Oil and gas field conventional workover operation working condition classification method

Abstract

The invention relates to the field of petroleum engineering, and provides an oil and gas field conventional workover operation working condition classification method, which uses a data processing technology based on deep learning to perform decomposition and semantic coding on text description of oil and gas field conventional workover operation. Secondly, extracting a process semantic feature corresponding to the ith process from the encoded set of process semantic features as a to-be-identified process semantic feature, and dividing the whole set into a set of process semantic features of a preamble and a postamble by taking the to-be-identified process semantic feature as a boundary; and performing context coding on the semantic features of the to-be-recognized process, the set of the semantic features of the preamble process and the set of the semantic features of the subsequent process, so as to intelligently judge the operation category of the to-be-recognized process according to positive and negative attention response representation between the context semantic features of the preamble and the subsequent process of the to-be-recognized process. Therefore, the efficiency and accuracy of process classification can be effectively improved.

Description

Technical Field

[0001] The present application relates to the field of petroleum engineering, and more specifically, to a method for classifying working conditions of conventional well repair operations in oil and gas fields. Background Art

[0002] Well repair operations in oil and gas fields are a key link in ensuring the daily production, management and optimization of oil, gas and water wells. These operations are not only crucial to maintaining and restoring production capacity, but also can effectively deal with oil and gas layer and wellbore structure problems, eliminate faults, and thus ensure stable and increased production of oil wells. With the continuous deepening of exploration and development, the technical complexity and diversity of well repair operations are also increasing, forming a series of process technologies for different well types and well conditions.

[0003] However, despite the wide variety of well repair operations and diverse processes, it is difficult to achieve efficient and targeted management in actual operations. Traditional oil and gas field well repair operation management mainly relies on manual experience to judge the working conditions, which is particularly insufficient when faced with the operational details and complex working conditions of different processes. In addition, manual identification and classification are not only inefficient, but also prone to misjudgment, resulting in waste of resources and increased potential risks.

[0004] Therefore, an optimized classification scheme for the working conditions of conventional well workover operations in oil and gas fields is needed. Summary of the invention

[0005] In view of the shortcomings of the prior art, the present application provides a method for classifying the working conditions of conventional well repair operations in oil and gas fields.

[0006] According to one aspect of the present application, a method for classifying the working conditions of conventional well workover operations in oil and gas fields is provided, which comprises: obtaining a text description of conventional well workover operations in oil and gas fields; decomposing and semantically encoding the text description of conventional well workover operations in oil and gas fields to obtain a set of semantic encoding vectors of conventional well workover operations in oil and gas fields; extracting the semantic encoding vector of the conventional well workover operations in oil and gas fields corresponding to the ith process from the set of semantic encoding vectors of the conventional well workover operations in oil and gas fields as the semantic encoding vector of the process to be identified; dividing the set of semantic encoding vectors of the conventional well workover operations in oil and gas fields into a set of semantic encoding vectors of the preceding conventional well workover operations in oil and gas fields and a set of semantic encoding vectors of the subsequent conventional well workover operations in oil and gas fields based on the semantic encoding vector of the process to be identified; respectively encoding the semantic encoding vectors of the preceding conventional well workover operations in oil and gas fields and the semantic encoding vectors of the subsequent conventional well workover operations in oil and gas fields; The set of vectors and the set of semantic coding vectors of conventional well repairing operations in oil and gas fields in the subsequent context are context-coded to obtain the pre-context semantic coding vector of the process to be identified and the post-context semantic coding vector of the process to be identified; the pre-context semantic coding vector of the process to be identified and the post-context semantic coding vector of the process to be identified are subjected to positive and negative attention field interactive response to obtain a bidirectional context semantic coding vector of the process to be identified, wherein the positive and negative attention field interactive response is achieved by constructing a positive and negative pre-context-post-context semantic bidirectional attention balance field of the process to be identified, and semantic mapping modulation and response coding are performed based on the constructed positive and negative pre-context-post-context semantic bidirectional attention balance field of the process to be identified; based on the bidirectional context semantic coding vector of the process to be identified, the recognition result of the job category label used to represent the i-th process is obtained.

[0007] The present application has significant technical effects due to the adoption of the above technical solutions: the present application provides a method for classifying the working conditions of conventional well repair operations in oil and gas fields, which uses data processing technology based on deep learning to decompose and semantically encode the text description of conventional well repair operations in oil and gas fields, and then extracts the process semantic features corresponding to the ith process from the set of encoded process semantic features as the process semantic features to be identified, and uses them as a dividing line to divide the entire set into a set of preceding and succeeding process semantic features, and then, the process semantic features to be identified are contextually encoded with the set of preceding process semantic features and the set of succeeding process semantic features, respectively, so as to intelligently judge the work category of the process to be identified based on the positive and negative attention response representations between the preceding and succeeding context semantic features of the process to be identified. In this way, the efficiency and accuracy of process classification can be effectively improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0008] By describing the embodiments of the present application in more detail in conjunction with the accompanying drawings, the above and other purposes, features and advantages of the present application will become more apparent. The accompanying drawings are used to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the accompanying drawings, the same reference numerals generally represent the same components or steps.

[0009] Figure 1 The present invention is a flowchart of a method for classifying working conditions of conventional well repair operations in oil and gas fields according to an embodiment of the present application.

[0010] Figure 2 This is a flowchart of S120 in the method for classifying working conditions of conventional well repair operations in oil and gas fields according to an embodiment of the present application.

[0011] Figure 3 This is a flowchart of S150 in the method for classifying working conditions of conventional well repair operations in oil and gas fields according to an embodiment of the present application.

[0012] Figure 4 This is a flowchart of S160 in the method for classifying working conditions of conventional well repair operations in oil and gas fields according to an embodiment of the present application. DETAILED DESCRIPTION

[0013] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described here.

[0014] Well repair operations are a vital part of oil and gas field production management, and are directly related to the daily operation and maintenance of oil, gas and water wells and the improvement of production capacity. This type of operation can not only effectively solve the problems of wellbore structure and oil and gas layers, but also eliminate faults and restore or improve production performance, thereby providing guarantees for the stable production and increased production of oil wells. With the deepening of oil and gas field exploration and development, the technical means of well repair operations have become increasingly rich, and their complexity and diversity have also increased significantly, and a series of process flows for different well conditions and well types have gradually been developed.

[0015] However, although well repair operations cover a variety of processes and operating procedures, their management efficiency and pertinence still face many challenges. In actual operations, the traditional well repair operation management model mainly relies on manual experience to judge the working status. This method often seems to be inadequate when faced with complex operating environments and diverse process details. Manual identification and classification are not only inefficient, but also prone to misjudgment, resulting in waste of resources, while increasing operational risks and hidden dangers.

[0016] In view of the above technical problems, the technical concept of the present application is to obtain the text description of conventional well repair operations in oil and gas fields, and use natural language processing and encoding technology based on deep learning to decompose and semantically encode the text description of conventional well repair operations in oil and gas fields. Then, the semantic features of the conventional well repair operations in oil and gas fields corresponding to the i-th process are extracted from the set of semantic features of the encoded conventional well repair operations in oil and gas fields as the semantic features of the process to be identified, and the entire set is divided into the preceding and following conventional well repair operations in oil and gas fields. The semantic features and sets of semantic features are then used as the dividing line. The semantic features of the process to be identified are respectively context-encoded with the set of semantic features of the preceding conventional well repair operations in oil and gas fields and the set of semantic features of the following conventional well repair operations in oil and gas fields, so as to intelligently judge the operation category of the process to be identified based on the positive and negative attention response representations between the preceding context semantic features of the process to be identified and the following context semantic features of the process to be identified. The present application can automatically parse the text description of conventional well repair operations in oil and gas fields and semantically encode them. This not only improves the speed of process classification, but also ensures the consistency and accuracy of the classification results, and avoids the subjective bias caused by manual judgment. Moreover, each process can be analyzed independently and in detail, which helps to achieve more refined management and optimization, especially when facing complex and changing working conditions.

[0017] Figure 1 Flow chart of the method for classifying the working conditions of conventional well repair operations in oil and gas fields according to an embodiment of the present application. Figure 1As shown, according to the working status classification method of conventional well repair operations in oil and gas fields according to the embodiment of the present application, the method includes: S110, obtaining a text description of conventional well repair operations in oil and gas fields; S120, decomposing and semantically encoding the text description of conventional well repair operations in oil and gas fields to obtain a set of semantic coding vectors of conventional well repair operations in oil and gas fields; S130, extracting the semantic coding vector of the conventional well repair operations in oil and gas fields corresponding to the ith process from the set of semantic coding vectors of the conventional well repair operations in oil and gas fields as the semantic coding vector of the process to be identified; S140, using the semantic coding vector of the process to be identified as a boundary, dividing the set of semantic coding vectors of the conventional well repair operations in oil and gas fields into the set of semantic coding vectors of the conventional well repair operations in oil and gas fields as the set of semantic coding vectors of the conventional well repair operations in oil and gas fields in the preceding text and a set of semantic coding vectors of conventional well servicing operations in oil and gas fields in the following text; S150, respectively performing context coding on the set of semantic coding vectors of conventional well servicing operations in oil and gas fields in the preceding text and the set of semantic coding vectors of conventional well servicing operations in oil and gas fields in the following text to obtain a preceding context semantic coding vector of the process to be identified and a succeeding context semantic coding vector of the process to be identified; S160, performing positive and negative attention field interactive responses on the preceding context semantic coding vector of the process to be identified and the succeeding context semantic coding vector of the process to be identified to obtain a bidirectional context semantic coding vector of the process to be identified; S170, based on the bidirectional context semantic coding vector of the process to be identified, obtaining a recognition result of the job category label for representing the i-th process.

[0018] In step S110, a text description of conventional well repair operations in oil and gas fields is obtained. It should be understood that the text description of conventional well repair operations in oil and gas fields mainly includes operation process information, operation equipment and tool information, etc. Specifically, the operation process information includes the specific content of each process, such as well pressure and well washing in flushing operations; casing milling and grinding in rotation operations; the specifications and operation methods of pipe string lifting and lowering in pipe string lifting and lowering operations, etc., and also includes the operation details of different process construction stages. The operation equipment and tool information involves the mud pump (used for flushing operations, circulation, and squeezing fluid into the well), rotary table, top drive, power drill, screw drill (used for rotation operations, rotating pipe strings and tools in the well), well rig, workover machine (used for lifting and lowering pipe strings and lifting and lowering special pipe strings) and other equipment used in the well repair operation, as well as various construction tools. The use of these equipment and tools is one of the important bases for judging the operation type. In general, the text description of conventional well repair operations in oil and gas fields is the original data source for identifying process operation category labels. By analyzing this data, we can understand the specific operation content and working conditions of each process, and thus establish a corresponding relationship with the preset operation category (such as flushing operation, rotation operation, etc.), thereby realizing the final identification of the operation category label.

[0019] In step S120, the text description of the conventional well repair operation in the oil and gas field is decomposed and semantically encoded to obtain a set of semantic encoding vectors of the conventional well repair operation process in the oil and gas field. Specifically, Figure 2 Flow chart of S120 in the method for classifying the working conditions of conventional well repair operations in oil and gas fields according to an embodiment of the present application. Figure 2 As shown, the step S120 includes: S121, performing process decomposition based on the construction stage on the text description of the conventional well workover operation in the oil and gas field to obtain a set of text descriptions of the conventional well workover operation process in the oil and gas field; S122, performing semantic coding on each text description of the conventional well workover operation process in the set of text descriptions of the conventional well workover operation process in the oil and gas field to obtain a set of semantic coding vectors of the conventional well workover operation process in the oil and gas field.

[0020] In step S121, the text description of the conventional well repair operation in the oil and gas field is decomposed based on the process of the construction stage to obtain a set of text descriptions of the process of the conventional well repair operation in the oil and gas field. Accordingly, considering that the well repair operation in the oil and gas field is a complex and highly specialized process, it involves multiple different types of operation links. For example, from pulling out the downhole pipe string, checking and repairing equipment, to re-entering the pipe string and debugging, etc., each large construction stage contains many specific processes. Therefore, in order to better sort out the operation process and clarify each specific work link, the present application decomposes the text description of the conventional well repair operation in the oil and gas field based on the process of the construction stage to decompose the overall text description of the well repair operation into the text description of the specific process, and obtains a set of text descriptions of the process of the conventional well repair operation in the oil and gas field. In this way, the detailed information of each operation link can be understood more carefully. For example, in the text description of the process of "oil pipe replacement", the specific operation steps such as oil pipe disassembly, new oil pipe inspection and installation can be clarified, so that the function and category of the process in the entire well repair operation can be accurately judged.

[0021] In step S122, each text description of the conventional well workover operation procedure in the oil and gas field in the set of text descriptions of the conventional well workover operation procedure in the oil and gas field is semantically encoded to obtain a set of semantic encoding vectors of the conventional well workover operation procedure in the oil and gas field. Specifically, in an embodiment of the present application, the step S122 includes: using a semantic encoder including an embedding layer to semantically encode each text description of the conventional well workover operation procedure in the oil and gas field in the set of text descriptions of the conventional well workover operation procedure in the oil and gas field to obtain a set of semantic encoding vectors of the conventional well workover operation procedure in the oil and gas field. It should be understood that in order to further capture and mine the specific semantics contained in each text description of the conventional well workover operation procedure in the oil and gas field, in the technical solution of the present application, each text description of the conventional well workover operation procedure in the oil and gas field in the set of text descriptions of the conventional well workover operation procedure in the oil and gas field is semantically encoded to accurately capture the deep meaning in the text and obtain a set of semantic encoding vectors of the conventional well workover operation procedure in the oil and gas field. In particular, in one embodiment of the present application, a semantic encoder including an embedding layer is used to semantically encode each text description of the conventional well repair operation process in the oil and gas field in the set of text descriptions of the conventional well repair operation process in the oil and gas field to map each word to a low-dimensional continuous vector space, and convert the high-dimensional sparse text data into a low-dimensional dense vector representation, thereby obtaining a set of semantic encoding vectors of the conventional well repair operation process in the oil and gas field. That is, through the combination of the embedding layer and the semantic encoder, the text description of the conventional well repair operation process in the oil and gas field can be converted into a vector containing rich semantic information. These vectors not only contain the semantics of a single word, but also capture the semantic relationships such as the order and combination between words in the text through the processing of the entire text sequence by the semantic encoder, which can provide high-quality feature representation for subsequent process analysis and classification.

[0022] The following is a detailed description of a specific implementation process of "using a semantic encoder containing an embedding layer to semantically encode each text description of the conventional well workover operation process in the set of text descriptions of the conventional well workover operation process in the oil and gas field to obtain a set of semantic encoding vectors of the conventional well workover operation process in the oil and gas field": First, it is necessary to perform text preprocessing on the set of text descriptions of the conventional well workover operation process in the oil and gas field, which directly affects the performance of the final model. In this process, each text description of the conventional well workover operation process in the oil and gas field needs to be segmented into separate words or phrases. This step is usually called word segmentation. By removing stop words, punctuation marks, and performing stem extraction or word form restoration and other technical means, the original text can be effectively purified to make it easier for computers to understand and process. Subsequently, a vocabulary is generated based on all text descriptions, and an index value is assigned to each unique word. This step not only simplifies the subsequent data processing process, but also lays the foundation for building an embedding layer.

[0023] Next comes the application phase of the embedding layer, which is one of the core parts of the entire implementation process. At this stage, word embedding techniques such as Word2Vec, GloVe or FastText will be used to convert each word into a low-dimensional dense vector representation. This vector can not only capture the semantic information of the word itself, but also reflect the relationship between words, which can provide strong support for subsequent semantic encoding. For the text description of each process, the corresponding word vectors are connected in series according to their original order to form a vector sequence representing the entire process. This process is essentially to convert human-readable text information into a form that can be understood and processed by computers.

[0024] The next step is to build a suitable semantic encoder model architecture. In this scenario, deep learning models such as recurrent neural networks (RNNs) and their variants (such as LSTMs and GRUs) can be selected as the basic architecture of the semantic encoder. These models have their own characteristics. For example, RNNs can effectively process sequence data, especially when long-distance dependencies are involved. LSTMs and GRUs perform well due to their special gating mechanisms. After the model is selected, it is trained using annotated datasets. During the training process, the cross entropy loss function or other suitable loss functions are used to optimize the model parameters. The purpose is to let the model learn how to convert the input word vector sequence into a vector representation that can better express the semantic features of the process. This stage may require a lot of computing resources and time, which directly determines the quality of the final model.

[0025] Finally, the set of semantic encoding vectors of conventional well repair operation procedures in oil and gas fields is output. The carefully trained model can receive a set of processed process text descriptions, convert them into a sequence of word vectors through the embedding layer, and then process them through the core architecture of the semantic encoder, and finally output a set of semantic encoding vectors of conventional well repair operation procedures in oil and gas fields. Each vector in the set is not only a highly condensed representation of the process itself, but also contains rich semantic information that can be used for the analysis of subsequent steps. Taking "oil pipe replacement" as an example, suppose there is a process text description: "Pull out the old oil pipe, check the new oil pipe, and install the new oil pipe". After the above steps, this description will be converted into a series of word vectors, and further generated through the semantic encoder to generate a vector that can represent the semantics of the entire process operation.

[0026] It is worth noting that throughout the implementation process, issues of data quality and model optimization need to be considered. High-quality datasets are the key to success, so it is important to ensure the integrity and accuracy of the prepared data. At the same time, as model training progresses, it is also necessary to continuously adjust and optimize model parameters. This includes but is not limited to adjusting hyperparameters such as learning rate, batch size, number of layers, and number of units per layer to find the configuration that best suits the task at hand. In addition, you can also explore the use of pre-trained models for transfer learning, which can alleviate the problem of insufficient data to a certain extent and improve the generalization ability of the model.

[0027] In step S130, the semantic coding vector of the conventional well repair operation process in oil and gas fields corresponding to the ith process is extracted from the set of semantic coding vectors of the conventional well repair operation process in oil and gas fields as the semantic coding vector of the process to be identified. Accordingly, considering that the conventional well repair operation in oil and gas fields includes multiple processes, each process has its unique operation content and characteristics. Therefore, in order to extract a single process vector from the overall process semantic coding vector set, so as to focus on the specific process for in-depth analysis, in the technical solution of the present application, the semantic coding vector of the conventional well repair operation process in oil and gas fields corresponding to the ith process is extracted from the set of semantic coding vectors of the conventional well repair operation process in oil and gas fields as the semantic coding vector of the process to be identified. In this way, it can ensure a meticulous analysis of the process, which is helpful for the subsequent accurate identification of the specific category label of each process.

[0028] In step S140, the set of semantic coding vectors of conventional well repair operation procedures in oil and gas fields is divided into a set of semantic coding vectors of conventional well repair operation procedures in oil and gas fields in the preceding text and a set of semantic coding vectors of conventional well repair operation procedures in oil and gas fields in the following text, taking the semantic coding vector of the procedure to be identified as a boundary. It should be understood that in natural language understanding and actual operation process understanding, relevant information before and after an event or operation will naturally be considered to fully understand its meaning. Similarly, in the analysis of well repair operation procedures in oil and gas fields, the meaning and characteristics of a procedure are often affected by the procedures before and after it. For example, before the "oil pipe replacement" procedure, there may be a "oil pipe disassembly" procedure, and there may be a "new oil pipe installation" procedure later. Therefore, in order to be able to understand the procedure in context, the present application divides the set of semantic coding vectors of conventional well repair operation procedures in oil and gas fields into a set of semantic coding vectors of conventional well repair operation procedures in oil and gas fields in the preceding text and a set of semantic coding vectors of conventional well repair operation procedures in oil and gas fields in the following text by taking the semantic coding vector of the procedure to be identified as a boundary. In this way, by constructing the semantic encoding vector sets of the preceding and following contexts of the conventional well repairing operation procedures in oil and gas fields, the contextual semantic information of the procedure to be identified can be obtained from two directions. Specifically, the preceding context set can reflect the preconditions and preparation operations that lead to the execution of the procedure, while the following context set can reflect the subsequent operations and impacts caused by the completion of the procedure. For example, by analyzing the preceding context set, it can be seen that the "tubing lifting" process was carried out before the "tubing cleaning" process, which provided conditions for tubing cleaning. By analyzing the following context set, it can be seen that the "tubing inspection" process was carried out after cleaning to check the cleaning effect.

[0029] In step S150, context coding is performed on the set of semantic coding vectors of the preceding conventional well repairing operation procedures in oil and gas fields and the set of semantic coding vectors of the following conventional well repairing operation procedures in oil and gas fields to obtain the preceding context semantic coding vector of the procedure to be identified and the following context semantic coding vector of the procedure to be identified. Specifically, Figure 3 Flow chart of S150 in the method for classifying the working conditions of conventional well repair operations in oil and gas fields according to an embodiment of the present application. Figure 3 As shown, the step S150 includes: S151, inputting the set of the semantic coding vector of the process to be identified and the semantic coding vector of the previous conventional well repair operation process in the oil and gas field into a unidirectional context encoder based on a forward LSTM model to obtain the previous context semantic coding vector of the process to be identified; S152, inputting the set of the semantic coding vector of the process to be identified and the semantic coding vector of the subsequent conventional well repair operation process in the oil and gas field into a unidirectional context encoder based on a backward LSTM model to obtain the subsequent context semantic coding vector of the process to be identified.

[0030] In step S151, the set of the semantic coding vector of the process to be identified and the semantic coding vector of the conventional well repair operation process in the previous text of the oil and gas field is input into a unidirectional context encoder based on a forward LSTM model to obtain the previous context semantic coding vector of the process to be identified. Accordingly, considering that the semantic features of the previous text of the semantic coding vector of the process to be identified express some preceding work information of the process to be identified, for example, the process of "installing the oil pipe" may be followed by the process of "checking the tubing string and preparing a new tubing string". Therefore, in order to fully understand the contextual semantic information between the various processes in the previous text, in the technical solution of the present application, the set of the semantic coding vector of the process to be identified and the semantic coding vector of the conventional well repair operation process in the previous text of the oil and gas field is input into a unidirectional context encoder based on a forward LSTM model to capture the dependency between these information and obtain the previous context semantic coding vector of the process to be identified. It should be understood that the forward LSTM model processes data sequentially from the beginning of the sequence, making full use of the information in the set of semantic encoding vectors of conventional well repair operations in oil and gas fields in the previous text, and gradually learning and memorizing the impact of the previous text on the current process to be identified along the time or process execution sequence. For the process to be identified, the information such as the purpose of operation and preparatory work contained in the previous text process is crucial to accurately understand the semantics of the process, and the forward LSTM model can effectively capture this information, so as to more comprehensively reflect the semantic features of the process to be identified based on the previous text in the entire operation process.

[0031] The following is a detailed description of a specific implementation process of "inputting the semantic encoding vector of the process to be identified and the set of semantic encoding vectors of the conventional well repair operation process in the previous text of the oil and gas field into the unidirectional context encoder based on the forward LSTM model to obtain the semantic encoding vector of the previous context of the process to be identified": First, there is the data preparation and format adjustment stage. In this stage, two key data parts are required. On the one hand, there is the semantic encoding vector of the process to be identified, which is obtained by semantically encoding the text description of the process to be identified in the conventional well repair operation in the oil and gas field. This vector contains the unique semantic information of the process itself. On the other hand, there is the set of semantic encoding vectors of the conventional well repair operation process in the previous text of the oil and gas field. These vectors are also obtained by semantically encoding the text descriptions of each process in the previous text. They represent the semantic features of a series of operation processes before the process to be identified. In order to allow these data to enter the forward LSTM model smoothly for processing, some necessary operations need to be performed on them. First, the order of the data must be determined. The vectors in the set of the previous process vectors must be arranged in the order of execution of the processes. This is because the forward LSTM model processes sequence data in sequence. Only by ensuring the correctness of the order can the model accurately capture the time dependency and context information between processes. At the same time, these vectors need to be adjusted to a format suitable for the input of the forward LSTM model. Generally, the input expected by the forward LSTM model is a three-dimensional tensor with a shape of (number of samples, time step, feature dimension). In the application scenario of this application, the number of samples is generally 1, because the context of a single process to be identified is processed. The time step is the number of previous processes, which reflects how many processes need to be considered before the process to be identified. The feature dimension is the dimension of each semantic encoding vector, which determines the information richness contained in each process vector.

[0032] Next is to initialize the forward LSTM model. The forward LSTM model is composed of a series of LSTM units. Each LSTM unit has its own unique structure and function, mainly including input gate, forget gate, output gate and cell state. These gating mechanisms enable the forward LSTM model to effectively process sequence data and capture long-term dependencies in the data. When defining the model structure, it is necessary to ensure that the input dimension of the model is consistent with the dimension of the semantic encoding vector. This is because only when the dimensions match can the model correctly receive the input data and effectively process the input process vector information. At the same time, it is also necessary to adjust the output dimension of the model according to actual needs. The selection of output dimension usually takes into account multiple factors, such as the complexity of subsequent tasks, the feature scale of the data, etc. In general, it is necessary to select a suitable dimension to represent the contextual semantic information so that this information can be better used for classification and other operations in the future.

[0033] When initializing model parameters, random initialization can be used. Common initialization methods include Xavier initialization or He initialization. The purpose of these initialization methods is to ensure that the model can converge stably in the early stages of training. By properly initializing weights and biases, problems such as gradient vanishing or gradient exploding can be avoided during the training process, allowing the model to more effectively learn the contextual relationship between processes.

[0034] After completing the model initialization, the sequence input and forward propagation phase begins. In this phase, the previous process vectors are sequentially input into the forward LSTM model. Starting from the first previous process vector, it is used as the input of the LSTM model at the first time step. The LSTM unit performs a series of complex calculations based on the current input and the hidden state of the previous time step. Specifically, the first is the calculation of the forget gate, which processes the input and the hidden state of the previous time step through a sigmoid function to obtain a vector between 0 and 1, which determines which information in the cell state needs to be forgotten. Next is the calculation of the input gate, which also uses the sigmoid function to determine which new information needs to be added to the cell state. At the same time, the candidate value of the cell state is calculated, and the input and the hidden state of the previous time step are transformed using the hyperbolic tangent function to obtain a new vector. Then, the cell state is updated according to the output of the forget gate and the input gate. The cell state is a key part of the LSTM model for transmitting long-term information, which can retain important information between different time steps. Finally, the hidden state of the current time step is obtained through the processing of the output gate and the cell state. The hidden state contains the context information up to the current time step, which will serve as an important basis for the calculation of the next time step.

[0035] After processing the first previous process vector, the subsequent previous process vectors are input into the LSTM model in turn, and the above calculation process is repeated. Each new process vector input will update the hidden state and cell state, allowing the model to gradually accumulate the context information of the previous process. In this way, the forward LSTM model can effectively capture the temporal dependency and contextual semantic information between processes, which can lay the foundation for the subsequent accurate representation of the previous context of the process to be identified.

[0036] After completing the processing of the previous process vector, fuse the process vector to be identified with the context information. First, it is necessary to obtain the hidden state of the last time step. This hidden state is obtained after processing all the previous process vectors, and it contains all the context information of the previous process. Then, fuse the semantic encoding vector of the process to be identified with the hidden state of the last time step. There are many ways to fuse, and the most common one is splicing on the feature dimension. Through splicing, a new vector can be obtained, which combines the information of the process to be identified itself and the context information of the previous text. The advantage of this is that it allows the model to not only focus on its own semantic features when considering the process to be identified, but also combine the background information of the previous process, so as to more comprehensively understand the position and significance of the process to be identified in the entire operation process.

[0037] Finally, the semantic encoding vector of the previous context of the process to be identified is generated. The fused vector is automatically input into a fully connected layer or subjected to other forms of linear transformation. The fully connected layer is a common neural network layer that can map the input vector to the required dimension. Through the processing of the fully connected layer, a new vector is obtained, which is the semantic encoding vector of the previous context of the process to be identified. It combines the context information of the previous process and part of the information of the process to be identified itself, and can better represent the semantic environment of the process to be identified in the entire operation process. This vector will serve as an important input for subsequent data processing steps, providing key feature information for accurately classifying the working conditions of conventional well repair operations in oil and gas fields.

[0038] In step S152, the semantic encoding vector of the process to be identified and the set of semantic encoding vectors of the conventional well repair operation process in the subsequent oil and gas field are input into a unidirectional context encoder based on a backward LSTM model to obtain the semantic encoding vector of the context of the process to be identified. It should be understood that in the well repair operation in the oil and gas field, the execution of a process often has an impact on the subsequent process, and this impact is crucial to fully understand the nature and function of the process. For example, the "oil pipe installation" process may be followed by the "sealing detection" process. Through the backward LSTM model, it is possible to mine out information such as how the "oil pipe installation" process creates conditions for the "sealing detection" process and the verification requirements of the "sealing detection" for the quality of the "oil pipe installation". Therefore, for example, to effectively capture the information contained in the process after the process to be identified, and their association with the process to be identified, the present application obtains the semantic encoding vector of the context of the process to be identified by inputting the set of semantic encoding vectors of the process to be identified and the conventional well repair operation process in the subsequent oil and gas field into a unidirectional context encoder based on a backward LSTM model to capture and mine the semantic associations between each process, and obtain the semantic encoding vector of the context of the process to be identified.

[0039] In step S160, the semantic encoding vector of the context preceding the process to be identified and the semantic encoding vector of the context following the process to be identified are subjected to positive and negative attention field interactive responses to obtain a bidirectional semantic encoding vector of the context of the process to be identified. Specifically, Figure 4 Flow chart of S160 in the method for classifying the working conditions of conventional well repair operations in oil and gas fields according to an embodiment of the present application. Figure 4 As shown, the step S160 includes: S161, performing homography projection transformation on the semantic coding vector of the preceding context of the process to be identified and the semantic coding vector of the succeeding context of the process to be identified to obtain the homography projection coding vector of the preceding context of the process to be identified and the homography projection coding vector of the succeeding context of the process to be identified; S162, constructing a positive and negative preceding-postponing context bidirectional attention balance field between the homography projection coding vector of the preceding context of the process to be identified and the homography projection coding vector of the succeeding context of the process to be identified; S163, based on the positive and negative preceding-postponing context bidirectional attention balance field, performing semantic mapping modulation and response coding on the homography projection coding vector of the preceding context of the process to be identified and the homography projection coding vector of the succeeding context of the process to be identified to obtain the bidirectional context semantic coding vector of the process to be identified.

[0040] It should be understood that the preceding context semantic coding vector and the succeeding context semantic coding vector of the process to be identified provide information about the process to be identified from the front and back directions respectively, and can only provide partial context information. In order to ensure that the model obtains comprehensive and integrated information, so as to more accurately capture the true situation of the current process, in the technical solution of the present application, the preceding context semantic coding vector of the process to be identified and the succeeding context semantic coding vector of the process to be identified are subjected to positive and negative attention field interactive responses to obtain a bidirectional context semantic coding vector of the process to be identified. In this way, the information from the preceding and succeeding contexts can be deeply integrated, so that the model can fully and effectively utilize all relevant information. For example, the preceding text may emphasize the preparatory work of the process to be identified, and the succeeding text focuses on the impact of the process on subsequent processes. Through interactive responses, these two aspects of information can be integrated to form a more complete understanding of the process, so as to more accurately judge the job category label of the process to be identified.

[0041] In detail, first, it is necessary to perform homography projection transformation on the semantic coding vector of the preceding context of the process to be identified and the semantic coding vector of the following context of the process to be identified to obtain the homography projection coding vector of the semantic coding vector of the preceding context of the process to be identified and the homography projection coding vector of the semantic coding vector of the following context of the process to be identified. The above process can be expressed as: ;in, is the semantic encoding vector of the context preceding the process to be identified, is the context semantic encoding vector of the process to be identified, is the homography matrix of the context semantic mapping of the previous process to be identified, is the homography matrix of the context semantic mapping of the process to be identified, is the semantic homography projection encoding vector of the previous context of the process to be identified, It is the semantic homography projection encoding vector of the context after the process to be recognized.

[0042] It should be understood that the homography projection transformation can map the contextual semantic encoding vectors of the preceding and following processes to be identified to a new feature space. In the oil and gas field well repair operation scenario, the information represented by the original contextual semantic encoding vectors may be scattered and complex, with different processes, equipment usage, operation status and other information intertwined. By projecting the information to the new feature space, the information can be reorganized and integrated, so that the model can process the information more systematically. And in the new feature space, the relative position relationship between the features is maintained. This means that the inherent association between the contextual information is preserved. For example, in the well repair operation, the preparation process described in the previous text and the subsequent influencing process described in the later text have a specific association in the original semantic encoding vector. After the homography projection transformation, this association still exists in the new feature space. This helps the model to capture the logical connection between the context in the new space, and the coherence of the information will not be destroyed due to the transformation of the space. In addition, the homography projection transformation also allows the model to observe features from different perspectives and scales. That is, the semantic homography projection encoding vectors of the previous and subsequent contexts obtained through the homography projection transformation contain the previous and subsequent context information observed from different perspectives and scales. This allows the model to comprehensively consider more factors when judging the category label of the process operation to be identified, and can reduce misjudgments caused by incomplete information or a single perspective.

[0043] Specifically, in the embodiment of the present application, the step S162 includes: calculating the forward process to be identified previous context-post context semantic attention score field of the process to be identified previous context semantic homography projection coding vector relative to the process to be identified post context semantic homography projection coding vector, and the process can be expressed as: ;in, is the semantic homography projection encoding vector of the previous context of the process to be identified, is the semantic homography projection encoding vector of the context of the process to be identified, yes The transposed vector of is matrix multiplication, yes Length, It is the semantic attention score field of the preceding and following contexts in the positive process to be recognized.

[0044] The reverse previous-post context semantic attention score field of the to-be-recognized process is calculated relative to the homography projection coding vector of the previous context semantic of the to-be-recognized process. The process can be expressed as: ;in, is the semantic homography projection encoding vector of the previous context of the process to be identified, is the semantic homography projection encoding vector of the context of the process to be identified, yes The transposed vector of is matrix multiplication, yes Length, It is the semantic attention score field of the previous and subsequent contexts in the reverse process to be recognized.

[0045] The feature concatenation of the preceding and following context semantic attention score field of the forward process to be identified and the preceding and following context semantic attention score field of the reverse process to be identified is input into a convolution layer based on a convolution kernel of 3×3 to obtain the preceding and following context semantic bidirectional attention balance field of the forward and reverse processes to be identified. The process can be expressed as: ;in, is the semantic attention score field of the preceding and following contexts in the positive process to be recognized, is the semantic attention score field of the previous and subsequent contexts in the reverse recognition process, is the feature concatenation operation, is a convolutional code with a convolution kernel of 3×3. It is a bidirectional attention balance field of the context semantics of the preceding and following texts in the positive and negative recognition processes.

[0046] It should be understandable that in complex well repair operation scenarios, text descriptions cover a lot of information, from equipment use, operating procedures to well condition details. By calculating the semantic attention score field of the context before and after the forward process to be identified, the model can simulate the human visual system and quickly focus on the part of the context before and after semantic homography projection encoding vector that is closely related to the vector after the text. For example, when judging whether a process belongs to the special string operation, the information about the string entry depth and special connection method in the after-text vector is the key basis for judgment. The forward attention mechanism allows the model to accurately locate the relevant content such as special string preparation and special tool equipment in the complex previous information, avoiding interference from other irrelevant information, thereby providing strong support for accurate identification of process categories. In addition, the text data of well repair operations may contain noise information, which is of no substantial help to process identification. The forward attention mechanism acts as an efficient filter. By quantifying the correlation between the previous and next encoding vectors, the noise information that is irrelevant to the current process judgment is filtered out, which can reduce the interference of noise on the recognition results and improve the accuracy and reliability of recognition.

[0047] Accordingly, considering that the semantic attention score field of the preceding and succeeding context of the positive process to be identified mainly focuses on the importance distribution of the preceding vector in the context of the succeeding vector, this may not be able to fully capture all the relationships between the preceding and succeeding contexts. In the text description of the process of the well repair operation, the succeeding vector also has an important correlation and influence on the preceding vector. This application can mine this information from the opposite direction by calculating the semantic attention score field of the preceding and succeeding context of the reverse process to be identified, supplementing the part that the semantic attention score field of the preceding and succeeding context of the positive process to be identified fails to cover, and avoiding information omissions. That is, through the supplementary information provided by the semantic attention score field of the preceding and succeeding context of the reverse process to be identified, the model can consider more comprehensive factors when identifying the job category label of the process, avoiding misjudgment caused by only considering the relationship in a single direction, thereby improving the accuracy of process identification.

[0048] It should be understood that in oil and gas field well repair operations, the judgment of the process needs to comprehensively consider various complex information in the context. The semantic attention score fields of the context of the forward and reverse processes to be identified reveal the relationship between the homography projection encoding vectors of the semantic context of the forward and reverse processes from different directions. First, through feature splicing, these information from different directions are completely integrated together. For example, when judging whether a process belongs to the operation of lifting and lowering special pipes, the forward attention score field may highlight the selection and preparation of special pipes in the previous preparation work, and the reverse attention score field may emphasize the impact of this process on the use of special pipes in subsequent processes. After splicing the two, the model can obtain both information at the same time and have a more comprehensive understanding of the process. Then, the convolution layer with a convolution kernel of 3×3 is used to process the spliced ​​features, which can extract more representative features and enhance the adaptability of the model to different data. That is, the generated bidirectional attention balance field of the context semantics of the positive and negative processes to be identified can enable the model to quickly adapt to the data characteristics and accurately analyze the relationship between the context when facing new well repair operation data, thereby improving the generalization ability of the model. This means that whether it is a regular well repair operation or an operation encountering special circumstances, the model can stably and accurately identify the operation category label of the process.

[0049] Specifically, in the embodiment of the present application, the step S163 includes: mapping the homography projection coding vector of the context semantics of the previous process to be identified and the homography projection coding vector of the context semantics of the subsequent process to be identified to the bidirectional attention balance field of the context semantics of the positive and negative processes to be identified to obtain the homography projection attention modulation coding vector of the context semantics of the previous process to be identified and the homography projection attention modulation coding vector of the context semantics of the subsequent process to be identified. The process can be expressed as: ;in, is the semantic homography projection encoding vector of the previous context of the process to be identified, is the semantic homography projection encoding vector of the context of the process to be identified, is matrix multiplication, It is the bidirectional attention balance field of the context semantics of the positive and negative processes to be recognized, is the semantic homography projection attention modulation encoding vector of the previous context of the process to be recognized, It is the context semantic homography projection attention modulation encoding vector after the process to be recognized.

[0050] The bidirectional context semantic coding vector of the process to be identified is obtained by dividing the position point by the homography projection attention modulation coding vector of the previous context semantics of the process to be identified and the homography projection attention modulation coding vector of the subsequent context semantics of the process to be identified. The process can be expressed as: ;in, is the semantic homography projection attention modulation encoding vector of the previous context of the process to be recognized, is the context semantic homography projection attention modulation encoding vector of the post-process to be recognized, is the bidirectional context semantic encoding vector of the process to be identified.

[0051] Accordingly, considering that well repair operations in oil and gas fields involve many complex processes, the characteristics of each process not only contain their own operation information, but are also closely related to the preceding and following processes. Although the original context semantic homography projection coding vectors of the preceding and following processes to be identified each carry certain information, it is difficult to fully reflect the complex interactive relationship between features. In order to capture the complex activities between features, the present application maps the preceding context semantic homography projection coding vector of the preceding process to be identified and the following context semantic homography projection coding vector of the following process to be identified to the preceding and following context semantic bidirectional attention balance field of the positive and negative processes to be identified, so as to comprehensively consider the characteristics of the preceding and following contexts with the help of the balance field, and then deeply explore the complex relationship between features. Through this process, the model can understand the feature relationship at a higher level of abstraction. It is no longer limited to the surface description of the process operation, but can grasp the deeper logical connection between the processes. For example, when judging which operation category a complex process belongs to, the model can make a more accurate judgment by comprehensively considering the preceding and following factors from the perspective of the overall operation process, which helps to improve the accuracy of process classification.

[0052] It should be understood that well repair operations involve a variety of complex processes. After modulation, the feature vectors of each process contain rich contextual information. The positional dot division operation can deeply mine this information and reveal new relationships between two modulated feature vectors. For example, when judging whether a process belongs to a pipe string operation or a special pipe string operation, the previous context vector may contain relevant information about the pipe string preparation work, and the subsequent context vector contains the operation details during the pipe string operation. Through the dot division operation, the degree of correlation between the context information at each position can be clearly presented, helping the model to more accurately judge the characteristics of the process. Specifically, this dot division operation reveals the subtle proportional relationship between features, which provides a key basis for model decision-making. In the well repair operation scenario, different processes differ in terms of equipment usage time, complexity of operation steps, etc., which are reflected in the different proportions of the element values ​​of each dimension in the feature vector. For example, when distinguishing between flushing operations and rotation operations, flushing operations may focus more on the proportional relationship between the working time of the mud pump and the fluid flow rate, while rotation operations focus more on the proportional relationship between the rotation speed of the pipe string and the torque. The proportional relationship obtained by the dot division operation allows the model to accurately capture these differences and thus more accurately identify the process category.

[0053] In step S170, based on the bidirectional context semantic coding vector of the process to be identified, an identification result of the job category label for representing the i-th process is obtained. Specifically, in an embodiment of the present application, the step S170 includes: inputting the bidirectional context semantic coding vector of the process to be identified into a process identifier based on a classifier to obtain the identification result, and the identification result is the job category label of the i-th process. That is, the bidirectional context semantic coding vector of the process to be identified obtained by interactively responding with the preceding context semantic coding vector of the process to be identified and the following context semantic coding vector of the process to be identified is classified and processed to intelligently determine the job category to be identified. In particular, the job category labels include flushing operations, rotation operations, lifting and lowering tubular operations, lifting and lowering special tubular operations, empty wellbore, cable perforation, etc. In this way, managers can better plan resources, arrange manpower and technical equipment, and ensure that each process can be executed efficiently. In particular, in a specific embodiment of the present application, the bidirectional context semantic coding vector of the process to be identified is input into a process identifier based on a classifier to obtain the identification result, and the identification result is the job category label of the i-th process, including: using the fully connected layer of the classifier to fully connect the bidirectional context semantic coding vector of the process to be identified to obtain a bidirectional context semantic fully connected coding classification feature vector of the process to be identified; inputting the bidirectional context semantic fully connected coding classification feature vector of the process to be identified into the Softmax classification function of the classifier to obtain the identification result.

[0054] Preferably, the bidirectional context semantic encoding vector of the process to be identified is input into a process identifier based on a classifier to obtain an identification result, which includes: determining the label probability value corresponding to each job category label of the i-th process obtained by inputting the bidirectional context semantic encoding vector of the process to be identified into the process identifier based on a classifier , and calculate the probability value of each label The square root of the sum of the squares of is used to obtain the bidirectional context semantic probability value of the process to be identified: ;in, Represents the label probability value corresponding to each job category label, The number of labels representing the job category labels, Represents the bidirectional context semantic probability value of the process to be identified.

[0055] For each label probability value , the label probability value Subtract the bidirectional context semantic probability value of the process to be identified and divide it by the label probability value , and for all label probability values Sum to get the bidirectional context semantic encoding density gradient value of the process to be identified: ;in, Represents the bidirectional context semantic encoding density gradient value of the process to be identified.

[0056] The feature mean of the bidirectional context semantic encoding vector of the process to be identified and characteristic variance Multiply the bidirectional context semantic coding density gradient value of the process to be identified respectively to obtain the bidirectional context semantic coding distribution field mapping value of the first process to be identified and the second step to be identified bidirectional context semantic encoding distribution field mapping value .

[0057] Calculate the point product of the bidirectional context semantic encoding vector of the process to be identified and the mapping value of the bidirectional context semantic encoding distribution field of the first process to be identified as the exponent, and use the natural constant as the base exponential function , and multiply it by the bidirectional context semantic encoding vector of the process to be identified to obtain the first bidirectional context semantic encoding fine-grained response vector of the process to be identified ,in, Represents dot product.

[0058] Calculate the point product of the bidirectional context semantic encoding vector of the process to be identified and the mapping value of the bidirectional context semantic encoding distribution field of the second process to be identified as the exponent, and use the natural constant as the base exponential function , and multiply it by the bidirectional context semantic encoding vector of the process to be identified to obtain the second bidirectional context semantic encoding fine-grained response vector of the process to be identified .

[0059] Calculate the weighted sum of the first process bidirectional context semantic coding fine-grained response vector and the second process bidirectional context semantic coding fine-grained response vector to obtain an optimized process bidirectional context semantic coding vector ,in, Indicates point addition, and represents different weight hyperparameters.

[0060] The optimized bidirectional context semantic encoding vector of the process to be identified is input into a process identifier based on a classifier to obtain an identification result.

[0061] Since the preceding context semantic coding vector of the process to be identified and the succeeding context semantic coding vector of the process to be identified respectively represent the short-range and long-range bidirectional context association coding semantic features of the ith process in the set of text descriptions of conventional well repair operations in oil and gas fields relative to the preceding and succeeding texts, when performing interactive responses based on feature positive and negative attention fields, the encoded semantic positive and negative attention field reinforcement feature representation between local semantic space domains will have a micro-macro interactive response representation deviation relative to the global semantic space domain, thereby causing a dynamic deviation in the mapping of features to class target probabilities when the bidirectional context semantic coding vector of the process to be identified is input into a process identifier based on a classifier, thereby reducing the accuracy of the obtained recognition results.

[0062] Therefore, by calculating the probability density gradient features of the target sequence corresponding to the bidirectional contextual semantic coding vector of the process to be identified and taking it as the input of the distributed local feature aggregation unit, the distribution field feature collaborative mapping of the bidirectional contextual semantic coding vector of the process to be identified can be constructed based on the heterogeneous core representation framework, thereby enhancing the responsiveness of the fine-grained pattern representation of the global semantic dependence of the bidirectional contextual semantic coding vector of the process to be identified to the class probability modeling, that is, by coordinating the global semantic space correlation of the different scale features of the bidirectional contextual semantic coding vector of the process to be identified, the nonlinear transformation mechanism of high-dimensional embedding into the probability space in the feature-class probability mapping optimization process is strengthened to ensure that on the feature space-class probability mapping path, the fine-grained pattern representation and the class probability distribution always maintain a dynamic balance of iterative optimization, thereby improving the accuracy of the recognition result obtained by inputting the bidirectional contextual semantic coding vector of the process to be identified into the process identifier based on the classifier.

[0063] In summary, a method for classifying the working conditions of conventional well repair operations in oil and gas fields based on the embodiment of the present application is explained, which uses data processing technology based on deep learning to decompose and semantically encode the text description of conventional well repair operations in oil and gas fields, and then extracts the process semantic features corresponding to the i-th process from the set of encoded process semantic features as the process semantic features to be identified, and uses it as a dividing line to divide the entire set into a set of preceding and succeeding process semantic features, and then, the process semantic features to be identified are contextually encoded with the set of preceding process semantic features and the set of succeeding process semantic features, respectively, so as to intelligently judge the work category of the process to be identified based on the positive and negative attention response representations between the preceding and succeeding context semantic features of the process to be identified. In this way, the efficiency and accuracy of process classification can be effectively improved.

Claims

1. A method for classifying working conditions of conventional well repair operations in oil and gas fields, characterized in that: include: Obtain a text description of conventional well repair operations in an oil and gas field; decompose and semantically encode the text description of the conventional well repair operations in an oil and gas field to obtain a set of semantic encoding vectors of conventional well repair operations in an oil and gas field; extract the semantic encoding vector of the conventional well repair operations in an oil and gas field corresponding to the ith process from the set of semantic encoding vectors of the conventional well repair operations in an oil and gas field as the semantic encoding vector of the process to be identified; using the semantic encoding vector of the process to be identified as a boundary, divide the set of semantic encoding vectors of the conventional well repair operations in an oil and gas field into a set of semantic encoding vectors of the preceding conventional well repair operations in an oil and gas field and a set of semantic encoding vectors of the subsequent conventional well repair operations in an oil and gas field; respectively A set of process semantic coding vectors is context-encoded to obtain a context semantic coding vector of a preceding process to be identified and a context semantic coding vector of a following process to be identified; a positive and negative attention field interactive response is performed on the preceding process to be identified and the context semantic coding vector of a following process to be identified to obtain a bidirectional context semantic coding vector of the process to be identified, wherein the positive and negative attention field interactive response is performed by constructing a positive and negative preceding and following process to be identified context semantic bidirectional attention balance field, and semantic mapping modulation and response coding are performed based on the constructed positive and negative preceding and following process to be identified context semantic bidirectional attention balance field; based on the bidirectional context semantic coding vector of the process to be identified, an identification result of a job category label representing the i-th process is obtained.

2. The method for classifying working conditions of conventional well repairing operations in oil and gas fields according to claim 1, characterized in that: The text description of the conventional well workover operation in the oil and gas field is decomposed and semantically encoded to obtain a set of semantic encoding vectors of the conventional well workover operation process in the oil and gas field, including: decomposing the text description of the conventional well workover operation in the oil and gas field based on the process of construction stage to obtain a set of text descriptions of the conventional well workover operation process in the oil and gas field; and semantically encoding each text description of the conventional well workover operation process in the set of text descriptions of the conventional well workover operation process in the oil and gas field to obtain a set of semantic encoding vectors of the conventional well workover operation process in the oil and gas field.

3. The method for classifying working conditions of conventional well repairing operations in oil and gas fields according to claim 2, characterized in that: The method comprises: using a semantic encoder including an embedding layer to semantically encode each text description of a conventional well workover operation procedure in the set of text descriptions of a conventional well workover operation procedure in the oil and gas field to obtain a set of semantic encoding vectors of the conventional well workover operation procedure in the oil and gas field.

4. The method for classifying working conditions of conventional well repairing operations in oil and gas fields according to claim 3 is characterized in that: The set of semantic coding vectors of the preceding conventional well repair operation procedures in oil and gas fields and the set of semantic coding vectors of the following conventional well repair operation procedures in oil and gas fields are respectively context-encoded to obtain the preceding context semantic coding vector of the procedure to be identified and the following context semantic coding vector of the procedure to be identified, including: inputting the set of semantic coding vectors of the procedure to be identified and the preceding semantic coding vector of the conventional well repair operation procedures in oil and gas fields into a unidirectional context encoder based on a forward LSTM model to obtain the preceding context semantic coding vector of the procedure to be identified; inputting the set of semantic coding vectors of the procedure to be identified and the following semantic coding vector of the conventional well repair operation procedures in oil and gas fields into a unidirectional context encoder based on a backward LSTM model to obtain the following context semantic coding vector of the procedure to be identified.

5. The method for classifying working conditions of conventional well repairing operations in oil and gas fields according to claim 4 is characterized in that: The positive and negative attention fields are interactively responded to the preceding context semantic coding vector of the process to be identified and the succeeding context semantic coding vector of the process to be identified to obtain the bidirectional context semantic coding vector of the process to be identified, including: performing homography projection transformation on the preceding context semantic coding vector of the process to be identified and the succeeding context semantic coding vector of the process to be identified to obtain the homography projection coding vector of the preceding context semantic of the process to be identified and the homography projection coding vector of the succeeding context semantic of the process to be identified; constructing a positive and negative preceding-postponing context semantic bidirectional attention balance field between the homography projection coding vector of the preceding context semantic of the process to be identified and the homography projection coding vector of the succeeding context semantic of the process to be identified; based on the positive and negative preceding-postponing context semantic bidirectional attention balance field, semantic mapping modulation and response coding are performed on the homography projection coding vector of the preceding context semantic of the process to be identified and the homography projection coding vector of the succeeding context semantic of the process to be identified to obtain the bidirectional context semantic coding vector of the process to be identified.

6. The method for classifying working conditions of conventional well repairing operations in oil and gas fields according to claim 5, characterized in that: Constructing a positive and negative bidirectional attention balance field between the homography projection coding vector of the previous context semantics of the process to be identified and the homography projection coding vector of the subsequent context semantics of the process to be identified, including: calculating the positive bidirectional attention score field of the previous context semantics of the process to be identified relative to the homography projection coding vector of the subsequent context semantics of the process to be identified; calculating the negative bidirectional attention score field of the previous context semantics of the process to be identified relative to the homography projection coding vector of the subsequent context semantics of the process to be identified; and feature splicing the positive bidirectional attention score field of the previous context semantics of the process to be identified and the negative bidirectional attention score field of the previous context semantics of the process to be identified into a convolution layer based on a convolution kernel of 3×3 to obtain the positive and negative bidirectional attention balance field of the previous context semantics of the process to be identified.

7. The method for classifying working conditions of conventional well repairing operations in oil and gas fields according to claim 6, characterized in that: Based on the positive and negative preceding and succeeding context semantics bidirectional attention balance fields of the processes to be identified, semantic mapping modulation and response coding are performed on the preceding context semantics homography projection coding vector of the processes to be identified and the succeeding context semantics homography projection coding vector of the processes to be identified to obtain the bidirectional context semantics coding vector of the processes to be identified, including: mapping the preceding context semantics homography projection coding vector of the processes to be identified and the succeeding context semantics homography projection coding vector of the processes to be identified to the positive and negative preceding and succeeding context semantics bidirectional attention balance fields of the processes to be identified respectively to obtain the preceding context semantics homography projection attention modulation coding vector of the processes to be identified and the succeeding context semantics homography projection attention modulation coding vector of the processes to be identified; calculating the position point-by-position division between the preceding context semantics homography projection attention modulation coding vector of the processes to be identified and the succeeding context semantics homography projection attention modulation coding vector of the processes to be identified to obtain the bidirectional context semantics coding vector of the processes to be identified.

8. The method for classifying working conditions of conventional well repairing operations in oil and gas fields according to claim 7, characterized in that: Based on the bidirectional context semantic coding vector of the process to be identified, an identification result for representing the job category label of the i-th process is obtained, including: inputting the bidirectional context semantic coding vector of the process to be identified into a classifier-based process identifier to obtain the identification result, and the identification result is the job category label of the i-th process.

Images