Drilling Risk Diagnosis Method and Computer Equipment Based on Knowledge Graph

Abstract

This specification relates to the technical field of oil and gas exploration, and particularly to a drilling risk diagnosis method and a computer device based on a knowledge graph. The method includes: constructing a knowledge graph ontology according to drilling data; extracting drilling risk instances according to drilling logs; constructing a drilling risk knowledge graph according to the knowledge graph ontology and the drilling risk instances; constructing a drilling risk question-answering system according to the drilling risk knowledge graph; wherein, the drilling risk question-answering system is used to determine corresponding drilling risk diagnosis results according to the input drilling risk questions. The embodiments of this specification can improve the accuracy of risk diagnosis.

Description

Technical Field

[0001] This specification relates to the technical field of oil and gas exploration, and particularly to a drilling risk diagnosis method and a computer device based on a knowledge graph. Background Art

[0002] With the continuous deepening of modern oil and gas exploration and development work, the geological conditions and reservoirs of exploration and development targets are becoming more and more complex, and the development difficulty is increasing day by day. This also makes more and more abnormal working conditions encountered during drilling, and the frequency of risks is getting higher and higher. Therefore, the demand for accurate diagnosis of abnormal working conditions and risks during drilling is becoming more and more urgent.

[0003] In recent years, significant progress has been made in oil and gas drilling technology, and the popularization of comprehensive logging technology has made it possible to analyze downhole working conditions in real time. Ground experts can analyze the abnormal state downhole in real time through logging data, effectively avoiding the occurrence of some risks and accidents. However, the timeliness of expert discrimination is poor, and the accuracy of manual discrimination is also low, and errors are prone to occur.

[0004] In addition, with the continuous development and application of artificial intelligence technology, more and more intelligent algorithm models such as support vector machines and artificial neural networks are applied to the oil and gas drilling process to predict abnormal working conditions and risks during drilling. However, most intelligent algorithm models are black-box attributes, with poor interpretability and accuracy. The actual application effect on site often fails to meet the requirements. Summary of the Invention

[0005] Embodiments of this specification provide a drilling risk diagnosis method and a computer device based on a knowledge graph to improve the accuracy of risk diagnosis.

[0006] Embodiments of this specification provide a drilling risk diagnosis method based on a knowledge graph, including:

[0007] Construct a knowledge graph ontology according to drilling data;

[0008] Extract drilling risk instances according to drilling logs;

[0009] Construct a drilling risk knowledge graph according to the knowledge graph ontology and drilling risk instances;

[0010] Construct a drilling risk Q&A system according to the drilling risk knowledge graph; wherein, the drilling risk Q&A system is used to determine corresponding drilling risk diagnosis results according to the input drilling risk questions.

[0011] Embodiments of this specification also provide another drilling risk diagnosis method based on a knowledge graph, including:

[0012] Receive drilling risk questions;

[0013] Determine the corresponding problem categories according to the drilling risk problems;

[0014] Determine the corresponding answer templates according to the problem categories;

[0015] Query the answers corresponding to the drilling risk problems and problem categories according to the drilling risk knowledge graph;

[0016] Generate the drilling risk diagnosis results according to the answers and answer templates.

[0017] The embodiment of this specification also provides another drilling risk diagnosis method based on a knowledge graph, including:

[0018] Generate drilling risk problems when abnormal mud logging data is monitored;

[0019] Query at least one candidate risk type corresponding to the drilling risk problem according to the drilling risk knowledge graph;

[0020] Match the on-site characterization data with the characterization data of the at least one candidate risk type;

[0021] Use the candidate risk type corresponding to the successfully matched characterization data as the target risk type, and query the solution measures corresponding to the target risk type according to the drilling risk knowledge graph.

[0022] The embodiment of this specification also provides a computer device, including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the computer program, the above-mentioned drilling risk diagnosis method is implemented.

[0023] The drilling risk diagnosis method in the embodiment of this specification can construct the knowledge graph ontology by sorting out the literature and materials related to drilling abnormal risks; can make full use of the on-site drilling logs to extract abnormal risk instances; can construct the drilling risk knowledge graph according to the knowledge graph ontology and drilling risk instances. In this way, the professional knowledge related to drilling risks and a large number of drilling abnormal risk instances can be constructed into a drilling risk knowledge graph containing a large amount of expert knowledge and actual cases. The drilling risk Q&A system can be constructed according to the drilling risk knowledge graph. It can greatly improve the cognitive ability of drilling risks. It provides technical support and important reference for accurately and quickly identifying drilling risks at the operation sites where drilling accidents are extremely likely to occur, such as deep water, deep layer, and unconventional operations.

[0024] The drilling risk diagnosis method in the embodiment of this specification can greatly improve the cognitive ability of drilling risks through the drilling risk knowledge graph, improve the accuracy of risk diagnosis, and provide technical support and important reference for accurately and quickly identifying drilling risks at the operation sites where drilling accidents are extremely likely to occur, such as deep water, deep layer, and unconventional operations. Brief Description of the Drawings

[0025] To more clearly illustrate the technical solutions in the embodiments of this specification or the prior art, the following will briefly introduce the accompanying drawings required for the description of the embodiments or the prior art. The accompanying drawings in the following description are only some embodiments recorded in this specification. For those of ordinary skill in the art, without creative efforts, other accompanying drawings can also be obtained based on these drawings.

[0026] Figure 1 It is a schematic flow chart of the drilling risk diagnosis method in the embodiments of this specification;

[0027] Figure 2 It is a schematic diagram of the drilling risk diagnosis process in the embodiments of this specification;

[0028] Figure 3 It is a schematic diagram of the knowledge graph ontology in the embodiments of this specification;

[0029] Figure 4 It is a schematic diagram of the named entity recognition process in the embodiments of this specification;

[0030] Figure 5 It is a schematic diagram of the relationship category determination process in the embodiments of this specification;

[0031] Figure 6 It is a schematic diagram of the drilling risk entity verification process in the embodiments of this specification;

[0032] Figure 7 It is a schematic diagram of the drilling risk knowledge graph in the embodiments of this specification;

[0033] Figure 8 It is a schematic diagram of the drilling risk knowledge graph in the embodiments of this specification;

[0034] Figure 9 It is a schematic functional structure diagram of the drilling risk Q&A system in the embodiments of this specification;

[0035] Figure 10 It is a schematic flow chart of the drilling risk diagnosis method in the embodiments of this specification;

[0036] Figure 11 It is a schematic diagram of the drilling risk diagnosis process in the embodiments of this specification;

[0037] Figure 12 It is a schematic flow chart of the drilling risk diagnosis method in the embodiments of this specification;

[0038] Figure 13 It is a schematic functional structure diagram of the drilling risk diagnosis device in the embodiments of this specification;

[0039] Figure 14This is a schematic diagram of the functional structure of the drilling risk diagnosis device in the embodiments of this specification;

[0040] Figure 15 This is a schematic diagram of the functional structure of the drilling risk diagnosis device in the embodiments of this specification. Detailed implementation manners

[0041] Next, the technical solutions in the embodiments of this specification will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of this specification. Obviously, the described embodiments are only a part of the embodiments of this specification, rather than all the embodiments. The specific embodiments described herein are only used to explain this disclosure, rather than limiting this disclosure. Based on the described embodiments of this disclosure, all other embodiments obtained by those of ordinary skill in the art belong to the scope of protection of this disclosure. In addition, relational terms such as "first" and "second" are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations.

[0042] Please refer to Figure 1 and Figure 2 . The embodiments of this specification provide a drilling risk diagnosis method based on a knowledge graph. The drilling risk diagnosis method can be applied to a computer device. The drilling risk diagnosis method specifically may include the following steps.

[0043] Step 11: Construct a knowledge graph ontology according to drilling data.

[0044] In some embodiments, the drilling data may include documents, books, data manuals, etc. related to drilling risks. The drilling data may include unstructured data and semi-structured data. The unstructured data may include text data in documents, books, data manuals, etc. The semi-structured data may include data such as tables and standardized charts in documents, books, data manuals, etc. Specifically, natural language processing methods can be used to extract drilling risk knowledge from the drilling data; the extracted drilling risk knowledge can be organized into a drilling risk knowledge system, thereby obtaining the knowledge graph ontology.

[0045] In some embodiments, drilling risk knowledge can be directly extracted from the drilling data. Alternatively, the drilling data can also be preprocessed, and drilling risk knowledge can be extracted from the preprocessed drilling data. For example, the drilling data can be converted into a file in a preset format convenient for natural language processing; the drilling risk knowledge can be extracted from the file in the preset format. The preset format may include a text format (such as txt format), a lightweight data exchange format (such as json format), etc.

[0046] In some embodiments, drilling risk knowledge can be extracted from drilling data through rule-based and machine learning-based methods. The drilling risk knowledge may include expert experience knowledge such as the definition of drilling construction, drilling anomalies, and drilling risks. Among them, the drilling risk knowledge extracted by the rule-based method may include equipment in the drilling process, operations in the drilling process, names of drilling real-time data parameters, names of normal and abnormal drilling conditions, etc. The rule-based method may include a method based on regular expressions and a method based on a dictionary. In practical applications, the method based on regular expressions can be combined with the method based on a dictionary. For example, keywords can be used as a dictionary, and key information can be extracted through regular expressions and spliced to obtain text for describing drilling risk knowledge. The drilling risk knowledge extracted by the machine learning-based method may include descriptions of drilling risk types and definitions, descriptions of drilling conditions and risk parameter characterizations, risk mitigation measures, etc. The machine learning-based method may include extracting drilling risk knowledge from drilling data based on models such as conditional random fields and SVMs.

[0047] The knowledge graph ontology can be a drilling risk knowledge system. The drilling risk knowledge system can be a tree structure for representing the hierarchical relationship between drilling risk knowledge. The drilling risk knowledge can be used to represent the definition of drilling risks, risk characterizations, causes and conditions of occurrence, occurrence frequencies, etc. The extracted drilling risk knowledge can be organized into a tree structure to obtain a framework of the drilling risk knowledge system; the framework of the drilling risk knowledge system can be used as the knowledge graph ontology. Specifically, the extracted drilling risk knowledge can be organized into a framework of the drilling risk knowledge system through a method combining top-down and bottom-up approaches. For example, the expert knowledge and domain knowledge of oil and gas drilling risks can be integrated, the basic concepts in the drilling risk knowledge system can be defined and refined first, and then the knowledge can be gradually enriched and refined and the ontology design can be improved. It should be noted that the framework of the drilling risk knowledge system may include multiple nodes. Specifically, the framework of the drilling risk knowledge system may include at least one root node and at least two leaf nodes subordinate to the root node. Each node can represent a type of drilling risk knowledge.

[0048] Please refer to Figure 3 。 Figure 3A constructed knowledge graph ontology is shown. The drilling risk knowledge in the knowledge graph ontology includes knowledge related to the occurrence of drilling risks and knowledge related to the prevention of drilling risks. The knowledge related to the occurrence of drilling risks may include types of drilling risks, risk occurrence probabilities, drilling risk signs, causes of abnormal risks, abnormal handling measures, risk regulation, etc. Among them, the types of drilling risks may include stuck pipe, overflow, drill string breakage, stick-slip, etc. The drilling risk signs may include various relevant representations before and during the occurrence of different risks, such as a sudden decrease in torque, a decrease in drilling speed, a decrease in rotary table speed, an increase in pump pressure, etc. The knowledge related to the prevention of drilling risks may include risk prevention measures, emergency plans, prevention cost analysis, etc.

[0049] Step 12: Extract drilling risk instances according to the drilling log.

[0050] In some embodiments, the drilling comprehensive logging system records the changes in real-time logging data during the drilling process. As a kind of time-series data, the real-time logging data records the change trends of logging parameters before, after, and during the occurrence of abnormalities and risks. The drilling log can be extracted from the real-time logging data. Specifically, as an important basis for reflecting the drilling state and analyzing drilling risks, the trends and changes of logging parameters when they are abnormal are recorded in the drilling log of the actual well site. The drilling log can be text data. Thus, drilling risk instances can be extracted from the drilling log through natural language processing methods. The extraction of drilling risk instances includes extracting drilling risk entities, entity categories, relationship categories between entities, etc.

[0051] In some embodiments, a language model can be specifically used to extract drilling risk instances from the drilling log. The language model may include the BERT pre-trained model based on the deep learning Transformer framework. In practical applications, a pre-trained language model can be obtained. Considering that the oil and gas exploration technology field belongs to a professional technical field and the vocabulary related to drilling risks belongs to professional vocabulary, if the pre-trained language model is directly used to extract drilling risk instances, the effect will be poor. For this reason, the pre-trained language model can be fine-tuned with the professional vocabulary in the oil and gas exploration technology field. For example, the pre-trained language model can be fine-tuned with the drilling risk vocabulary extracted from drilling materials through rule-based and machine learning methods. The fine-tuned language model can be used to extract drilling risk instances from the drilling log, thereby improving the extraction effect of drilling risk instances.

[0052] In some embodiments, named entity recognition can be performed on the drilling log to obtain drilling risk entities and their corresponding entity categories. The drilling risk entities are used to represent drilling risk knowledge. The entity category is an entity category related to drilling risks. Table 1 shows some drilling risk entities and their corresponding entity categories.

[0053] Table 1

[0054]

[0055] Please refer to Figure 4 。The language model may include a named entity recognition model. The drilling log can be tokenized; the parts of speech of the tokens can be labeled; the tokens and their corresponding parts of speech can be input into the named entity recognition model to obtain the drilling risk entities output by the named entity recognition model and their corresponding entity categories. The drilling risk entities include tokens for describing real-world entities, etc. For example, the named entity recognition model may include a softmax classification layer, and the softmax classification layer is used to output the drilling risk entities and their corresponding entity categories. The parts of speech may include adjectives, conjunctions, adverbs, interjections, locative words, numerals, nouns, etc. Specifically, all the tokens in the drilling log and their corresponding parts of speech can be input into the named entity recognition model together. Or, the drilling log may include at least one sentence. The tokens and their corresponding parts of speech in a sentence can also be input into the named entity recognition model together.

[0056] In some embodiments, please refer to Figure 5 。The language model may include a relationship recognition model. Multiple drilling risk entities and the association data between the multiple drilling risk entities can be input into the relationship recognition model to obtain the output of the relationship recognition model. The output is used to indicate that there is no relationship between the multiple drilling risk entities. Or, the output is used to indicate the relationship category between the multiple drilling risk entities. The association data between the drilling risk entities can be obtained from the drilling log. The relationship category may include belonging to, containing, causality, solution measures, characterization, etc. For example, the relationship category between the drilling risk entity "differential sticking" and the drilling risk entity "sticking" is belonging to. Another example is that the relationship category between the drilling risk entity "jarring to release stuck pipe" and the drilling risk entity "sticking" is solution measures. Another example is that the relationship category between the drilling risk entity "wellbore pressure is greater than formation pressure" and the drilling risk entity "lost circulation" is causality. Another example is that the relationship between the drilling risk entity "imbalance between inlet and outlet flow rates" and the drilling risk entity "lost circulation" is characterization.

[0057] In practical applications, the extracted drilling risk entities can be combined in pairs to obtain multiple groups of drilling risk entities. Each group of drilling risk entities includes 2 drilling risk entities. For each group of drilling risk entities, the 2 drilling risk entities in the group of drilling risk entities and the association data between the 2 drilling risk entities can be input into a relationship recognition model to obtain the output of the relationship recognition model. The output can be used to indicate that there is no relationship between the 2 drilling risk entities. Alternatively, the output can also be used to indicate the relationship category between the 2 drilling risk entities. For example, a statement containing 2 drilling risk entities can be obtained from a drilling log; special symbols can be added on both sides of the drilling risk entities in the statement for marking; delimiters can be added at the beginning and / or end of the statement; and then the statement can be input into the relationship recognition model. The relationship recognition model can include an embedding layer, a linear layer, a softmax classification layer, etc. The embedding representation (Embedding) of the drilling risk entities in the statement and the embedding representation of the association data between the drilling risk entities can be obtained through the embedding layer. Multiple embedding representations can be concatenated through the linear layer. The concatenated result can be processed through the softmax classification layer to obtain the output of the relationship recognition model.

[0058] In some embodiments, refer to Figure 6 . Logging data matching the drilling log can be obtained; the logging data and the drilling log can be time-aligned; the drilling risk entities can be verified using the aligned logging data; for the drilling risk entities that pass the verification, the attribute data of the drilling risk entities can be extracted from the aligned logging data.

[0059] The drilling comprehensive logging system records the real-time change of logging data during the drilling process. The real-time logging data can be a type of drilling time-series data, which records the data change trends before, during, and after the occurrence of anomalies and risks. As an important basis for reflecting the drilling state and analyzing drilling risks, the trends and changes of logging data when anomalies occur are recorded in the drilling log. However, the numerical values and time records in the drilling log are generally approximate values. Therefore, the time nodes recorded in the drilling log can be corrected by the detailed time recorded by the logging system. The relevant descriptions in the drilling log can be supplemented with more detailed parameter changes to achieve mutual supplementation and alignment with the drilling log, improving the authenticity and accuracy of drilling anomalies and risk entities. The specific alignment method can include: using the drilling parameter change points as nodes, and the change trends, change amplitudes, and durations as features to align with the drilling log.

[0060] The aligned mud logging data can be used to verify the drilling risk entities and obtain the verification results. The verification results are used to indicate whether the drilling risk entities are correct, existent, incorrect, accurate, etc. For non-existent drilling risk entities, deletion processing can be performed. For incorrect or inaccurate drilling risk entities, the aligned mud logging data can be used for correction. For the verified drilling risk entities and the corrected drilling risk entities, the attribute data of the drilling risk entities can be extracted from the aligned mud logging data to enrich and supplement the drilling risk entities.

[0061] Step 13: Construct a drilling risk knowledge graph based on the knowledge graph ontology and drilling risk instances.

[0062] In some embodiments, a drilling risk knowledge graph can be constructed based on the knowledge graph ontology and drilling risk instances. In this way, the drilling risk knowledge extracted from drilling data according to natural language processing techniques, supplemented by the drilling risk instances extracted from on-site drilling logs, is used to construct a drilling risk knowledge graph containing a large amount of expert knowledge and actual cases. This effectively utilizes the drilling data and on-site drilling logs, which helps to enhance the understanding of the drilling process and the interpretability of drilling risk diagnosis.

[0063] The drilling risk knowledge in the knowledge graph ontology can be used as drilling risk entities. The drilling risk instances can include the drilling risk entities extracted from the drilling logs. Therefore, entity disambiguation and / or co-reference resolution can be performed based on the knowledge graph ontology and drilling risk instances. Specifically, the ambiguity of a drilling risk entity refers to the situation where one entity can correspond to multiple real-world entities. Entity disambiguation can refer to determining the real-world entity pointed to by an entity. Co-reference refers to different expressions of a real-world entity, which thus correspond to multiple drilling risk entities. For example, drilling risk entities such as "mud", "cement slurry", "drilling fluid", "return slurry", etc., all refer to the drilling fluid used in the drilling process. A pre-trained language model can be used to convert the co-reference resolution task into a word-filling task, thereby achieving the unification of co-reference resolution and real-world entities.

[0064] In some embodiments, nodes in the knowledge graph can be constructed based on drilling risk entities; the drilling risk knowledge represented by the drilling risk entities and the corresponding attribute data of the drilling risk entities can be used as the attribute data of the nodes; the entity category corresponding to the drilling risk entities can be used as the entity category of the nodes; and the edges between the nodes can be determined according to the relationship categories between the drilling risk entities, thereby constructing a knowledge graph. The constructed knowledge graph can include at least one node and the edges between the nodes. Each node represents a type of drilling risk entity. Each node can also have an entity category and attribute data. The knowledge graph can include multiple types of edges, and each type of edge corresponds to a relationship category.

[0065] In some embodiments, the drilling risk knowledge graph can be saved in a graph database, and the drilling risk knowledge graph can be displayed to show the nodes and edges in the drilling risk knowledge graph. Figure 7 Some risk nodes and characterization nodes in the drilling risk knowledge graph are shown. Figure 8 The edges between the risk nodes and symptom nodes in the drilling risk knowledge graph are shown.

[0066] Step 14: Construct a drilling risk question-answering system according to the drilling risk knowledge graph.

[0067] In some embodiments, a drilling risk question-answering system can be constructed according to the drilling risk knowledge graph. Thus, the knowledge graph reasoning can be carried out by using the drilling risk question-answering system. Knowledge Graph Reasoning (KGR) aims to utilize the existing knowledge in the knowledge graph to obtain new knowledge through reasoning. In practical applications, the drilling risk question-answering system can be used to determine the corresponding drilling risk diagnosis result according to the input drilling risk question. Among them, the drilling risk question can be a question input by a user. Or, the drilling risk question can also be generated by a computer device. The drilling risk question can include questions related to drilling risks for inquiry. For example, the drilling risk question can include: What are the solutions to the jar release? What are the characterizations that will occur in the case of lost circulation? etc. The drilling risk diagnosis result can include the characterizations of drilling risks, the causes of risk occurrence, preventive measures, solutions, etc.

[0068] In some embodiments, please refer to Figure 9 . The drilling risk question-answering system can include an input module, a classification module, a query module, and an output module. The input module can receive drilling risk questions. The classification module can determine the corresponding question category according to the drilling risk question. The query module is used to query the answers corresponding to the drilling risk question and the question category according to the drilling risk knowledge graph. The output module is used to determine the corresponding answer template according to the question category, and generate the drilling risk diagnosis result according to the answer and the answer template. In this way, the drilling risk question-answering system can realize real-time risk diagnosis, solution recommendation, risk knowledge query, etc. The real-time risk diagnosis can be carried out by inputting the changes of on-site real-time drilling parameters into the knowledge graph for query to diagnose possible drilling risks. After diagnosing the drilling risk, subsequent abnormal development process analysis, solution recommendation, risk information update, etc. can be carried out. The solution recommendation can be carried out by inputting the risk characterizations into the knowledge graph for query to recommend the corresponding solutions and further update the knowledge graph cases.

[0069] The drilling risk diagnosis method according to the embodiments of this specification can construct a knowledge graph ontology by sorting out literature and materials related to abnormal drilling risks; can fully utilize on-site drilling logs to extract abnormal risk instances; and can construct a drilling risk knowledge graph based on the knowledge graph ontology and drilling risk instances. In this way, professional knowledge related to drilling risks and a large number of abnormal drilling risk instances can be constructed into a drilling risk knowledge graph containing a large amount of expert knowledge and actual cases. A drilling risk Q&A system can be constructed based on the drilling risk knowledge graph. This can greatly improve the cognitive ability of drilling risks. It provides technical support and important reference for accurately and quickly identifying drilling risks at the operation sites of deep water, deep layer, unconventional, etc., where drilling accidents are extremely likely to occur.

[0070] The said drilling risk knowledge graph can be used in multiple aspects such as abnormal risk diagnosis, recommendation of solution measures, and query training of abnormal risk knowledge, solving problems such as poor timeliness of on-site expert risk judgment and poor interpretability of intelligent model judgment. By effectively using text data such as drilling logs, it helps to enhance the on-site understanding of the drilling process and the interpretability of drilling risk diagnosis, so as to more reasonably and clearly solve problems related to drilling risks and facilitate safe and efficient drilling. In addition, it can effectively solve problems such as complex characterization of on-site drilling abnormal risks, low accuracy of single-parameter judgment, and low efficiency of manual judgment, improving the accuracy, speed, and credibility of drilling risk identification. The drilling risk Q&A system formed based on the knowledge graph can effectively improve the perception and cognitive level of the current on-site drilling team regarding drilling risks. Its solution recommendation function helps the on-site to make quick decisions and take relevant solution measures, minimizing drilling risks and thus avoiding the occurrence of drilling complex accidents.

[0071] The embodiments of this specification also provide another drilling risk diagnosis method. The said drilling risk diagnosis method can be applied to computer devices such as servers. The said drilling risk diagnosis method can be implemented based on the drilling risk Q&A system constructed according to the drilling risk knowledge graph. For the drilling risk knowledge graph and the drilling risk Q&A system, reference can be made to the previous embodiments.

[0072] Please refer to Figure 10 and Figure 11 . The said drilling risk diagnosis method can include the following steps.

[0073] Step 21: Receive a drilling risk question.

[0074] In some embodiments, the said drilling risk question can be a question input by the user. Or, the said drilling risk question can also be generated by the computer device. The said drilling risk question can include questions related to drilling risks for inquiry. For example, the said drilling risk question can include: What are the solution measures for the shock absorber to release stuck? What are the manifestations of lost circulation?

[0075] In some embodiments, the terminal device may receive a drilling risk problem input by a user and send the drilling risk problem to the server. The server may receive the drilling risk problem. Alternatively, the server may also be connected to the on-site logging system. The server may receive logging data sent by the on-site logging system, perform real-time monitoring on the logging data, and generate a drilling risk problem when it detects an abnormality in the logging data. For example, the logging data may include various logging parameters. When the server detects an abnormality in one or more of the logging parameters in the logging data, it may convert the numerical change of the abnormal logging parameter into text to obtain a drilling risk problem.

[0076] Among them, the abnormality of the logging parameter may include: the value of the logging parameter is greater than a certain threshold, the value of the logging parameter is less than a certain threshold, the change range of the logging parameter is greater than a certain threshold, etc.

[0077] Step 22: Determine the corresponding problem category according to the drilling risk problem.

[0078] In some embodiments, a problem category set may be obtained. The problem category set may include at least one problem category, and each problem category may correspond to at least one keyword. The drilling risk problem may be segmented to obtain the segmentation of the drilling risk problem. The segmentation may be matched with the keywords corresponding to the problem category. The problem category to which the matched keyword belongs may be used as the problem category corresponding to the drilling risk problem. Specifically, the number of segmentations of the drilling risk problem may be one or more. The one or more segmentations may be respectively matched with the keywords of each problem category to obtain the successfully matched keywords. The number of successfully matched keywords for each problem category may be counted. Thus, the problem category corresponding to the largest number of keywords may be used as the problem category to which the drilling risk problem belongs. Among them, the successful match may include: the segmentation is the same as the keyword, or the segmentation has the same semantics as the keyword. Of course, a machine learning method may also be used to determine the problem category corresponding to the drilling risk problem. For example, the drilling risk problem may be input into a problem category prediction model to obtain the output of the problem category prediction model. The output represents the probability that the drilling risk problem belongs to each problem category. The problem category with the highest probability may be selected as the problem category to which the drilling risk problem belongs. Table 2 shows some problem categories.

[0079] Table 2

[0080]

[0081]

[0082] Step 23: Determine the corresponding answer template according to the problem category.

[0083] In some embodiments, an answer template set can be obtained. The answer template set can include at least one answer template. Each answer template can correspond to a question category. Then, according to the question category corresponding to the drilling risk question, the corresponding answer template can be selected from the answer template set. Table 3 shows the answer templates corresponding to some question categories.

[0084] Table 3

[0085]

[0086] Step 24: Query the answer corresponding to the drilling risk question and the question category according to the drilling risk knowledge graph.

[0087] In some embodiments, according to the drilling risk question, the corresponding first node can be matched in the drilling risk knowledge graph; the second node having a connection relationship with the first node can be obtained in the drilling risk knowledge graph; the answer can be determined according to the second node. The answer can include keywords. For example, the feature representation of the drilling risk question can be obtained, and the feature representation of the attribute data of each node in the drilling risk knowledge graph can be obtained; the distance between the feature representation of the drilling risk question and the attribute data of each node in the drilling risk knowledge graph can be calculated. The feature representation can include an embedding representation. The feature representation can include a feature vector. The smaller the distance, the more similar the drilling risk question is to the node attribute data. The node corresponding to the minimum distance can be selected as the first node. Of course, at least one distance less than or equal to a certain threshold can also be selected, and the nodes corresponding to the at least one distance can be used as the first nodes. For the first node, a second node having a connection relationship with the first node and the relationship category of the edge matching the question category can be obtained in the drilling risk knowledge graph. The relationship category matching the question category can include: the relationship category is the same as the question category. The attribute data of the second node can express the drilling risk knowledge. Thus, the attribute data of the second node can be obtained as the answer.

[0088] Step 25: Generate a drilling risk diagnosis result according to the answer and the answer template.

[0089] In some embodiments, the answer can be incorporated into the answer template to obtain a drilling risk diagnosis result. In this way, the drilling risk diagnosis result can include a risk knowledge answer text that is convenient for operators to understand, thus facilitating understanding.

[0090] In some scenario examples, through the drilling risk diagnosis method of the embodiments of this specification, three main functions such as real-time risk diagnosis, solution recommendation, and risk knowledge query training can be realized.

[0091] (1) Real-time risk diagnosis function.

[0092] When abnormal mud logging data is detected, drilling risk problems can be generated; according to the drilling risk knowledge graph, at least one candidate risk type corresponding to the drilling risk problem can be queried; according to the drilling risk knowledge graph, the characterization data corresponding to each candidate risk type can be queried respectively; the on-site characterization data can be matched with the characterization data of the at least one candidate risk type; if the match is successful, the candidate risk type corresponding to the successfully matched characterization data can be used as the target risk type; according to the drilling risk knowledge graph, the solution corresponding to the target risk type can be queried; the queried solution can be output. If the match fails, new drilling risk problems can be generated based on the on-site characterization data; according to the drilling risk knowledge graph, similar solution cases for the new drilling risk problems can be queried; the queried similar solution cases can be output.

[0093] The queried solution can be directly output. Or, when abnormal mud logging data is detected, drilling risk problems can be generated; according to the drilling risk problems, the corresponding problem categories can be determined; according to the problem categories, the corresponding answer templates can be determined; the queried solution can be incorporated into the answer template to obtain the drilling risk diagnosis result; the drilling risk diagnosis result can be output. In addition, the queried similar solution cases can be directly output. Or, according to the new drilling risk problems, the corresponding problem categories can be determined; according to the problem categories, the corresponding answer templates can be determined; the queried similar solution cases can be incorporated into the answer template to obtain the drilling risk diagnosis result; the drilling risk diagnosis result can be output.

[0094] The on-site characterization data can be the characterization data that appears at the drilling site. The on-site characterization data can include one or more sub-characterization data. The characterization data of the candidate risk type can include one or more sub-characterization data. Then, the successful match between the on-site characterization data and the characterization data of the candidate risk type can include: for each sub-characterization data in the on-site characterization data, there is a corresponding sub-characterization data in the characterization data of the candidate risk type. Among them, the existence of a corresponding relationship between the sub-characterization data can include: the sizes of the sub-characterization data are equal, or the difference between the sub-characterization data is less than or equal to the set threshold.

[0095] Of course, the successful match between the on-site characterization data and the characterization data of the candidate risk type can also include: the number of sub-characterization data in the on-site characterization data is less than or equal to the number of sub-characterization data in the characterization data of the candidate risk type.

[0096] For example, by connecting to the on-site mud logging system, the real-time changes in mud logging parameters can be converted into natural language to obtain drilling risk problems. At this time, the problem category corresponding to the drilling risk problem can be "risk query type problem". The representation corresponding to the drilling risk problem can be queried by querying the drilling risk knowledge graph. If it strictly conforms to the representation of a certain risk, "xx risk may occur currently" can be directly output. If the representation does not fully conform to the relevant risk representation in the knowledge graph, the case that best conforms to this representation is found, and "Based on the similarity between the representation at this time and that of the xx accident / risk that occurred in Well xx in the xx block previously, xx risk may occur. Please pay close attention" is output.

[0097] (2) The solution recommendation function is that the system automatically classifies it as a measure query type problem according to the input drilling risk, and gives the most similar successful solution case in the graph by analyzing the current state and on-site actual conditions.

[0098] (3) Risk knowledge query training refers to on-site operators and users querying the knowledge related to abnormal risks stored in the current graph, and learning the risk representation, risk occurrence reasons, preventive measures, etc.

[0099] The drilling risk diagnosis method of the embodiments of this specification can receive drilling risk problems; can determine the corresponding problem category according to the drilling risk problems; can determine the corresponding answer template according to the problem category; can query the answer corresponding to the drilling risk problem and the problem category according to the drilling risk knowledge graph; can generate the drilling risk diagnosis result according to the answer and the answer template. In this way, through the drilling risk knowledge graph, the cognitive ability of drilling risks can be greatly improved, the accuracy of risk diagnosis can be improved, and technical support and important reference are provided for accurately and quickly discriminating drilling risks at the current operation sites where drilling accidents are extremely likely to occur, such as deep water, deep layer, and unconventional operations.

[0100] The embodiments of this specification also provide another drilling risk diagnosis method. The drilling risk diagnosis method can be applied to computer devices such as servers. The drilling risk diagnosis method can be implemented based on a drilling risk question-answering system constructed according to the drilling risk knowledge graph. For the drilling risk knowledge graph and the drilling risk question-answering system, reference can be made to the previous embodiments.

[0101] Please refer to Figure 12 . The drilling risk diagnosis method can include the following steps.

[0102] Step 31: Generate a drilling risk problem when it is monitored that the mud logging data is abnormal.

[0103] In some embodiments, it can be connected to the on-site mud logging system; it can receive the mud logging data sent by the on-site mud logging system; it can monitor the mud logging data in real time; when it detects that the mud logging data is abnormal, it can generate a drilling risk problem. For example, the mud logging data can include various mud logging parameters. When it detects that one or more of the mud logging parameters in the mud logging data are abnormal, it can convert the numerical change of the abnormal mud logging parameter into text, so as to obtain a drilling risk problem.

[0104] Step 32: According to the drilling risk knowledge graph, query at least one candidate risk type corresponding to the drilling risk problem.

[0105] In some embodiments, according to the drilling risk problem, one or more first nodes can be matched in the drilling risk knowledge graph; one or more second nodes having a connection relationship with each first node can be obtained in the drilling risk knowledge graph; according to each second node, the corresponding candidate risk type can be determined.

[0106] Specifically, according to the drilling risk knowledge graph, at least one candidate risk type corresponding to the drilling risk problem can be directly queried. Alternatively, the corresponding problem category can also be determined according to the drilling risk problem. Then, according to the drilling risk knowledge graph, the candidate risk type corresponding to the drilling risk problem and the problem category can be queried. For example, the feature representation of the drilling risk problem can be obtained, and the feature representation of the attribute data of each node in the drilling risk knowledge graph can be obtained; the distance between the feature representation of the drilling risk problem and the attribute data of each node in the drilling risk knowledge graph can be calculated. The feature representation can include a feature vector. The smaller the distance, the more similar the drilling risk problem is to the node attribute data. The node corresponding to the minimum distance can be selected as the first node. Of course, at least one distance less than or equal to a certain threshold can also be selected, and the at least one node corresponding to the at least one distance can be used as the first node. For each first node, at least one second node having a connection relationship with the first node and the relationship category of the edge matching the problem category can be obtained in the drilling risk knowledge graph. The attribute data of the second node can be obtained as the candidate risk type.

[0107] Step 33: According to the drilling risk knowledge graph, query the characterization data corresponding to each candidate risk type respectively.

[0108] In some embodiments, corresponding third nodes may be matched in the drilling risk knowledge graph according to the candidate risk types; fourth nodes having a connection relationship with the third nodes may be obtained in the drilling risk knowledge graph; and corresponding characterization data may be determined according to the fourth nodes. Specifically, the characterization data corresponding to the candidate risk types may be directly queried according to the drilling risk knowledge graph. Alternatively, new drilling risk problems may be generated respectively according to each candidate risk type; new problem categories may be determined according to the new drilling risk problems. The characterization data corresponding to the candidate risk types and the new problem categories may be queried according to the drilling risk knowledge graph. The specific process may refer to the previous step S32 and will not be elaborated here.

[0109] Step 34: Match the on-site characterization data with the characterization data of the at least one candidate risk type.

[0110] In some embodiments, the on-site characterization data may be the characterization data that appears at the drilling site. The risk types queried through the knowledge graph may not actually match the drilling site. Therefore, it is necessary to match the on-site characterization data with the characterization data of the at least one candidate risk type to find out the risk types that actually match the drilling site.

[0111] Specifically, the on-site characterization data includes one or more sub-characterization data, and the characterization data of the candidate risk types includes one or more sub-characterization data. The matching of the on-site characterization data with the characterization data of the candidate risk types includes: for each sub-characterization data in the on-site characterization data, determining whether there is a corresponding sub-characterization data in the characterization data of the candidate risk types. Of course, the matching of the on-site characterization data with the characterization data of the candidate risk types may also include: determining whether the number of sub-characterization data in the on-site characterization data is less than or equal to the number of sub-characterization data in the characterization data of the candidate risk types.

[0112] Step 35: Use the candidate risk type corresponding to the successfully matched characterization data as the target risk type, and query the solution measures corresponding to the target risk type according to the drilling risk knowledge graph.

[0113] In some embodiments, if the matching is successful, it indicates that the target risk type actually matches the drilling site, so that the solution measures of the target risk type can be obtained to cope with the abnormal condition of the logging data.

[0114] The successful matching of the on-site characterization data and the candidate risk type characterization data may include: for each sub-characterization data in the on-site characterization data, there is a corresponding sub-characterization data in the candidate risk type characterization data. Of course, the successful matching of the on-site characterization data and the candidate risk type characterization data may also include: the number of sub-characterization data in the on-site characterization data is less than or equal to the number of sub-characterization data in the candidate risk type characterization data. Among them, the existence of a corresponding relationship between sub-characterization data may include: the sizes of the sub-characterization data are equal, or the difference between the sub-characterization data is less than or equal to a set threshold.

[0115] In some embodiments, if the matching is successful, the candidate risk type corresponding to the successfully matched characterization data can be used as the target risk type; the solution corresponding to the target risk type can be queried according to the drilling risk knowledge graph. Specifically, the solution corresponding to the target risk type can be directly queried according to the drilling risk knowledge graph. Or, a new drilling risk problem can also be generated according to the target risk type; a new problem category can be determined according to the new drilling risk problem. The solution corresponding to the target risk type and the new problem category can be queried according to the drilling risk knowledge graph. The specific process can refer to the previous step S32 and will not be elaborated here.

[0116] In some embodiments, the queried solution can be directly output. Or, when it is detected that the mud logging data is abnormal, a drilling risk problem can be generated; a corresponding problem category can be determined according to the drilling risk problem; a corresponding answer template can be determined according to the problem category; the queried solution can be incorporated into the answer template to obtain a drilling risk diagnosis result; the drilling risk diagnosis result can be output.

[0117] In some embodiments, if the matching fails, it indicates that the queried candidate risk type does not match the actual drilling site. For this reason, a new drilling risk problem can be generated according to the on-site characterization data; similar solution cases for the new drilling risk problem can be queried according to the drilling risk knowledge graph; the queried similar solution cases can be output. This can recommend similar solution cases, through the cases that have occurred before and their corresponding measures, so as to help on-site staff make decisions.

[0118] Specifically, similar solution cases for the new drilling risk problem can be directly queried according to the drilling risk knowledge graph. Or, a corresponding problem category can also be determined according to the new drilling risk problem; similar solution cases corresponding to the new drilling risk problem and the new problem category can be queried according to the drilling risk knowledge graph as similar solution cases.

[0119] The specific process can refer to the previous step S32 and will not be elaborated here.

[0120] In some embodiments, the retrieved similar solution cases can be directly output. Alternatively, according to the new drilling risk problem, the corresponding problem category can be determined; according to the problem category, the corresponding answer template can be determined; the retrieved similar solution cases can be incorporated into the answer template to obtain the drilling risk diagnosis result; and the drilling risk diagnosis result can be output.

[0121] The drilling risk diagnosis method according to the embodiments of this specification can receive the drilling risk problem; can determine the corresponding problem category according to the drilling risk problem; can determine the corresponding answer template according to the problem category; can query the answers corresponding to the drilling risk problem and the problem category according to the drilling risk knowledge graph; and can generate the drilling risk diagnosis result according to the answers and the answer template. In this way, through the drilling risk knowledge graph, the cognitive ability of drilling risks can be greatly improved, the accuracy of risk diagnosis can be enhanced, and technical support and important references can be provided for accurately and quickly identifying drilling risks at the current operation sites where drilling accidents are extremely likely to occur, such as deep water, deep layer, and unconventional operations.

[0122] The following introduces a scenario example. Taking a certain well in a certain block as an example, during the drilling process, the pump pressure increased significantly. The comprehensive logging system detected the significant increase in pump pressure in real time and transmitted it to the drilling risk diagnosis system.

[0123] ① The drilling risk diagnosis system needs to be started in advance;

[0124] ② The logging system senses the increase in pump pressure and transmits it to the drilling risk diagnosis system;

[0125] ③ The main program automatically converts the pump pressure parameter increase signal into natural language text for query;

[0126] ④ Subsequently, the possible risk types for the increase in pump pressure are retrieved: differential pressure, sand bridge, and stuck pipe due to collapse;

[0127] ⑤ Traversal consultation is started in the risk type set. Among them, the characteristics of differential pressure sticking are most similar to the on-site characteristics, so it is determined as differential pressure sticking;

[0128] ⑥ After determining it as differential pressure sticking, query the solutions and measures for differential pressure sticking;

[0129] ⑦ The well site starts to adjust to the coming risk and eliminates the risk.

[0130] Please refer to Figure 13 This specification embodiment provides a drilling risk diagnosis device, including the following units.

[0131] A construction unit 41, configured to construct a knowledge graph ontology according to drilling data;

[0132] The first construction unit 42 is used to extract drilling risk instances according to the drilling log;

[0133] The second construction unit 43 is used to construct a drilling risk knowledge graph according to the knowledge graph ontology and drilling risk instances;

[0134] The third construction unit 44 is used to construct a drilling risk Q&A system according to the drilling risk knowledge graph; wherein, the drilling risk Q&A system is used to determine the corresponding drilling risk diagnosis result according to the input drilling risk question.

[0135] Please refer to Figure 14 This embodiment of the specification also provides another drilling risk diagnosis device, including the following units.

[0136] The receiving unit 51 is used to receive drilling risk questions;

[0137] The first determination unit 52 is used to determine the corresponding question category according to the drilling risk question;

[0138] The second determination unit 53 is used to determine the corresponding answer template according to the question category;

[0139] The query unit 54 is used to query the answers corresponding to the drilling risk question and the question category according to the drilling risk knowledge graph;

[0140] The generation unit 55 is used to generate a drilling risk diagnosis result according to the answer and the answer template.

[0141] Please refer to Figure 15 This embodiment of the specification also provides another drilling risk diagnosis device, including the following units.

[0142] The first generation unit 61 is used to generate drilling risk questions when it is monitored that the logging data is abnormal;

[0143] The first query unit 62 is used to query at least one candidate risk type corresponding to the drilling risk question according to the drilling risk knowledge graph;

[0144] The second query unit 63 is used to query the characterization data corresponding to each candidate risk type according to the drilling risk knowledge graph respectively;

[0145] The matching unit 64 is used to match the on-site characterization data with the characterization data of the at least one candidate risk type;

[0146] The third query unit 65 is used to use the candidate risk type corresponding to the successfully matched characterization data as the target risk type, and query the solution measures corresponding to the target risk type according to the drilling risk knowledge graph.

[0147] An embodiment of this specification also provides a computer device, including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the computer program, the above-mentioned drilling risk diagnosis method is implemented.

[0148] An embodiment of this specification also provides a computer-readable storage medium. The computer-readable storage medium stores a computer program. When the computer program is executed by a processor, the above-mentioned drilling risk diagnosis method is implemented.

[0149] An embodiment of this specification also provides a computer program product. The computer program product includes a computer program. When the computer program is executed by a processor, the above-mentioned drilling risk diagnosis method is implemented.

[0150] Those skilled in the art can understand that this specification can be provided as a method, a system, or a computer program product. Therefore, this specification can take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, this specification can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0151] This specification is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products of the embodiments of this specification. It should be understood that each flow and / or block in the flowcharts and / or block diagrams, as well as the combination of flows and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. The computer can be a personal computer, a laptop computer, a cellular phone, a camera phone, a smart phone, a personal digital assistant, a media player, a navigation device, an email device, a game console, a tablet computer, a wearable device, or any combination of these devices.

[0152] Each functional unit in the embodiments of this specification can be integrated into a processing unit, or each functional unit can exist physically alone, or two or more functional units can be integrated into a processing unit.

[0153] Those skilled in the art can understand that the descriptions of the embodiments in this specification each have their own emphases. For the parts not detailed in a certain embodiment, reference can be made to the relevant descriptions of other embodiments. Additionally, it can be understood that after reading this specification document, those skilled in the art can, without creative work, think of combining some or all of the embodiments listed in this specification arbitrarily, and these combinations are also within the scope of disclosure and protection of this specification.

[0154] Although this specification is described through examples, those of ordinary skill in the art know that the above examples are only used to help understand the core idea of this specification. Those skilled in the art can understand that this specification has many variations and changes. It is hoped that the appended claims will cover these variations and changes without departing from the spirit of this specification.

Claims

1. A drilling risk diagnosis method based on a knowledge graph, characterized in that Including: Construct a knowledge graph ontology based on drilling data; Extract drilling risk instances according to the drilling log; Construct a drilling risk knowledge graph based on the knowledge graph ontology and drilling risk instances; Construct a drilling risk Q&A system based on the drilling risk knowledge graph; wherein, the drilling risk Q&A system is used to determine the corresponding drilling risk diagnosis result according to the input drilling risk question; The extraction of drilling risk instances includes: performing named entity recognition on the drilling log to obtain drilling risk entities and their corresponding entity categories; inputting multiple drilling risk entities and the association data between the multiple drilling risk entities into a language model to obtain the relationship categories between the multiple drilling risk entities; The method further includes: obtaining logging data matching the drilling log; aligning the logging data with the drilling log in terms of time; verifying the drilling risk entities by using the aligned logging data; for the drilling risk entities that pass the verification, extracting the attribute data of the drilling risk entities from the aligned logging data; the construction of the drilling risk knowledge graph includes: integrating the drilling risk entities and their corresponding entity categories, the relationship categories between the drilling risk entities, and the attribute data of the drilling risk entities into the knowledge graph ontology to obtain the drilling risk knowledge graph.

2. The method according to claim 1, wherein The construction of the knowledge graph ontology includes: Extracting drilling risk knowledge from drilling data through natural language processing methods; Organizing the extracted drilling risk knowledge into a drilling risk knowledge system to obtain the knowledge graph ontology.

3. A drilling risk diagnosis method based on a knowledge graph, characterized in that, The drilling risk knowledge graph is as described in any one of claims 1-2, and the method includes: Receiving a drilling risk question; Determining the corresponding question category according to the drilling risk question; Determining the corresponding answer template according to the question category; Querying the answer corresponding to the drilling risk question and the question category according to the drilling risk knowledge graph; Generating a drilling risk diagnosis result according to the answer and the answer template.

4. The method according to claim 3, wherein The receiving of the drilling risk question includes: Receiving the drilling risk question input by the user; or, Performing real-time monitoring on the logging data, and generating a drilling risk question when it is monitored that the logging data is abnormal.

5. The method according to claim 3, wherein The determination of the corresponding question category includes: Obtaining a question category set, which includes question categories, and each question category corresponds to keywords; Performing word segmentation on the drilling risk question to obtain the word segmentation of the drilling risk question; Matching the word segmentation with the keywords corresponding to the question categories; Taking the question category corresponding to the matched keyword as the question category corresponding to the drilling risk question; The determination of the corresponding answer template includes: Obtaining an answer template set, which includes answer templates, and each answer template corresponds to a question category; Selecting the corresponding answer template from the answer template set according to the question category corresponding to the drilling risk question.

6. A drilling risk diagnosis method based on a knowledge graph, characterized in that, The drilling risk knowledge graph is as described in any one of claims 1-2, and the method includes: Generating a drilling risk question when it is monitored that the logging data is abnormal; Querying at least one candidate risk type corresponding to the drilling risk question according to the drilling risk knowledge graph; According to the drilling risk knowledge graph, query the characterization data corresponding to each candidate risk type respectively; Match the on-site characterization data with the characterization data of the at least one candidate risk type; Take the candidate risk type corresponding to the successfully matched characterization data as the target risk type, and according to the drilling risk knowledge graph, query the corresponding solution measures for the target risk type.

7. A computer device, characterized in that, Comprising: A memory for storing computer instructions; A processor for executing the computer instructions to implement the method according to any one of claims 1-6.

Images