Training method and device for urban underground gas leak identification model

Abstract

The present disclosure proposes a training method, device, electronic device and storage medium for an urban underground gas leak identification model, wherein the method comprises: based on an acquired first methane concentration sequence to be labeled, determining a target segment methane concentration sequence corresponding to an abnormal change in methane concentration, extracting features of the target segment methane concentration sequence, obtaining methane concentration change features, matching the corresponding target real methane concentration sequence from a real gas leak case library, and determining label data of the target segment methane concentration sequence, adding the target segment methane concentration sequence and the corresponding label data to the real gas leak case library and training the gas leak identification model, thereby training the gas leak identification model based on the target segment real methane concentration sequence and the corresponding label data, improving the recognition accuracy of gas leaks, expanding the labels of the real gas leak case library, and reducing the cost of manual labeling.

Description

Technical Field

[0001] The present disclosure relates to the fields of artificial intelligence Internet of Things and gas safety, and in particular to a training method, device, electronic equipment and storage medium for an urban underground gas leak identification model. Background Art

[0002] At present, underground gas pipelines are still an important part of cities, and intelligent monitoring of urban gas is crucial. In related technologies, gas leaks are identified by analyzing the correlation between methane concentration and temperature, for example, the correlation coefficient between methane concentration and temperature, but this method will result in a large number of false positives and false negatives. Summary of the invention

[0003] The purpose of the present disclosure is to solve one of the above technical problems at least to some extent.

[0004] To this end, the first purpose of the present invention is to propose a training method for an urban underground gas leak identification model. Based on the acquired first methane concentration sequence to be labeled, the target segment methane concentration sequence corresponding to the abnormal change of methane concentration is determined, so as to extract features of the target segment methane concentration sequence and obtain the methane concentration change features. Then, the corresponding target real methane concentration sequence is matched from the real gas leakage case library, and the label data of the target segment methane concentration sequence is determined, so as to add the target segment methane concentration sequence and the corresponding label data to the real gas leakage case library and train the gas leakage identification model. Thus, based on the target segment real methane concentration sequence and the corresponding label data, the gas leakage identification model is trained, the recognition accuracy of gas leakage is improved, the labels of the real gas leakage case library are expanded, and the cost of manual labeling is reduced.

[0005] The second objective of the present disclosure is to provide a training device for an urban underground gas leak identification model.

[0006] A third objective of the present disclosure is to provide an electronic device.

[0007] A fourth objective of the present disclosure is to provide a computer-readable storage medium.

[0008] A fifth object of the present disclosure is to provide a computer program product.

[0009] To achieve the above-mentioned purpose, the first aspect of the present disclosure proposes a training method for an urban underground gas leak identification model, comprising: obtaining a first methane concentration sequence to be labeled from a time series database; determining a target segment methane concentration sequence corresponding to an abnormal change in methane concentration from the first methane concentration sequence; performing feature extraction on the target segment methane concentration sequence to obtain a methane concentration change feature; obtaining a target real methane concentration sequence matching the methane concentration change feature from a real gas leak case library; using the gas leak label corresponding to the target real methane concentration sequence as label data of the target segment methane concentration sequence; adding the target segment methane concentration sequence and the corresponding label data to the real gas leak case library to obtain an updated real gas leak case library; and training the gas leak identification model according to each methane concentration sequence and the corresponding label data in the updated real gas leak case library.

[0010] The training method of the urban underground gas leakage identification model of the embodiment of the present disclosure determines the target segment methane concentration sequence corresponding to the abnormal change of methane concentration based on the acquired first methane concentration sequence to be labeled, so as to extract features of the target segment methane concentration sequence and obtain the methane concentration change features, and then matches the corresponding target real methane concentration sequence from the real gas leakage case library, and determines the label data of the target segment methane concentration sequence, so as to add the target segment methane concentration sequence and the corresponding label data to the real gas leakage case library and train the gas leakage identification model. Thus, the gas leakage identification model is trained based on the target segment real methane concentration sequence and the corresponding label data, thereby improving the recognition accuracy of gas leakage, expanding the labels of the real gas leakage case library, and reducing the cost of manual labeling.

[0011] To achieve the above-mentioned purpose, the second embodiment of the present disclosure proposes a training device for an urban underground gas leak identification model, comprising: a first acquisition module, used to acquire a first methane concentration sequence to be labeled from a time series database; a determination module, used to determine the target segment methane concentration sequence corresponding to the abnormal change of methane concentration from the first methane concentration sequence; an extraction module, used to extract features from the target segment methane concentration sequence to obtain the methane concentration change feature; a second acquisition module, used to acquire a target real methane concentration sequence matching the methane concentration change feature from a real gas leak case library; a generation module, used to use the gas leak label corresponding to the target real methane concentration sequence as the label data of the target segment methane concentration sequence; an adding module, used to add the target segment methane concentration sequence and the corresponding label data to the real gas leak case library to obtain an updated real gas leak case library; a first training module, used to train the gas leak identification model according to each methane concentration sequence and the corresponding label data in the updated real gas leak case library.

[0012] The training device of the urban underground gas leakage identification model of the embodiment of the present disclosure determines the target segment methane concentration sequence corresponding to the abnormal change of methane concentration based on the acquired first methane concentration sequence to be labeled, so as to extract features of the target segment methane concentration sequence and obtain the methane concentration change features, and then matches the corresponding target real methane concentration sequence from the real gas leakage case library, and determines the label data of the target segment methane concentration sequence, so as to add the target segment methane concentration sequence and the corresponding label data to the real gas leakage case library and train the gas leakage identification model. Thus, the gas leakage identification model is trained based on the target segment real methane concentration sequence and the corresponding label data, thereby improving the recognition accuracy of gas leakage, expanding the labels of the real gas leakage case library, and reducing the cost of manual labeling.

[0013] To achieve the above-mentioned purpose, the third aspect embodiment of the present disclosure proposes an electronic device, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that when the processor executes the program, the training method of the urban underground gas leak identification model as described above is implemented.

[0014] To achieve the above-mentioned purpose, the fourth aspect of the present disclosure proposes a computer-readable storage medium on which a computer program is stored, and when the program is executed by a processor, the training method of the urban underground gas leak identification model as described above is implemented.

[0015] The fifth aspect of the present disclosure provides a computer program product, including a computer program, which, when executed by a processor, implements the training method of the urban underground gas leak identification model as described above.

[0016] Additional aspects and advantages of the present disclosure will be given in part in the following description and in part will be obvious from the following description or learned through practice of the present disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] The above and / or additional aspects and advantages of the present disclosure will become apparent and easily understood from the following description of the embodiments in conjunction with the accompanying drawings, in which:

[0018] Figure 1 A flowchart of a method for training a city underground gas leakage identification model according to an embodiment of the present disclosure;

[0019] Figure 2 A technical flow chart of gas monitoring data stream access according to an embodiment of the present disclosure;

[0020] Figure 3 This is an example diagram of a unit fragment of abnormal changes in methane concentration according to an embodiment of the present disclosure;

[0021] Figure 4 This is an example diagram of a real gas leak slice according to an embodiment of the present disclosure;

[0022] Figure 5 A flowchart of urban underground gas leakage analysis and identification based on AIoT technology according to an embodiment of the present disclosure;

[0023] Figure 6 A flowchart of a method for training a city underground gas leakage identification model according to another embodiment of the present disclosure;

[0024] Figure 7 A schematic diagram of a flow chart of a training device for a city underground gas leakage identification model according to an embodiment of the present disclosure;

[0025] Figure 8 The figure is a schematic diagram of the structure of an electronic device according to an embodiment of the present disclosure. DETAILED DESCRIPTION

[0026] Embodiments of the present disclosure are described in detail below, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present disclosure, and should not be construed as limiting the present disclosure.

[0027] The following describes the training method, device, electronic device and storage medium of the urban underground gas leak identification model of the embodiment of the present disclosure with reference to the accompanying drawings.

[0028] Figure 1 FIG. 1 is a flow chart of a method for training a city underground gas leak identification model according to an embodiment of the present disclosure. Figure 1 As shown, it mainly includes the following steps:

[0029] Step 101: Obtain a first methane concentration sequence to be annotated from a time series database.

[0030] Among them, the time series database can store gas-related data collected by the intelligent hardware devices of the city’s underground gas pipeline network.

[0031] As an exemplary implementation, intelligent hardware equipment can be installed in a manhole near a gas pipe section in a city's underground gas network, and the gas conditions around the gas pipe section can be monitored by the intelligent hardware equipment, and the gas-related data can be sent to the gas big data basic platform, which saves the received gas-related data in the time series database TDengine.

[0032] The gas-related data may include gas dynamic monitoring data and gas static data, wherein the gas dynamic monitoring data may include time, methane concentration, temperature, humidity, and equipment status but is not limited thereto. The gas static data may include manhole number, manhole type, manhole address, equipment number, and installation date but is not limited thereto.

[0033] Specifically, Figure 2 As shown, the local .xls file, local data CSV file and local Kafaka data in the gas dynamic monitoring data can be input into the high-throughput distributed publish-subscribe messaging system Kafka cluster for processing, and the gas dynamic monitoring data can be converted into data of the same data type and transmitted to the real-time computing framework Flink for calculation, so as to be stored in the relational database Mysql and the time series database. At the same time, the historical data of the time series database TDengine in the gas static monitoring data can be input into the offline computing framework flink for processing, so as to transmit and store the processed gas static data in the time series database, and the historical data of TDengine can also be combined with the remote dictionary service Redis, and the historical data of TDengine can be calculated through the real-time computing framework Flink to be stored in the relational database Mysql.

[0034] Step 102: Determine, from the first methane concentration sequence, a target segment methane concentration sequence corresponding to an abnormal change in methane concentration.

[0035] In the embodiment of the present disclosure, the first methane concentration sequence is segmented according to the methane concentration range corresponding to the preset risk level to obtain multiple methane concentration sequences, and combined with the risk level change of each methane concentration sequence segment, it is determined whether the methane concentration has changed, and when it is determined that the methane concentration has changed, the target segment methane concentration sequence corresponding to the abnormal methane change can be determined.

[0036] Among them, the alarm level corresponding to the methane concentration reaching a certain risk level, for example, the sequence of methane concentration below 1% can be set to no alarm, the sequence of methane concentration between 1%-4% can be set to third-level alarm, the sequence of methane concentration between 4%-10% can be set to second-level alarm, and the sequence of methane concentration above 10% can be set to first-level alarm. Specifically, the first methane concentration value can be replaced according to the intervals of [0-1%], [1-4%], [4-10%], [10%--], and replaced with different risk levels such as 0, 1, 2, 3, etc. in turn, and the encoded data column is obtained, so as to segment different encoded data columns according to different risk levels to obtain sub-segments with risk levels of 0, 1, 2, 3 respectively.

[0037] Specifically, for sub-segments of 0, 1, 2, and 3 of different line division levels, sub-segments of different levels are divided respectively. For example, the process of dividing sub-segments can be to first determine whether there are burrs in each sub-segment. If the number of burrs is less than a threshold, these burrs are classified into risk levels corresponding to most sampling points. If the number of burrs is greater than the threshold, the risk level corresponding to the burrs is used as an independent sub-segment. For example, the threshold for the number of burrs can be 5, but is not limited to this.

[0038] Among them, based on the fact that most of the sampling points have methane concentration values ​​at the same risk level, a small number of sampling points with values ​​that do not belong to this risk level can be determined as burrs.

[0039] It can be understood that after determining the independent sub-slice corresponding to the burr, each sub-slice is divided. Specifically, the values ​​at different risk levels are divided according to the risk level, and the values ​​at the same risk level are divided according to the time threshold. If the time interval between adjacent sampling points is greater than the time threshold, they are divided into different sub-slices, otherwise they are processed as one sub-slice, where the time threshold can be 1 day, but is not limited to this.

[0040] In summary, by dividing the first methane concentration sequence into multiple segments and performing different risk levels on the methane concentration sequences, and determining the sub-slices corresponding to different risk levels, the end time of the first candidate segment methane concentration sequence corresponding to the highest risk level, and the start time of the second candidate segment methane concentration sequence with a risk level of zero and a start time before the start time of the first target segment methane concentration sequence and closest to the start time of the first candidate segment methane concentration sequence, an example diagram of the unit segment of abnormal methane concentration change is obtained, as shown in FIG. Figure 3 As shown, a segment of the unit segment with abnormal methane concentration changes can be used as the target segment methane concentration sequence. Figure 3 The methane concentration sequence corresponding to the medium risk is used as the target segment methane concentration sequence, but is not limited to this.

[0041] Step 103: extracting features of the target segment methane concentration sequence to obtain methane concentration variation features.

[0042] Step 104: Obtain a target real methane concentration sequence that matches the methane concentration variation characteristics from a real gas leakage case library.

[0043] In the embodiment of the present disclosure, any real methane concentration sequence in the real gas leakage case library can be obtained by obtaining a second methane concentration different from the first methane concentration sequence from a time series database, and segmenting the second methane concentration sequence according to the methane concentration interval corresponding to the preset risk level to obtain multiple methane concentration sequences. For the multiple methane concentration sequences, the time point of the gas leakage confirmed by the maintenance personnel for the methane concentration sequence is obtained, and the time point is used as the end time corresponding to the abnormal change in methane concentration, and from the multiple methane concentration sequences, a third candidate methane concentration sequence with a risk level of zero, located before the end time, and the start time closest to the end time is obtained, and each methane concentration sequence between the start time and the end time of the third candidate methane sequence in the multiple methane concentration sequences is used as the real methane concentration sequence.

[0044] Among them, the time points confirmed by maintenance personnel for the methane concentration sequence, such as Figure 4 As shown, Figure 4 The scattered points in the figure are the methane concentration values ​​corresponding to different time points, and the vertical lines are the time points when the maintenance personnel confirmed the gas leak.

[0045] In other embodiments of the present disclosure, in order to improve the efficiency of obtaining a target real methane concentration sequence that matches the methane concentration change characteristics, for any real methane concentration sequence in a real gas leakage case library, feature extraction can be performed on the real methane concentration sequence, and the extracted methane concentration change characteristics can be saved in a feature library corresponding to the real gas leakage case library.

[0046] In some embodiments, the feature library may also include business features, device personalization indicators, and corresponding basic features and coding features corresponding to the real methane concentration sequence, so as to more accurately match the target real methane concentration sequence with the same methane concentration change characteristics.

[0047] In some embodiments, the first methane concentration sequence may be combined to determine basic features corresponding to the first methane concentration sequence, wherein the basic features corresponding to the first methane concentration sequence include the maximum value, average value and quantile of the methane concentration.

[0048] In some implementations, the time series database also stores temperature sequences and humidity sequences corresponding to the actual methane concentration sequences.

[0049] In order to facilitate processing based on the features corresponding to the temperature sequence and the humidity sequence, the feature library may also store the basic features corresponding to the temperature sequence and the humidity sequence.

[0050] Among them, the basic characteristics of the temperature series may include: maximum temperature, mean value, and quantile, but are not limited to these.

[0051] The basic characteristics of humidity series may include: humidity maximum value, mean value, and quantile, but are not limited to these.

[0052] In some embodiments, the above-mentioned real methane concentration sequence is composed of each methane concentration sequence between the start time and the end time of the third candidate methane sequence. Feature extraction of each methane concentration sequence constituting the real methane concentration sequence can also be performed to obtain the encoding features of the real methane concentration sequence.

[0053] Among them, the coding features of the real methane concentration sequence may include the number of sub-slice segments, segment risk level, minimum segment duration, maximum segment duration, concentration level corresponding to the minimum segment duration, concentration level corresponding to the maximum segment duration, maximum concentration level, segment volatility and trend, whether there is a pulse, and the degree of pulse, but are not limited to these.

[0054] The feature library may also store business features and device personalized indicators to facilitate subsequent query and use of business features, device personalized indicators and other information.

[0055] Among them, business characteristics may include periodicity, for example, methane concentration levels during the day and at night, methane concentration levels in the morning, noon, evening and other time periods, and methane concentration levels in different months.

[0056] Equipment personalized indicators may include, but are not limited to, the number of gas leaks that have occurred in different manholes in history, the number of biogas development processes, and the number of abnormal changes in methane concentrations in different manholes.

[0057] Step 105: Use the gas leakage label corresponding to the target real methane concentration sequence as label data of the target segment methane concentration sequence.

[0058] The gas leakage label corresponding to the target real methane concentration sequence may be manually labeled on the target real methane concentration sequence.

[0059] Among them, leaks in the target real methane concentration series can also be manually marked, for example Figure 4 As shown,

[0060] Step 106: Add the target segment methane concentration sequence and the corresponding label data to the real gas leakage case library to obtain an updated real gas leakage case library.

[0061] In this embodiment, corresponding to the target segment methane concentration sequence to be labeled, the label data corresponding to the target segment methane concentration sequence is accurately determined in combination with the gas leakage label corresponding to the target real methane concentration sequence that matches the target segment methane concentration sequence in the real gas leakage case library. Therefore, there is no need to manually label the target segment methane concentration sequence to be labeled, and the target segment methane concentration sequence and the corresponding label data can be accurately determined. This achieves the expansion of the real gas leakage case library while reducing the cost of manual labeling.

[0062] Step 107: training the gas leak recognition model according to each methane concentration sequence and corresponding label data in the updated real gas leak case library.

[0063] The present disclosure proposes a training method for an urban underground gas leakage identification model. Based on an acquired first methane concentration sequence to be labeled, a target segment methane concentration sequence corresponding to an abnormal change in methane concentration is determined, so as to extract features of the target segment methane concentration sequence and obtain methane concentration change features. Then, the corresponding target real methane concentration sequence is matched from a real gas leakage case library, and label data of the target segment methane concentration sequence is determined, so as to add the target segment methane concentration sequence and the corresponding label data to the real gas leakage case library and train the gas leakage identification model. Thus, the gas leakage identification model is trained based on the target segment real methane concentration sequence and the corresponding label data, thereby improving the recognition accuracy of gas leakage, expanding the labels of the real gas leakage case library, and reducing the cost of manual labeling.

[0064] Based on the above embodiments, in order to further improve the efficiency of obtaining a gas leakage model that meets the requirements, before the real gas leakage case library is expanded, the gas leakage recognition model can also be trained in combination with the real gas leakage case library that has been manually annotated in the real gas leakage case library. That is to say, before the gas leakage recognition model is trained according to each methane concentration sequence and the corresponding label data in the updated real gas leakage case library, the gas leakage recognition model can also be trained according to each methane concentration sequence and the corresponding label data in the real gas leakage case library.

[0065] In the disclosed embodiment, in order to accurately identify urban underground gas leaks, after the gas leak identification model is trained, for each monitoring point in the urban underground, the methane concentration sequence monitored by the monitoring point can be input into the gas leak identification model for identification, and whether the monitoring point is in a gas leak state can be determined based on the identification result. In other words, whether a gas leak occurs at the monitoring point can be determined based on the identification result.

[0066] In some embodiments, the training method for the urban underground gas leak recognition model based on artificial intelligence & Internet of Things (AIoT) can be as follows: Figure 5 As shown below, combined Figure 5 An example description of the training process is as follows:

[0067] Specifically, the data source of intelligent hardware equipment monitoring in the manholes near the gas pipe sections in the urban underground gas pipeline network can be obtained from the gas big data basic platform, and the time series data and label data in the data source can be transmitted to the kafka cluster for data type conversion, and then the data of the same data type is input into the offline computing framework flink for processing, and then stored in the time series database TDengine, and then the historical time series data and historical label data are obtained from the time series database TDengine. According to the risk level corresponding to the methane concentration, the historical time series data and historical label data are encoded and segmented according to the risk level to determine the corresponding event library and case library. Based on deep learning, feature extraction is performed on the quantile, mean and other features of the historical time series data to determine the methane concentration change characteristics corresponding to the historical time series data. Based on the machine learning classification algorithm, the historical annotated data with the same methane concentration change characteristics as the historical time series data is matched, and the label data corresponding to the historical annotated data is obtained, and the label data is used as the pseudo-label data corresponding to the historical time series data to add the label data to the case library. Finally, based on the semi-supervised learning algorithm, the historical annotated data and the corresponding label data are used to train the gas leak recognition model to accurately identify whether the historical time series data is leaked, increase the pseudo-labels of the case library, and reduce the cost of manual labeling.

[0068] In order to accurately determine the target segment methane concentration sequence corresponding to the abnormal change of methane concentration, such as Figure 6 As shown, Figure 6 FIG. 4 is a flow chart of a method for training a city underground gas leakage identification model according to another embodiment of the present disclosure. Figure 6 As shown, the method may also include the following steps:

[0069] Step 601: Obtain a first methane concentration sequence to be annotated from a time series database.

[0070] Step 602: segment the first methane concentration sequence according to the methane concentration intervals corresponding to the preset risk levels to obtain multiple methane concentration sequences.

[0071] In the embodiments of the present disclosure, since the methane concentration changes from a safe state to a dangerous state and the duration is not exactly the same, it is necessary to identify the methane concentration development intervals in which the methane concentration has an upward trend or level change. The methane concentration intervals can be segmented according to the preset risk level of the gas alarm service, but are not limited to this.

[0072] The preset risk level may be adjusted according to actual business conditions, and this embodiment does not specifically limit this.

[0073] Step 603 : For the multiple methane concentration sequences, a first candidate methane concentration sequence corresponding to the highest risk level is obtained, and an end time of a first target methane concentration sequence is obtained.

[0074] In the embodiment of the present disclosure, the highest risk level may be medium risk, but is not limited thereto.

[0075] In the embodiment of the present disclosure, the end time of the first target segment methane concentration sequence may be the time when the highest risk level decreases, but is not limited thereto.

[0076] Step 604 , obtaining, from the multiple methane concentration sequences, a second candidate methane concentration sequence with a risk level of zero, a start time before the start time of the first target methane concentration sequence, and a start time closest to the first candidate methane concentration sequence.

[0077] In the embodiment of the present disclosure, when the risk level is zero, it may be a risk level when the methane concentration is zero, but it is not limited thereto.

[0078] Step 605: each methane concentration sequence between the start time of the second candidate methane concentration sequence and the end time of the first candidate methane concentration sequence in the multiple methane concentration sequences is used as a target methane concentration sequence corresponding to the abnormal change in methane concentration.

[0079] In the embodiment of the present disclosure, each methane concentration sequence between the start time of the second candidate methane concentration sequence and the end time of the first candidate methane concentration sequence may include a methane concentration sequence corresponding to each of no risk, low risk, and medium risk. That is, based on a methane concentration sequence corresponding to each of no risk, low risk, and medium risk, a target methane concentration sequence corresponding to an abnormal change in methane concentration is formed.

[0080] Step 606: extract features of the target segment methane concentration sequence to obtain methane concentration variation features.

[0081] It should be noted that, for the specific implementation of steps 605 to 606, reference may be made to the relevant description in the above embodiments.

[0082] Step 607: Obtain a target real methane concentration sequence that matches the methane concentration variation feature from a real gas leakage case library.

[0083] In an embodiment of the present disclosure, an implementation method for obtaining a target real methane concentration sequence that matches the methane concentration change characteristics from a real gas leakage case library may be that, for any real methane concentration sequence in the real gas leakage case library, the methane concentration change characteristics corresponding to the real methane concentration sequence are matched with the target methane concentration change characteristics corresponding to the target segment methane concentration sequence. If the degree of match between the methane concentration change characteristics corresponding to the real methane concentration sequence and the target methane concentration change characteristics corresponding to the target segment methane concentration sequence is greater than a preset matching degree threshold, it is determined that the real methane concentration sequence and the target segment methane concentration sequence have the same methane concentration change law, and the real methane concentration sequence is used as the target real methane concentration sequence that matches the methane concentration change characteristics.

[0084] If the degree of match between the methane concentration change characteristics corresponding to the actual methane concentration sequence and the target methane concentration change characteristics corresponding to the target segment methane concentration sequence is less than a preset matching threshold, further judgment can be made manually.

[0085] Step 608: Use the gas leakage label corresponding to the target real methane concentration sequence as label data of the target segment methane concentration sequence.

[0086] Step 609: Add the target segment methane concentration sequence and the corresponding label data to the real gas leakage case library to obtain an updated real gas leakage case library.

[0087] Step 610: training a gas leak recognition model according to each methane concentration sequence and corresponding label data in the updated real gas leak case library.

[0088] The present disclosure proposes a training method for an urban underground gas leakage identification model. Based on a first methane concentration sequence to be labeled, the first methane concentration sequence is segmented according to a methane concentration interval corresponding to a preset risk level to obtain multiple methane concentration sequences, and a first candidate segment methane concentration sequence corresponding to the highest risk level is obtained, and an end time of a first target segment methane concentration sequence is obtained, and a second candidate segment methane concentration sequence with a risk level of zero and a start time before the start time of the first target segment methane concentration sequence and closest to the start time of the first candidate segment methane concentration sequence is obtained, so as to separate each segment between the start time of the second candidate segment methane concentration sequence and the end time of the first candidate segment methane concentration sequence. The methane concentration sequence is used as the target segment methane concentration sequence corresponding to the abnormal change of methane concentration, so as to extract features of the target segment methane concentration sequence and obtain the methane concentration change features, and then match the corresponding target real methane concentration sequence from the real gas leakage case library, and determine the label data of the target segment methane concentration sequence, so as to add the target segment methane concentration sequence and the corresponding label data to the real gas leakage case library and train the gas leakage recognition model. Thus, based on the target segment real methane concentration sequence and the corresponding label data, the gas leakage recognition model is trained, and the methane concentration sequence is split into fine granularity to achieve accurate determination of the target segment real methane concentration sequence, thereby improving the recognition accuracy of gas leakage.

[0089] Figure 7 FIG. 1 is a schematic diagram of a training device for a city underground gas leak identification model according to an embodiment of the present disclosure. Figure 7 As shown, the training device 700 for the urban underground gas leakage identification model includes: a first acquisition module 701, a determination module 702, an extraction module 703, a second acquisition module 704, a generation module 705, an addition module 706 and a first training module 707.

[0090] The first acquisition module 701 is used to acquire a first methane concentration sequence to be labeled from a time series database.

[0091] The determination module 702 is used to determine, from the first methane concentration sequence, a target segment methane concentration sequence corresponding to the abnormal change in methane concentration.

[0092] The extraction module 703 is used to extract features of the target segment methane concentration sequence to obtain methane concentration variation features.

[0093] The second acquisition module 704 is used to acquire a target real methane concentration sequence matching the methane concentration variation characteristics from a real gas leakage case library.

[0094] The generating module 705 is used to use the gas leakage label corresponding to the target real methane concentration sequence as the label data of the target segment methane concentration sequence.

[0095] The adding module 706 is used to add the target segment methane concentration sequence and the corresponding label data to the real gas leakage case library to obtain an updated real gas leakage case library.

[0096] The first training module 707 is used to train the gas leakage recognition model according to each methane concentration sequence and the corresponding label data in the updated real gas leakage case library.

[0097] As a possible implementation of the embodiment of the present disclosure, the device further includes:

[0098] The second training module is used to train the gas leak recognition model according to the methane concentration sequences and corresponding label data in the real gas leak case library.

[0099] As a possible implementation of the embodiment of the present disclosure, the determination module 702 is specifically configured to:

[0100] The methane concentration sequence is segmented according to the methane concentration range corresponding to the preset risk level to obtain multiple methane concentration sequences.

[0101] For multiple methane concentration sequences, the first candidate methane concentration sequence corresponding to the highest risk level is obtained, and the end time of the first target methane concentration sequence is obtained.

[0102] From the multiple methane concentration sequences, a second candidate methane concentration sequence is obtained, which has a risk level of zero, a start time before the start time of the first target methane concentration sequence, and is closest to the start time of the first candidate methane concentration sequence.

[0103] Each methane concentration sequence between the start time of the second candidate methane concentration sequence and the end time of the first candidate methane concentration sequence in the multiple methane concentration sequences is used as the target methane concentration sequence corresponding to the abnormal change of methane concentration.

[0104] As a possible implementation of the embodiment of the present disclosure, any target real methane concentration sequence in the real gas leakage case library is obtained by the following method:

[0105] A second methane concentration sequence different from the first methane concentration sequence is obtained from the time series database.

[0106] The second methane concentration sequence is segmented according to the methane concentration range corresponding to the preset risk level to obtain multiple methane concentration sequences.

[0107] For multiple methane concentration sequences, the time point of the gas leakage confirmed by the maintenance personnel for the methane concentration sequence is obtained, and the time point is used as the end time corresponding to the abnormal change of the methane concentration.

[0108] From the multiple methane concentration sequences, a third candidate methane concentration sequence with a risk level of zero, located before the end time, and with a start time closest to the end time is obtained.

[0109] Each methane concentration sequence between the start time and the end time of the third candidate methane concentration sequence in the multiple methane concentration sequences is used as the target true methane concentration sequence.

[0110] As a possible implementation of the embodiment of the present disclosure, the second acquisition module 704 is specifically configured to:

[0111] For any real methane concentration sequence in the real gas leakage case library, the methane concentration change characteristics corresponding to the real methane concentration sequence are matched with the target methane concentration change characteristics corresponding to the target segment methane concentration sequence.

[0112] If the degree of match between the methane concentration change characteristics corresponding to the real methane concentration sequence and the target methane concentration change characteristics corresponding to the target segment methane concentration sequence is greater than a preset matching degree threshold, it is determined that the methane concentration change law of the real methane concentration sequence and the target segment methane concentration sequence is the same, and the real methane concentration sequence is used as the target real methane concentration sequence that matches the methane concentration change characteristics.

[0113] The present disclosure proposes a training device for an urban underground gas leakage identification model. Based on an acquired first methane concentration sequence to be labeled, a target segment methane concentration sequence corresponding to an abnormal change in methane concentration is determined, so as to extract features of the target segment methane concentration sequence and obtain methane concentration change features. Then, the corresponding target real methane concentration sequence is matched from a real gas leakage case library, and label data of the target segment methane concentration sequence is determined, so as to add the target segment methane concentration sequence and the corresponding label data to the real gas leakage case library and train the gas leakage identification model. Thus, the gas leakage identification model is trained based on the target segment real methane concentration sequence and the corresponding label data, thereby improving the recognition accuracy of gas leakage, expanding the labels of the real gas leakage case library, and reducing the cost of manual labeling.

[0114] In order to implement the above embodiment, the present disclosure also provides an electronic device, Figure 8 1 is a schematic diagram of the structure of an electronic device according to an embodiment of the present disclosure. The electronic device includes:

[0115] A memory 801 , a processor 802 , and a computer program stored in the memory 801 and executable on the processor 802 .

[0116] When the processor 802 executes the program, the training method of the urban underground gas leakage identification model provided in the above embodiment is implemented.

[0117] Furthermore, the electronic device further comprises:

[0118] The communication interface 803 is used for communication between the memory 801 and the processor 802 .

[0119] The memory 801 is used to store computer programs that can be executed on the processor 802 .

[0120] The memory 801 may include a high-speed RAM memory, and may also include a non-volatile memory (non-volatile memory), such as at least one disk memory.

[0121] The processor 802 is used to implement the training method of the urban underground gas leakage identification model described in the above embodiment when executing the program.

[0122] If the memory 801, the processor 802 and the communication interface 803 are implemented independently, the communication interface 803, the memory 801 and the processor 802 can be connected to each other through a bus and communicate with each other. The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component (PCI) bus or an Extended Industry Standard Architecture (EISA) bus. The bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 8 Only one thick line is used in the diagram, but this does not mean that there is only one bus or only one type of bus.

[0123] Optionally, in a specific implementation, if the memory 801, the processor 802 and the communication interface 803 are integrated on a chip, the memory 801, the processor 802 and the communication interface 803 can communicate with each other through an internal interface.

[0124] The processor 802 may be a central processing unit (CPU), or an application specific integrated circuit (ASIC), or one or more integrated circuits configured to implement the embodiments of the present disclosure.

[0125] In order to implement the above-mentioned embodiments, the present disclosure proposes a computer-readable storage medium on which a computer program is stored. When the program is executed by a processor, the training method of the urban underground gas leakage identification model as described above is implemented.

[0126] In order to implement the above embodiments, the present disclosure proposes a computer program product, including a computer program, which, when executed by a processor, implements the above-mentioned training method for the urban underground gas leakage identification model.

[0127] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present disclosure. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine the different embodiments or examples described in this specification and the features of the different embodiments or examples, without contradiction.

[0128] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of the features. In the description of the present disclosure, "plurality" means at least two, such as two, three, etc., unless otherwise clearly and specifically defined.

[0129] Any process or method description in a flowchart or otherwise described herein may be understood to represent a module, segment or portion of code that includes one or more executable instructions for implementing the steps of a custom logical function or process, and the scope of the preferred embodiments of the present disclosure includes additional implementations in which functions may not be performed in the order shown or discussed, including performing functions in a substantially simultaneous manner or in reverse order depending on the functions involved, which should be understood by technicians in the technical field to which the embodiments of the present disclosure belong.

[0130] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute the instructions), or in combination with these instruction execution systems, devices or apparatuses. For the purpose of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in combination with these instruction execution systems, devices or apparatuses. More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection with one or more wires (electronic device), a portable computer disk box (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), a fiber optic device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium and then editing, interpreting or processing in other suitable ways if necessary, and then stored in a computer memory.

[0131] It should be understood that the various parts of the present disclosure can be implemented in hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, multiple steps or methods can be implemented in software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, any one of the following technologies known in the art or a combination thereof can be used to implement: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0132] A person skilled in the art may understand that all or part of the steps in the method for implementing the above-mentioned embodiment may be completed by instructing related hardware through a program, and the program may be stored in a computer-readable storage medium, which, when executed, includes one or a combination of the steps of the method embodiment.

[0133] In addition, each functional unit in each embodiment of the present disclosure may be integrated into a processing module, or each unit may exist physically separately, or two or more units may be integrated into one module. The above-mentioned integrated module may be implemented in the form of hardware or in the form of a software functional module. If the integrated module is implemented in the form of a software functional module and sold or used as an independent product, it may also be stored in a computer-readable storage medium.

[0134] The storage medium mentioned above may be a read-only memory, a disk or an optical disk, etc. Although the embodiments of the present disclosure have been shown and described above, it can be understood that the above embodiments are exemplary and cannot be understood as limitations of the present disclosure. A person of ordinary skill in the art may change, modify, replace and modify the above embodiments within the scope of the present disclosure.

Claims

1. A training method for an urban underground gas leak identification model, characterized in that: include: Obtaining the first methane concentration sequence to be annotated from the time series database; Determining, from the first methane concentration sequence, a target segment methane concentration sequence corresponding to when the methane concentration changes abnormally; Extracting features of the target segment methane concentration sequence to obtain methane concentration variation features; Obtaining a target real methane concentration sequence matching the methane concentration variation characteristics from a real gas leakage case library; Using the gas leakage label corresponding to the target real methane concentration sequence as the label data of the target segment methane concentration sequence; Adding the target segment methane concentration sequence and the corresponding label data to the real gas leakage case library to obtain an updated real gas leakage case library; The gas leak recognition model is trained according to each methane concentration sequence and corresponding label data in the updated real gas leak case library.

2. The method according to claim 1, characterized in that Before training the gas leak recognition model according to each methane concentration sequence and corresponding label data in the updated real gas leak case library, the method further includes: The gas leak recognition model is trained according to each methane concentration sequence and corresponding label data in the real gas leak case library.

3. The method according to claim 1, characterized in that The step of determining, from the first methane concentration sequence, a target segment methane concentration sequence corresponding to when the methane concentration changes abnormally comprises: According to the methane concentration ranges corresponding to the preset risk levels, the first methane concentration sequence is segmented to obtain multiple methane concentration sequences; For the multiple methane concentration sequences, obtaining a first candidate methane concentration sequence corresponding to the highest risk level, and obtaining an end time of the first candidate methane concentration sequence; From the multiple methane concentration sequences, obtaining a second candidate methane concentration sequence with a risk level of zero, a start time before the start time of the first candidate methane concentration sequence, and a start time closest to the first candidate methane concentration sequence; Each methane concentration sequence segment among the multiple methane concentration sequences, which is located between the start time of the second candidate methane concentration sequence segment and the end time of the first candidate methane concentration sequence segment, is used as a target methane concentration sequence segment corresponding to when the methane concentration changes abnormally.

4. The method according to claim 1, characterized in that: Any real methane concentration sequence in the real gas leakage case database is obtained in the following way: Acquire, from the time series database, a second methane concentration sequence different from the first methane concentration sequence; According to the methane concentration ranges corresponding to the preset risk levels, the second methane concentration sequence is segmented to obtain multiple methane concentration sequences; For the multiple methane concentration sequences, obtaining the time point confirmed by the maintenance personnel for the methane concentration sequence, and using the time point as the end time corresponding to the abnormal change of the methane concentration; From the multiple methane concentration sequences, obtain a third candidate methane concentration sequence with a risk level of zero, located before the end time, and with a start time closest to the end time; Each methane concentration sequence segment among the multiple methane concentration sequences, which is located between the start time and the end time of the third candidate methane concentration sequence segment, is used as the true methane concentration sequence.

5. The method according to claim 1, characterized in that The step of obtaining a target real methane concentration sequence matching the methane concentration variation characteristic from a real gas leakage case library includes: For any real methane concentration sequence in the real gas leakage case library, matching the methane concentration change feature corresponding to the real methane concentration sequence with the methane concentration change feature corresponding to the target segment methane concentration sequence; If the degree of match between the methane concentration change characteristics corresponding to the real methane concentration sequence and the methane concentration change characteristics corresponding to the target segment methane concentration sequence is greater than a preset matching degree threshold, it is determined that the methane concentration change law of the real methane concentration sequence and the target segment methane concentration sequence is the same, and the real methane concentration sequence is used as the target real methane concentration sequence that matches the methane concentration change characteristics.

6. A training device for an urban underground gas leak identification model, characterized in that: include: A first acquisition module is used to acquire a first methane concentration sequence to be marked from a time series database; A determination module, used to determine, from the first methane concentration sequence, a target segment methane concentration sequence corresponding to when the methane concentration changes abnormally; An extraction module, used for extracting features of the target segment methane concentration sequence to obtain methane concentration variation features; A second acquisition module is used to acquire a target real methane concentration sequence matching the methane concentration variation characteristics from a real gas leakage case library; A generating module, used to use the gas leakage label corresponding to the target real methane concentration sequence as the label data of the target segment methane concentration sequence; An adding module, used for adding the target segment methane concentration sequence and the corresponding label data to the real gas leakage case library to obtain an updated real gas leakage case library; The first training module is used to train the gas leakage recognition model according to each methane concentration sequence and corresponding label data in the updated real gas leakage case library.

7. The device according to claim 6, characterized in that The device also includes: The second training module is used to train the gas leakage recognition model according to each methane concentration sequence and corresponding label data in the real gas leakage case library.

8. The device according to claim 6, characterized in that The determination module is specifically used for: According to the methane concentration ranges corresponding to the preset risk levels, the first methane concentration sequence is segmented to obtain multiple methane concentration sequences; For the multiple methane concentration sequences, obtaining a first candidate methane concentration sequence corresponding to the highest risk level, and obtaining an end time of the first candidate methane concentration sequence; From the multiple methane concentration sequences, obtaining a second candidate methane concentration sequence with a risk level of zero, a start time before the start time of the first candidate methane concentration sequence, and a start time closest to the first candidate methane concentration sequence; Each methane concentration sequence segment among the multiple methane concentration sequences, which is located between the start time of the second candidate methane concentration sequence segment and the end time of the first candidate methane concentration sequence segment, is used as a target methane concentration sequence segment corresponding to when the methane concentration changes abnormally.

9. The device according to claim 6, characterized in that Any real methane concentration sequence in the real gas leakage case database is obtained in the following way: Acquire, from the time series database, a second methane concentration sequence different from the first methane concentration sequence; According to the methane concentration ranges corresponding to the preset risk levels, the second methane concentration sequence is segmented to obtain multiple methane concentration sequences; For the multiple methane concentration sequences, obtaining the time point confirmed by the maintenance personnel for the methane concentration sequence, and using the time point as the end time corresponding to the abnormal change of the methane concentration; From the multiple methane concentration sequences, obtain a third candidate methane concentration sequence with a risk level of zero, located before the end time, and with a start time closest to the end time; Each methane concentration sequence segment among the multiple methane concentration sequences, which is located between the start time and the end time of the third candidate methane concentration sequence segment, is used as the true methane concentration sequence.

10. The device according to claim 6, characterized in that The second acquisition module is specifically used for: For any real methane concentration sequence in the real gas leakage case library, matching the methane concentration change feature corresponding to the real methane concentration sequence with the methane concentration change feature corresponding to the target segment methane concentration sequence; If the degree of match between the methane concentration change characteristics corresponding to the real methane concentration sequence and the methane concentration change characteristics corresponding to the target segment methane concentration sequence is greater than a preset matching degree threshold, it is determined that the methane concentration change law of the real methane concentration sequence and the target segment methane concentration sequence is the same, and the real methane concentration sequence is used as the target real methane concentration sequence that matches the methane concentration change characteristics.

11. An electronic device, characterized in that: include: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the training method for the urban underground gas leak identification model as described in any one of claims 1 to 5 is implemented.

12. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, it implements the training method of the urban underground gas leakage identification model as described in any one of claims 1-5.

Images