A method and device for intelligently identifying dangerous sources of external damage hazards in distribution networks

Abstract

The present invention discloses a method and device for intelligently identifying the source of external damage hidden dangers in a distribution network, and relates to the technical field of power grid safety management. The method comprises: using a millimeter wave radar to dynamically monitor the spatial environment of a predetermined distribution network, and obtaining real-time external damage hidden danger detection information; extracting the first real-time hidden danger information in the real-time external damage hidden danger detection information, and judging whether it meets the predetermined tracking constraint; if it meets the requirement, activating the laser radar to track and monitor according to the analyzed monitoring strategy, and obtaining the first tracking monitoring information; reading the predetermined external damage hidden danger characteristics, matching the characteristics in the detection information and the tracking monitoring information, and generating hidden danger characteristic parameters; inputting the characteristic parameters into an integrated intelligent model, and identifying the danger category; combining the hidden danger characteristic parameters and the prediction category to generate the danger warning information, and start the emergency warning. Thus, the technical effect of improving the recognition accuracy and real-time performance is achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of power grid safety management, and in particular to a method and device for intelligently identifying dangerous sources of external damage hazards in a distribution network. Background Art

[0002] In the operation of distribution networks, external damage (external destruction) hazards are one of the main causes of power supply interruptions and equipment damage. Existing hidden danger detection methods usually rely on manual inspections or simple sensor monitoring, which have technical problems such as untimely response and low identification and classification accuracy. Summary of the invention

[0003] The present invention provides a method and device for intelligently identifying dangerous sources of external damage hazards in a distribution network, so as to solve the technical problems of untimely response and low identification and classification accuracy in the prior art, and achieve the technical effect of improving identification accuracy and real-time performance.

[0004] In a first aspect, the present invention provides a method for intelligently identifying a potential danger source of external damage in a distribution network, wherein the method comprises:

[0005] Acquire real-time external failure hidden danger detection information, the real-time external failure hidden danger detection information refers to information obtained by dynamically detecting the predetermined spatial environment of the predetermined power distribution network through the millimeter wave radar; extract the first detection information of the first real-time hidden danger in the real-time external failure hidden danger detection information, and judge whether the first detection information meets the predetermined tracking constraints; if it does, activate the laser radar to track and monitor the first real-time hidden danger according to the first predetermined monitoring strategy to obtain the first tracking monitoring information, wherein the first predetermined monitoring strategy is a monitoring strategy determined by analyzing the first detection information; read the predetermined external failure hidden danger characteristics, and perform feature matching in the first detection information and the first tracking monitoring information based on the predetermined external failure hidden danger characteristics to obtain the first hidden danger feature parameter; input the first hidden danger feature parameter into the hazard category integrated identification model to obtain the first predicted hazard category, wherein the hazard category integrated identification model is an integrated model of multiple intelligent models; generate the first hazard warning information in combination with the first hazard feature parameter and the first predicted hazard category, and perform emergency warning on the first hazard warning information.

[0006] In a second aspect, the present invention further provides a device for intelligently identifying dangerous sources of external damage hazards in a distribution network, wherein the device comprises:

[0007] An information acquisition module, wherein the information acquisition module is used to obtain real-time external damage hidden danger detection information, wherein the real-time external damage hidden danger detection information refers to information obtained by dynamically detecting a predetermined spatial environment of a predetermined distribution network through a millimeter wave radar.

[0008] A tracking and distinguishing module is used to extract first detection information of a first real-time hidden danger in the real-time external failure hidden danger detection information, and determine whether the first detection information meets the predetermined tracking constraint.

[0009] A tracking and monitoring module is used to activate the laser radar to track and monitor the first real-time hidden danger according to a first predetermined monitoring strategy if it meets the requirements, and obtain first tracking and monitoring information, wherein the first predetermined monitoring strategy is a monitoring strategy determined after analyzing the first detection information.

[0010] A feature matching module is used to read a predetermined external damage hidden danger feature, and perform feature matching in the first detection information and the first tracking and monitoring information based on the predetermined external damage hidden danger feature to obtain a first hidden danger feature parameter.

[0011] A hazard identification module is used to input the first hidden danger characteristic parameter into a hazard category integrated identification model to obtain a first predicted hazard category, wherein the hazard category integrated identification model is an integrated model of multiple intelligent models.

[0012] An early warning module is used to generate first danger early warning information by combining the first hidden danger characteristic parameter with the first predicted danger category, and to perform emergency early warning on the first danger early warning information.

[0013] The present invention discloses a method and device for intelligently identifying the source of hidden dangers of external failure of distribution network, including: obtaining real-time detection information of hidden dangers of external failure, which is obtained by dynamically detecting the spatial environment of a predetermined distribution network through a millimeter wave radar; extracting the detection information of the first real-time hidden danger, and judging whether it meets the predetermined tracking constraint; if it meets the requirement, activating the laser radar to track and monitor the hidden danger according to the first predetermined monitoring strategy, and obtaining the first tracking monitoring information, wherein the first predetermined monitoring strategy is obtained by analyzing the first detection information; reading the characteristics of the predetermined hidden dangers of external failure, and performing feature matching in the first detection information and the first tracking monitoring information, and obtaining the first hidden danger characteristic parameter; inputting the first hidden danger characteristic parameter into the integrated identification model of the hazard category, which is integrated by multiple intelligent models, and obtaining the first predicted hazard category; combining the first hidden danger characteristic parameter and the first predicted hazard category to generate the first hazard warning information, and carrying out emergency warning. The method and device for intelligently identifying the source of hidden dangers of external failure of distribution network disclosed by the present invention solve the technical problems of untimely response and low accuracy of identification and classification, and achieve the technical effect of improving the recognition accuracy and real-time performance. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 It is a flow chart of a method for intelligently identifying dangerous sources of external damage hazards in a distribution network according to the present invention;

[0015] Figure 2 It is a structural schematic diagram of a device for intelligently identifying dangerous sources of external damage hazards in a distribution network according to the present invention.

[0016] Explanation of the reference numerals: information acquisition module 11 , tracking and identification module 12 , tracking and monitoring module 13 , feature matching module 14 , danger identification module 15 , and early warning module 16 . DETAILED DESCRIPTION

[0017] The technical solution provided in the embodiments of the present invention is to solve the technical problems of untimely response and low recognition and classification accuracy in the prior art. The overall idea adopted is as follows:

[0018] First, obtain real-time external damage hidden danger detection information, which is information obtained by dynamically detecting the predetermined spatial environment of the predetermined distribution network through the millimeter wave radar. Then, extract the first detection information of the first real-time hidden danger in the real-time external damage hidden danger detection information, and judge whether the first detection information meets the predetermined tracking constraint. Then, if the first detection information meets the predetermined tracking constraint, activate the laser radar to track and monitor the first real-time hidden danger according to the first predetermined monitoring strategy to obtain the first tracking monitoring information, wherein the first predetermined monitoring strategy is a monitoring strategy determined after analyzing the first detection information. Next, read the predetermined external damage hidden danger feature, and perform feature matching in the first detection information and the first tracking monitoring information based on the predetermined external damage hidden danger feature to obtain the first hidden danger feature parameter. Then, input the first hidden danger feature parameter into the hazard category integrated identification model to obtain the first predicted hazard category, wherein the hazard category integrated identification model is an integrated model of multiple intelligent models. Finally, combine the first hidden danger feature parameter with the first predicted hazard category to generate the first hazard warning information, and perform emergency warning on the first hazard warning information.

[0019] The above technical solution will be described in detail below in conjunction with the accompanying drawings and specific implementation methods of the specification to better understand the above technical solution. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all of the embodiments of the present invention. It should be understood that the present invention is not limited to the example embodiments used only to explain the present invention. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention. In addition, it should be noted that, for the convenience of description, only the parts related to the present invention are shown in the drawings, rather than all of them.

[0020] Embodiment 1

[0021] Figure 1 The present invention is a flow chart of a method for intelligently identifying a potential source of damage to a distribution network, the method comprising:

[0022] Real-time external damage hidden danger detection information is obtained, where the real-time external damage hidden danger detection information refers to information obtained by dynamically detecting a predetermined spatial environment of a predetermined power distribution network through a millimeter wave radar.

[0023] Specifically, first, the millimeter-wave radar is used to detect the relative position information between other objects in the space surrounding the target power grid and the power grid, and the output is real-time external damage hidden danger detection information. Specifically, the millimeter-wave radar is used to detect other objects in the space surrounding the target power grid that may pose external damage risks to the power grid, including trees, buildings, birds, etc.

[0024] Optionally, the real-time external damage potential detection information includes the location, shape, size, distance, speed and other information of multiple external damage potential objects in a predetermined spatial environment of a predetermined distribution network.

[0025] For example, first, millimeter wave radars are deployed around the predetermined distribution network. The deployment location and number of millimeter wave radars need to be determined according to the scale and environment of the power grid. Then, the millimeter wave radar is activated to dynamically detect the predetermined space environment with a predetermined detection accuracy and detection frequency. By sending millimeter wave signals and receiving reflected signals, the distance, position, speed, shape and other information of the external damage potential objects in the predetermined space environment are analyzed.

[0026] The deployment of millimeter-wave radar can help monitor environmental changes around the power grid in real time, obtain information on external damage risks in a timely manner, and provide an immediate and accurate information basis for subsequent analysis and identification.

[0027] Extract the first detection information of the first real-time hidden danger in the real-time external failure hidden danger detection information, and determine whether the first detection information meets the predetermined tracking constraint.

[0028] Specifically, the real-time external failure hidden danger detection information includes multiple real-time hidden dangers, corresponding to multiple external failure hidden danger objects in a predetermined spatial environment of a predetermined distribution network, and the first real-time hidden danger is any one of the multiple real-time hidden dangers, which is obtained by random extraction from the real-time external failure hidden danger detection information.

[0029] Specifically, the predetermined tracking constraints involve multiple constraint dimensions, including distance constraints (the minimum safe distance between the hidden danger object and the distribution network equipment or lines), speed constraints (the maximum allowable speed of the hidden danger object), and size constraints (the maximum allowable size of the hidden danger object).

[0030] Optionally, determining whether the first detection information meets the predetermined tracking constraints includes: calculating the distance from the distribution network equipment. If the first distance is less than the set minimum safety distance, the hidden danger object may contact the distribution network equipment or line, and further tracking and monitoring are required. Compare the speed of the hidden danger object with the set maximum allowable speed. If the first speed is greater than the maximum allowable speed, the hidden danger object moves too fast, which may increase the danger and needs to be tracked. Compare the size of the hidden danger object with the set maximum allowable size. If the first size is greater than the maximum allowable size, the volume of the hidden danger object is too large, which may cause safety problems and needs to be tracked.

[0031] In some embodiments, the first detection information includes a first distance, a first speed, and a first size of the first real-time hidden danger;

[0032] A hidden danger risk assessment function is introduced to analyze the first detection information to obtain a first risk, wherein the hidden danger risk assessment function is expressed as follows:

[0033] ;

[0034] in, Refers to the first real-time hidden danger The first risk level, , , are respectively the first distance, the first speed, and the first size, They refer to the first risk factor, the second risk factor, and the third risk factor respectively, and ;

[0035] A predetermined risk level threshold in the predetermined tracking constraint is extracted, and when the first risk level reaches the predetermined risk level threshold, the first detection information complies with the predetermined tracking constraint.

[0036] Specifically, the detection information of the first real-time hidden danger is extracted. The first real-time hidden danger includes: a first distance, that is, the distance from the first hidden danger object to the distribution network equipment or line; a first speed, that is, the moving speed of the first hidden danger object, the first speed is a vector, the direction of the vector represents the direction of movement, and the magnitude of the vector represents the speed of movement; a first size, that is, the volume or size of the first hidden danger object.

[0037] Specifically, the closer the first hidden danger object is to the distribution network equipment or line, the faster the approach or movement speed, and the larger the size, the greater the threat to the predetermined distribution network. Through the above-mentioned hidden danger risk assessment function, combined with the first risk coefficient , the second risk factor , the third risk factor By adjusting the weights, the risk level of the first detection information corresponding to the first hidden danger object can be comprehensively evaluated and quantified, so as to facilitate the determination of the objects that need to be tracked and monitored.

[0038] Specifically, the predetermined danger level threshold is a pre-set value used to determine when the potential danger object needs to be tracked and monitored. If the first danger level is greater than or equal to the predetermined danger level threshold, it is considered that the potential danger of the potential danger object reaches the standard that requires tracking and monitoring.

[0039] Optionally, the predetermined danger threshold can be an empirical coefficient determined by experts or engineers based on long-term accumulated practical operation experience or a statistical parameter (such as confidence) based on historical external failure events. By adopting a combination of empirical coefficients and statistical parameters, a more reliable and flexible predetermined danger threshold can be provided for external failure hazard monitoring in different scenarios, thereby better protecting the safety of the distribution network.

[0040] Exemplarily, to determine the predetermined risk threshold based on a statistical model, first, collect historical data on external damage events, including relevant features of each event (such as distance, speed, size, etc.). Then, use the statistical model to analyze the above-mentioned relevant features, determine their relevance to the external damage event, and calculate the risk. Finally, based on a set confidence level (such as 95% or 99%), determine the confidence interval of the risk, so as to set a reasonable predetermined risk threshold.

[0041] Through the above steps, real-time external hidden danger detection information can be systematically processed to ensure efficient and accurate identification of hidden danger objects, thereby excluding low-threat and non-threat objects and improving the accuracy and efficiency of identifying hidden danger objects.

[0042] If it is met, activate the laser radar to track and monitor the first real-time hidden danger according to a first predetermined monitoring strategy to obtain first tracking and monitoring information, wherein the first predetermined monitoring strategy is a monitoring strategy determined after analyzing the first detection information.

[0043] Specifically, when it is determined that the first detection information meets the predetermined tracking constraint (the risk level reaches the threshold), the laser radar system is activated to prepare for monitoring and real-time capture of the target's location information, movement path, and its relative relationship with the distribution network equipment. Among them, the laser radar monitoring is used to obtain tracking monitoring information with a high-precision sampling rate, thereby providing higher-precision and higher-quality information for subsequent analysis and classification.

[0044] In some embodiments, the first predetermined monitoring strategy is determined according to a first monitoring space range generated based on the first distance.

[0045] Optionally, determine a first predetermined monitoring strategy, first analyze the first detection information (first distance, first speed, first size), evaluate the potential degree of danger and the area where hidden dangers may occur, and generate a first monitoring space range. The first monitoring space range refers to the monitoring space range that the laser radar needs to cover, which is a dynamically adjusted area.

[0046] Optionally, a corresponding first predetermined monitoring strategy is formulated in combination with the monitoring spatial range and other parameters of the detection information (such as speed). For example, if the target is close, the monitoring frequency needs to be increased; if the target speed is fast, the spatial resolution of the monitoring needs to be improved.

[0047] The predetermined external failure hidden danger feature is read, and based on the predetermined external failure hidden danger feature, feature matching is performed in the first detection information and the first tracking and monitoring information to obtain a first hidden danger feature parameter.

[0048] Specifically, the predetermined external damage hazard features are a set of key features obtained through analysis that have a high correlation with the external damage hazard, including but not limited to the shape, size, distance, speed, etc. of the object.

[0049] Exemplarily, first, read the predetermined external damage hidden danger feature. The external damage hidden danger feature may be preset based on historical data and professional knowledge, or may be selected through correlation analysis or principal component analysis. Then, feature matching is performed from the first detection information and the first tracking monitoring information obtained by the millimeter wave radar through formula recognition or machine learning algorithm. The feature matching with the predetermined external damage hidden danger feature is obtained and set as the first hidden danger feature parameter.

[0050] The first hidden danger characteristic parameter is input into a hazard category integrated identification model to obtain a first predicted hazard category, wherein the hazard category integrated identification model is an integrated model of multiple intelligent models.

[0051] In some embodiments, constructing a hazard category integrated identification model includes:

[0052] Extract the first hazard category from the predetermined hazard categories; traverse the first historical record of the first hazard category in the historical external breach hazard source identification record; obtain the first historical hazard characteristic parameter in the first historical record, the first historical hazard characteristic parameter including the first historical detection hazard characteristic parameter and the first historical monitoring hazard characteristic parameter; perform supervised learning on a first training data set based on the first historical detection hazard characteristic parameter and the first hazard category to obtain a first initial recognition model; perform supervised learning on a second training data set based on the first historical monitoring hazard characteristic parameter and the first hazard category to obtain a second initial recognition model; form a third training data set based on the initial recognition results of the first initial recognition model and the second initial recognition model and the first hazard category, and perform supervised learning on the third training data set to obtain a meta-recognition model; build the first initial recognition model, the second initial recognition model and the meta-recognition model to obtain the hazard category integrated recognition model.

[0053] Specifically, the predetermined hazard categories include vegetation hazard sources (short circuits caused by rapid growth of trees or broken branches, overheating of equipment caused by excessive growth of weeds), man-made hazard sources (construction activities, accidents, human damage, illegal operations, etc.) and animal hazard sources.

[0054] Specifically, in the existing external damage hidden danger source identification records, the historical records related to the first hazard category are traversed to extract the first historical records, which include the first historical detection hidden danger characteristic parameters and the first historical monitoring hidden danger characteristic parameters.

[0055] Specifically, based on the first historical detection hidden danger characteristic parameters and the first hazard category, a first training data set is constructed, and the data set is used for model training to obtain a first initial recognition model. Based on the first historical monitoring hidden danger characteristic parameters and the first hazard category, a second training data set is constructed, and the data set is used for model training to obtain a second initial recognition model.

[0056] Furthermore, based on the initial recognition results of the first initial recognition model and the second initial recognition model, combined with the first hazard category, a third training data set is constructed. The third training data set is studied to obtain a higher-level meta-recognition model, which can be considered as a student model obtained by fusing the first initial recognition model and the second initial recognition model for knowledge distillation, and has higher recognition accuracy and generalization ability.

[0057] Furthermore, the model integrates the first initial recognition model, the second initial recognition model and the meta-recognition model through ensemble learning to construct the final hazard category integrated recognition model. Model integration helps to improve the accuracy and robustness of overall recognition and achieve more efficient hazard category recognition.

[0058] Through the above steps, the constructed integrated hazard category identification model can comprehensively utilize historical data and hidden danger characteristic information at different stages to improve the accuracy of identifying potential hazard categories.

[0059] The first hidden danger characteristic parameter and the first predicted danger category are combined to generate first danger warning information, and an emergency warning is performed on the first danger warning information.

[0060] Specifically, the first hazard warning information is generated by combining the first hidden danger characteristic parameter and the first predicted hazard category. The first hazard warning information includes the hazard category, the location of the hazard source (hidden danger object), the speed of the hazard source, the countdown of the dangerous event, etc. Emergency warning is performed on the first hazard warning information. This may include sending warning information to relevant personnel, starting emergency response procedures, dispatching maintenance personnel to conduct inspections and repairs, etc.

[0061] In some implementations, the first initial recognition model, the second initial recognition model, and the meta-recognition model are constructed to obtain the hazard category integrated recognition model, including:

[0062] The first initial recognition model and the second initial recognition model are used as the first recognition layer of the hazard category integrated recognition model, and the meta recognition model is used as the second recognition layer of the hazard category integrated recognition model.

[0063] Specifically, the first recognition layer of the hazard category integrated recognition model is set by the first initial recognition model and the second initial recognition model. The first recognition layer is a 1PASS recognition layer constructed based on ensemble learning, which is used for first-layer recognition. The second recognition layer of the hazard category integrated recognition model configured based on the meta-recognition model is a 2PASS recognition layer that plays a verification role. By setting up a dual recognition layer, the accuracy and confidence of hazard category recognition can be further improved.

[0064] Furthermore, the method further comprises:

[0065] Read predetermined environmental characteristics; obtain an environmental monitoring log of the predetermined spatial environment, the environmental monitoring log including first environmental monitoring information in a first time period; perform feature traversal on the first environmental monitoring information based on the predetermined environmental characteristics to obtain first target environmental characteristic parameters in the first time period; construct a non-contact accident tree based on historical non-contact external failure hazard records; perform risk analysis on the first target environmental characteristic parameters in combination with the historical non-contact minimum cut set set of the non-contact accident tree to obtain second hazard warning information.

[0066] Specifically, the predetermined environmental characteristics refer to a variety of environmental characteristic indicator types that need to be obtained, in other words, a variety of environmental monitoring dimensions that need to be collected, including, for example, ambient temperature, ambient humidity, radiation intensity, convection intensity, etc.

[0067] Specifically, the content of the environmental monitoring log includes a variety of external environmental factors, and the richness of the monitoring dimensions reflected by the environmental monitoring log is greater than the above-mentioned predetermined environmental features. In other words, the above-mentioned predetermined environmental features are a subset of the detection dimensions contained in the environmental monitoring log. Exemplarily, the content of the environmental monitoring log may include temperature, humidity, air quality, concentration of corrosive substances, etc.

[0068] Specifically, the monitoring log records the environmental monitoring information of multiple time periods, among which the environmental monitoring information in the first time period (such as environmental data of a certain day or a certain week) is the environmental monitoring information in the current analysis time window. The current analysis time window is determined based on the analysis requirements of the target scenario to ensure that the acquired environmental monitoring information has good timeliness and data quality, and to avoid the impact of outdated data on subsequent analysis.

[0069] Specifically, non-contact external damage risk records correspond to non-contact external damage hazards, including non-contact external damage hazards caused by natural disaster risk sources and non-contact external damage hazards caused by environmental risk sources. Among them, natural disaster risk sources include: severe convection, floods, snowstorms, lightning strikes, earthquakes, etc. Environmental risk sources include: chemical pollution, dust pollution, dust concentration, etc.

[0070] Specifically, the non-contact accident tree includes multiple historical non-contact minimum cut sets corresponding to multiple non-contact accidents, wherein the historical non-contact minimum cut set is a set of basic events that lead to the occurrence of non-contact accidents.

[0071] Specifically, characteristic parameters in the current environment, such as temperature, humidity, and equipment status, are obtained as the first target environment characteristic parameters, and the first target environment characteristic parameters are matched with the event characteristics in the historical non-contact minimum cut set, and the similarity between the current environment characteristic parameters and historical accidents is analyzed. If the parameter combination in the current environment is similar to the combination in the historical minimum cut set, the risk is high. Then, based on the degree of matching and the scope of influence, a second hazard warning information is generated to indicate the potential risk of non-contact accidents.

[0072] Optionally, the second hazard warning information includes identified high-risk factors, possible accident types, and recommended preventive measures, etc. By combining the historical non-contact minimum cut set set to perform risk analysis on the first target environment characteristic parameters, the occurrence of potential accidents can be predicted more accurately.

[0073] In some embodiments, a non-contact accident tree is constructed based on historical non-contact external failure risk records, including:

[0074] According to the first non-contact accident record in the historical non-contact external damage hazard record, the first non-contact accident record is used as the first top event; the first accident factor of the first top event is analyzed in combination with the predetermined environmental characteristics; according to the first top event and the first accident factor, a first minimum cut set is determined, and the historical non-contact minimum cut set set is formed.

[0075] Specifically, the non-contact accident tree includes multiple top events, each top event corresponds to a non-contact accident, and the multiple top events are determined based on multiple groups of non-contact accident records.

[0076] Specifically, the predetermined environmental features related to the first top event are analyzed to identify the accident factors that lead to the top event. The accident factors refer to the basic events or conditions that directly or indirectly lead to the top event, and the predetermined environmental features include weather conditions, equipment status, external interference, etc.

[0077] Specifically, based on the relationship between the top event and the accident factor, the first minimum cut set leading to the top event is analyzed and determined. The minimum cut set refers to the minimum basic event set leading to the top event, representing the minimum condition combination for the accident to occur.

[0078] Furthermore, the identified first minimum cut set is added to the historical non-contact minimum cut set set. The historical non-contact minimum cut set set includes all the identified minimum cut sets that lead to non-contact accidents, reflecting the basic conditions and patterns of historical accidents.

[0079] Through the above steps, the non-contact accident tree formed can effectively reflect the pattern of historical accidents and provide a scientific basis for current and future risk management.

[0080] In summary, the method for intelligently identifying the source of external damage hazards in a distribution network provided by the present invention has the following technical effects:

[0081] By acquiring real-time external damage hidden danger detection information, which is obtained by dynamically detecting the spatial environment of the predetermined distribution network through millimeter wave radar; extracting the detection information of the first real-time hidden danger, and judging whether it meets the predetermined tracking constraint; if it meets the requirement, activating the laser radar to track and monitor the hidden danger according to the first predetermined monitoring strategy, and obtaining the first tracking monitoring information, wherein the first predetermined monitoring strategy is obtained by analyzing the first detection information; reading the characteristics of the predetermined external damage hidden danger, and matching the characteristics in the first detection information and the first tracking monitoring information, and obtaining the first hidden danger characteristic parameter; inputting the first hidden danger characteristic parameter into the hazard category integrated identification model, which is integrated by multiple intelligent models, and obtaining the first predicted hazard category; combining the first hidden danger characteristic parameter and the first predicted hazard category to generate the first hazard warning information, and conducting an emergency warning. Thus, the technical effect of improving the recognition accuracy and real-time performance is achieved.

[0082] Embodiment 2

[0083] Figure 2 This is a schematic diagram of the structure of a device for intelligently identifying dangerous sources of external damage to a distribution network according to the present invention. Figure 1 The flow chart of the intelligent identification method of the potential danger source of the external failure of the distribution network in the present invention can be shown as follows: Figure 2 The structure shown is implemented.

[0084] Based on the same concept as the method for intelligently identifying external damage sources of distribution network damage in the above embodiment, the present invention also provides an intelligent device for identifying external damage sources of distribution network damage, including:

[0085] The information acquisition module 11 is used to acquire real-time external damage hidden danger detection information, where the real-time external damage hidden danger detection information refers to information obtained by dynamically detecting a predetermined spatial environment of a predetermined power distribution network through a millimeter wave radar.

[0086] The tracking and judging module 12 is used to extract the first detection information of the first real-time hidden danger in the real-time external failure hidden danger detection information, and judge whether the first detection information meets the predetermined tracking constraint.

[0087] The tracking and monitoring module 13 is used to activate the laser radar to track and monitor the first real-time hidden danger according to the first predetermined monitoring strategy if it meets the requirements, and obtain first tracking and monitoring information, wherein the first predetermined monitoring strategy is a monitoring strategy determined after analyzing the first detection information.

[0088] The feature matching module 14 is used to read the predetermined external damage hidden danger feature, and perform feature matching in the first detection information and the first tracking and monitoring information based on the predetermined external damage hidden danger feature to obtain a first hidden danger feature parameter.

[0089] The hazard identification module 15 is used to input the first hidden danger characteristic parameter into the hazard category integrated identification model to obtain a first predicted hazard category, wherein the hazard category integrated identification model is an integrated model of multiple intelligent models.

[0090] The warning module 16 is used to generate first danger warning information by combining the first hidden danger characteristic parameter with the first predicted danger category, and to perform emergency warning on the first danger warning information.

[0091] In some implementations, the first detection information in the tracking and distinguishing module 12 includes a first distance, a first speed, and a first size of the first real-time hidden danger, and the tracking and distinguishing module 12 includes:

[0092] The risk calculation unit is used to introduce a hidden danger risk assessment function to analyze the first detection information to obtain a first risk, wherein the hidden danger risk assessment function is expressed as follows:

[0093] ;

[0094] in, Refers to the first real-time hidden danger The first risk level, , , are respectively the first distance, the first speed, and the first size, They refer to the first risk factor, the second risk factor, and the third risk factor respectively, and .

[0095] The threshold determination unit is used to extract a predetermined risk threshold in the predetermined tracking constraint. When the first risk reaches the predetermined risk threshold, the first detection information meets the predetermined tracking constraint.

[0096] In some embodiments, the execution step of the device includes: determining the first predetermined monitoring strategy according to a first monitoring space range generated based on the first distance.

[0097] In some embodiments, the hazard identification module 15 includes:

[0098] The category traversal unit is used to extract the first hazard category from the predetermined hazard categories.

[0099] The historical record traversal unit is used to traverse the first historical record of the first hazard category in the historical external breach potential hazard source identification record.

[0100] The characteristic parameter acquisition unit is used to acquire the first historical hidden danger characteristic parameter in the first historical record, where the first historical hidden danger characteristic parameter includes the first historical detection hidden danger characteristic parameter and the first historical monitoring hidden danger characteristic parameter.

[0101] The first training data set learning unit is used to perform supervised learning on a first training data set formed based on the first historical hidden danger detection characteristic parameters and the first hazard category to obtain a first initial recognition model.

[0102] The second training data set learning unit is used to perform supervised learning on a second training data set formed based on the first historical monitoring hidden danger characteristic parameters and the first hazard category to obtain a second initial recognition model.

[0103] The meta-recognition model training unit is used to form a third training data set based on the initial recognition results of the first initial recognition model and the second initial recognition model and the first hazard category, and to supervise the learning of the third training data set to obtain a meta-recognition model.

[0104] The integrated recognition model building unit is used to build the first initial recognition model, the second initial recognition model and the meta-recognition model to obtain the hazard category integrated recognition model.

[0105] In some embodiments, the integrated recognition model building unit in the hazard recognition module 15 is used to:

[0106] The first initial recognition model and the second initial recognition model are used as the first recognition layer of the hazard category integrated recognition model, and the meta recognition model is used as the second recognition layer of the hazard category integrated recognition model.

[0107] In some embodiments, the apparatus further comprises:

[0108] The environmental characteristic reading unit is used to read the predetermined environmental characteristic.

[0109] A monitoring data extraction unit, configured to obtain an environmental monitoring log of the predetermined space environment, wherein the environmental monitoring log includes first environmental monitoring information in a first time period;

[0110] The target environment characteristic parameter traversal unit is used to perform characteristic traversal on the first environment monitoring information based on the predetermined environment characteristic to obtain the first target environment characteristic parameter in the first time period.

[0111] The non-contact accident tree construction unit is used to construct a non-contact accident tree based on historical non-contact external failure hazard records.

[0112] The risk analysis unit is used to perform risk analysis on the first target environment characteristic parameter in combination with the historical non-contact minimum cut set set of the non-contact accident tree to obtain second danger warning information.

[0113] In some embodiments, the non-contact fault tree construction unit includes:

[0114] The top event determination unit is used to determine the first non-contact accident record in the historical non-contact external damage risk record and use the first non-contact accident record as the first top event.

[0115] An accident factor analysis unit is used to analyze the first accident factor of the first top event in combination with the predetermined environmental characteristics.

[0116] A minimum cut set determination unit is used to determine a first minimum cut set according to the first top event and the first accident factor, and form the historical non-contact minimum cut set set.

[0117] It should be understood that the embodiments mentioned in this specification focus on their differences from other embodiments. The specific embodiments in the aforementioned embodiment one are also applicable to the intelligent identification device for hidden danger sources of external damage in the distribution network described in embodiment two. For the sake of brevity of the specification, they will not be further elaborated here.

[0118] It should be understood that the embodiments disclosed in the present invention and the above description can enable those skilled in the art to use the present invention to implement the present invention. At the same time, the present invention is not limited to the above-mentioned embodiments. It should be understood that those skilled in the art can still modify the technical solutions recorded in the above-mentioned embodiments, or replace some of the technical features therein by equivalents; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should be included in the protection scope of the present invention.

Claims

1. A method for intelligently identifying dangerous sources of external damage hazards in a distribution network, characterized in that: include: Acquire real-time external damage hidden danger detection information, wherein the real-time external damage hidden danger detection information refers to information obtained by dynamically detecting a predetermined spatial environment of a predetermined distribution network through a millimeter wave radar; Extracting first detection information of a first real-time hidden danger from the real-time external failure hidden danger detection information, and determining whether the first detection information meets a predetermined tracking constraint; If it is in compliance, activating the laser radar to perform tracking and monitoring of the first real-time hidden danger according to a first predetermined monitoring strategy to obtain first tracking and monitoring information, wherein the first predetermined monitoring strategy is a monitoring strategy determined after analyzing the first detection information; Reading a predetermined external failure hidden danger feature, and performing feature matching in the first detection information and the first tracking and monitoring information based on the predetermined external failure hidden danger feature to obtain a first hidden danger feature parameter; Inputting the first hidden danger characteristic parameter into a hazard category integrated identification model to obtain a first predicted hazard category, wherein the hazard category integrated identification model is an integrated model of multiple intelligent models; generating first hazard warning information by combining the first hidden danger characteristic parameter with the first predicted hazard category, and performing emergency warning on the first hazard warning information; Wherein, the first detection information includes a first distance, a first speed and a first size of the first real-time hidden danger; A hidden danger risk assessment function is introduced to analyze the first detection information to obtain a first risk, wherein the hidden danger risk assessment function is expressed as follows: ； in, Refers to the first real-time hidden danger The first risk level, , , are respectively the first distance, the first speed, and the first size, , , They refer to the first risk factor, the second risk factor, and the third risk factor respectively, and ; A predetermined risk level threshold in the predetermined tracking constraint is extracted, and when the first risk level reaches the predetermined risk level threshold, the first detection information complies with the predetermined tracking constraint.

2. According to claim 1, a method for intelligently identifying dangerous sources of external damage hazards in a distribution network is characterized in that: The first predetermined monitoring strategy is determined according to a first monitoring space range generated based on the first distance.

3. According to claim 1, a method for intelligently identifying dangerous sources of external damage hazards in a distribution network is characterized in that: include: extracting a first hazard category from among the predetermined hazard categories; Traversing the first historical records of the first hazard category in the historical external failure potential hazard source identification records; Acquire first historical hidden danger characteristic parameters in the first historical record, where the first historical hidden danger characteristic parameters include first historical hidden danger detection characteristic parameters and first historical hidden danger monitoring characteristic parameters; Performing supervised learning on a first training data set formed based on the first historical hidden danger detection characteristic parameter and the first hazard category to obtain a first initial recognition model; Performing supervised learning on a second training data set formed based on the first historical monitoring hidden danger characteristic parameters and the first hazard category to obtain a second initial recognition model; Based on the initial recognition results of the first initial recognition model and the second initial recognition model and the first hazard category, a third training data set is formed, and supervised learning of the third training data set is performed to obtain a meta-recognition model; The first initial recognition model, the second initial recognition model and the meta-recognition model are constructed to obtain the hazard category integrated recognition model.

4. According to claim 3, a method for intelligently identifying dangerous sources of external damage hazards in a distribution network is characterized in that: include: The first initial recognition model and the second initial recognition model are used as the first recognition layer of the hazard category integrated recognition model, and the meta recognition model is used as the second recognition layer of the hazard category integrated recognition model.

5. According to claim 1, a method for intelligently identifying dangerous sources of external damage hazards in a distribution network is characterized in that: include: Reading predetermined environmental characteristics; Acquire an environmental monitoring log of the predetermined space environment, wherein the environmental monitoring log includes first environmental monitoring information in a first time period; Performing feature traversal on the first environment monitoring information based on the predetermined environment feature to obtain a first target environment feature parameter in the first time period; Construct a non-contact accident tree based on historical non-contact external failure hazard records; A risk analysis is performed on the first target environment characteristic parameter in combination with the historical non-contact minimum cut set set of the non-contact accident tree to obtain second hazard warning information.

6. According to claim 5, a method for intelligently identifying dangerous sources of external damage hazards in a distribution network is characterized in that: include: According to the first non-contact accident record in the historical non-contact external damage risk record, the first non-contact accident record is used as the first top event; Analyzing a first accident factor of the first overhead event in combination with the predetermined environmental feature; A first minimum cut set is determined according to the first top event and the first accident factor, and the historical non-contact minimum cut set set is formed.

7. An intelligent identification device for dangerous sources of external damage to the distribution network, characterized in that: The device is used to execute a method for intelligently identifying a source of hidden danger of external damage in a distribution network according to any one of claims 1 to 6, and the device comprises: An information acquisition module, the information acquisition module is used to obtain real-time external damage hidden danger detection information, the real-time external damage hidden danger detection information refers to information obtained by dynamically detecting a predetermined spatial environment of a predetermined power distribution network through a millimeter wave radar; A tracking and distinguishing module, the tracking and distinguishing module is used to extract first detection information of a first real-time hidden danger in the real-time external failure hidden danger detection information, and determine whether the first detection information meets the predetermined tracking constraint; A tracking and monitoring module, wherein if the condition is met, the laser radar is activated to perform tracking and monitoring of the first real-time hidden danger according to a first predetermined monitoring strategy to obtain first tracking and monitoring information, wherein the first predetermined monitoring strategy is a monitoring strategy determined after analyzing the first detection information; a feature matching module, the feature matching module being used to read a predetermined external failure hidden danger feature, and perform feature matching in the first detection information and the first tracking and monitoring information based on the predetermined external failure hidden danger feature to obtain a first hidden danger feature parameter; A hazard identification module, wherein the hazard identification module is used to input the first hidden danger characteristic parameter into a hazard category integrated identification model to obtain a first predicted hazard category, wherein the hazard category integrated identification model is an integrated model of multiple intelligent models; An early warning module is used to generate first danger early warning information by combining the first hidden danger characteristic parameter with the first predicted danger category, and to perform emergency early warning on the first danger early warning information.

Images