Intelligent temperature loss early warning monitoring device for underground steam pipe network based on ad hoc network

Abstract

The application discloses an underground steam pipe network intelligent temperature loss early warning monitoring device based on a self-organizing network, and relates to the technical field of pipe network monitoring.The method comprises the following steps: determining a protection pipe diagram, establishing a temperature loss data acquisition system; obtaining temperature loss early warning historical data, extracting a text feature vector and a structured field to establish a temperature loss word table; establishing an alarm type classification identification model based on the temperature loss word table, obtaining an alarm type sequence, an associated temperature loss cause, and an evaluation hazard level; establishing an association graph; determining an alarm type, inputting the association graph to obtain a temperature loss cause optimal solution, and early warning based on the optimal solution and a key node position.The technical problems of single feature information, high false alarm rate and insufficient alarm accuracy of the existing underground steam pipe network temperature loss early warning system are solved, the accuracy and reliability of underground steam pipe network temperature loss early warning are improved, potential temperature loss problems are accurately found, false alarms are reduced, and temperature loss causes are quickly located, and the technical effect of guaranteeing safe and stable operation of the pipe network is achieved.

Description

TECHNICAL FIELD

[0001] The application relates to the technical field of pipe network monitoring, in particular to an underground steam pipe network intelligent temperature loss early warning monitoring device based on an ad hoc network. BACKGROUND

[0002] In the underground steam pipe network operation scenario, the temperature loss problem is particularly prominent, and the demand for efficient and accurate temperature loss early warning monitoring is also more prominent. Therefore, realizing intelligent temperature loss early warning monitoring of the underground steam pipe network has become a crucial link to ensure the safe operation of the pipe network. Traditional underground steam pipe network monitoring is often limited and incomplete, mainly relies on manual inspection and limited sensor data, and lacks the use and in-depth analysis of unstructured data. It is difficult to accurately determine the temperature loss situation, and the accuracy and reliability of the temperature loss early warning are insufficient, and the false positive rate is high. The determination of the temperature loss reason is relatively single, and it is difficult to cope with the complex and variable actual operation of the pipe network.

[0003] In the related art at present, the underground steam pipe network temperature loss early warning system has the technical problems of single feature information, high false positive rate and insufficient alarm accuracy. SUMMARY

[0004] The application provides an underground steam pipe network intelligent temperature loss early warning monitoring device based on an ad hoc network, which determines the underground steam pipe network protection pipe map, analyzes the key node position and establishes a temperature loss data acquisition system; obtains temperature loss early warning historical data, extracts text feature vectors and structured fields to establish a temperature loss word table; establishes an alarm type classification recognition model based on the temperature loss word table, obtains an alarm type sequence and associates temperature loss reasons and hazard levels; calculates the alarm type similarity to establish an association graph; analyzes the key node alarm information to determine the alarm type, inputs the association graph to obtain the optimal solution of the temperature loss reason, and early warns based on the optimal solution and the key node position. Through the introduction of unstructured data analysis, damage process time feature and transaction logic chain data filtering and multi-node multi-dimensional alarm prediction mutual verification and other means, the accuracy and reliability of the underground steam pipe network temperature loss early warning are improved, potential temperature loss problems are accurately found, false positives are reduced, and the temperature loss reason is quickly located, thereby ensuring the safe and stable operation of the pipe network.

[0005] The application provides an underground steam pipe network intelligent temperature loss early warning monitoring device based on an ad hoc network, which includes:

[0006] The temperature loss data acquisition system establishment module is configured to determine a protection pipeline map of the underground steam pipe network, analyze a key node position of the protection pipeline map, and establish a temperature loss data acquisition system of a pipeline key node; the temperature loss vocabulary construction module is configured to obtain temperature loss early warning historical data, obtain a text feature vector and a structured field based on the temperature loss early warning historical data, obtain the text feature vector through an unstructured field, analyze the text feature vector and the structured field, and establish a limited set temperature loss vocabulary; the classification and identification model construction module is configured to establish a self-supervised alarm type classification and identification model based on the limited set temperature loss vocabulary, obtain an alarm type sequence, associate each alarm type in the alarm type sequence with a plurality of temperature loss causes, and evaluate a hazard level of each alarm type; the temperature loss cause association graph construction module is configured to obtain the alarm type sequence, the hazard level, and the plurality of temperature loss causes, calculate an alarm type similarity, and establish an alarm type temperature loss cause association graph; the temperature loss cause optimal solution acquisition module is configured to analyze a plurality of key node early warning information based on the self-supervised alarm type classification and identification model, determine a plurality of alarm types, input the plurality of alarm types into the alarm type temperature loss cause association graph, and obtain a temperature loss cause optimal solution; and the early warning module is configured to perform early warning based on the temperature loss cause optimal solution and the key node position.

[0007] The underground steam pipe network intelligent temperature loss early warning monitoring device based on a self-organizing network provided in the present application first determines a protection pipeline map of an underground steam pipe network, analyzes a key node position, and establishes a temperature loss data acquisition system; obtains temperature loss early warning historical data, extracts a text feature vector and a structured field to establish a temperature loss vocabulary; establishes an alarm type classification and identification model based on the temperature loss vocabulary, obtains an alarm type sequence, associates temperature loss causes, and evaluates a hazard level; calculates an alarm type similarity to establish an association graph; analyzes key node alarm information to determine alarm types, inputs the alarm types into the association graph to obtain a temperature loss cause optimal solution, and performs early warning based on the optimal solution and the key node position. Through the introduction of unstructured data analysis, damage process time characteristics, transaction logic chain data filtering, and multi-node multi-dimensional alarm prediction mutual verification, the precision and reliability of underground steam pipe network temperature loss early warning are improved, potential temperature loss problems are accurately found, false positives are reduced, and temperature loss causes are quickly located, thereby ensuring the safe and stable operation of the pipe network. BRIEF DESCRIPTION OF DRAWINGS

[0008] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings of the embodiments of the present application will be briefly introduced below, and the flowcharts are used to illustrate the operations performed by the system according to the embodiments of the present application in the present application. It should be understood that the foregoing or the following operations are not necessarily performed in sequence. On the contrary, various steps can be processed in reverse order or simultaneously according to needs. Meanwhile, other operations can be added to these processes, or one or more steps of operations can be removed from these processes.

[0009] Figure 1 A structure schematic diagram of the underground steam pipe network intelligent temperature loss early warning monitoring device based on ad hoc network provided for the embodiments of the present application is shown in the figure.

[0010] Figure 2 A structure schematic diagram of the temperature loss reason correlation graph construction module 40 of the underground steam pipe network intelligent temperature loss early warning monitoring device based on ad hoc network provided for the embodiments of the present application is shown in the figure.

[0011] Label explanation: temperature loss data acquisition system establishment module 10, temperature loss word table construction module 20, classification recognition model construction module 30, temperature loss reason correlation graph construction module 40, temperature loss reason optimal solution acquisition module 50, early warning module 60. DETAILED DESCRIPTION

[0012] The foregoing description is only a summary of the technical solutions of the present application, in order to more clearly understand the technical means of the present application, the content of the specification can be implemented, and in order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the following specific embodiments of the present application are described.

[0013] In order to make the purposes, technical solutions and advantages of the present application more clear, the present application will be further described in detail below with reference to the drawings, and the described embodiments should not be regarded as limiting the present application, all other embodiments obtained by those skilled in the art without creative labor belong to the scope of protection of the present application.

[0014] In the following description, "some embodiments" are referred to, which describe a subset of all possible embodiments, but it can be understood that "some embodiments" can be the same subset or different subsets of all possible embodiments, and can be combined with each other without conflict, and the term "first\second" referred to is only to distinguish similar objects, and does not represent a specific order for the objects. The terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusion, for example, a process, method, system, product or server including a series of steps or units does not have to be limited to those steps or units clearly listed, but can include other steps or modules not clearly listed or inherent to these processes, methods, products or devices. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as understood by those skilled in the art to which the present application belongs. The terms used herein are only for the purpose of describing the embodiments of the present application.

[0015] The embodiments of the present application provide an underground steam pipe network intelligent temperature loss early warning monitoring device based on an ad hoc network, as shown in Figure 1 The device comprises:

[0016] The temperature loss data acquisition system establishment module 10 is used to determine the protection pipe diagram of the underground steam pipe network, analyze the key node positions of the protection pipe diagram, and establish the temperature loss data acquisition system of the pipe key nodes. Specifically, the temperature loss data acquisition system establishment module needs to collect relevant information of the underground steam pipe network and conduct field investigation, then draw the protection pipe diagram, determine the key node types, analyze the pipe characteristics combined with the protection pipe diagram and consider the environmental factors, so as to determine the key node positions, select appropriate sensors to be installed at the key nodes, and establish a data transmission network to transmit the collected data to the data processing center in real time for processing and storage, so as to provide accurate data basis for the intelligent temperature loss early warning monitoring of the underground steam pipe network.

[0017] In a possible implementation, the temperature loss data acquisition system establishment module 10 further comprises a data acquisition unit, which is used to acquire real-time data and historical data of the key nodes. Specifically, the key node positions in the underground steam pipe network are determined, the key nodes are usually parts prone to temperature loss problems, such as pipe connections, elbows, etc., and corresponding data acquisition devices such as temperature sensors and pressure sensors are installed to monitor various parameters of the key nodes in real time, and the historical data of the key nodes are collected from the database or storage device, which can include temperature and pressure change conditions in the past period of time, and temperature loss event records, etc.

[0018] A data filtering unit is configured to obtain a damage process time feature of the key node based on historical data of the key node, and filter the real-time data based on the damage process time feature. Specifically, the historical data of the key node is analyzed to understand the time regularity and development process of the temperature damage, for example, to determine the change trend within a period of time after the temperature damage occurs and the time range in which the temperature damage is indicated. The damage process time feature obtained based on the historical data is used to filter the real-time data collected. If the currently collected data reflects a sharp change within a short period of time, which is inconsistent with the normal development process of the temperature damage according to the historical data, the system will regard it as abnormal data and filter it out, thereby excluding abnormal data caused by sudden interference or incorrect measurement and improving the accuracy and reliability of the data.

[0019] A transaction logic chain construction unit is configured to construct a transaction logic chain of the temperature damage of the underground steam pipe network, and filter the real-time data based on the transaction logic chain. Specifically, the transaction logic chain of the temperature damage of the underground steam pipe network is constructed, including collecting path information of steam flow, operation state data of each key node, and maintenance records, etc. The path information of steam flow can help determine the direction and influence range of the temperature damage propagation. The operation state data of each key node can reflect the overall operation of the pipe network. The maintenance records can provide information about the past maintenance of the pipe network and potential problems. The real-time data collected is compared with the transaction logic chain. If abnormal temperature damage occurs suddenly in a key node without any operation, which is inconsistent with the normal operation logic according to the transaction logic chain, the system can determine that the data is a false alarm and filter it out. In this way, the consistency and continuity of the data can be ensured, and the quality of the data can be improved.

[0020] A real-time data output unit is configured to output the real-time data of the key node. Specifically, after the data collection, filtering, and comparison with the transaction logic chain, the filtered real-time data of the key node is output. The data can be transmitted to a monitoring center, a data analysis system, or other related equipment for further processing and analysis. The output data can be presented in the form of charts, reports, etc., to help management personnel understand the operation state and temperature damage of the underground steam pipe network in a timely manner, and provide a basis for taking appropriate preventive and maintenance measures.

[0021] In a possible implementation, the transaction logic chain building unit further includes a structure model building sub-unit, configured to build a structure model of the underground steam pipe network. Specifically, various information of the underground steam pipe network is collected to build the structure model, including determining the layout of the pipelines through engineering drawings, field survey, etc., determining the position distribution of each node in space, recording the pipe diameter size and the material type used at different parts, and the connection mode between the pipelines, such as welding, flange connection, etc., arranging the historical maintenance records of each node, including the time, reason, and measures taken for maintenance, etc., and integrating the information to form a structure model that comprehensively describes the physical structure and maintenance history of the underground steam pipe network.

[0022] An event association sub-unit is configured to acquire the temperature damage characteristics of the underground steam pipe network and the corresponding key nodes and time, analyze the key nodes and time to associate the temperature damage characteristics with actual temperature damage events, and build the multi-stage transaction logic chain. Specifically, the possible temperature damage characteristics of the underground steam pipe network are determined, such as pipeline aging, which is manifested as brittle pipeline material and color change; temperature and pressure fluctuations can cause uneven pipeline expansion and contraction; in terms of environmental factors, soil humidity changes can cause pipeline corrosion, and geological activities can cause pipeline deformation; improper maintenance can cause improper sealing and installation errors, etc. When a temperature damage event is detected, the corresponding key nodes and time are determined, the position and characteristics of the key nodes are analyzed, and various factors at the time nodes are analyzed, the temperature damage characteristics are associated with the actual temperature damage events, based on the association relationship, the multi-stage transaction logic chain is built, for example, if the soil humidity in a certain area is high for a long time, the pipeline at the corresponding node can first show corrosion signs, which is the initial stage of temperature damage; as time goes by, corrosion worsens, the pipeline wall thins, which can cause temperature fluctuations to increase, entering the middle stage; if it continues to develop, the pipeline can have fine cracks, pressure abnormalities, and further deterioration; finally, the cracks expand to cause pipeline rupture, reduced flow, or increased leakage, which is the final damage stage.

[0023] The logic chain adjustment subunit is used for dynamically adjusting the multi-stage transaction logic chain based on the real-time data and historical data. Specifically, real-time data of the underground steam pipe network is continuously collected, including changes in parameters such as temperature, pressure, flow rate, etc., the development process and treatment results of similar temperature loss events in the historical data are referred to, and if the real-time data shows that the development of the temperature loss does not conform to the preset transaction logic chain, for example, the performance characteristics of a certain stage appear too early or too late, or new abnormal situations appear, the multi-stage transaction logic chain needs to be dynamically adjusted. The adjustment includes modifying the performance characteristic description of a certain stage, adjusting the transition conditions between stages, adding or deleting certain stages, etc. Through continuous adjustment according to real-time data and historical data, the transaction logic chain can more accurately reflect the actual development of the underground steam pipe network temperature loss, and the judgment and prediction ability of the system for temperature loss is improved.

[0024] The warm damage word table construction module 20 is used for obtaining warm damage early warning historical data, obtaining a text feature vector and a structured field based on the warm damage early warning historical data, the text feature vector being obtained through an unstructured field, analyzing the text feature vector and the structured field, and establishing a limited set warm damage word table. Specifically, the scope of the warm damage early warning historical data is determined. The data not only includes specific data when the warm damage occurs, such as temperature abnormal change value, pressure fluctuation condition, etc., but also includes the processing flow corresponding to the warm damage event. The historical data collected from the monitoring system database of the underground steam pipe network, the maintenance record archives and the related management information system is preliminarily sorted and classified to facilitate subsequent analysis, for example, classified according to the time, place, severity, etc. of the warm damage occurrence, to prepare for subsequent extraction of the feature vector and the structured field. For the unstructured field in the warm damage early warning historical data, such as the descriptive words in the processing flow and the remarks in the maintenance record, etc., the word / sentence vector embedding method is used to extract the text feature vector, to convert the natural language text into a fixed-length feature vector, which can effectively represent the meaning of the text. The words in the unstructured field are converted into vector representation, and then the feature vector of the sentence is obtained through the combination and processing of the word vectors in the sentence. The effective information is extracted from the unstructured field such as the alarm load, to supplement new valuable features for the alarm evaluation. The structured field, such as the time, place, warm damage degree, etc. of the warm damage occurrence, can be directly used for data analysis and model construction. The obtained text feature vector and structured field are comprehensively analyzed. Statistical analysis, cluster analysis, etc. are used to find out the key features and patterns related to the warm damage in them. For example, the similarity between the text feature vectors is calculated to classify similar texts into a category, so as to discover the potential relationship between different warm damage events. Combined with the time, place, etc. information in the structured field, the occurrence rule and trend of the warm damage event are analyzed. In the analysis process, machine learning algorithms such as principal component analysis, etc. can be introduced to reduce the dimension of the feature vector, remove redundant information, and improve the analysis efficiency and accuracy. According to the analysis result, the limited set warm damage word table is established, which contains the key words and phrases related to the warm damage, as well as their corresponding feature vectors and structured field information. For example, the word table may include keywords such as “pipe rupture”, “temperature anomaly”, “pressure fluctuation”, etc., as well as their corresponding text feature vectors and structured fields such as occurrence time, place, severity, etc. After the word table is established, it can be continuously optimized and updated. With the occurrence and processing of new warm damage events, new keywords and feature information are continuously added to improve the accuracy and completeness of the word table. The warm damage word table construction module can fully utilize the unstructured field and structured field in the warm damage early warning historical data, extract valuable feature information, establish a limited set warm damage word table, and provide strong support for the intelligent warm damage early warning monitoring of the underground steam pipe network.

[0025] In a possible implementation, the temperature loss vocabulary construction module 20 further includes an unstructured field acquisition unit, configured to acquire unstructured fields based on the temperature loss early warning historical data. Specifically, the temperature loss early warning historical data is comprehensively combed. The historical data contains both structured data, such as time, location, temperature loss degree, and explicit numerical information, and unstructured descriptive text content, such as word description in a fault report and notes in a maintenance record. The unstructured fields are screened from the historical data. The fields are usually in the form of natural language text and contain rich detailed information about temperature loss events, but lack explicit structure and format.

[0026] The text feature vector acquisition unit is configured to process the unstructured fields based on a natural language processing technology to obtain the text feature vector. Specifically, the natural language processing technology is used to process the unstructured fields. First, text cleaning is performed to remove irrelevant punctuation marks, special characters, and the like. Word embedding technologies such as Word2Vec and GloVe are used to convert words in the unstructured fields into vector representations. The vectors can capture semantic information of the words, and in the vector space, words with similar semantics will have similar vector representations. For sentence-level unstructured fields, the feature vector of the sentence can be obtained by combining and processing the word vectors in the sentence. For example, the word vectors in the sentence are combined into a fixed-length vector by using methods such as average pooling and maximum pooling, as the feature vector representation of the sentence.

[0027] The vector preprocessing unit is configured to preprocess the text feature vector. Specifically, the operation of removing noise words is performed. Common stop words such as “of”, “is”, and “in” have a high frequency of occurrence in the text but contribute less to the core meaning of the text. These words are removed from the text feature vector to reduce noise interference. The operation of unifying word forms is performed, including stem extraction and lemma reduction. The stem extraction is to remove affixes from a word to obtain the stem part of the word. The lemma reduction is to reduce the word to its basic form to ensure that different forms of the word can be correctly identified and processed. The operation of standardization is performed, such as logarithmic scaling, normalization, and the like. The logarithmic scaling can reduce the value of the feature vector to a certain range to avoid the influence of excessively large values on subsequent analysis. The normalization can normalize the value of the feature vector to a specific interval, such as [0, 1] or [-1, 1], so that different feature vectors are comparable.

[0028] a clustering analysis unit configured to perform clustering analysis on the preprocessed text feature vectors to obtain a plurality of groups of text feature vectors. Specifically, the preprocessed text feature vectors are grouped using a clustering analysis algorithm. During the clustering process, the text feature vectors are grouped according to semantic similarity, word frequency co-occurrence, and other information. Feature vectors with high semantic similarity are grouped into the same class. If two words are semantically similar, their corresponding feature vectors in the vector space will also be close in distance. If two words often appear in the same text context, they may have similar semantics and thus be grouped into the same class. Each class is assigned a preliminary label to facilitate subsequent identification and analysis of different vector groups. For example, a vector group can be named and labeled based on the frequently occurring words or topics in the group.

[0029] a vector grouping unit configured to establish a limited set of temperature loss word table based on the plurality of groups of text feature vectors. Specifically, a limited set of temperature loss word table is established based on the results of the plurality of groups of text feature vectors. The word table contains various temperature loss-related keywords and their corresponding feature vectors. The keywords, which have been preliminarily screened, represent common types of underground steam pipe network temperature loss and their performance characteristics. For example, the keywords include "pipe rupture", "temperature anomaly", "pressure fluctuation", and their corresponding feature vectors after clustering analysis. To enhance the accuracy of the word table, expert knowledge or existing literature can be used to proofread and correct the keywords in the word table. Experts can evaluate the accuracy and representativeness of the keywords based on their experience and professional knowledge and provide suggestions for modification. Related research results in existing literature can also be used as a reference to further improve the content of the word table.

[0030] The classification recognition model construction module 30 is configured to establish a self-supervised alarm type classification recognition model based on the limited set of temperature loss words, obtain an alarm type sequence, associate each alarm type in the alarm type sequence with multiple temperature loss causes, and evaluate the hazard level of each alarm type. Specifically, the content and structure of the limited set of temperature loss words are determined, including various temperature loss-related keywords and their corresponding feature vectors, representing common types of underground steam pipe network temperature loss and their characteristics, providing basic data for establishing a self-supervised alarm type classification recognition model. Further sorting and analysis of the limited set of temperature loss words is performed to understand the temperature loss implications and characteristics represented by different keywords, taking into account the frequency and importance of keywords in actual temperature loss events to prepare for subsequent model establishment and training. An appropriate machine learning algorithm and model architecture are selected to establish a self-supervised alarm type classification recognition model using a neural network model in deep learning. The input layer, hidden layer, and output layer of the model are designed based on the feature vectors in the temperature loss word table and the characteristics of the alarm types. The input layer receives the keyword feature vectors in the temperature loss word table, the hidden layer abstracts and extracts features through a multi-layer neural network structure, and the output layer outputs the classification results of the alarm types. During model establishment, a self-supervised learning method can be used, i.e., using unlabeled data for learning. For example, by randomly masking or replacing keywords in the temperature loss word table, the model can predict the masked or replaced parts, thereby learning the semantic relationships between keywords and the characteristics of alarm types. Actual temperature loss data can be input into the established self-supervised alarm type classification recognition model. The data can be real-time monitoring temperature loss data or part of historical temperature loss data. The model processes and analyzes the input data, outputs an alarm type sequence, and represents different alarm types at different time points or different key nodes, such as "temperature anomaly alarm," "pressure fluctuation alarm," "pipeline leakage alarm," etc. The alarm type sequence is formed in chronological order. Different alarm types are analyzed in depth to determine the possible multiple temperature loss causes for each alarm type. This is achieved through analysis of historical temperature loss data, introduction of expert knowledge, and understanding of the physical structure and operation principle of underground steam pipe networks. For example, if a "temperature anomaly alarm" occurs, possible temperature loss causes include pipeline insulation damage, abnormal steam flow, and external environmental temperature changes. If a "pressure fluctuation alarm" occurs, possible causes include pipeline blockage, valve failure, and unstable steam generator. Each alarm type is associated with multiple temperature loss causes to establish a mapping relationship table, allowing quick determination of possible temperature loss causes based on alarm types in actual applications. The hazard level of each alarm type is evaluated based on the severity of temperature loss, the impact on underground steam pipe network operation, and possible consequences, etc. A combination of quantitative and qualitative methods is used for evaluation.The quantitative evaluation can consider specific indexes such as a temperature change amplitude caused by the temperature damage, a pressure change value, a leakage amount, and the like; and the qualitative evaluation can consider influences of the temperature damage on safety, reliability, and economy of the pipe network, and divide a hazard level of the alarm type into different levels, such as high, medium, and low three levels. In actual application, a priority of processing can be determined according to the hazard level of the alarm type, and alarm types with high hazard levels are processed preferentially, so that influences of the temperature damage on the underground steam pipe network are reduced to the greatest extent.

[0031] In a possible implementation, the classification and identification model construction module 30 further includes a structured field acquisition unit configured to acquire a structured field based on the temperature damage early warning historical data. Specifically, the temperature damage early warning historical data is combed and analyzed. The historical data contains various information, including unstructured text description and structured data field. The structured field is extracted from the historical data. The field is usually data with clear format and meaning, such as temperature damage occurrence time, location, and specific numerical value of temperature damage degree. The structured field can be directly used for data analysis and model construction, and provides basic data for subsequent steps.

[0032] The initial self-supervised alarm type classification and identification model construction unit is configured to establish an initial self-supervised alarm type classification and identification model. Specifically, a suitable machine learning algorithm and model architecture are selected to construct an initial self-supervised alarm type classification and identification model. A neural network model in deep learning, such as a convolutional neural network (CNN), a recurrent neural network (RNN), or a long short-term memory network (LSTM), is considered. The input layer, hidden layer, and output layer of the model are designed according to the task requirements of the alarm type classification and identification. The input layer can receive the text comprehensive feature vector generated in the subsequent step. The hidden layer abstracts and extracts the features through a multi-layer neural network structure. The output layer outputs the classification result of the alarm type.

[0033] The text comprehensive feature vector acquisition unit is configured to acquire a text comprehensive feature vector based on the structured field and the text feature vector. Specifically, the structured field and the text feature vector acquired previously are fused. A plurality of methods are used for fusion, such as inputting the structured field as an additional feature into the model and processing the text feature vector through the neural network. The structured field and the text feature vector are spliced into a longer feature vector through feature splicing. The text comprehensive feature vector is obtained through the fusion operation. The feature vector contains clear information of the structured data and semantic features of the unstructured text, and can more comprehensively describe the characteristics of the temperature damage alarm, and provide more abundant information for the alarm type classification and identification.

[0034] The model training unit is configured to construct positive sample pairs and negative sample pairs based on the text comprehensive feature vector, input the initial self-supervised alarm type classification and recognition model, perform training, and obtain the self-supervised alarm type classification and recognition model. Specifically, the positive sample pairs and the negative sample pairs are constructed based on the text comprehensive feature vector. According to the method of self-supervised learning, virtual alarm pairs can be generated through data enhancement technology. For example, some temperature damage alarm data is subjected to data masking, rotation, cropping and other operations to generate a group of enhanced data of similar alarms, and the enhanced data is marked as a positive sample pair. An unrelated alarm pair, such as an alarm pair from different times, places or with different temperature damage characteristics, is marked as a negative sample pair. The positive sample pairs and the negative sample pairs are input into the initial self-supervised alarm type classification and recognition model for training. The model continuously adjusts parameters so that the similarity score of the positive sample pairs is higher than that of the negative sample pairs. In the training process, the model learns the inherent characteristics of the alarm type and gradually improves the classification accuracy of the alarm type. After multiple iterations of training, when the performance of the model on the validation set reaches a certain standard, the training is stopped, and the final self-supervised alarm type classification and recognition model is obtained. The model can be used for accurate alarm type classification and recognition of new temperature damage alarm data.

[0035] The temperature damage reason correlation graph construction module 40 is configured to acquire the alarm type sequence, the hazard level, and the plurality of temperature damage reasons, calculate alarm type similarity, and establish an alarm type temperature damage reason correlation graph. Specifically, the alarm type sequence, the hazard level of each alarm type, and the plurality of temperature damage reasons related to the alarm type are collected from the previous modules. The alarm type sequence is obtained by the classification recognition model construction module and records alarm types occurring at different time points or different key nodes. The hazard level is determined after evaluating the impact of each alarm type. The plurality of temperature damage reasons are obtained by analyzing the alarm type in depth, correlating historical data and professional knowledge, and determining possible reasons for the alarm type. The cosine similarity calculation method is used to evaluate the similarity between different alarm types. For example, for alarm types based on text features, the cosine similarity between text feature vectors can be used to determine their similarity. If the text descriptions of two alarm types are relatively close in semantics, their similarity will be higher. In addition to text features, other factors such as the time interval of alarm occurrence and the location relationship of key nodes can be combined to calculate the similarity of alarm types. If two alarm types occur close in time or involve adjacent key nodes, they may have some similarity. According to the calculated alarm type similarity and the obtained hazard level and plurality of temperature damage reasons, an alarm type temperature damage reason correlation graph is constructed. The correlation graph can display different alarm types, temperature damage reasons, and their relationships in a graphical manner. The nodes in the correlation graph can represent alarm types and temperature damage reasons. For alarm type nodes, information such as the name and hazard level can be labeled. For temperature damage reason nodes, specific temperature damage situations such as pipeline aging and temperature and pressure fluctuations can be described in detail. The edges represent the correlation between alarm types and temperature damage reasons. If an alarm type is highly likely to be associated with a temperature damage reason, they are connected by an edge in the correlation graph, and the edge can be assigned different weights according to the similarity to represent the strength of the correlation. By establishing the correlation graph, the relationship between different alarm types and temperature damage reasons can be intuitively observed, providing strong support for quickly determining temperature damage reasons and taking appropriate measures.

[0036] In a possible implementation manner, as Figure 2As shown, the temperature damage cause correlation graph construction module 40 further comprises a correlation graph structure acquisition unit, which is configured to acquire a preliminary correlation relationship between the alarm types and the temperature damage causes based on the alarm type sequence and the temperature damage causes, and establish a basic structure of the alarm type-temperature damage cause correlation graph based on the preliminary correlation relationship. Specifically, the alarm type sequence is analyzed to understand various alarm types occurring at different time points, the temperature damage causes are sorted out to determine various factors that can cause temperature damage, and the preliminary correlation relationship between the alarm types and the temperature damage causes is established through analysis of historical data, introduction of expert knowledge, and logical reasoning. For example, if a certain alarm type appears simultaneously with a specific temperature damage cause in historical data multiple times, it can be preliminarily considered that they are correlated. Based on the preliminary correlation relationship, the basic structure of the alarm type-temperature damage cause correlation graph is constructed, which can be represented in a graphical manner, where the alarm types and the temperature damage causes are nodes, and the preliminary correlation relationship is an edge connecting the nodes.

[0037] A weight adjustment unit is configured to dynamically adjust the weight value of the preliminary correlation relationship based on the alarm type similarity. Specifically, the similarity between different alarm types is calculated, such as using a cosine similarity algorithm to measure their similarity based on feature vectors and other information of the alarm types. The weight value of the preliminary correlation relationship is dynamically adjusted based on the alarm type similarity. If two alarm types have a high similarity and they are preliminarily correlated with the same temperature damage cause, the weight of the correlation relationship between the temperature damage cause and the two alarm types can be appropriately increased. Conversely, if the alarm type similarity is low, the weight of the corresponding correlation relationship can be reduced to make the correlation graph more accurately reflect the actual correlation strength between different alarm types and temperature damage causes.

[0038] A multi-level correlation relationship construction unit is configured to construct a multi-level correlation relationship based on the correlation type and the preliminary correlation relationship. Specifically, the preliminary correlation relationship is further analyzed to determine different correlation types, which can include direct correlation and indirect correlation. For example, certain alarm types can be directly caused by a specific temperature damage cause, which is direct correlation. Some alarm types can be associated with temperature damage causes through intermediate links, which is indirect correlation. Based on the correlation type and the preliminary correlation relationship, a multi-level correlation relationship is constructed. In the correlation graph, different levels of nodes and edges can be used to represent the multi-level correlation relationship. For example, direct correlation can be represented at a lower level, and indirect correlation can be represented at a higher level through the connection of intermediate nodes, thereby forming a more complex and comprehensive correlation network.

[0039] The temperature damage reason correlation graph construction unit is configured to establish the alarm type temperature damage reason correlation graph based on the multi-level correlation relationship. Specifically, the multi-level correlation relationship is constructed, and a complete alarm type temperature damage reason correlation graph is finally established. The correlation graph clearly shows different alarm types, temperature damage reasons, and complex correlation relationships therebetween. The correlation graph is visualized to enable a user to more intuitively understand and analyze the relationship between the alarm types and the temperature damage reasons. The correlation graph is continuously updated and optimized. With the addition of new data and knowledge, the accuracy and practicality of the correlation graph are further improved.

[0040] In a possible implementation, the correlation graph structure acquisition unit further includes a correlation graph composition unit configured to: the alarm type temperature damage reason correlation graph includes correlation graph nodes and correlation graph edges, the correlation graph nodes are the alarm types and the temperature damage reasons, and the correlation graph edges are the preliminary correlation relationships between the alarm types and the temperature damage reasons. Specifically, for the correlation graph nodes, the alarm types and the temperature damage reasons are determined. On one hand, various data in the operation process of the underground steam pipe network are analyzed, including historical alarm records and monitoring data, to identify different alarm types, such as a temperature anomaly alarm, which is manifested as temperature exceeding a normal range at a key node, and a pressure fluctuation alarm, which is manifested as unstable pressure in a pipeline. On the other hand, various reasons that may cause temperature damage are studied in depth, including analysis of pipe material properties, such as pipe aging, which may be caused by long-term use of the pipe material, consideration of external environmental factors, such as corrosion caused by soil humidity changes and the influence of geological activities on the pipe, and improper maintenance, such as nonstandard installation and untimely repair. The correlation graph edges represent the preliminary correlation relationships between the alarm types and the temperature damage reasons. The correlation relationships are determined in multiple ways. One is to mine historical data. If a certain alarm type frequently appears simultaneously with a specific temperature damage reason in past cases, it can be preliminarily considered that they are correlated. Two is to combine expert knowledge. Industry experts judge, based on experience, which temperature damage reasons may cause certain alarm types. Three is to perform logical reasoning to analyze the operation principle and structural characteristics of the underground steam pipe network and deduce the potential relationship between the alarm types and the temperature damage reasons. After the preliminary correlation relationship is determined, the correlation graph edges are used to connect the corresponding alarm type nodes and temperature damage reason nodes, thereby constructing a complete alarm type temperature damage reason correlation graph to more intuitively analyze and understand the complex relationship between the alarm types and the temperature damage reasons and provide strong support for temperature damage early warning and processing of the underground steam pipe network.

[0041] The warm damage reason optimal solution acquisition module 50 is used for analyzing the alarm information uploaded by a plurality of key nodes based on the self-supervised alarm type classification identification model, determining a plurality of alarm types, inputting the plurality of alarm types into the alarm type warm damage reason association graph, and acquiring a warm damage reason optimal solution. Specifically, the alarm information uploaded by a plurality of key nodes is acquired. The key node is a part that is prone to warm damage problems in the underground steam pipe network, such as a pipe connection, a bend, and the like, which is determined through analysis. The alarm information uploaded includes temperature abnormalities, pressure fluctuations, flow changes, and the like. The self-supervised alarm type classification identification model is used to analyze the alarm information. The model is trained and can determine the corresponding alarm type according to the alarm information features input, for example, the model can determine whether it is a temperature abnormality alarm, a pressure fluctuation alarm, or other types of alarms according to the temperature change amplitude, the pressure fluctuation frequency, and the like. Through the analysis of the model, a plurality of alarm types are determined to provide a basis for subsequent determination of the warm damage reason. Once the plurality of alarm types are determined, the alarm types are input into the alarm type warm damage reason association graph. The association graph is constructed through the previous steps and shows the relationship between different alarm types and warm damage reasons. The association graph analyzes and reasons through its internal logic and association relationship according to the input alarm types. For example, if the input alarm types include temperature abnormality alarms and pressure fluctuation alarms, the association graph will comprehensively analyze the alarm types according to the association strength between the two alarm types and different warm damage reasons. Through the analysis of the association graph, a warm damage reason optimal solution is finally acquired. The optimal solution is the most likely combination of warm damage reasons after considering a plurality of alarm types. For example, it can be concluded that the warm damage reason is the combined effect of pipe aging and external environmental influence. The warm damage reason optimal solution can provide accurate guidance for the pipe network management personnel to take targeted maintenance and prevention measures to reduce the impact of warm damage on the underground steam pipe network.

[0042] The warning module 60 is used for warning based on the warm damage reason optimal solution and the key node position. Specifically, the warning module first receives the warm damage reason optimal solution from the warm damage reason optimal solution acquisition module and determines the key node position where the warm damage occurs. Then, according to the warm damage reason optimal solution, the influence degree and the emergency degree that may be caused are analyzed, and the importance of the key node position is considered, a suitable warning mode such as an audible and light alarm, an SMS notification, a system pop-up window, and the like is selected, and accurate and complete warning information including the warm damage reason, the damage degree, the key node position, and the recommended measures is issued. After the warning is issued, the warm damage situation and the key node change are continuously monitored. If the situation deteriorates or new alarm information is obtained, the warning content is updated in time and the warning mode and range are adjusted to ensure the safe operation of the underground steam pipe network.

[0043] In a possible implementation, the early warning module 60 further comprises: a warm damage treatment decision tree construction unit, configured to establish a warm damage treatment decision tree based on historical early warning data and a preset treatment scheme. Specifically, historical early warning data of the underground steam pipe network is collected, and the data includes information such as the alarm type, the warm damage reason, the key node position, and the damage degree when a warm damage event occurred in the past. Through analysis of the historical data, the characteristics and treatment methods of different types of warm damage events can be understood, and the preset treatment scheme can be determined. The preset treatment scheme can be formulated by experts according to experience, or can be derived by summarizing historical successful treatment cases. The scheme includes maintenance methods, emergency measures, resource allocation, and the like for different warm damage reasons. Based on the historical early warning data and the preset treatment scheme, the warm damage treatment decision tree is established. Each node of the decision tree represents a decision point, and branches to different treatment schemes according to different conditions. For example, the root node of the decision tree can be the severity of the warm damage, which is divided into three branches of high, medium, and low according to the severity. Each branch is further subdivided according to other factors such as the warm damage reason and the key node position, and finally reaches a specific treatment scheme node.

[0044] A treatment scheme production unit is configured to input the entered early warning information into the warm damage treatment decision tree to generate a treatment scheme. Specifically, when new early warning information is generated, the early warning information is entered into the treatment scheme production unit. The early warning information can include the alarm type, the warm damage reason, the key node position, and the damage degree of the current warm damage event. The treatment scheme production unit inputs the entered early warning information into the warm damage treatment decision tree. The decision tree starts from the root node and makes judgments according to the branch conditions of the decision tree based on the characteristics of the early warning information, and iterates downward step by step until a specific treatment scheme node is reached, generating a treatment scheme. This treatment scheme is determined according to the judgment result of the decision tree, and includes specific maintenance measures, emergency actions, resource requirements, and the like.

[0045] The processing result data acquisition unit is configured to execute the processing scheme, acquire processing result data, and update the thermal damage processing decision tree according to the processing result data. Specifically, the processing scheme is executed. Relevant personnel take actions according to the generated processing scheme to carry out maintenance and processing of thermal damage. During the execution, the process and result data of the processing are recorded, including the actual measures taken, processing time, resource consumption, thermal damage repair situation, and the like. The processing result data is acquired. After the processing is completed, the processing result is evaluated and analyzed to acquire the processing result data. The processing result data can reflect the effectiveness and efficiency of the processing scheme. The thermal damage processing decision tree is updated according to the processing result data. The processing result data is fed back to the thermal damage processing decision tree construction unit. The decision tree is adjusted and optimized according to the actual processing result. If the processing scheme successfully solves the thermal damage problem, the weight of the scheme in the decision tree can be increased; if the processing scheme is not effective, the reasons can be analyzed, and the branch conditions of the decision tree and the processing scheme are adjusted to improve the accuracy and practicability of the decision tree.

[0046] The embodiment of the application adopts a method for determining a protection pipeline map of an underground steam pipe network, analyzing a key node position, and establishing a thermal damage data acquisition system; acquiring thermal damage early warning historical data, extracting a text feature vector and a structured field to establish a thermal damage word table; establishing an alarm type classification recognition model based on the thermal damage word table, acquiring an alarm type sequence, and associating a thermal damage cause and an evaluation hazard level; calculating an alarm type similarity to establish an association graph; analyzing key node alarm information to determine an alarm type, inputting the association graph to acquire a thermal damage cause optimal solution, and early warning based on the optimal solution and the key node position. The technical effects of improving the accuracy and reliability of thermal damage early warning of the underground steam pipe network, accurately discovering potential thermal damage problems, reducing false positives, and quickly locating thermal damage causes, and guaranteeing safe and stable operation of the pipe network are achieved by introducing unstructured data analysis, damage process time feature and transaction logic chain data filtering, and mutual verification of multi-node and multi-dimensional alarm prediction.

[0047] Although the present application makes various references to certain modules in the system according to the embodiments of the present application, however, any number of different modules can be used and run on the user terminal and / or server, and the various units and modules are only divided according to functional logic, but are not limited to the above division, as long as the corresponding functions can be realized; in addition, the specific names of the functional units are only for easy mutual differentiation, and do not limit the protection scope of the present application.

[0048] The foregoing DETAILED DESCRIPTION, including the above section titled "Detailed Description," is not to be taken as limiting the scope of the application. Various modifications, combinations, and equivalents can be apparent to those skilled in the art and can be made once the nature of the application is understood. Any modification, combination, or equivalent, which falls within the principles and the scope of the present application, is intended to be included in the present application. In some instances, the actions or steps can be performed in different order from those described herein, and still achieve desirable results. Additionally, the process depicted in the figures can not necessarily require the particular order shown or sequential order to achieve the desired results. In certain implementations, multitasking and parallel processing can be advantageous.

Claims

1. The underground steam pipe network intelligent temperature loss early warning monitoring device based on ad hoc network, characterized in that, The system comprises: A temperature loss data acquisition system establishment module, which is used to determine a protection pipeline map of an underground steam pipe network, analyze the key node positions of the protection pipeline map, and establish a temperature loss data acquisition system of the key nodes; A temperature loss vocabulary construction module, which is used to obtain temperature loss early warning historical data, obtain a text feature vector and a structured field based on the temperature loss early warning historical data, analyze the text feature vector and the structured field, and establish a limited set temperature loss vocabulary, wherein the text feature vector is obtained through an unstructured field; A classification and recognition model construction module, which is used to establish a self-supervised alarm type classification and recognition model based on the limited set temperature loss vocabulary, obtain an alarm type sequence, associate each alarm type in the alarm type sequence with a plurality of temperature loss causes, and evaluate the harm level of each alarm type; A temperature loss cause association graph construction module, which is used to obtain the alarm type sequence, the harm level, and the plurality of temperature loss causes, calculate an alarm type similarity, and establish an alarm type temperature loss cause association graph; A temperature loss cause optimal solution acquisition module, which is used to analyze a plurality of alarm information of the key nodes based on the self-supervised alarm type classification and recognition model, determine a plurality of alarm types, input the plurality of alarm types into the alarm type temperature loss cause association graph, and obtain a temperature loss cause optimal solution; An early warning module, which is used to perform early warning based on the temperature loss cause optimal solution and the key node positions; The temperature loss vocabulary construction module comprises: A data acquisition unit, which is used to acquire real-time data and historical data of the key nodes; A data filtering unit, which is used to obtain a damage process time feature of the key nodes based on the historical data of the key nodes, and filter the real-time data based on the damage process time feature; A transaction logic chain construction unit, which is used to construct a transaction logic chain of temperature loss damage of the underground steam pipe network, and filter the real-time data based on the transaction logic chain; A real-time data output unit, which is used to output the real-time data of the key nodes; The transaction logic chain construction unit comprises: A structure model construction subunit, which is used to construct a structure model of the underground steam pipe network; An event association subunit, which is used to obtain temperature loss damage features of the underground steam pipe network and corresponding key nodes and times, analyze the key nodes and times to associate the temperature loss damage features with actual temperature loss events, and construct a plurality of stages of the transaction logic chain; A logic chain adjustment subunit, which is used to dynamically adjust the plurality of stages of the transaction logic chain based on the real-time data and the historical data; The temperature loss vocabulary construction module comprises: An unstructured field acquisition unit is configured to acquire unstructured fields based on the temperature damage early warning historical data; A text feature vector acquisition unit is configured to process the unstructured fields based on natural language processing technology to obtain the text feature vector; A vector preprocessing unit is configured to preprocess the text feature vector; A clustering analysis unit is configured to perform clustering analysis on the preprocessed text feature vector to obtain a plurality of groups of text feature vector groupings; A vector grouping unit is configured to establish a limited set of temperature damage word tables based on the plurality of groups of text feature vector groupings.

2. The self-organizing network-based underground steam pipe network intelligent temperature damage early warning monitoring device according to claim 1, wherein, The classification and recognition model construction module includes: A structured field acquisition unit is configured to acquire structured fields based on the temperature damage early warning historical data; An initial self-supervised alarm type classification and recognition model construction unit is configured to establish an initial self-supervised alarm type classification and recognition model; A text comprehensive feature vector acquisition unit is configured to acquire a text comprehensive feature vector based on the structured fields and the text feature vector; A model training unit is configured to construct positive sample pairs and negative sample pairs based on the text comprehensive feature vector, input the initial self-supervised alarm type classification and recognition model, and perform training to obtain the self-supervised alarm type classification and recognition model. 3.The self-organizing network based intelligent temperature damage early warning and monitoring device for underground steam pipe network according to claim 1, characterized in that, The temperature damage cause association graph construction module includes: An association graph structure acquisition unit is configured to acquire a preliminary association relationship between the alarm types and the temperature damage causes based on the alarm type sequence and the temperature damage causes, and establish a basic structure of an alarm type temperature damage cause association graph based on the preliminary association relationship; A weight adjustment unit is configured to dynamically adjust the weight value of the preliminary association relationship based on the alarm type similarity; A multi-level association relationship construction unit is configured to construct a multi-level association relationship based on the association type and the preliminary association relationship; A temperature damage cause association graph construction unit is configured to establish the alarm type temperature damage cause association graph based on the multi-level association relationship.

4. The self-organizing network-based underground steam pipe network intelligent temperature damage early warning monitoring device according to claim 3, characterized in that, The alarm type temperature damage cause association graph includes association graph nodes and association graph edges, the association graph nodes are the alarm types and the temperature damage causes, and the association graph edges are the preliminary association relationships between the alarm types and the temperature damage causes. 5.The self-organizing network based intelligent temperature damage early warning and monitoring device for underground steam pipe network according to claim 1, wherein, The early warning module includes: A temperature damage handling decision tree construction unit is configured to establish a temperature damage handling decision tree based on historical early warning data and a preset handling scheme; A handling scheme production unit is configured to input the entered early warning information into the temperature damage handling decision tree to generate a handling scheme; A processing result data acquisition unit is configured to execute the processing scheme, acquire processing result data, and update the warm-deformation processing decision tree according to the processing result data.

Images