Transmission line hidden danger event perception method and system integrating social information

Abstract

A method and system for perceiving transmission line hidden danger events that integrate social information. The method includes collecting social information data from multiple social information data sources, converting the social information data into a standardized text sequence; encoding the standardized text sequence that integrates the social information data word by word and sentence by sentence to convert it into a text vector representation; constructing a hidden danger event vocabulary, matching the text vector representation with the keyword vector in the hidden danger event vocabulary to determine the transmission line hidden danger event corresponding to the standardized text sequence, extracting key information of the transmission line hidden danger event; and performing alarm and intervention processing according to the transmission line hidden danger event. The scheme of the present invention enables the system to promptly discover and intervene in transmission line channel hidden dangers from multiple perspectives, thereby improving the accuracy and timeliness of hidden danger event perception.

Description

Technical Field

[0001] The present invention belongs to the field of natural language processing and risk identification, and in particular relates to a method and system for perceiving transmission line hidden danger events integrating social information. Background Art

[0002] With the development of new power systems, the requirements for power operation are getting higher and higher. The safety of power systems is related to social and economic stability, especially the early detection and handling of hidden dangers. The existing power transmission line safety monitoring focuses on real-time monitoring of the daily operation and maintenance of transmission lines through daily inspections of power systems, internal and external network sensor data, and transmission line channel video monitoring. The relevant technology mainly relies on the openness of data interfaces.

[0003] On the technical level, in many sensitive departments and different platforms, data interfaces are not open, making it impossible to import and process the full amount of data. Traditional methods have problems with poor recognition accuracy and slow response speed, and cannot meet actual application needs. On the operation and maintenance level, in traditional transmission line hidden danger event perception systems, they can usually only rely on limited internal monitoring data sources of the transmission line system for event perception, and often cannot detect hidden dangers in a timely and accurate manner. Summary of the invention

[0004] In order to solve the deficiencies in the prior art, the present invention provides a method and system for perceiving transmission line hidden danger events by integrating social information, so as to solve the technical problems that the existing transmission line monitoring data sources are relatively single and the technology is relatively rigid.

[0005] In order to solve the above technical problems, the present invention adopts the following technical solutions.

[0006] The present invention first discloses a method for perceiving transmission line hidden danger events integrating social information, the method comprising the following steps:

[0007] Step 1: Collecting social information data from multiple social information data sources, and converting the social information data into a standardized text sequence;

[0008] Step 2: Encode the standardized text sequence that integrates social information data word by word and sentence by sentence to convert it into a text vector representation;

[0009] Step 3: construct a hidden danger event vocabulary, match the text vector representation with the keyword vector in the hidden danger event vocabulary, so as to determine the transmission line hidden danger event corresponding to the standardized text sequence, and extract key information of the transmission line hidden danger event;

[0010] Step 4: Perform alarm and intervention processing according to the transmission line hidden danger event.

[0011] The present invention further includes the following preferred embodiments:

[0012] The collecting of social information data from multiple social information data sources further includes:

[0013] When the social information data source is the transmission line monitoring system information, screenshots and related description information in the transmission line monitoring system are obtained, keyword screening is performed according to the type of hidden dangers in the system, the time range for data collection and the frequency of video surveillance image data rotation are pre-limited, and risky video surveillance data are extracted in the preprocessing stage.

[0014] The collecting of social information data from multiple social information data sources further includes:

[0015] When the social information data source is social media data, the API interface provided by the social media platform is used to obtain data, and keywords related to hidden danger events are set in the keyword screening, the keywords are monitored in real time, and the text content related to the keywords is filtered; the time range for data collection is limited; in the preprocessing process, the obtained text data is cleaned to remove advertisements, irrelevant content and duplicate content; and the web crawler technology is used to regularly capture relevant reports on news websites.

[0016] The collecting of social information data from multiple social information data sources further includes:

[0017] When the social information data source is a public video surveillance screen, obtain screenshots and related description information from the public surveillance, filter the events, and only collect surveillance screens related to hidden danger events; limit the time range for data collection, perform frame capture on the video screen during preprocessing, and obtain key information frames.

[0018] The step of encoding the standardized text sequence integrating social information data word by word and sentence by sentence to convert it into a text vector representation further includes:

[0019] Step 2.1: Perform text preprocessing on the standardized text sequence, including:

[0020] Perform word segmentation on the text sequence and divide the text into independent words;

[0021] Remove stop words from the text;

[0022] Convert the segmented text into word vectors and use the pre-trained word vector model to generate word vector representations;

[0023] Step 2.2: Based on the BiLSTM-CRF model input layer, the preprocessed word vector is input into the BiLSTM model. In the BiLSTM layer, the bidirectional long short-term memory network is used to encode the input word vector to generate a word vector representation containing context information. In the output layer, the output vector of the BiLSTM layer is used as the final text vector representation.

[0024] The step of constructing a hidden danger event word library further includes:

[0025] Collect hidden danger keywords related to transmission line channels, and define multiple hidden danger event categories based on the risk type, time and location information of hidden danger events; collect keywords for each type of hidden danger event and build a keyword thesaurus; use the knowledge graph algorithm to associate the keywords of hidden danger events with the relevant time and location information to obtain a complete hidden danger event thesaurus.

[0026] The alarm and intervention processing according to the transmission line hidden danger event further includes:

[0027] According to the severity of hidden danger events, hidden danger events are divided into different risk levels, and hidden danger events are classified by type;

[0028] Generate alarm information based on the risk level, time and location information of hidden danger events, and send the alarm information to relevant departments or personnel;

[0029] Track the potential danger events that have been alarmed, record the progress and results of the processing, feedback the processing results to the system, and update the potential danger event database.

[0030] The present invention also discloses a system for sensing hidden danger events of power transmission lines integrating social information using the aforementioned method for sensing hidden danger events of power transmission lines integrating social information, comprising:

[0031] A data collection module, used to collect social information data from multiple social information data sources and convert the social information data into a standardized text sequence;

[0032] A text encoding module is used to encode the standardized text sequence that integrates social information data word by word and sentence by sentence to convert it into a text vector representation;

[0033] An event matching module is used to build a hidden danger event vocabulary, match the text vector representation with the keyword vector in the hidden danger event vocabulary, so as to determine the transmission line hidden danger event corresponding to the standardized text sequence, and extract key information of the transmission line hidden danger event;

[0034] The alarm processing module is used to perform alarm and intervention processing according to the hidden danger event of the transmission line.

[0035] Accordingly, the present application also discloses a terminal, including a processor and a storage medium;

[0036] The storage medium is used to store instructions;

[0037] The processor is used to operate according to the instructions to execute the steps of the power transmission line hidden danger event perception method based on the aforementioned social information integration.

[0038] Correspondingly, the present application also discloses a computer-readable storage medium on which a computer program is stored. When the program is executed by a processor, the steps of the aforementioned method for perceiving transmission line hidden danger events integrating social information are implemented.

[0039] The beneficial effect of the present invention is that, compared with the prior art, the present invention provides a method and system for perceiving hidden danger events of power transmission lines that integrate social information. By collecting and integrating social information data, the hidden danger event perception of sensitive data and non-open interface platforms is realized. The new information source after integrating social information provides more perspectives and information support for the perception of hidden danger events. By encoding the collected social information data word by word and sentence by sentence, a vector representation that can be recognized by the model is generated, thereby improving the accuracy and processability of the text representation. The hidden danger event vocabulary is used for matching, and the hidden danger event information in the text sequence is perceived in real time, thereby improving the accuracy and response speed of risk perception. Alarm processing is performed based on the perceived hidden danger event information to ensure that the hidden danger event can be discovered and handled in a timely manner, and the type, time, location and other information of the hidden danger event are perceived in real time, thereby improving the accuracy and timeliness of event perception. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 It is a framework diagram of the power transmission line hidden danger event perception system integrating social information in the present invention.

[0041] Figure 2 It is a schematic diagram of a risk-free picture taken by a fixed camera in the present invention.

[0042] Figure 3 It is a schematic diagram of the social system data collection process related to the power transmission channel in the present invention.

[0043] Figure 4 It is a schematic diagram of the subject perception process of hidden danger events based on BiLSTM-CRF in the present invention.

[0044] Figure 5 It is a structure diagram of the hidden danger event network structure restoration model and traceability system based on compressed sensing in the present invention. DETAILED DESCRIPTION

[0045] In order to make the purpose, technical solution and advantages of the present invention more clear, the technical solution of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention.

[0046] The embodiments described in this application are only some embodiments of the present invention, not all embodiments. Based on the spirit of the present invention, other embodiments obtained by ordinary technicians in this field without creative work are all within the protection scope of the present invention.

[0047] As a new source of information, social information contains rich social system and event information, which can provide more perspectives and information support for the perception of hidden danger events. The relevant information of the social system also reflects the safety hazards of the transmission line to a certain extent. For example, social information such as disaster information, extreme weather, road construction, etc. released by social media, news reports, and government announcements is another important source of information on transmission line safety hazards. On the existing basis, it is necessary to integrate the social system data of the external network into the transmission line safety monitoring system to perceive the hidden dangers of the transmission line more comprehensively.

[0048] Based on the above analysis, the present invention proposes a method and system for perceiving hidden danger events in power transmission lines that integrate social information, which is used for the collection and processing of social information data to achieve accurate perception and rapid alarm of hidden danger events. First, the text sequence after the integration of social information data is encoded word by word and sentence by sentence, and the text sequence is converted into a vector representation that can be recognized by the model, and matched with the hidden danger event vocabulary, so as to perceive the risk type, time and location and other information of the implicit content in the massive text sequence in real time. The new information source after the integration of social information enables the system to promptly discover and intervene in the hidden dangers of the transmission line channel from multiple perspectives, thereby improving the accuracy and timeliness of the perception of hidden danger events.

[0049] The method for perceiving transmission line hidden danger events integrating social information disclosed in the present invention specifically comprises the following steps:

[0050] Step 1: Collect social information data from multiple social information data sources, and convert the social information data into a standardized text sequence.

[0051] See also Figure 1The system includes a data acquisition module, a text encoding module, an event matching module and an alarm processing module. Among them, the data acquisition module is used to collect social information data and convert it into a text sequence. The text encoding module encodes the text sequence word by word and sentence by sentence to generate a vector representation that can be recognized by the model. The event matching module matches the vector representation with the hidden danger event vocabulary to perceive the hidden danger event information in the massive text sequence in real time; the alarm processing module performs alarm processing based on the perceived hidden danger event information. Through screenshots, data including the transmission line monitoring system, sensor system, social media, official announcements, etc. are collected and converted into text sequences, the vector representation is matched with the hidden danger event vocabulary, and the hidden danger event rating in the massive text sequence is perceived in real time.

[0052] Specifically, the data collection module collects social information data from multiple sources such as social media, news reports, and public video surveillance images, and converts them into standardized text sequences. The data collection process does not rely on data interfaces of sensitive departments or closed interface platforms, thereby avoiding the impact on the system and the risk of network attacks. Multiple social information data sources (such as social media, news reports, government announcements, etc.) are integrated to form a more comprehensive risk information data source.

[0053] In a preferred embodiment, after collecting relevant text data from multiple channels such as social media and news reports, preprocessing is first performed, including word segmentation, removal of stop words, etc.

[0054] The main data sources include:

[0055] 1) Transmission line monitoring system information, including transmission line channel video monitoring.

[0056] 2) Transmission line sensor data.

[0057] 3) Social media information, including social media, news reports, and government announcements.

[0058] 4) Public video surveillance information: such as screenshots and descriptions of videos from public cameras in cities.

[0059] Among them, the information of the transmission line monitoring system can be collected by obtaining screenshots and related description information in the transmission line monitoring system. Keyword screening mainly involves the types of hidden dangers in the system, such as "crane construction" and "kite entanglement". The time range of data collection and the frequency of video monitoring picture data rotation are pre-defined, such as refreshing once every 15 minutes / (adjustable). Risky video monitoring data is extracted in the preprocessing stage. The data type is defined as team name + voltage level + transmission line name + tower number.

[0060] For the data collection of the transmission line sensors, there is no need for the sensor monitoring system to provide an interface. Take the hardware sensor as an example. The data information required to be collected by the hardware temperature sensor is as follows:

[0061] Data identifier: Model LTM-50;

[0062] Data format: binary message;

[0063] Data size: 26 bytes;

[0064] Data source (data acquisition sensor name): temperature sensor;

[0065] Data sampling time: real-time sampling;

[0066] Sampling interval: 10min.

[0067] Data format example: 2024-01-02 13:20:35

[0068] Data collection location (transmission line name): Duxing 4235, 4236, 4239, 4240 lines, Sitong 4125, 4216 lines.

[0069] Measuring content: -55℃-+150℃

[0070] Other information (such as storage type): FLASH, RAM, EEPROM;

[0071] For social media data collection, data is obtained using the API interface provided by the social media platform, which must match the platform's usage policy and privacy regulations.

[0072] In the keyword screening, set keywords related to hidden danger events, such as "mechanical construction", "tank truck fire", "truck rollover", "road construction", "fire", etc. Monitor keywords in real time and filter text content related to keywords, such as news reports. Limit the time range of data collection, such as content published in the past 1 hour, to ensure the timeliness of the data. Prioritize the collection of content published by certified users (such as certified official releases) to improve the credibility of the data. In the preprocessing process, clean the acquired text data to remove redundant information such as advertisements, irrelevant content, and duplicate content. Give priority to authoritative reports to ensure the authority and accuracy of the data. Use web crawler technology to regularly capture relevant reports on news websites.

[0073] For the collection of public video surveillance images, data entry does not require an authorized interface, and the screenshots and related description information in the public surveillance are obtained. Events are screened and only surveillance images related to hidden danger events are collected, such as "mechanical construction", "tank truck fire", "truck rollover", "road construction", "fire", etc. The time range of data collection is limited to ensure the timeliness of monitoring data. In the preprocessing, the video screen is framed to obtain key information frames.

[0074] In the process of converting to a standardized text sequence, the processed text is converted into a unified format and a timestamp and data source label are added. Text data from different sources are converted into a unified standardized format to ensure the consistency of the data format. Specifically, first add a timestamp to each piece of data to record the time of data collection for subsequent processing and analysis. Then add data source labels such as "social media", "news reports", "public video surveillance", "power transmission monitoring", etc. for classification and processing.

[0075] Through the above data collection process, the present invention can efficiently and accurately collect social information data from multiple sources and convert it into a standardized text sequence, providing reliable data support for subsequent hidden danger event perception and alarm processing.

[0076] Step 2: Encode the standardized text sequence that integrates social information data word by word and sentence by sentence to convert it into a text vector representation.

[0077] Specifically, the text encoding module uses a pre-trained natural language processing model to encode the text sequence that integrates social information data word by word and sentence by sentence, converting the text sequence into a word or sentence vector representation that can be recognized by the model, thereby enhancing the accuracy and processability of the text representation.

[0078] In a preferred embodiment, a Fast-CRNN fast text localization method based on image content is used to efficiently extract text information from multi-source information images such as visual monitoring of power transmission channels in complex environments, government microblogs, social media, and Internet portals; a carousel data recognition model based on a CRNN deep network is constructed, an end-to-end variable-length text sequence recognition algorithm is proposed, CNN is combined with RNN, and CTCloss is introduced to solve the problem of repeated sequence recognition, improve the text recognition rate, and realize real-time, automatic collection and fusion of social system (government microblogs, social media, Internet portals, etc.) data and channel monitoring data, such as Figure 3 shown.

[0079] The specific implementation steps are as follows:

[0080] Step 2.1: Perform text preprocessing on the standardized text sequence, including:

[0081] Tokenize the text sequence and split the text into individual words.

[0082] Remove stop words (such as "的", "了", etc.) from the text to reduce noise data.

[0083] Convert the tokenized text into word vectors and generate word vector representations using pre-trained word vector models (such as Word2Vec, GloVe).

[0084] Step 2.2: In the input layer of the BiLSTM-CRF (Bidirectional Long Short-Term Memory - Conditional Random Field) model, input the preprocessed word vectors into the BiLSTM model. In the BiLSTM layer, use the bidirectional long short-term memory network to encode the input word vectors and generate word vector representations containing context information. In the output layer, use the output vector of the BiLSTM layer as the final text vector representation.

[0085] In the input layer, not only use word vectors but also introduce character-level vector representations to enhance the model's ability to handle out-of-vocabulary (OOV) words. Assume that each word is composed of a character sequence Then the vector representation corresponding to each character is .

[0086] For the word , its mixed vector representation is:

[0087] ,

[0088] where represents the vector representation obtained by processing the character sequence through a convolutional neural network (CNN).

[0089] In the BiLSTM layer, a dynamic weight allocation mechanism is introduced to better fuse the joint output of the forward and backward LSTMs:

[0090] .

[0091] where is the weight dynamically calculated through the self-attention mechanism:

[0092] ,

[0093] where is a parameter to be learned, and respectively represent the hidden states at the th position in the forward LSTM and the backward LSTM.

[0094] In the output layer, the loss function of the CRF layer is also improved, and a regularization term is introduced to prevent overfitting:

[0095] ,

[0096] in, Indicates The label or state of each time step, Indicates that the BiLSTM network is The hidden state of the time step contains the output of the forward and backward LSTM. The length of the input sequence, i.e. the number of time steps or the number of data points.

[0097] For transmission line monitoring, Indicates a specific state or event at a certain time step, such as a fault, hidden danger type, or abnormal situation of monitoring data.

[0098] Optionally, during the data collection process, the usage policies and privacy regulations of each data source platform must be followed to ensure the legality and security of the data. Data involving personal privacy should be anonymized to remove sensitive information and protect user privacy.

[0099] Step 3: Construct a hidden danger event vocabulary, match the text vector representation with the keyword vector in the hidden danger event vocabulary to determine the transmission line hidden danger event corresponding to the standardized text sequence, and extract key information of the transmission line hidden danger event.

[0100] Specifically, the event matching module constructs a hidden danger event vocabulary, which contains key information such as risk type, time and location, matches the encoded text sequence vector with the keyword vector in the hidden danger event vocabulary, determines whether there is hidden danger-related content in the text sequence, and perceives the implicit risk information in the text sequence in real time.

[0101] In the process of establishing the hidden danger event word library, we collect hidden danger keywords related to the transmission line channel, such as "mechanical construction", "tank truck fire", "truck rollover", "road construction", "fire", etc. According to the risk type, time and location of the hidden danger event, define multiple hidden danger event categories. Collect keywords for each type of hidden danger event and build a keyword word library. Use the knowledge graph algorithm to associate the keywords of the hidden danger event with the relevant time and location information to build a complete hidden danger event word library.

[0102] Through the metadata and description information of hidden danger events, the subjects involved and the types of hidden danger events are matched again, the subjects involved are accurately located, and the types of hidden danger events and subjects are accurately matched, achieving the improvement and effective balance of accuracy and recall rate. Figure 4shown.

[0103] The vector representation generated by the text encoding module is matched with the keyword vector in the hidden danger event vocabulary to calculate the similarity. Based on the similarity, it is determined whether there is hidden danger event information in the text, and key information such as risk type, time and location of the hidden danger event is extracted.

[0104] The conditional random field (CRF) model is used to optimize the matching results and improve the accuracy of hidden danger event identification.

[0105] Step 4: Perform alarm and intervention processing according to the transmission line hidden danger event.

[0106] Specifically, the alarm processing module realizes hidden danger perception and intervention. When matching the content related to hidden dangers, the system perceives the type, time and location of the hidden dangers in real time, and takes timely intervention measures, such as early warning notification, line scheduling, etc.

[0107] A network knowledge graph structure model based on hidden danger event information is constructed, and the attitude descriptions of ordinary users and professional users of power transmission operation and maintenance towards specific keywords are calculated through the compressed sensing method, so as to quickly restore the network correlation structure between hidden danger event information; a hidden danger event tracing method based on graph path analysis and similarity calculation is proposed, and spatiotemporal calculations are performed in combination with time information and geographic positioning information. It can timely determine the specific correlation relationship between hidden danger events and channel equipment, and realize the rapid positioning of hidden danger events around the channel.

[0108] The specific implementation steps are as follows:

[0109] Classify hidden danger events. First, classify hidden danger events into different risk levels, such as low, medium, and high, according to the severity of the hidden danger events. Classify hidden danger events by type, such as "mechanical construction", "tank truck fire", "truck rollover", "road construction", and "fire" to facilitate subsequent processing.

[0110] Generate alarm information. Generate detailed alarm information based on the risk level, time and location of the hidden danger event. Send the alarm information to relevant departments or personnel, such as the public security system, emergency management department, etc., to ensure that the hidden danger event can be discovered and handled in a timely manner.

[0111] Track the potential danger events that have been reported, record the progress and results of the processing, feed back the processing results to the system, update the potential danger event database, and optimize the system's alarm mechanism.

[0112] The beneficial effect of the present invention is that, compared with the prior art, the present invention provides a method and system for perceiving hidden danger events of power transmission lines that integrate social information. By collecting and integrating social information data, the hidden danger event perception of sensitive data and non-open interface platforms is realized. The new information source after integrating social information provides more perspectives and information support for the perception of hidden danger events. By encoding the collected social information data word by word and sentence by sentence, a vector representation that can be recognized by the model is generated, thereby improving the accuracy and processability of the text representation. The hidden danger event vocabulary is used for matching, and the hidden danger event information in the text sequence is perceived in real time, thereby improving the accuracy and response speed of risk perception. Alarm processing is performed based on the perceived hidden danger event information to ensure that the hidden danger event can be discovered and handled in a timely manner, and the type, time, location and other information of the hidden danger event are perceived in real time, thereby improving the accuracy and timeliness of event perception.

[0113] The present invention may be a system, a method and / or a computer program product. The present invention also discloses a transmission line hidden danger event perception system integrating social information based on the above-mentioned transmission line hidden danger event perception method integrating social information, comprising:

[0114] A data collection module, used to collect social information data from multiple social information data sources and convert the social information data into a standardized text sequence;

[0115] A text encoding module is used to encode the standardized text sequence that integrates social information data word by word and sentence by sentence to convert it into a text vector representation;

[0116] An event matching module is used to build a hidden danger event vocabulary, match the text vector representation with the keyword vector in the hidden danger event vocabulary, so as to determine the transmission line hidden danger event corresponding to the standardized text sequence, and extract key information of the transmission line hidden danger event;

[0117] The alarm processing module is used to perform alarm and intervention processing according to the hidden danger event of the transmission line.

[0118] Based on the spirit of the present invention, those skilled in the art can easily think that a computer program product can be obtained based on the aforementioned method for perceiving transmission line hidden danger events integrating social information. The computer program product may include a computer-readable storage medium, which carries computer-readable program instructions for enabling a processor to implement various aspects of the present disclosure. That is, the present application also includes a terminal, including a processor and a storage medium; the storage medium is used to store instructions; the processor is used to operate according to the instructions to execute the steps of the method for perceiving transmission line hidden danger events integrating social information.

[0119] Computer readable storage medium can be a tangible device that can keep and store the instructions used by the instruction execution device. Computer readable storage medium can be, for example, - but not limited to - electrical storage device, magnetic storage device, optical storage device, electromagnetic storage device, semiconductor storage device or any suitable combination of the above. More specific examples (non-exhaustive list) of computer readable storage medium include: portable computer disk, hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), static random access memory (SRAM), portable compact disk read-only memory (CD-ROM), digital versatile disk (DVD), memory stick, floppy disk, mechanical encoding device, such as punch card or groove protrusion structure with instructions stored thereon and any suitable combination of the above. Computer readable storage medium used here is not interpreted as instantaneous signal itself, such as radio wave or other free propagating electromagnetic wave, electromagnetic wave propagated by waveguide or other transmission medium (for example, light pulse by optical fiber cable) or electrical signal transmitted by wire.

[0120] The computer-readable program instructions described herein can be downloaded from a computer-readable storage medium to each computing / processing device, or downloaded to an external computer or external storage device via a network, such as the Internet, a local area network, a wide area network, and / or a wireless network. The network can include copper transmission cables, optical fiber transmissions, wireless transmissions, routers, firewalls, switches, gateway computers, and / or edge servers. The network adapter card or network interface in each computing / processing device receives the computer-readable program instructions from the network and forwards the computer-readable program instructions for storage in the computer-readable storage medium in each computing / processing device.

[0121] The computer program instructions for performing the operations of the present disclosure may be assembly instructions, instruction set architecture (ISA) instructions, machine instructions, machine-dependent instructions, microcode, firmware instructions, state setting data, or source code or object code written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Smalltalk, C++, etc., and conventional procedural programming languages ​​such as "C" language or similar programming languages. The computer-readable program instructions may be executed entirely on the user's computer, partially on the user's computer, as a separate software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer through any type of network including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., using an Internet service provider to connect through the Internet). In some embodiments, by using the state information of the computer-readable program instructions to personalize an electronic circuit, such as a programmable logic circuit, a field programmable gate array (FPGA) or a programmable logic array (PLA), the electronic circuit may execute the computer-readable program instructions, thereby implementing various aspects of the present disclosure.

[0122] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the relevant field should understand that the specific implementation methods of the present invention can still be modified or replaced by equivalents, and any modifications or equivalent replacements that do not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A method for perceiving transmission line hidden danger events integrating social information, characterized in that: The following steps are involved: Step 1: Collecting social information data from multiple social information data sources, and converting the social information data into a standardized text sequence; Step 2: Encode the standardized text sequence that integrates social information data word by word and sentence by sentence to convert it into a text vector representation; Step 3: construct a hidden danger event vocabulary, match the text vector representation with the keyword vector in the hidden danger event vocabulary, so as to determine the transmission line hidden danger event corresponding to the standardized text sequence, and extract key information of the transmission line hidden danger event; Step 4: Performing alarm and intervention processing according to the transmission line hidden danger event; The step of encoding the standardized text sequence integrating social information data word by word and sentence by sentence to convert it into a text vector representation further includes: Step 2.1: Perform text preprocessing on the standardized text sequence, including: Perform word segmentation on the text sequence and divide the text into independent words; Remove stop words from the text; Convert the segmented text into word vectors and use the pre-trained word vector model to generate word vector representations; Step 2.2: Based on the BiLSTM-CRF model input layer, the preprocessed word vector is input into the BiLSTM model. In the BiLSTM layer, the input word vector is encoded using the bidirectional long short-term memory network to generate a word vector representation containing context information. In the output layer, the output vector of the BiLSTM layer is used as the final text vector representation. In the input layer, word vectors are used and character-level vector representations are introduced; assuming that each word By character sequence Each character The corresponding vector is expressed as ; For words , whose mixed vector is expressed as: ; in, Represents the vector representation obtained by processing the character sequence through the convolutional neural network CNN; In the BiLSTM layer, a dynamic weight allocation mechanism is introduced to fuse the joint outputs of the forward and backward LSTMs: ; in, are weights dynamically calculated by the self-attention mechanism: ,in, are the parameters that need to be learned. and Respectively represent the first The hidden state of each position; In the output layer, a regularization term is introduced into the loss function of the CRF layer: , in, Indicates The label or state of each time step, Indicates that the BiLSTM network is The hidden state of the time step contains the output of the forward and backward LSTM. The length of the input sequence, i.e. the number of time steps or the number of data points.

2. The method for perceiving transmission line hidden danger events integrating social information according to claim 1 is characterized in that: The collecting of social information data from multiple social information data sources further includes: When the social information data source is the transmission line monitoring system information, screenshots and related description information in the transmission line monitoring system are obtained, keyword screening is performed according to the type of hidden dangers in the system, the time range for data collection and the frequency of video surveillance image data rotation are pre-limited, and risky video surveillance data are extracted in the preprocessing stage.

3. The method for perceiving transmission line hidden danger events integrating social information according to claim 2 is characterized in that: The collecting of social information data from multiple social information data sources further includes: When the social information data source is social media data, the API interface provided by the social media platform is used to obtain data, and keywords related to hidden danger events are set in the keyword screening, the keywords are monitored in real time, and the text content related to the keywords is filtered; the time range for data collection is limited; in the preprocessing process, the obtained text data is cleaned to remove advertisements, irrelevant content and duplicate content; and the web crawler technology is used to regularly capture relevant reports on news websites.

4. The method for perceiving transmission line hidden danger events integrating social information according to claim 3 is characterized in that: The collecting of social information data from multiple social information data sources further includes: When the social information data source is a public video surveillance screen, obtain screenshots and related description information from the public surveillance, filter the events, and only collect surveillance screens related to hidden danger events; limit the time range for data collection, perform frame capture on the video screen during preprocessing, and obtain key information frames.

5. The method for perceiving transmission line hidden danger events integrating social information according to claim 4 is characterized in that: The step of constructing a hidden danger event word library further includes: Collect hidden danger keywords related to transmission line channels, and define multiple hidden danger event categories based on the risk type, time and location information of hidden danger events; collect keywords for each type of hidden danger event and build a keyword thesaurus; use the knowledge graph algorithm to associate the keywords of hidden danger events with the relevant time and location information to obtain a complete hidden danger event thesaurus.

6. The method for perceiving transmission line hidden danger events integrating social information according to claim 5 is characterized in that: The alarm and intervention processing according to the transmission line hidden danger event further includes: According to the severity of hidden danger events, hidden danger events are divided into different risk levels, and hidden danger events are classified by type; Generate alarm information based on the risk level, time and location information of hidden danger events, and send the alarm information to relevant departments or personnel; Track the potential danger events that have been alarmed, record the progress and results of the processing, feedback the processing results to the system, and update the potential danger event database.

7. A power transmission line hidden danger event perception system integrating social information, characterized in that: include: A data collection module, used to collect social information data from multiple social information data sources and convert the social information data into a standardized text sequence; A text encoding module is used to encode the standardized text sequence that integrates social information data word by word and sentence by sentence to convert it into a text vector representation; An event matching module is used to build a hidden danger event vocabulary, match the text vector representation with the keyword vector in the hidden danger event vocabulary, so as to determine the transmission line hidden danger event corresponding to the standardized text sequence, and extract key information of the transmission line hidden danger event; An alarm processing module, used for performing alarm and intervention processing according to the transmission line hidden danger event; The text encoding module is further used for: Perform text preprocessing on the standardized text sequence, including: Perform word segmentation on the text sequence and divide the text into independent words; Remove stop words from the text; Convert the segmented text into word vectors and use the pre-trained word vector model to generate word vector representations; The preprocessed word vector is input into the BiLSTM model based on the input layer of the BiLSTM-CRF model. The input word vector is encoded using a bidirectional long short-term memory network in the BiLSTM layer to generate a word vector representation containing context information. The output vector of the BiLSTM layer is used as the final text vector representation in the output layer. In the input layer, word vectors are used and character-level vector representations are introduced; assuming that each word By character sequence Each character The corresponding vector is expressed as ; For words , whose mixed vector is expressed as: ; in, Represents the vector representation obtained by processing the character sequence through the convolutional neural network CNN; In the BiLSTM layer, a dynamic weight allocation mechanism is introduced to fuse the joint outputs of the forward and backward LSTMs: ; in, are weights dynamically calculated by the self-attention mechanism: ,in, are the parameters that need to be learned. and Respectively represent the first The hidden state of each position; In the output layer, a regularization term is introduced into the loss function of the CRF layer: , in, Indicates The label or state of each time step, Indicates that the BiLSTM network is The hidden state of the time step contains the output of the forward and backward LSTM. The length of the input sequence, i.e. the number of time steps or the number of data points.

8. A terminal comprising a processor and a storage medium; characterized in that: The storage medium is used to store instructions; The processor is used to operate according to the instructions to execute the steps of the method for perceiving power transmission line hidden danger events integrating social information according to any one of claims 1-6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps of the method for perceiving transmission line hidden danger events integrating social information as described in any one of claims 1 to 6 are implemented.

Images