Task adaptive processing-oriented oil and gas intelligent application platform construction method and device

Abstract

The invention provides a task adaptive processing-oriented oil and gas intelligent application platform construction method and device, belongs to the field of oil and gas industry intelligent technologies, and is used for solving the problems of poor task processing flexibility, inaccurate matching and incomplete optimization in related technologies. According to the method, intelligent disassembly, intelligent operator matching and intelligent workflow construction and selection can be carried out on tasks by utilizing an intelligent agent, an operator and a workflow, so that intelligent processing of the tasks is realized, and efficient and intelligent processing of the tasks in the oil and gas industry is facilitated; the problems of poor task processing flexibility, inaccurate matching and incomplete optimization in the prior art are solved.

Description

Technical Field

[0001] This application relates to the field of intelligent technologies in the oil and gas industry, and particularly to a method and device for constructing an oil and gas intelligent application platform for task adaptive processing. Background Art

[0002] In the actual operation of the oil and gas industry, task processing faces many challenges. On the one hand, oil and gas tasks are complex and diverse, involving knowledge in multiple fields such as geology, engineering, and management. During the task execution process, various constraints such as underlying characteristics, equipment performance, cost budget, safety, and environmental protection need to be comprehensively considered. On the other hand, traditional task processing methods mainly rely on manual experience and simple rule systems, and have the following problems: Lack of flexibility in task decomposition: Existing methods mostly adopt fixed task decomposition modes, which are difficult to flexibly adjust according to the dynamic changes and complex constraints of actual tasks, and cannot fully adapt to the complex and changeable operation environment of the oil and gas industry.

[0003] Insufficient accuracy in operator matching: The operator matching process often only considers a single factor, such as function matching, while ignoring important dimensions such as historical execution effects and resource consumption, resulting in low efficiency or inability to meet task requirements when the matched operator is actually executed.

[0004] Incomplete optimization of the workflow: Workflow generation usually only focuses on one aspect of task execution efficiency or cost, lacks comprehensive optimization of multiple objectives, and is difficult to achieve the overall optimum of task processing.

[0005] With the in-depth application of artificial intelligence technology in the oil and gas industry, there is an urgent need for a method and system that can achieve task adaptive intelligent processing to improve the efficiency of oil and gas operations, reduce costs, and ensure operation safety. Summary of the Invention

[0006] This application provides a method and device for constructing an oil and gas intelligent application platform for task adaptive processing, which realizes efficient and intelligent processing of oil and gas industry tasks through intelligent decomposition of tasks, accurate operator matching, and multi-objective optimized workflow generation, and solves the problems of poor flexibility, inaccurate matching, and incomplete optimization in task processing in related technologies.

[0007] In the first aspect, this application provides a method for constructing an oil and gas intelligent application platform for task adaptive processing. The method includes: Based on a pre-constructed intelligent agent and pre-acquired task-related information, decompose the task into multiple subtasks, and determine the task constraint information and the dependency relationships between the subtasks; Determine the dimensional evaluation values of multiple dimensions of each operator relative to each subtask and the comprehensive historical execution effect score based on a pre-built operator library and historical task processing records, and screen the optional matching operators for each subtask; Based on the task constraint information and the dependency relationship between subtasks, traverse and form a feasible workflow structure according to the optional matching operators of all subtasks; For each feasible workflow structure, determine the preference degree of the feasible workflow structure based on the subtask adaptation degree between the subtasks and the optional matching operators in the feasible workflow structure, the operator coupling degree of each group of adjacent coupled optional matching operators, and the pre-obtained execution efficiency data and execution risk data of the feasible workflow structure. Call the optional workflow structure with the highest preference degree as the selected workflow structure to generate task plan information, and the task plan information is used to guide the implementation of the task.

[0008] By adopting the above technical solution, it is possible to use agents, operators, and workflows to perform intelligent decomposition, intelligent operator matching, and intelligent workflow construction and selection of tasks, thereby realizing intelligent processing of tasks, which is beneficial to realizing efficient and intelligent processing of tasks in the oil and gas industry, and solving the problems of poor flexibility, inaccurate matching, and incomplete optimization in task processing in the prior art.

[0009] Furthermore, the decomposing the task into multiple subtasks based on the pre-built agent and the pre-obtained task-related information, and determining the task constraint information and the dependency relationship between subtasks includes: The agent includes an oil and gas knowledge graph. The agent decomposes the task into subtasks based on the pre-built oil and gas knowledge graph using the recursive analysis method according to the task-related information, determines the set of constraint conditions as the task constraint information, determines the relationship between subtasks as the dependency relationship, and the subtasks cannot be further divided.

[0010] Furthermore, the determining the dimensional evaluation values of multiple dimensions of each operator relative to each subtask and the comprehensive historical execution effect score based on the pre-built operator library and historical task processing records, and screening the optional matching operators for each subtask includes: Determine the description information of the operator and the subtask based on the pre-built oil and gas knowledge graph model, and calculate the semantic similarity between the description information of the operator and the subtask as the function matching degree of the operator to the subtask; Calculate the dimensional evaluation values of multiple dimensions of the operator to the subtask based on the pre-obtained historical task processing records, and determine the historical execution effect score by synthesizing all dimensional evaluation values; Screen the optional matching operators for the subtask based on the dimensional evaluation value and the historical execution effect score. The historical execution effect score of the optional matching operator is higher than the effect score threshold, and the dimensional evaluation value of the specified dimension is higher than the preset dimensional evaluation threshold of the corresponding dimension; The dimension evaluation value includes one or more of an execution accuracy evaluation value, an execution efficiency evaluation value, and a resource consumption evaluation value. The historical task processing record includes a historically selected workflow structure and a workflow operation record of the historically selected workflow structure. The workflow operation record of the historically selected workflow structure includes the operation time consumed by each historical operator therein. The method for determining the execution accuracy evaluation value includes: Let the execution accuracy evaluation value for the subtask facing be . There are historical subtasks in the historical task processing record that have the same description information as the subtask. Among these historical subtasks, there are historical operators whose description information is the same as that of the operator . Then . The method for determining the execution efficiency evaluation value includes: Let the execution efficiency evaluation value for the subtask facing be . Determine the historical operators in the historical task processing record whose semantic similarity to the description information of the operator is higher than the first preset similarity threshold as the similar operators of the operator . And determine the average operation time consumed by the similar operators of the operator among the historical operators to be . Determine the historical operators in the historical task processing information whose description information is the same as that of the operator as the homogeneous operators of the operator . And determine the average time consumed by the homogeneous operators of the operator among the historical operators to be . Then . The method for determining the resource consumption evaluation value includes: Let the resource consumption evaluation value for the subtask facing be . The resource occupancy value of the operator is . The preset resource limit is . Then .

[0011] Furthermore, determining the preference degree of the feasible workflow structure based on the subtask adaptation degree between the subtasks and the optional matching operators in the feasible workflow structure, the operator coupling degree of each group of adjacent coupled optional matching operators, and the pre-obtained execution efficiency data and execution risk data of the feasible workflow structure includes: Determine the sub - task fitness between sub - tasks and optional matching operators in the feasible workflow structure, the operator coupling degree of each group of adjacent coupled optional matching operators, and the execution efficiency data and execution risk data of the pre - obtained feasible workflow structure; Determine the preference degree based on the sub - task fitness between all sub - tasks and optional matching operators in the feasible workflow structure, the operator coupling degree of all groups of adjacent coupled optional matching operators, and the execution efficiency data and execution risk data of the pre - obtained feasible workflow structure. The preference degree is positively correlated with the sub - task fitness, the operator coupling degree, and the execution efficiency data, and negatively correlated with the execution risk data.

[0012] Furthermore, the method for determining the sub - task fitness between a sub - task and an optional matching operator includes: ; In the formula, is the sub - task fitness of operator facing sub - task , is the function matching degree between operator and sub - task , is the historical execution effect score of operator facing sub - task , is the resource fitness, which is equal to the pre - obtained operator resource occupancy value divided by the pre - obtained total available resource amount, are all pre - designed calculation weights greater than zero; Let the total sub - task fitness of the feasible workflow structure be , then , m is the number of sub - tasks, and the preference degree of the feasible workflow structure is positively correlated with the total sub - task fitness.

[0013] Furthermore, the method for determining the operator coupling degree of adjacent coupled optional matching operators includes: Let the previous operator in a group of adjacent coupled operators be , and the subsequent operator be . The output of the previous operator is used as the input of the subsequent operator . Pre - obtain the data compatibility , process connection degree and collaborative gain effect value between adjacent coupled operators. The comprehensive matching degree between adjacent coupled operators is , then ; In the formula, are all pre - designed calculation weights greater than zero; Among them, the data compatibility degree is obtained based on a pre-configured operator compatibility degree comparison table; The process connection degree is determined based on the dependency relationship between subtasks. When the subtask corresponding to the subsequent operator depends on the previous operator, the process connection degree is 1; otherwise, the process connection degree is 0. The collaborative gain effect value is determined based on pre-labeled historical task processing records. The historical task processing records are pre-labeled with the execution effect value and running time of each historical operator. The average value of the execution effect values when the previous operator is executed alone is , and the average value of the running times is . The average value of the execution effect values when the subsequent operator is executed alone is , and the average value of the running times is . The average values of the execution effects when the previous operator input operator and the subsequent operator are coupled and executed are respectively , and the average values of the running times are respectively . Then , where are all pre-designed calculation weights greater than zero; The sum of the coupling degrees between operators in a feasible workflow structure is . Then , where is the set of edges in the feasible workflow structure.

[0014] Furthermore, the method for obtaining the execution efficiency data includes: The historical task processing records are pre-labeled with the task plan content and historical implementation plan similarity of the historically selected workflow structure, and the historical construction period of the historical task implementation plan. Screen the historical task processing records for the historically selected workflow structure identical to the feasible workflow structure and the historical construction period of the historical task implementation plan. Based on the plan similarity, the weighted average value of the historical construction periods of the historical task implementation plans corresponding to all identical selected workflow structures is determined as the expected construction period, and the result of dividing the preset benchmark construction period by the expected construction period is used as the execution efficiency data.

[0015] Furthermore, the method for obtaining the execution risk data includes: Determine the task risk value of each subtask based on the oil and gas knowledge graph; In the historical task processing record, the similarity between the task plan content pre-labeled with the historical selected workflow structure and the historical implementation plan, and the historical risk frequency of each subtask in the historical task implementation plan. The weighted sum of the historical risk frequencies of the corresponding subtasks in all historical task implementation plans based on the plan similarity for each subtask is the total risk frequency of the subtask; For each operator, let the task risk value of the subtask corresponding to the operator in the feasible workflow structure be and the total risk frequency be . The operator risk score is . Then , where e is the natural constant, is the pre-designed calculation weight and ; The execution risk data of the feasible workflow structure is . Then .

[0016] Further, the preference degree is determined based on the subtask adaptation degree between all subtasks in the comprehensive feasible workflow structure and the subtasks of the optional matching operators, the operator coupling degree of all groups of adjacent coupled optional matching operators, and the pre-obtained execution efficiency data and execution risk data of the feasible workflow structure. The preference degree is positively correlated with the subtask adaptation degree, operator coupling degree, and execution efficiency data, including: Let the total subtask adaptation degree of the feasible workflow structure be , the total operator coupling degree be , the execution efficiency data be , the execution risk data be , and the preference degree be . Then ; where are all calculation coefficients greater than zero, is the pre-obtained calculation influence coefficient, and e is the natural constant; The method for obtaining the calculation influence coefficient includes: Define multiple adjacent time windows with the same width in the pre-labeled historical task processing record. Determine the historical task processing records with the same historical selected workflow structure and the target feasible workflow structure in each time window as reference records. Determine the average execution efficiency and the average execution risk of the reference records in each time window. x represents the sequence number of the time window. Calculate the ratio of the average execution efficiency and the average execution risk of each time window as the reference value of the reference record. Using the sequence number as the abscissa and the reference value as the ordinate, fit a straight line to minimize the sum of the squares of the perpendicular distances from all coordinate points to the straight line. Let the slope of the straight line be , then , where is the pre-designed calculation weight and .

[0017] In a second aspect, the present application provides an apparatus for constructing an intelligent application platform for task-oriented adaptive processing. The apparatus applies any one of the methods described in the first aspect above.

[0018] In summary, the present application at least includes the following beneficial effects: A method and an apparatus for constructing an intelligent oil and gas application platform for task-oriented adaptive processing are provided, which can realize intelligent decomposition of oil and gas industry tasks, accurate operator matching, and generation of a workflow for multi-objective optimization, and achieve efficient and intelligent processing of oil and gas industry tasks; The self-developed operator matching and multi-objective optimization workflow determination and evaluation selection model makes the operator matching more accurate and the workflow selection more accurate.

[0019] It should be understood that the content described in the invention content part is not intended to limit the key or important features of the embodiments of the present application, nor is it used to limit the scope of the present application. Other features of the present application will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] In combination with the accompanying drawings and with reference to the following detailed description, the above and other features, advantages and aspects of the embodiments of the present application will become more obvious. In the drawings, the same or similar reference numerals represent the same or similar elements, where: Figure 1 The flowchart of a method for constructing an intelligent oil and gas application platform for task-oriented adaptive processing in an embodiment of the present application is shown. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0021] To make the objectives, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are some, but not all, of the embodiments of the present application. All other embodiments obtained by those of ordinary skill in the art without creative efforts based on the embodiments of the present application belong to the scope of protection of the present application.

[0022] In addition, the term "and / or" in this article is only a description of the association relationship of associated objects, indicating that three relationships may exist. For example, A and / or B may represent: A exists alone, A and B exist simultaneously, and B exists alone. In addition, the character " / " in this article generally represents an "or" relationship between the associated objects before and after.

[0023] The present application provides a method and apparatus for constructing an oil and gas intelligent application platform for task-adaptive processing, which can achieve intelligent decomposition of oil and gas industry tasks, accurate operator matching, and generation of workflows with multi-objective optimization, so as to realize efficient and intelligent processing of oil and gas industry tasks.

[0024] In a first aspect, an embodiment of the present application discloses a method for constructing an oil and gas intelligent application platform for task-adaptive processing. This method can be executed by a server.

[0025] Figure 1 The flowchart of a method for constructing an oil and gas intelligent application platform for task-adaptive processing in an embodiment of the present application is shown.

[0026] Referring to Figure 1 , the method specifically includes the following steps: S110: Based on a pre-constructed intelligent agent and pre-acquired task-related information, decompose the task into multiple subtasks, and determine the task constraint information and the dependency relationship between the subtasks.

[0027] The method of this step specifically includes: The intelligent agent includes an oil and gas knowledge graph. The intelligent agent uses the pre-constructed oil and gas knowledge graph and the recursive analysis method to decompose the task into subtasks according to the task-related information, and determines the set of constraint conditions as the task constraint information, and determines the relationship between the subtasks as the dependency relationship. The subtasks cannot be further divided.

[0028] In an example, the task-related information is a natural language description, which can be an input text document, a collected conversation, or an obtained text command, etc., and specifically includes some descriptions of the task. The oil and gas knowledge graph is pre-constructed, which includes various general tasks in various scenarios, various conventional decomposition methods of various general tasks (generally, only one decomposition method is included for each typical task, but multiple methods can also be included), the description information of the general tasks and each level of subtasks, the tree-like relationship between the general tasks and each level of subtasks (including the smallest subtasks, that is, atomic tasks), the general constraint conditions of each level of subtasks, the dependency relationship between each level of subtasks (such as a certain subtask needs to be completed before a certain subtask, or a certain subtask needs to be completed within a specified time after another subtask is completed, etc.), the description information of the subtasks, etc. The intelligent agent decomposes the task based on the interpretation of the task-related information, the invocation of the oil and gas knowledge graph, and the recursive analysis method model, and obtains a series of atomic tasks that cannot be further divided based on the knowledge graph. Subsequent subtasks are all in this concept. The intelligent agent contains a large model. The intelligent agent uses the large model, the oil and gas knowledge graph, and the recursive analysis method to decompose and extract the task-related information described in natural language into the general application methods of the large model and the proprietary knowledge graph. The specific working principle here is not elaborated, and it is only necessary to be able to clarify the task constraint information within the task and the dependency relationship between the subtasks and the subtasks based on the task-related information.

[0029] The tasks after decomposition can be modeled and defined here. A directed acyclic graph definition of the task is established based on subtasks, task constraint information, and the dependencies between subtasks. The task is defined as a triple , where represents the nodes corresponding to all subtasks, m is the number of subtasks, is the dependency relationship between subtasks, means needs to be completed before , represents all the constraint conditions of the task constraint information, represents the task dependency matrix, if and only if .

[0030] S120: Determine the dimension evaluation values of multiple dimensions of each operator relative to each subtask and the comprehensive historical execution effect score based on a pre-built operator library and historical task processing records, and screen the optional matching operators for each subtask.

[0031] The method of this step specifically includes: determining the description information of the operator and the subtask based on a pre-built oil and gas knowledge graph model, and calculating the semantic similarity between the description information of the operator and the subtask as the function matching degree of the operator to the subtask; calculating the dimension evaluation values of multiple dimensions of the operator for the subtask based on pre-obtained historical task processing records, and determining the historical execution effect score by synthesizing all dimension evaluation values; screening the optional matching operators for the subtask based on the dimension evaluation values and the historical execution effect score, where the historical execution effect score of the optional matching operator is higher than the effect score threshold and the dimension evaluation value of the specified dimension is higher than the preset dimension evaluation threshold of the corresponding dimension.

[0032] Here, the description information of the operator includes the function, processing means, input requirements, output format, etc. of the operator. It is a natural language description, which may be represented as a description or a series of labels. The description information of the subtask is similar, and also includes the general task objective, general work content, general input data requirements, general output data requirements, etc. of the subtask, and can also be represented as a description or a series of labels. The two have corresponding forms. When the semantic distance of the description information is represented as "a description" for both, the overall semantic similarity is calculated as the semantic distance between the two; if the two are represented as "a series of labels", then all the labels of the two are combined in a given order, and the semantic similarity between the combined results of the labels is calculated.

[0033] In the method of this step, the dimension evaluation value includes one or more of an execution accuracy evaluation value, an execution efficiency evaluation value, and a resource consumption evaluation value. The historical task processing record includes the historical selected workflow structure and the workflow operation record of the historical selected workflow structure. The workflow operation record of the historical selected workflow structure includes the operation time consumption of each historical operator therein. The method for determining the execution accuracy evaluation value includes: Let the execution accuracy evaluation value for the subtask be . There are historical subtasks in the historical task processing record that have the same description information as the subtask. Among these historical subtasks, there are historical operators corresponding to the historical subtasks in the historical task processing record that have the same description information as the operator . Then ; The method for determining the execution efficiency evaluation value includes: Let the execution efficiency evaluation value for the subtask be . Determine the historical operators in the historical task processing record whose semantic similarity to the description information of the operator is higher than the first preset similarity threshold as the similar operators of the operator . And determine the average operation time consumption of the similar operators of the operator in the historical operators as . Determine the historical operators in the historical task processing information that have the same description information as the operator as the homogeneous operators of the operator . And determine the average time consumption of the homogeneous operators of the operator in the historical operators as . Then ; The method for determining the resource consumption evaluation value includes: Let the resource consumption evaluation value for the subtask be . The resource occupancy value of the operator is . The preset resource limit is . Then ; .

[0034] In an example, the historical execution effect score , where are all preset calculation weights greater than zero, and specifically can be sequentially taken as .

[0035] Theoretically speaking, the historical task processing record contains the selected workflow structure obtained from the historical task-related information according to this method, as well as the task plan content and task implementation plan determined based on the selected workflow structure (some content of the task plan information of the selected workflow structure may be adjusted during the implementation process). The task implementation plan is the plan executed during the actual task implementation process, and its implementation process is recorded and marked in the historical task processing record. For example, the adjusted part of the plan during the implementation process, the duration of each subtask during the implementation process, risk events, etc. Of course, the process of calling the selected workflow structure to generate the task plan content will also be recorded, such as the running time of each operator. Based on this, the dimension evaluation value of the established evaluation dimension can be determined, and the dimension evaluation value of each dimension can be analyzed.

[0036] The evaluation dimension can also only consider one or two of the foregoing, or consider other evaluation dimensions, which will not be listed one by one here. Other mathematical models for determining the historical execution effect score based on the dimension evaluation value of the evaluation dimension can also be selected, as long as they conform to the general evaluation logic of the evaluation dimension.

[0037] In one example, using the filtering condition in the operator library to screen the optional matching operators for each subtask. Here is the experience threshold, which can be determined and optimized based on experience data.

[0038] The filtering condition can also be determined as other, as long as it can screen out the operators that are more suitable for the subtask in terms of the specified evaluation dimension and the historical execution effect score direction according to the general logic of the evaluation dimension.

[0039] S130: Based on the task constraint information and the dependency relationship between subtasks, traverse all the optional matching operators of all subtasks to form a feasible workflow structure.

[0040] Since the dependency relationship between subtasks is determined, a directed acyclic graph of subtasks can be drawn under the restriction of the dependency relationship. Each node of the directed acyclic graph represents a subtask, and each node can be filled with an optional matching operator for the subtask. In this way, all possible directed acyclic graphs and all possibilities of traversing each node to select each optional matching operator can be obtained, and all optional workflow structures can be obtained.

[0041] Of course, other methods can also be used, such as Petri nets and path search.

[0042] S140: For each feasible workflow structure, determine the preference degree of the feasible workflow structure based on the subtask adaptation degree between the subtasks in the feasible workflow structure and the optional matching operators, the operator coupling degree of each group of adjacent coupled optional matching operators, and the pre-obtained execution efficiency data and execution risk data of the feasible workflow structure. Call the optional workflow structure with the highest preference degree as the selected workflow structure to generate task plan information, and the task plan information is used to guide the implementation of the task.

[0043] The method of this step specifically includes: determining the subtask adaptation degree between the subtasks in the feasible workflow structure and the optional matching operators, the operator coupling degree of each group of adjacent coupled optional matching operators, and the pre-obtained execution efficiency data and execution risk data of the feasible workflow structure; comprehensively determining the preference degree based on the subtask adaptation degree between all subtasks and optional matching operators in the feasible workflow structure, the operator coupling degree of all groups of adjacent coupled optional matching operators, and the pre-obtained execution efficiency data and execution risk data of the feasible workflow structure. The preference degree is positively correlated with the subtask adaptation degree, the operator coupling degree, and the execution efficiency data, and negatively correlated with the execution risk data.

[0044] In the method of this step, the method for determining the subtask adaptation degree between the subtasks and the optional matching operators includes: ; In the formula, is the subtask adaptation degree of the operator facing the subtask , is the function matching degree between the operator and the subtask , is the historical execution effect score of the operator facing the subtask , is the resource adaptation degree, which is equal to the pre-obtained operator resource occupancy value divided by the pre-obtained total available resource amount, are all pre-designed calculation weights greater than zero; Let the total subtask adaptation degree of the feasible workflow structure be , then , m is the number of subtasks, and the preference degree of the feasible workflow structure is positively correlated with the total subtask adaptation degree. Of course, the subtask matching degree of a single operator can also consider one or two of the above factors, or expand to consider other factors, as long as it can conform to the operator and subtask adaptation evaluation logic, The higher it is, the better the adaptation between the operator and the subtask.

[0045] The method for determining the operator coupling degree of adjacent coupled optional matching operators includes: Let the operator in front in a group of adjacent coupled operators be and the operator behind be . The output of the operator in front is used as the input of the operator behind . It is intended to obtain the data compatibility degree , process connection degree and collaborative gain effect value defined between adjacent coupled operators. The comprehensive matching degree between adjacent coupled operators is . Then ; In the formula, are all pre-designed calculation weights greater than zero; Among them, the data compatibility degree is obtained based on a pre-configured operator compatibility comparison table; The process connection degree is determined based on the dependency relationship between subtasks. When the subtask corresponding to the operator behind depends on the operator in front, the process connection degree is 1; otherwise, the process connection degree is 0; The collaborative gain effect value is determined based on pre-labeled historical task processing records. The historical task processing records are pre-labeled with the execution effect value and running time consumption of each historical operator. The average value of the execution effect values when the operator in front is executed alone is , and the average value of the running time consumption is . The average value of the execution effect values when the operator behind is executed alone is , and the average value of the running time consumption is . The average values of the execution effects when the input operator of the operator in front and the operator behind are coupled and executed are respectively, and the average values of the running time consumption are respectively. Then . In the formula, are all pre-designed calculation weights greater than zero; The sum of the coupling degrees between operators in a feasible workflow structure is . Then . In the formula, is the set of edges in the feasible workflow structure.

[0046] The method for obtaining the execution efficiency data includes: The historical task processing records are pre-labeled with the task plan content and historical implementation plan similarity of the historically selected workflow structure (the optional workflow structure that is called), and the historical construction period of the historical task implementation plan. Screen the historical task processing records for those that are similar to the feasible workflow structure For the same historical selection workflow structure and the historical duration of the historical task implementation plan, based on the plan similarity, determine the weighted average of the historical durations of the historical task implementation plans corresponding to all the same selected workflow structures as the expected duration, and calculate the result of dividing the preset benchmark duration by the expected duration as the execution efficiency data.

[0047] Similarly, only one or two of the aforementioned data compatibility, process connection degree, and collaborative gain effect value can be considered, or other indicators characterizing the coupling degree of operators can be considered for extension, without listing them one by one.

[0048] The method for obtaining the execution risk data includes: determining the task risk value of each subtask based on the oil and gas knowledge graph, and the task risk value of the subtask is included in the description information of the subtask in the oil and gas knowledge graph; The plan similarity between the task plan content with the historical selected workflow structure pre-marked in the historical task processing record and the historical implementation plan, and the historical risk frequency of each subtask occurring in the historical task implementation plan. For each subtask, calculate the weighted sum of the historical risk frequencies of the corresponding subtasks occurring in all historical task implementation plans based on the plan similarity as the total risk frequency of the subtask; For each operator, let the task risk value of the subtask corresponding to the operator in the feasible workflow structure be and the total risk frequency be , and the operator risk score be , then , where e is the natural constant, is the pre-designed calculation weight and ; The execution risk data of the feasible workflow structure is , then .

[0049] Based on the foregoing content, the total subtask adaptation degree of the feasible workflow structure is , the total sum of the coupling degrees between operators is , the execution efficiency data is , the execution risk data is , and the preference is , then ; In the formula, are all calculation coefficients greater than zero, is a pre-acquired computational influence coefficient, and e is the natural constant. Of course, the preference is determined by considering four factors: the sub-task adaptation degree of all sub-tasks and optional matching operators within the feasible workflow structure, the operator coupling degree of all groups of adjacent coupled optional matching operators, and the execution efficiency data and execution risk data of the pre-acquired feasible workflow structure. Similarly, it can also be determined by only using three, two, or one of these factors, or by expanding to consider other factors, which will not be listed one by one here.

[0050] Here, the method for obtaining the computational influence coefficient includes: Define multiple adjacent time windows with the same width in the pre-labeled historical task processing records. Determine the historical task processing records with the same historical selected workflow structure as the target feasible workflow structure in each time window as reference records, and determine the average execution efficiency and the average execution risk in each time window. Let x represent the sequence number of the time window. Calculate the ratio of the average execution efficiency and the average execution risk in each time window as the reference value of the reference record. With the sequence number as the abscissa and the reference value as the ordinate, fit a straight line to minimize the sum of the squares of the perpendicular distances from all coordinate points to the line. Let the slope of the line be , then , where is the pre-designed computational weight and .

[0051] In summary, this method can intelligently disassemble tasks, intelligently and accurately match operators for each sub-task, construct an optional workflow structure based on the operators, then intelligently determine the priority of the workflow structure, and call the selected workflow structure to generate task plan information for implementation. In this way, an intelligent solution for oil and gas tasks can be determined, which is beneficial to the efficient and intelligent processing of oil and gas tasks.

[0052] It should be noted that for the foregoing method embodiments, for the sake of simple description, they are all expressed as a series of action combinations. However, those skilled in the art should know that this application is not limited by the described action sequence, because according to the embodiments of this application, certain steps can be performed in other sequences or simultaneously. Secondly, those skilled in the art should also know that the embodiments described in the specification are all optional embodiments, and the actions and modules involved are not necessarily essential to this application.

[0053] In a second aspect, an apparatus for constructing an intelligent application platform for task adaptive processing according to an embodiment of the present application is disclosed. The apparatus can be implemented as a server or included in a server. The apparatus is used to execute the method disclosed in the first aspect above.

[0054] Those skilled in the art can clearly understand that for the convenience and conciseness of description, the specific working process of the described device can refer to the corresponding process in the foregoing method embodiment, which will not be elaborated herein.

[0055] In summary, the present application at least includes the following beneficial effects: A method and device for constructing an intelligent oil and gas application platform for task-oriented adaptive processing are provided, which can realize intelligent decomposition of oil and gas industry tasks, accurate operator matching, and generation of workflows with multi-objective optimization, and achieve efficient and intelligent processing of oil and gas industry tasks; 2. The self-developed operator matching and multi-objective optimization workflow determination and evaluation selection model makes the operator matching more accurate and the workflow selection more accurate.

[0056] The above description is only a preferred embodiment of the present application and an explanation of the applied technical principles. Those skilled in the art should understand that the scope of disclosure involved in the present application is not limited to the technical solutions formed by the specific combination of the above technical features, and should also cover other technical solutions formed by any combination of the above technical features or their equivalent features without departing from the foregoing disclosure concept. For example, the technical solutions formed by mutually replacing the above features with the (but not limited to) technical features with similar functions disclosed in the present application.

Claims

1. A method for constructing an oil and gas intelligent application platform for task adaptive processing, characterized in that: include: Based on the pre-built agent and pre-acquired task-related information, the task is broken down into multiple subtasks, and the task constraint information and the dependencies between the subtasks are determined; Based on the pre-built operator library and historical task processing records, determine the dimensional evaluation value of each operator relative to each subtask in multiple dimensions and the comprehensive historical execution effect score, and screen the optional matching operators for each subtask; Based on the task constraint information and the dependencies between subtasks, a feasible workflow structure is formed by traversing the optional matching operators of all subtasks; For each feasible workflow structure, the preference of the feasible workflow structure is determined based on the subtask fitness between the subtask and the optional matching operator in the feasible workflow structure, the operator coupling degree of each group of adjacent coupled optional matching operators, and the pre-acquired execution efficiency data and execution risk data of the feasible workflow structure, and the optional workflow structure with the highest preference is called as the selected workflow structure to generate task plan information, and the task plan information is used to guide the implementation of the task.

2. The method according to claim 1, characterized in that The process of decomposing a task into multiple subtasks based on a pre-built agent and pre-acquired task-related information, and determining task constraint information and dependencies between subtasks includes: The intelligent agent includes an oil and gas knowledge graph. Based on the pre-built oil and gas knowledge graph, the intelligent agent uses a recursive analysis method to decompose the task into subtasks according to task-related information, and determines a set of constraint conditions as the task constraint information, determines the relationship between the subtasks as a dependency relationship, and the subtasks cannot be divided any further.

3. The method according to claim 1, characterized in that The method of determining the dimensional evaluation values ​​of each operator relative to multiple dimensions of each subtask and the comprehensive historical execution effect score based on the pre-built operator library and historical task processing records, and screening the optional matching operators for each subtask includes: Determine the description information of operators and subtasks based on the pre-built oil and gas knowledge graph model, and calculate the semantic similarity between the description information of operators and subtasks as the functional matching degree of operators to subtasks; Calculate the dimensional evaluation values ​​of the operator for multiple dimensions of the subtask based on the pre-acquired historical task processing records, and determine the historical execution effect score by combining all dimensional evaluation values; Filtering the optional matching operators of the subtask based on the dimension evaluation value and the historical execution effect score, wherein the historical execution effect score of the optional matching operator is higher than the effect score threshold and the dimension evaluation value of the specified dimension is higher than the dimension evaluation threshold preset for the corresponding dimension; The dimension evaluation value includes one or more of an execution accuracy evaluation value, an execution efficiency evaluation value, and a resource consumption evaluation value, the historical task processing record includes a historically selected workflow structure and a workflow running record of a historically selected workflow structure, and the workflow running record of a historically selected workflow structure includes the running time of each historical operator therein; The method for determining the execution accuracy evaluation value comprises: setting a characterization operator Subtask oriented The execution accuracy evaluation value is , the historical task processing records include A historical subtask with the same description information as the subtask. The historical operators corresponding to the historical subtasks in the historical task processing records are AND Operator The description information is the same, then ; The method for determining the execution efficiency evaluation value comprises: setting a characterization operator Subtask oriented The execution efficiency evaluation value is , determine the historical task processing records and the operator The historical operator whose semantic similarity of the description information is higher than the first preset similarity threshold is the operator and determine the operators in the historical operators The average running time of similar operators is , determine the historical task processing information and the operator The historical operators with the same description information are operators Homogeneous operators, and determine the operators in the historical operators The average time consumption of homogeneous operators is ,but ; The method for determining the resource consumption evaluation value comprises: setting a characterization operator Subtask oriented The resource consumption evaluation value is , operator The resource usage is , the default resource limit is ,but .

4. The method according to claim 1, characterized in that The determining of the preferred degree of the feasible workflow structure based on the subtask adaptation degree of the subtask and the optional matching operator in the feasible workflow structure and the operator coupling degree of each group of adjacent coupled optional matching operators and the pre-acquired execution efficiency data and execution risk data of the feasible workflow structure comprises: Determine the subtask fitness of the subtask and the optional matching operator in the feasible workflow structure and the operator coupling degree of each group of adjacent coupled optional matching operators, as well as the pre-acquired execution efficiency data and execution risk data of the feasible workflow structure; The preference is determined by comprehensively considering the subtask fitness of all subtasks in the feasible workflow structure and the optional matching operators, the operator coupling of all groups of adjacent coupled optional matching operators, and the pre-acquired execution efficiency data and execution risk data of the feasible workflow structure. The preference is positively correlated with the subtask fitness, operator coupling and execution efficiency data, and negatively correlated with the execution risk data.

5. The method according to claim 4, characterized in that Methods for determining the compatibility of a subtask with a subtask of an optional matching operator include: ； In the formula, For operator Subtask oriented The subtask fitness of For operator With subtasks Functional matching degree, For operator Subtask oriented The historical execution performance score of is the resource adaptation degree, which is equal to the pre-acquired operator resource occupancy value divided by the pre-acquired total available resources. All are preset calculation weights greater than zero; Assume that the total subtask fitness of the feasible workflow structure is ,but , m is the number of subtasks, and the preference of the feasible workflow structure is positively correlated with the total subtask fitness.

6. The method according to claim 4, characterized in that The method for determining the operator coupling degree of the adjacent coupled optional matching operator includes: Suppose the preceding operator in a set of adjacent coupled operators is , and the subsequent operator is , in front of the operator The output of Input, pre-acquire the data compatibility defined between adjacent coupling operators , Process connectivity and synergy gain value , the comprehensive matching degree between adjacent coupling operators is ,but ； In the formula, All are preset calculation weights greater than zero; Wherein, the data compatibility is obtained based on a preconfigured operator compatibility comparison table; The process connection degree is determined based on the dependency relationship between subtasks. When the subtask corresponding to the subsequent operator depends on the previous operator, the process connection degree is 1, otherwise the process connection degree is 0; The synergy gain effect value is determined based on the pre-marked historical task processing records. The historical task processing records are pre-marked with the execution effect value and running time of each historical operator. The average value of the execution effect when executed separately is , average running time , statistics after operator The average value of the execution effect when executed separately is The average running time is , statistics first operator Input Operator and Post Operator The average values ​​of the execution effects during coupled execution are The average running time is ,but , where All are preset calculation weights greater than zero; The sum of the coupling degrees between operators of a feasible workflow structure is ,but , where is the set of edges in the feasible workflow structure.

7. The method according to claim 4, characterized in that The method for acquiring the execution efficiency data comprises: The historical task processing records are pre-marked with the similarity between the content of the task plan with the historical selected workflow structure and the plan of the historical implementation plan, and the historical duration of the historical task implementation plan. The historical duration of the same historical selected workflow structure and historical task implementation plan is determined based on the similarity of the plans. The weighted average of the historical duration of all historical task implementation plans corresponding to the same selected workflow structure is determined as the expected duration, and the result of dividing the preset benchmark duration by the expected duration is calculated as the execution efficiency data.

8. The method according to claim 4, characterized in that The method for acquiring the execution risk data includes: Determine the task risk value of each subtask based on the oil and gas knowledge graph; The similarity between the task solution content and the historical implementation plan of the task solution pre-marked with the historical selected workflow structure in the historical task processing record, and the historical risk frequency of each subtask in the historical task implementation plan, and the weighted sum of the historical risk frequencies of the corresponding subtasks in all historical task implementation plans calculated based on the solution similarity for each subtask is the total risk frequency of the subtask; For each operator, let the task risk value of the subtask corresponding to the operator in the feasible workflow structure be The total frequency of risk is , the operator risk score is ,but , where e is a natural constant, Calculate the weights for the preset and ; The execution risk data of the feasible workflow structure is ,but .

9. The method according to claim 4, characterized in that The optimization degree is determined by combining the subtask fitness of all subtasks and the optional matching operators in the feasible workflow structure, the operator coupling degree of all groups of adjacent coupled optional matching operators, and the execution efficiency data and execution risk data of the feasible workflow structure acquired in advance. The optimization degree is positively correlated with the subtask fitness, operator coupling and execution efficiency data, including: Assume that the total subtask fitness of the feasible workflow structure is The sum of the coupling degree between operators is , the execution efficiency data is , execution risk data is , the preference is ,but ； In the formula, are all calculation coefficients greater than zero. is the pre-acquired calculated influence coefficient, e is a natural constant; Calculate the influence coefficient The methods for obtaining include: Define multiple adjacent time windows of the same width in the pre-marked historical task processing records, determine the historical task processing records in each time window with the same historical selected workflow structure as the target feasible workflow structure as the reference record, and determine the average execution efficiency of the reference records in each time window and average execution risk , x represents the sequence number of the time window. The ratio of the average execution efficiency and the average execution risk of each time window is calculated as the reference value of the reference record. The sequence number is used as the horizontal coordinate and the reference value is used as the vertical coordinate. The straight line is fitted to minimize the sum of the squares of the vertical distances from all coordinate points to the straight line. The slope of the straight line is set to ,but , where is the preset calculation weight and .

10. A device for constructing a task-adaptive processing-oriented intelligent application platform, characterized in that: The device applies the method according to any one of claims 1-9.

Images