for text answer content, the scoring of the recommended treatment measure answer item is to first train the text in the standard answer library using the Word2Vector model to establish the mapping relationship between the text and the numerical vector. When scoring, the standard answer and the student answer are converted from the word vector to the numerical vector through the trained Word2Vector model. The cosine distance of the converted numerical vector is calculated to score the student answer; S43, comprehensive scoring is performed based on the numerical and text answer calculation scores.

8. The use of a mud logging anomaly identification and early warning training system according to claim 7, characterized in that: In step S41, the scoring rules for well depth, abnormal start time, prediction time, and parameter change are as follows: a). The well depth item scoring calculation rule satisfies the relationship of formula (1): （1）， In formula (1), S D denotes the score of the student in the well-depth answer item, dimensionless; D stu denotes the answer filled in by the student in the well-depth answer item, m ; D sta denotes the standard answer in the well-depth answer item, m ; D rang is the error value allowed in the well-depth answer item, m ; b). For the abnormal start time item, the specific scoring calculation rule satisfies the following formula (2): First, convert the time format to seconds, and the calculation formula is as follows: BT = H *3600+ M *60+ S （2）， In formula (2), BT represents the number of seconds from the start of the anomaly, s H is the number of hours in the time format, dimensionless; M is the number of minutes in the time format, dimensionless; S is the number of seconds in the time format, dimensionless;​ Then, based on the time format converted to seconds, calculate the scoring of the abnormal start time item, and the specific scoring rule satisfies the following formula (3): （3）， In formula (3), S BT denotes the score of the student in the abnormal start time item, dimensionless; BT stu denotes the answer filled in by the student in the abnormal start time answer item, s ; BT sta denotes the standard answer in the abnormal start time answer item, s; BT rang is the error value allowed in the abnormal start time answer item, s; c). For the prediction time item, the specific scoring rule satisfies the following relationship formula (4): First, convert the time format to seconds, and the calculation formula is as follows: WT = H *3600+ M *60+ S （4）， In formula (4), WT represents the number of seconds of the prediction time, s ; H is the hour value in the time format; M is the minute value in the time format; S is the second value in the time format; Then, based on the time format converted to seconds, calculate the scoring of the prediction time item, and the specific scoring rule satisfies the following formula (5): （5）， In formula (5), S WT denotes the score of the student in the prediction time item, dimensionless; WT stu denotes the answer filled in by the student in the prediction time answer item, s ; WT sta denotes the standard answer in the prediction time answer item, s; WT rang is the error value allowed in the prediction time answer item, s ; d). The specific scoring rule for the parameter change item satisfies the following formula (6): （6）， In formula (6), represents the student score in the parameter variation term, dimensionless; i The number of parameters, the value range is [1, M], M represents the total number of change parameters considered in single prediction; P stu The answer filled in by the student in the parameter change item; P sta The standard answer in the parameter change item answer; P rang The error value allowed in the parameter change item answer.

9. The use of a mud logging anomaly identification and early warning training system according to claim 8, characterized in that: In step S42, the method for obtaining the text answer content is as follows: i). First, divide the standard answer in the answer library into words. When dividing the text description type, do not consider irrelevant words and punctuation marks that do not have semantic meaning; ii). Then, encode all the words in the answer library using one-hot encoder. The length of the encoding is the total number of words in the answer library. Based on the encoding, perform word vector mapping that contains semantic information, and use the word2vector model to train the word embedding word vector; iii). The scoring rule of the recommended treatment measure answer item satisfies the following formula (7) relationship: （7）； In formula (7), S W Score for the answer item of the suggested handling measure, dimensionless Value of the number of the word embedding encoding of the position i in which the answer filled in by the student in the answer item of the suggested handling measure, dimensionless Value of the number of the word embedding encoding of the position i in which the standard answer in the answer item of the suggested handling measure, dimensionless V Total number of words in the answer library N Length of the output word vector.

10. The use of a mud logging anomaly identification and early warning training system according to claim 9, characterized in that: In step S43, on the basis of the numerical and text answer score calculation, comprehensive scoring is performed, and the scoring method satisfies the following formula (8) and (9) relationships: （8）； （9）； In formula (8) and formula (9), S is the score of the student, dimensionless; w1 is the weight of the depth item answer score, dimensionless; w2 is the weight of the abnormal start time item answer score, dimensionless; w3 is the weight of the abnormal early warning time item answer score, dimensionless; w4 is the weight of the parameter change item answer score, dimensionless; w5 is the weight of the recommended treatment measure item answer score, dimensionless.

Images