Lightning early warning system for port dangerous cargo major dangerous source tank field

Abstract

The invention provides a lightning early warning system for a port dangerous cargo major dangerous source tank field, and relates to the field of lightning early warning. The lightning early warning system uses a data processing technology based on artificial intelligence to carry out data set splitting on a time sequence data set of lightning characteristic parameters; performing relative position prediction based on time sequence characteristics on the time sequence data set of the split lightning position, and performing message constraint transfer based on the time sequence characteristics on the time sequence data set of the atmospheric electric field intensity value at the same time; therefore, the lightning early warning level is automatically determined according to the predicted value of the lightning relative position data and the atmospheric electric field intensity time sequence fluctuation coding characteristics after time sequence transmission. Therefore, by monitoring the change of the intensity of the atmospheric electric field in real time and accurately capturing the position and the moving path of the lightning, the occurrence time and the intensity of the lightning can be predicted more accurately, and then a reliable safety guarantee can be provided for a dangerous cargo tank field of a port.

Description

Technical Field

[0001] The present application relates to the field of lightning warning, and more specifically, to a lightning warning system for a major hazard source tank area for dangerous goods at a port. Background Art

[0002] In the past, lightning warnings mainly relied on simple meteorological observation data and empirical judgment. This method lacks accurate monitoring of the core physical process of lightning occurrence, and it is difficult to accurately capture the rapid changes in the atmospheric electric field intensity when lightning approaches, as well as the specific location and movement trajectory of lightning. Moreover, traditional methods cannot effectively process massive, dynamically changing monitoring data, and the prediction accuracy of the time, location and intensity of lightning is low. In the face of complex and changing port environments, it is impossible to provide reliable early warning information for dangerous goods tank areas with major hazard sources in a timely and accurate manner, and it is difficult to meet the high requirements of tank areas for lightning protection.

[0003] Therefore, an optimized lightning warning scheme for the tank area with major hazard sources of dangerous goods in ports is desired. Summary of the invention

[0004] In order to solve the above technical problems, the present application is proposed. The embodiment of the present application provides a lightning warning system for a major hazard source tank area of ​​dangerous goods in a port.

[0005] According to one aspect of the present application, a lightning warning system for a major hazard source tank area of ​​dangerous goods in a port is provided, which comprises:

[0006] A lightning parameter acquisition module, used to obtain a time series data set of lightning characteristic parameters acquired by the lightning detection module, wherein the lightning characteristic parameters include an atmospheric electric field intensity value acquired by an atmospheric electric field instrument and a lightning position acquired by a lightning location system;

[0007] A data splitting module, used for splitting the time series data set of the lightning characteristic parameters to obtain a time series data set of the atmospheric electric field intensity value and a time series data set of the lightning position;

[0008] An analysis module is used to determine the lightning warning level based on the time series characteristics of the time series data set of the atmospheric electric field strength values ​​and the predicted relative position of the time series data set of the lightning positions, wherein the analysis module includes: a coding unit, used to perform electric field strength spatial constraint time series propagation and relative position prediction on the time series data set of the atmospheric electric field strength values ​​and the time series data set of the lightning positions, respectively, to obtain the atmospheric electric field strength time series fluctuation coding characteristics and the predicted value of the lightning relative position data; a level determination unit, used to determine the lightning warning level based on the predicted value of the lightning relative position data and the atmospheric electric field strength time series fluctuation coding characteristics.

[0009] Compared with the prior art, the lightning warning system for the major hazard source tank area of ​​dangerous goods in the port provided by the present application uses artificial intelligence-based data processing technology to split the time series data set of lightning characteristic parameters, and then performs relative position prediction based on the time series characteristics on the split time series data set of lightning positions, and at the same time performs message constraint transmission based on the time series characteristics on the time series data set of atmospheric electric field strength values, so as to automatically determine the lightning warning level according to the predicted value of the lightning relative position data and the time series fluctuation coding characteristics of the atmospheric electric field strength after the time series transmission. In this way, by real-time monitoring of the changes in the atmospheric electric field strength and accurately capturing the location and movement path of lightning, the occurrence time and intensity of lightning can be predicted more accurately, thereby providing reliable safety protection for the port dangerous goods tank area. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the prior art descriptions. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work. In the drawings:

[0011] Figure 1 This is a system block diagram of a lightning warning system for a tank area with major hazard sources for dangerous goods at a port according to an embodiment of the present application.

[0012] Figure 2 It is a block diagram of the analysis module in the lightning warning system for the tank area of ​​dangerous goods in the port, which is a major hazard source, according to an embodiment of the present application.

[0013] Figure 3 It is a block diagram of the coding unit in the lightning warning system of the tank area of ​​the major hazard source of dangerous goods in the port according to the embodiment of the present application.

[0014] Figure 4 It is a block diagram of the atmospheric electric field strength encoding subunit in the lightning warning system of the tank area of ​​major hazard sources of dangerous goods in the port according to an embodiment of the present application. DETAILED DESCRIPTION

[0015] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described here.

[0016] In the modern logistics system, ports are important hubs for the transportation of dangerous goods, and their tank areas store a large number of flammable, explosive, toxic and hazardous chemicals. The strong current, high temperature and shock waves brought by the natural phenomenon of lightning pose a direct threat to port facilities, especially in tank areas. Lightning may not only cause equipment damage and material leakage, but also damage the electrical system through induced current, increasing the risk of safety accidents. Therefore, establishing an efficient and accurate lightning warning system is crucial to ensure the safe operation of ports.

[0017] However, traditional lightning warning mainly relies on simple meteorological observation data and empirical judgment. This method is difficult to accurately monitor the core physical process of lightning, and cannot capture the rapid changes in the intensity of the atmospheric electric field and the specific location and trajectory of lightning. In addition, traditional methods are inefficient in processing massive dynamic monitoring data, and the prediction accuracy of the time, location and intensity of lightning is insufficient. It is difficult to cope with the complex and changing port environment, and it is difficult to provide timely and reliable warning information for dangerous goods tank areas, and it is difficult to meet the high standards of lightning protection in tank areas.

[0018] In response to the above technical problems, the technical concept of the present application is to obtain a time series data set of lightning characteristic parameters (atmospheric electric field strength values ​​and lightning positions) collected by a lightning detection module, and use artificial intelligence-based data analysis and coding technology to split the time series data set of the lightning characteristic parameters, and then perform relative position prediction based on the time series characteristics on the split time series data set of the lightning position, and at the same time perform message constraint transmission based on the time series characteristics on the time series data set of the atmospheric electric field strength value, so as to automatically determine the lightning warning level according to the predicted value of the lightning relative position data and the coding characteristics of the atmospheric electric field strength time series fluctuation after the time series transmission. The present application can monitor the changes in the atmospheric electric field strength in real time, and accurately capture the position and movement path of lightning, so as to more accurately predict the occurrence time and intensity of lightning, thereby providing reliable safety protection for the port's dangerous goods tank area.

[0019] Figure 1 1 is a system block diagram of a lightning warning system for a major hazard source tank area of ​​dangerous goods in a port according to an embodiment of the present application. Figure 1As shown, in the lightning warning system 100 for the tank area of ​​dangerous goods of a port with major hazardous sources, it includes: a lightning parameter acquisition module 110, which is used to obtain a time series data set of lightning characteristic parameters collected by a lightning detection module, wherein the lightning characteristic parameters include an atmospheric electric field strength value collected by an atmospheric electric field meter and a lightning position collected by a lightning positioning system; a data splitting module 120, which is used to perform data set splitting on the time series data set of the lightning characteristic parameters to obtain a time series data set of atmospheric electric field strength values ​​and a time series data set of lightning positions; an analysis module 130, which is used to determine a lightning warning level based on the time series characteristics of the time series data set of the atmospheric electric field strength values ​​and the predicted relative position of the time series data set of the lightning positions.

[0020] In the embodiment of the present application, the lightning parameter acquisition module 110 is used to obtain a time series data set of lightning characteristic parameters collected by the lightning detection module, and the lightning characteristic parameters include the atmospheric electric field strength value collected by the atmospheric electric field instrument and the lightning position collected by the lightning positioning system. It should be understood that the atmospheric electric field is an electric field existing in the earth's atmosphere, and under normal circumstances, there is a relatively stable background value. When lightning is about to occur, the atmospheric electric field strength will change significantly. For example, during the formation and development of thunderstorm clouds, the separation and accumulation of charges will lead to an increase in the electric field strength. The time series data set of the collected atmospheric electric field strength values ​​can reflect the fluctuation of the electric field strength over time, including information such as the rise, fall, and peak of the electric field strength. These fluctuation characteristics can reflect the different stages of lightning activity, such as the electric field strength slowly rises in the initial charge accumulation stage of thunderstorm clouds, and may change sharply when lightning is about to occur. The time series data set of the lightning position records the information of the lightning position within a period of time. The lightning position is a two-dimensional spatial coordinate information (longitude, latitude), and the time series data set is formed by arranging multiple such position information in chronological order. Through this dataset, we can understand the changes in the scope and movement path of lightning activities. In general, changes in the intensity of the atmospheric electric field are important precursors to lightning. When the atmospheric electric field intensity reaches a certain threshold and continues to change, it means that the possibility of lightning increases. Lightning location information is crucial to determining whether the tank area of ​​major hazardous sources of dangerous goods is within the threat range of lightning. The time series dataset of the lightning location can track the activity trajectory of lightning, and combined with the location data of the hazardous source irrigation area for analysis, it can predict whether lightning will approach the tank area. In other words, the atmospheric electric field intensity value provides information on the possibility of lightning, and the lightning location provides spatial information on the degree of threat of lightning to the tank area. Together, they provide a comprehensive basis for determining the lightning warning level.

[0021] The following is a detailed description of a specific implementation process of "obtaining a time series data set of lightning characteristic parameters collected by a lightning detection module":

[0022] The equipment installation link is the basis for obtaining data. In the major hazardous source tank area of ​​dangerous goods in the port, the atmospheric electric field instrument should be scientifically arranged according to the actual situation of the tank area, such as the size of the area, topography and surrounding environment. The edge of the tank area, the central area and the location with high lightning risk, such as near open space or near tall buildings, are all key installation areas to ensure that the atmospheric electric field changes in the tank area and the surrounding area can be fully and without blind spots. After the atmospheric electric field instrument is installed in place, strict calibration and debugging must be carried out to ensure the accuracy of its measurement of the atmospheric electric field strength value. At the same time, a stable data transmission link is established. The atmospheric electric field instrument can be connected to a data acquisition terminal with a stable power supply and reliable communication function by means of wired Ethernet or wireless 4G / 5G network and other communication methods. In this process, the communication network should be fully tested to avoid data loss or delayed transmission, and ensure that the data can be transmitted to the acquisition terminal in real time and accurately.

[0023] The acquisition of lightning location data requires the help of a third-party lightning location system. Careful screening is required among the many third-party service providers. Key indicators such as the accuracy of lightning location, coverage, data update frequency, and compatibility of data interfaces should be considered, and priority should be given to suppliers with a good reputation in the field of meteorological monitoring, high positioning accuracy, and frequent data updates. After selecting a supplier, a detailed data service agreement should be signed with it to clarify the rights and obligations of both parties, including important contents such as the scope of data use, acquisition methods, fee payment, and data security and confidentiality clauses. Subsequently, based on the data interface specifications provided by the third party, a special data access program is developed to obtain lightning location data from the third-party platform and effectively integrate it with the data collected by the atmospheric electric field instrument. When accessing data, the data format should be converted and strictly verified to ensure the consistency and accuracy of the data, and then stored in the local database.

[0024] Time synchronization is a key step to ensure data accuracy. The premise for accurate correspondence between atmospheric electric field intensity values ​​and lightning location data is time synchronization. To this end, a high-precision clock source, such as an atomic clock, is required to calibrate the time of the atmospheric electric field instrument and the third-party lightning location system. The time error is controlled within a very small range, generally requiring millisecond-level or even higher accuracy to ensure the consistency of all data in the time dimension.

[0025] After completing data collection and time synchronization, enter the data set generation stage. The collected atmospheric electric field strength values ​​and lightning location data are stored in a special database table established in a database (such as MySQL, Oracle, etc.) in chronological order. The table structure should be carefully designed, including key fields such as timestamp, atmospheric electric field strength value, lightning location (longitude, latitude), etc. The stored data should be sorted and cleaned regularly, and abnormal data and duplicate data should be deleted in time to ensure the high quality of the data set. Finally, data is extracted from the stored data at appropriate time intervals such as every second and every minute to generate a time series data set of lightning characteristic parameters. It is worth noting that during the generation process, the data needs to be standardized, such as normalization operations, so that different types of data are comparable, laying a solid foundation for subsequent data splitting and analysis.

[0026] In an embodiment of the present application, the data splitting module 120 is used to split the time series data set of the lightning characteristic parameters to obtain the time series data set of the atmospheric electric field strength value and the time series data set of the lightning position. Accordingly, considering that the atmospheric electric field strength value and the lightning position are two types of data of different nature, they have different physical meanings and change laws. The atmospheric electric field strength value reflects the change in the intensity of the electric field, which is a scalar data. Its change is mainly related to factors such as charge distribution and accumulation in the cloud layer. It may show a specific fluctuation law in the time series, such as there may be a trend of gradual strengthening or intensified fluctuations before the lightning occurs. The lightning position is spatial position information, usually expressed in the form of coordinates such as longitude and latitude, and belongs to vector data. Its change in time series is reflected in the movement trajectory of lightning in space, which is closely related to factors such as atmospheric convection movement and weather systems. In order to facilitate a clearer analysis of the respective change characteristics and laws, the present application splits the time series data set of the lightning characteristic parameters to obtain the time series data set of the atmospheric electric field strength value and the time series data set of the lightning position. That is, by splitting the data set, we can conduct in-depth analysis on the time series data of the atmospheric electric field strength value and the lightning position. For the time series data set of the atmospheric electric field strength value, we can study its changing trend on different time scales and find characteristic patterns related to the occurrence of lightning, such as sudden changes in the electric field strength and periodic fluctuations, so as to establish a relationship model between the atmospheric electric field strength and the probability of lightning occurrence. For the time series data set of the lightning position, we can analyze the temporal and spatial distribution of lightning, such as the density of lightning and the regularity of the moving path, to provide a basis for predicting the possible location and direction of lightning.

[0027] In the embodiment of the present application, the analysis module 130 is used to determine the lightning warning level based on the time series characteristics of the time series data set of the atmospheric electric field strength value and the predicted relative position of the time series data set of the lightning position. Specifically, Figure 2FIG. 1 is a block diagram of an analysis module in a lightning warning system for a tank area of ​​a dangerous cargo port with a major hazard source according to an embodiment of the present application. Figure 2 As shown, the analysis module 130 includes: a coding unit 131, which is used to perform electric field strength spatial constraint time series propagation and relative position prediction on the time series data set of the atmospheric electric field strength value and the time series data set of the lightning position, respectively, to obtain the atmospheric electric field strength time series fluctuation coding characteristics and the predicted value of the lightning relative position data; a level determination unit 132, which is used to determine the lightning warning level based on the predicted value of the lightning relative position data and the atmospheric electric field strength time series fluctuation coding characteristics.

[0028] In the embodiment of the present application, the encoding unit 131 is used to perform electric field intensity spatial constraint time series propagation and relative position prediction on the time series data set of the atmospheric electric field intensity value and the time series data set of the lightning position, respectively, to obtain the atmospheric electric field intensity time series fluctuation coding features and the predicted value of the lightning relative position data. Specifically, Figure 3 FIG. 1 is a block diagram of a coding unit in a lightning warning system for a tank area with major hazard sources for dangerous goods in a port according to an embodiment of the present application. Figure 3 As shown, the encoding unit 131 includes: a relative position time series data set construction subunit 1311, which is used to obtain the position data of the hazardous source irrigation area, and construct a time series data set of lightning relative position data based on the time series data set of the lightning position and the position data of the hazardous source irrigation area; an atmospheric electric field strength encoding subunit 1312, which is used to perform message dynamic propagation based on local time series coding on the time series data set of the atmospheric electric field strength value to obtain the atmospheric electric field strength time series fluctuation coding characteristics; a relative position prediction subunit 1313, which is used to perform time series encoding and decoding on the time series data set of the lightning relative position data to obtain the predicted value of the lightning relative position data.

[0029] In the embodiment of the present application, the relative position time series data set construction subunit 1311 is used to obtain the position data of the dangerous source irrigation area, and construct the time series data set of the lightning relative position data based on the time series data set of the lightning position and the position data of the dangerous source irrigation area. It should be understood that the lightning position data reflects the dynamic changes of lightning in space, while the position data of the dangerous source irrigation area is fixed geographical information. In the scenario of the major dangerous source tank area of ​​dangerous goods in the port, people are not only concerned about the position of lightning, but more importantly, the relative position of lightning and the tank area. Because the tank area stores dangerous goods, once affected by lightning, it may cause serious safety accidents. Therefore, in the technical solution of the present application, the position data of the dangerous source irrigation area is obtained, and the time series data set of lightning relative position data is constructed based on the time series data set of the lightning position and the position data of the dangerous source irrigation area. In this way, by constructing the time series data set of lightning relative position data, the relative distance of lightning from the dangerous source irrigation area at different times and other information can be intuitively understood. By analyzing these data, the threat level of lightning to the tank area can be more accurately assessed, and more specific and useful information can be provided for subsequent lightning warnings.

[0030] The following is a detailed description of a specific implementation process of "obtaining the location data of the hazardous source irrigation area, and constructing a time series data set of lightning relative position data based on the time series data set of the lightning location and the location data of the hazardous source irrigation area":

[0031] Obtaining irrigation area location data is the basis of the entire implementation process. Geographic Information System (GIS) is an important way to obtain this data. With the help of professional GIS software and database, high-precision location information of major hazardous sources of dangerous goods in ports can be obtained. When obtaining data, it is important to ensure the currentness and accuracy of the data, as this is directly related to the reliability of subsequent analysis. For example, if the irrigation area has been expanded or relocated recently, outdated data will lead to deviations in the analysis results, so it is particularly important to obtain updated geographic data in a timely manner.

[0032] When it is not possible to obtain appropriate GIS data, or when existing data needs to be verified and supplemented, field measurement and positioning becomes an effective means of obtaining irrigation area location data. Use global positioning system (GPS) equipment, such as high-precision GPS receivers, to measure key locations in the irrigation area. These key locations include the boundary points of the irrigation area and the locations of important facilities. During the measurement process, special attention should be paid to the impact of the measurement environment on the GPS signal, ensuring that the GPS device is in an open space to avoid the signal being blocked or interfered with by objects such as buildings and trees, so as to obtain accurate longitude and latitude coordinates.

[0033] After obtaining the irrigation area location data, the next step is to construct a time series data set of lightning relative position data based on the time series data set of lightning location and the irrigation area location data. First, the data format must be unified and preprocessed. Lightning location data generally contains information such as longitude, latitude, and time, while irrigation area location data is mainly longitude and latitude coordinates. They need to be unified into a format that is easy to calculate and process, such as expressing longitude and latitude in decimal form. At the same time, the data is comprehensively preprocessed, the integrity and accuracy of the data are carefully checked, and outliers and erroneous data are removed. For example, if there are longitude and latitude values ​​in the lightning location data that are significantly deviated from the normal range, these data may be erroneous data caused by measurement errors or other reasons, and need to be verified and corrected. If verification and correction cannot be achieved, they should be deleted resolutely to avoid affecting the accuracy of subsequent calculations and analysis.

[0034] After completing the data preprocessing, the relative position calculation stage begins. For each lightning position record in the time series data set of the lightning position, the position of the lightning relative to the irrigation area is accurately calculated based on its corresponding time and the location data of the hazardous source irrigation area. Usually, a spatial distance formula (such as the Euclidean distance formula) is used to calculate the distance between the lightning and the boundary point or center point of the irrigation area to determine the relative distance between the lightning and the irrigation area. At the same time, the direction of the lightning relative to the irrigation area is determined by calculating the azimuth. For example, taking the center point of the irrigation area as the reference point, mathematical methods such as trigonometric functions are used to calculate the azimuth of the line connecting the lightning position and the point, so as to clearly determine in which direction of the irrigation area the lightning is. This step requires precise calculations and rigorous logic to ensure that the relative position information obtained is accurate and reliable.

[0035] The last step is to construct a time series data set. In chronological order, the calculated lightning relative position information (including relative distance and direction) is systematically sorted to construct a time series data set of lightning relative position data. In order to facilitate storage and subsequent analysis, these data can be stored in a database table. When designing the database table structure, the characteristics of the data and the analysis requirements should be fully considered, including key fields such as timestamp, relative distance, and relative direction. Through such a time series data set, the relative position relationship between lightning and the hazardous source irrigation area at different times can be intuitively presented, providing strong data support for the subsequent assessment of the threat level of lightning to the tank area.

[0036] In the embodiment of the present application, the atmospheric electric field strength encoding subunit 1312 is used to perform dynamic message propagation based on local time series coding on the time series data set of the atmospheric electric field strength value to obtain the atmospheric electric field strength time series fluctuation coding characteristics. Specifically, Figure 4 The block diagram of the atmospheric electric field intensity encoding subunit in the lightning warning system for the major hazard source tank area of ​​dangerous goods in the port according to the embodiment of the present application. Figure 4As shown, the atmospheric electric field strength encoding subunit 1312 includes: an atmospheric electric field strength time series encoding secondary subunit 13121, which is used to perform one-dimensional convolution encoding on the time series data set of the atmospheric electric field strength value to obtain a sequence distribution of the atmospheric electric field strength local time series fluctuation encoding feature vector; an atmospheric electric field strength time series fluctuation encoding secondary subunit 13122, which is used to perform atmospheric electric field strength local constraint time dimension message transmission on the sequence distribution of the atmospheric electric field strength local time series fluctuation encoding feature vector to obtain the atmospheric electric field strength time series fluctuation encoding feature vector as the atmospheric electric field strength time series fluctuation encoding feature.

[0037] In an embodiment of the present application, the atmospheric electric field intensity time series coding secondary subunit 13121 is used to perform one-dimensional convolution coding on the time series data set of the atmospheric electric field intensity value to obtain the sequence distribution of the local time series fluctuation coding feature vector of the atmospheric electric field intensity. Accordingly, considering that in the time series data of the atmospheric electric field intensity value, there is often a close association between the data of adjacent time points, which contains rich local feature information. Therefore, the present application performs one-dimensional convolution coding on the time series data set of the atmospheric electric field intensity value to capture the local patterns and features in the time series data, such as fluctuations and trend changes in a short period of time, and obtains the sequence distribution of the local time series fluctuation coding features of the atmospheric electric field intensity. In particular, the one-dimensional convolution coding can perform convolution operations by sliding the convolution kernel on the time series data, effectively capturing the time series fluctuation characteristics within these local ranges, such as rapid changes and small fluctuations in the electric field intensity in a short period of time, and these local features may be of great significance for analyzing the occurrence mechanism and trend of lightning. Converting the time series data of atmospheric electric field strength values ​​into a sequence distribution of local time series fluctuation coding features can more clearly show the fluctuation characteristics and changing patterns in the data. These coding features can highlight the changes in electric field strength on different time scales, such as the sharp rise or increased fluctuation of atmospheric electric field strength before lightning occurs, which helps to more accurately identify and analyze the change patterns of electric field strength related to lightning.

[0038] In an embodiment of the present application, the atmospheric electric field intensity time series fluctuation coding secondary subunit 13122 is used to perform time dimension message transmission of atmospheric electric field intensity local constraints on the sequence distribution of the atmospheric electric field intensity local time series fluctuation coding feature vector to obtain the atmospheric electric field intensity time series fluctuation coding feature vector as the atmospheric electric field intensity time series fluctuation coding feature. Specifically, in an embodiment of the present application, the atmospheric electric field intensity time series fluctuation coding secondary subunit includes: an atmospheric electric field intensity anchoring tertiary subunit, used to perform tail constraint anchoring and axis constraint anchoring on the sequence distribution of the atmospheric electric field intensity local time series fluctuation coding feature vector respectively to obtain the atmospheric electric field intensity local time series message transmission space tail constraint anchoring coding vector and the atmospheric electric field intensity time series message transmission space axis constraint anchoring coding vector; a constraint factor calculation tertiary subunit, used to calculate the atmospheric electric field intensity local time series message transmission space tail constraint anchoring coding vector and the atmospheric electric field intensity time series message transmission space axis constraint anchoring coding vector based on the atmospheric electric field intensity local time series message transmission space tail constraint anchoring coding vector and the atmospheric electric field intensity time series fluctuation coding feature. The message transmission space axis constraint anchors the coding vector, calculates the atmospheric electric field strength message transmission constraint factor of each atmospheric electric field strength local time series fluctuation coding feature vector in the sequence distribution of the atmospheric electric field strength local time series fluctuation coding feature vector; the atmospheric electric field strength time series fluctuation coding feature generates a three-level sub-unit, which is used to perform message dynamic constraint transmission coding on the sequence distribution of the atmospheric electric field strength local time series fluctuation coding feature vector based on the atmospheric electric field strength message transmission constraint factor of each atmospheric electric field strength local time series fluctuation coding feature vector to obtain the atmospheric electric field strength time series fluctuation coding feature vector.

[0039] It should be understood that the change of atmospheric electric field intensity is not isolated in the time series, and there are complex dependencies between adjacent time points. Although the local temporal fluctuation coding features of atmospheric electric field intensity capture local features, they may not fully take into account the dependency information in these time dimensions. Therefore, in order to better understand and analyze the characteristics of the overall dynamic fluctuation change of atmospheric electric field intensity in the time dimension, in the technical solution of the present application, the sequence distribution of the local temporal fluctuation coding feature vector of the atmospheric electric field intensity is subjected to the time dimension message transmission of the local constraint of the atmospheric electric field intensity to interact and integrate the information at different time points, so that the model can better understand the dynamic change process of atmospheric electric field intensity in time, dig out richer temporal dependency features, and obtain the temporal fluctuation coding feature vector of atmospheric electric field intensity. This method realizes efficient semantic coding by combining independence constraints with dynamic message transmission mechanism. In this way, it is possible to integrate time dimension information, dig deep semantics, and make the encoded feature vector more comprehensive and accurate, thereby improving the ability to characterize changes in atmospheric electric field intensity and the accuracy of lightning warning.

[0040] Specifically, in the embodiment of the present application, the atmospheric electric field strength anchoring three-level subunit is used to: extract the last atmospheric electric field strength local time series fluctuation coding feature vector from the sequence distribution of the atmospheric electric field strength local time series fluctuation coding feature vector as the atmospheric electric field strength local time series message transmission space tail end constraint anchor coding vector, and the process can be expressed by the formula:

[0041] O={v1,v2,...,v i ,...,v t}

[0042] vt ail =v t

[0043] Where O is the sequence distribution of the characteristic vector encoding the local temporal fluctuation of the atmospheric electric field intensity, v1, v2, v i and v t are the first, second, i-th and t-th atmospheric electric field intensity local time series fluctuation encoding feature vectors in the sequence distribution of the atmospheric electric field intensity local time series fluctuation encoding feature vectors, respectively, v tail is the spatial tail constraint anchor coding vector of the local temporal message delivery of the atmospheric electric field intensity;

[0044] Cluster analysis is performed on the sequence distribution of the atmospheric electric field intensity local time series fluctuation coding feature vector to obtain the atmospheric electric field intensity time series message transmission space axis constraint anchor coding vector. The process can be expressed by the formula:

[0045]

[0046] Among them, v1, v2, v i and v t are the first, second, i-th and t-th atmospheric electric field intensity local time series fluctuation encoding feature vectors in the sequence distribution of the atmospheric electric field intensity local time series fluctuation encoding feature vectors, Cluster is a clustering analysis operation, max(v i ) and min(v i ) are respectively taken as v i The maximum and minimum values ​​of η are the adjustment parameters, a i is the i-th atmospheric electric field intensity local time series characteristic constraint anchor value in the sequence distribution of atmospheric electric field intensity local time series characteristic constraint anchor value, softmax is a normalization function, e i is the anchor weight value of the local temporal feature constraint of the atmospheric electric field intensity in the time series set of the anchor weight value of the local temporal feature constraint of the atmospheric electric field intensity, t is the number of vectors in O, v axis It is the spatial axis constraint anchor coding vector of the atmospheric electric field intensity temporal message transmission.

[0047] It should be understood that there is a complex dependency relationship between the atmospheric electric field intensity and the last eigenvector in the sequence distribution of the local temporal fluctuation encoding feature vector of the atmospheric electric field intensity integrates part of the information of the previous time points, which can reflect the comprehensive situation of the change of the atmospheric electric field intensity to the current state. For example, in the process of lightning development, the atmospheric electric field intensity is constantly changing, and its local temporal fluctuation encoding feature vector sequence also changes accordingly. The last eigenvector reflects the atmospheric electric field intensity characteristics corresponding to the current lightning development stage. By taking the last atmospheric electric field intensity local temporal fluctuation encoding feature vector from the sequence distribution as the spatial tail constraint anchor encoding vector of the atmospheric electric field intensity local temporal message transmission, a clear reference point can be provided for the atmospheric electric field intensity characteristic time dimension message transmission, thereby constraining the direction and range of the message transmission, so that the model can ensure that it is in the direction related to the current state of the atmospheric electric field intensity when analyzing the relevant feature data. That is, this step can prevent the model from diverging or being unstable when processing the feature sequence distribution, and ensure that the model can effectively integrate the information at different time points, so as to better understand the dynamic changes of the atmospheric electric field intensity in time.

[0048] Correspondingly, the sequence distribution of the characteristic vector of the local temporal fluctuation encoding of the atmospheric electric field intensity contains a large amount of complex data information, which reflects the changing characteristics of the atmospheric electric field intensity at different time points. In order to extract the main mode (i.e., representative feature) in the sequence distribution from these complex characteristic data, the sequence distribution of the characteristic vector of the local temporal fluctuation encoding of the atmospheric electric field intensity can be clustered and analyzed in this application to distinguish the patterns with similar characteristics, find the most representative pattern, and help the model better understand the law of atmospheric electric field intensity change. In particular, when the atmospheric electric field intensity is transmitted in the time dimension to mine its dynamic change characteristics, a global guide is needed to ensure the effectiveness and accuracy of the information transmission process. The generated atmospheric electric field intensity temporal message transmission space axis constraint anchor coding vector can provide such a global structural constraint. This global constraint can be analogous to regularization, which can determine a main direction for message transmission and prevent the information transmission process from being able to proceed along the main distribution direction of the data. When the model processes the time series information of atmospheric electric field intensity, this axis constraint anchoring coding vector can help the model avoid falling into local and unrepresentative details during the information transmission process, thereby effectively improving the efficiency and quality of information transmission.

[0049] Specifically, in the embodiment of the present application, the constraint factor calculation three-level subunit includes: an atmospheric electric field strength tail end message transmission constraint factor calculation four-level subunit, which is used to calculate the atmospheric electric field strength tail end message transmission constraint factor of each atmospheric electric field strength local time series fluctuation coding feature vector in the sequence distribution of the atmospheric electric field strength local time series fluctuation coding feature vector relative to the atmospheric electric field strength local time series message transmission space tail end constraint anchor coding vector; an atmospheric electric field strength axis line message transmission constraint factor calculation four-level subunit, which is used to calculate each large atmospheric electric field strength in the sequence distribution of the atmospheric electric field strength local time series fluctuation coding feature vector The atmospheric electric field intensity axis line message transmission constraint factor of the atmospheric electric field intensity local time series fluctuation coding feature vector relative to the atmospheric electric field intensity time series message transmission space axis constraint anchor coding vector; the atmospheric electric field intensity message transmission constraint factor calculation four-level subunit is used to calculate the atmospheric electric field intensity message transmission constraint factor of each atmospheric electric field intensity local time series fluctuation coding feature vector based on the atmospheric electric field intensity tail message transmission constraint factor and the atmospheric electric field intensity axis line message transmission constraint factor of each atmospheric electric field intensity local time series fluctuation coding feature vector in the sequence distribution of the atmospheric electric field intensity local time series fluctuation coding feature vector.

[0050] Next, the atmospheric electric field strength tail end message transmission constraint factor of each atmospheric electric field strength local time series fluctuation coding feature vector in the sequence distribution of the atmospheric electric field strength local time series fluctuation coding feature vector relative to the atmospheric electric field strength local time series message transmission space tail end constraint anchor coding vector is calculated. The above process can be expressed by the formula:

[0051]

[0052] Among them, f(v i ,v tail ) is for v i and v tail To anchor the tail end, v i The eigenvalue of the jth position, v tail The eigenvalue of the jth position, log2 is the logarithmic function value with base 2, and n is v i The number of eigenvalues ​​in , exp is the exponential function value with the natural constant e as the base, α i Yes i The corresponding atmospheric electric field strength tail end message transmission constraint factor.

[0053] It should be understood that in the sequence distribution of the local temporal fluctuation encoding feature vector of the atmospheric electric field intensity, the relationship between each vector and the tail constraint anchor encoding vector of the local temporal message transmission space of the atmospheric electric field intensity is crucial to understanding the time series change of the atmospheric electric field intensity. In order to describe this relationship more accurately, it is necessary to quantify the similarity between each feature vector and the tail constraint anchor encoding vector by calculating the tail message transmission constraint factor. For example, in different stages of lightning development, the change pattern of the atmospheric electric field intensity is different, and the relationship between the local temporal fluctuation encoding feature vector corresponding to each time point and the current state (tail vector) is also different. By quantifying this relationship, the model can better grasp the evolution of the atmospheric electric field intensity in the time series. Moreover, in the process of message transmission, in order to enable the model to focus on the features related to the current lightning state, a mechanism is needed to adjust the importance of each feature vector. The calculated tail message transmission constraint factor of the atmospheric electric field intensity can be used as such a mechanism. By assigning a local constraint weight to each feature vector, the model can prioritize the feature vectors closely related to the current lightning state according to this weight, thereby more accurately mining the dynamic change characteristics of the atmospheric electric field intensity in the time dimension.

[0054] Then, the atmospheric electric field intensity axis message transmission constraint factor of each atmospheric electric field intensity local time series fluctuation coding feature vector in the sequence distribution of the atmospheric electric field intensity local time series fluctuation coding feature vector relative to the atmospheric electric field intensity time series message transmission space axis constraint anchor coding vector is calculated. The above process can be expressed by the formula:

[0055]

[0056] Among them, f(v i ,v axis ) is for v i and v axis To anchor the axis constraint, ‖·‖2 is the Euclidean norm of the calculated vector, arccosh is the inverse hyperbolic cosine function, β i Yes i Corresponding atmospheric electric field strength axis message transmission constraint factor.

[0057] It should be understood that the spatial axis constraint anchor coding vector of the atmospheric electric field intensity time series message transmission represents the main mode of the sequence distribution of the characteristic vector sequence of the atmospheric electric field intensity local time series fluctuation coding. By calculating the atmospheric electric field intensity axis message transmission constraint factor of each atmospheric electric field intensity local time series fluctuation coding characteristic vector relative to the axis constraint anchor coding vector, the similarity between each characteristic vector and the main mode can be quantified. For example, in the complex atmospheric electric field change process, the electric field intensity change characteristics at different time points are different. By calculating this quantitative relationship, the position and correlation degree of the electric field intensity at each time point in the overall atmospheric electric field intensity change pattern can be clearly understood. Similar to the atmospheric electric field intensity tail end message transmission constraint factor, calculating the atmospheric electric field intensity axis message transmission constraint factor can assign a global constraint weight to each characteristic vector. In the message transmission process, the model can adjust the information transmission priority and mode of the characteristic vector according to this weight. Characteristic vectors with high consistency with the global structure (i.e., high weight) will receive more attention in message transmission, thereby optimizing the entire message transmission process and making the information transmission more consistent with the global pattern of atmospheric electric field intensity change. This will help to more accurately grasp the overall changing trend of the atmospheric electric field intensity and provide a more reliable basis for lightning warning.

[0058] More specifically, in an embodiment of the present application, the atmospheric electric field strength message transmission constraint factor calculation four-level sub-unit is used to: use the standard space constraint matrix to perform linear transverse covariance on the atmospheric electric field strength tail message transmission constraint factor and the atmospheric electric field strength axis message transmission constraint factor corresponding to the atmospheric electric field strength local time series fluctuation coding feature vector to obtain the atmospheric electric field strength tail message transmission covariance constraint factor and the atmospheric electric field strength axis message transmission covariance constraint factor; based on the atmospheric electric field strength tail message transmission covariance constraint factor and the atmospheric electric field strength axis message transmission covariance constraint factor, perform global consistency processing on the atmospheric electric field strength tail message transmission constraint factor and the atmospheric electric field strength axis message transmission constraint factor to obtain the optimized atmospheric electric field strength tail message transmission constraint factor and the optimized atmospheric electric field strength axis message transmission constraint factor; perform weighted processing on the optimized atmospheric electric field strength tail message transmission constraint factor and the optimized atmospheric electric field strength axis message transmission constraint factor based on the sigmoid function to obtain the atmospheric electric field strength message transmission constraint factor corresponding to the atmospheric electric field strength local time series fluctuation coding feature vector. The above process can be expressed as:

[0059]

[0060] y i =Sigmoid(ω1·α i ′+ω2·β i ′)

[0061] Among them, δ i and ε i They are α i and β i The covariance constraint factor of the tail message transmission of the atmospheric electric field intensity after covariation and the covariance constraint factor of the axis message transmission of the atmospheric electric field intensity, T represents the transposition operation, |·| is the absolute value operation, α i ′ and β i ′ are respectively v i The corresponding optimized atmospheric electric field strength tail message transmission constraint factor and the optimized atmospheric electric field strength axis message transmission constraint factor, ω1 and ω2 are weighting parameters, Sigmoid is the weight mapping function, y i Yes i The corresponding atmospheric electric field strength message transmission constraint factor.

[0062] In particular, considering that in lightning monitoring scenarios, the temporal fluctuations of electric field intensity are easily affected by local disturbances in the lateral transfer field (such as non-uniform distribution of cloud charge, ground electric field distortion, etc.), if we only rely on the global regularization of the message transmission covariance constraint factor of the atmospheric electric field intensity axis, the adaptive expression of local dynamics may be ignored due to overfitting of the main mode; conversely, if we only emphasize the instantaneous state anchoring of the message transmission covariance constraint factor at the tail end of the atmospheric electric field intensity, the transfer path may diverge due to local noise accumulation. To this end, a dynamic balance between the two can be achieved through a transversely closed covariant mechanism: first, the two types of constraint factors are laterally coupled based on the standard space constraint matrix. For example, in the identification of lightning development stages, the current electric field mutation characteristics represented by the tail vector (such as the polarity reversal signal in the electrification stage of thunderstorm clouds) and the long-term evolution trend represented by the axis vector (such as the exponential growth pattern of charge accumulation) are aligned in feature space projection through matrix operations to eliminate the redundant conflicts between the two in dimensional distribution; then, the component rotation matrix of the reversible space metric is introduced to perform spinor reconstruction on the coupled covariant representation. For example, in the lightning imminent warning, the time series fluctuation characteristics of the electric field intensity are mapped to a finite-dimensional compact space through rotation operations, so that the local sensitivity of the tail constraint and the global stability of the axis constraint complement each other in the orthogonal subspace. This double closure mechanism can not only suppress the weight offset caused by local electric field transients during the transmission process, but also avoid over-reliance on historical patterns and loss of current state specificity, thereby enabling the optimized atmospheric electric field intensity tail message transmission constraint factor and the optimized atmospheric electric field intensity axis message transmission constraint factor to achieve the unification of transmission constraints within the lateral structural field in the global structure of message transmission.

[0063] It should be understood that the optimized atmospheric electric field intensity tail message transmission constraint factor reflects the relationship between the atmospheric electric field intensity local time series fluctuation encoding feature vector and the local current state (the atmospheric electric field intensity local time series message transmission space tail constraint anchor encoding vector), focusing on capturing the local dynamic changes in the time series and the information closely related to the current moment; while the atmospheric electric field intensity axis message transmission constraint factor reflects the similarity or distance between the atmospheric electric field intensity local time series fluctuation encoding feature vector and the global structure (the atmospheric electric field intensity time series message transmission space axis constraint anchor encoding vector), focusing on the overall atmospheric electric field intensity change pattern. In the actual atmospheric electric field intensity analysis, the importance and role of the atmospheric electric field intensity characteristics in message transmission cannot be fully and accurately described based on local or global constraints alone. For example, only considering local constraints may ignore the long-term trend and overall pattern of atmospheric electric field intensity, and only focusing on global constraints may miss some key details related to the atmospheric electric field intensity at the current moment. Therefore, it is necessary to merge these two constraint factors to comprehensively consider the local and global impacts. In the process of message transmission, the model can more reasonably determine the importance and role of each eigenvector based on this comprehensive weight. For example, when analyzing the time series of atmospheric electric field strength to predict lightning, those eigenvectors that are closely related to the change of atmospheric electric field strength at the current moment (high local constraint factor) and highly consistent with the overall atmospheric electric field strength change pattern (high global constraint factor) will obtain a higher comprehensive weight, thereby playing a greater role in message transmission and the final lightning warning judgment. In other words, this comprehensive constraint mechanism enables the model to learn richer and more representative features, thereby improving the ability to characterize changes in atmospheric electric field strength.

[0064] Finally, based on the atmospheric electric field strength message transmission constraint factors of the atmospheric electric field strength local time series fluctuation encoding feature vectors, the sequence distribution of the atmospheric electric field strength local time series fluctuation encoding feature vectors is subjected to message dynamic constraint transmission encoding to obtain the atmospheric electric field strength time series fluctuation encoding feature vectors. The above process can be expressed by the formula:

[0065]

[0066] Among them, y i Yes i The corresponding atmospheric electric field strength message transmission constraint factor, z is the encoding characteristic vector of the atmospheric electric field strength time series fluctuation.

[0067] It should be understood that although the original atmospheric electric field intensity local time series fluctuation encoding feature vector sequence distribution contains certain atmospheric electric field intensity change information, it lacks comprehensive consideration of local and global constraints. The atmospheric electric field intensity message transmission constraint factor combines the tail and axis message transmission constraint factors and contains local and global constraint information. Through message dynamic constraint transmission coding, the atmospheric electric field intensity message transmission constraint factor is applied to the original feature vector sequence distribution, which can integrate local and global constraint information into the feature encoding process, and weight each feature vector according to its importance (determined by the atmospheric electric field intensity message transmission constraint factor), so as to more accurately capture the semantic relationship and dynamic change pattern between different time points in the sequence. That is, the generated atmospheric electric field intensity time series fluctuation encoding feature vector contains richer and more accurate atmospheric electric field intensity change semantic information, which can better reflect the dynamic change characteristics of atmospheric electric field intensity in time series, and better reflect the nature and law of atmospheric electric field intensity change, which can provide more powerful support for subsequent lightning warning.

[0068] In the embodiment of the present application, the relative position prediction subunit 1313 is used to perform time series encoding and decoding on the time series data set of the lightning relative position data to obtain the predicted value of the lightning relative position data. Specifically, in the embodiment of the present application, the relative position prediction subunit is used to: perform time series encoding based on a bidirectional recurrent neural network on the time series data set of the lightning relative position data to obtain a time series encoding feature vector of the lightning relative position data; input the time series encoding feature vector of the lightning relative position data into a relative position predictor based on a decoder to obtain the predicted value of the lightning relative position data.

[0069] Accordingly, considering that the position of lightning will continue to change at different times, it is affected by various factors such as atmospheric movement and weather systems. In order to effectively capture the dependency and change trend of data in the time dimension, so as to make a reasonable prediction of the relative position of lightning in the future, the present application performs time series encoding and decoding on the time series data set of the lightning relative position data to obtain the predicted value of the lightning relative position data. In this way, these dynamic change patterns and laws can be learned based on the historical lightning relative position data, and the future position can be predicted based on this knowledge. In particular, in a specific embodiment of the present application, first of all, considering that the lightning relative position data has a time series attribute, its position information at different times has a front-to-back dependency relationship. That is, the time series characteristics of the lightning relative position data determine that there is a close connection between the data at different time points. The past position information can reflect the movement trend of lightning, and the future position information (to a certain extent, the relevant features can be obtained through back propagation) is also helpful to understand the background of the current position change. Based on this, the time series data set of the lightning relative position data is subjected to time series encoding based on a bidirectional recurrent neural network to obtain a time series encoding feature vector of the lightning relative position data. That is, the bidirectional recurrent neural network can simultaneously consider the past and future information of the data, and through propagation in both forward and backward directions, it can comprehensively capture the bidirectional dependency of lightning positions in time series, thereby more comprehensively understanding the movement state of lightning and more accurately representing the inherent patterns and laws of lightning relative position data.

[0070] Afterwards, the lightning relative position data time-series encoding feature vector is input into the decoder-based relative position predictor to obtain the predicted value of the lightning relative position data. In other words, the decoder can convert these abstract feature vectors into specific and understandable predicted values, namely lightning relative position data, to achieve mapping from feature space to actual position space. By predicting the relative position of lightning, the model can more accurately determine whether the major hazard source tank area of ​​dangerous goods in the port is on a potential lightning strike path, which is conducive to improving the accuracy of lightning warning.

[0071] The following is a detailed description of a specific implementation process of "inputting the lightning relative position data time series encoding feature vector into a decoder-based relative position predictor to obtain a predicted value of the lightning relative position data":

[0072] First, we need to determine the architecture of the decoder-based relative position predictor. According to the characteristics of the lightning relative position data, such as the length of the time series of the data, the complexity of the data changes, and other factors, we can reasonably select an appropriate decoder architecture. If the data presents a complex long-sequence dependency, the Transformer-based decoder can effectively capture the correlation information between each time point in the sequence with its self-attention mechanism, and accurately grasp the changing pattern of the relative position of lightning over a long time span. On the contrary, if the data has more complex dynamic changes and the sequence length is relatively short, variants of recurrent neural networks (RNNs), such as long short-term memory networks (LSTMs) or gated recurrent units (GRUs), may have more advantages. They perform well in dealing with short-term and long-term dependencies in time series and can better adapt to such data changes.

[0073] Next, enter the model training phase. In this phase, the parameters of the selected decoder-based relative position predictor must be initialized. The initialization method varies depending on the model architecture. You can use random initialization to assign initial values ​​to the model parameters, or you can use the weights of the pre-trained model to initialize. For example, for a Transformer-based model, initializing with pre-trained weights can speed up the convergence of the model and improve training efficiency. At the same time, carefully set training parameters such as learning rate, number of iterations, and batch size. The learning rate determines the step size of the model parameter update during training. A too large learning rate may cause the model training to be unstable and difficult to converge to the optimal solution; while a too small learning rate will make the training process extremely slow and time-consuming. The number of iterations controls the number of rounds of learning of the model for the training data, and needs to be set reasonably according to the actual situation to avoid overtraining or undertraining. The batch size affects the number of data samples used in each training. A suitable batch size helps improve the stability and efficiency of model training. Through multiple trials and adjustments, find the best combination of these parameters to achieve a balance between model training speed and accuracy.

[0074] During the training process, the data in the training set are sequentially input into the decoder-based relative position predictor, and the model will output the predicted value. The error between the predicted value and the actual relative position data is calculated, and the parameters of the model are adjusted according to the error with the help of the back-propagation algorithm, so that the error gradually decreases. In this process, pay close attention to the performance of the model on the validation set. Use the validation set to evaluate the model regularly. If the performance of the model on the validation set is found to be declining, this may be a signal that the model is overfitting. At this time, it is necessary to adopt corresponding regularization methods, such as L1 or L2 regularization, Dropout, etc., to prevent the model from overfitting the training data and enhance the generalization ability of the model. On the contrary, if the performance of the model on the validation set has not been effectively improved, it may mean that the model structure is too simple and cannot fully learn the complex patterns in the data, or the training is not sufficient. It is necessary to optimize the model structure, increase the complexity of the model, or further increase the number of training rounds to allow the model to better learn the data characteristics.

[0075] After sufficient training, when the model performance meets the requirements, it enters the prediction stage. First, load the trained model to ensure that the model parameters are in the optimal state. Input the time-series encoded feature vector of the lightning relative position data to be predicted into the loaded model. The decoder-based relative position predictor processes the input feature vector, combines the mapping mode learned through its own training, and outputs the predicted value of the lightning relative position data. These predicted values ​​can reflect key information such as the distance and direction of the lightning relative to the irrigation area at a certain moment in the future.

[0076] In an embodiment of the present application, the level determination unit 132 is used to determine the lightning warning level based on the predicted value of the lightning relative position data and the atmospheric electric field intensity time series fluctuation coding characteristics. Specifically, in an embodiment of the present application, the level determination unit is used to: add the predicted value of the lightning relative position data to the tail of the atmospheric electric field intensity time series fluctuation coding feature vector and then input it into the level analyzer based on the classifier to obtain the lightning warning level. It should be understood that the occurrence and degree of harm of lightning are not determined by a single factor. The relative position of lightning directly reflects the spatial relationship between lightning and the major hazard source tank area of ​​dangerous goods in the port, and its predicted value can know in advance the location and range of the threat that lightning may pose to the tank area. Therefore, in order to more comprehensively and accurately assess the lightning risk, the present application determines the lightning warning level by based on the predicted value of the lightning relative position data and the atmospheric electric field intensity time series fluctuation coding characteristics. In particular, in a specific example of the present application, the predicted value of the lightning relative position data is added to the tail of the atmospheric electric field intensity time series fluctuation coding feature vector and then input into the level analyzer based on the classifier to obtain the lightning warning level. That is, by integrating the predicted value of lightning relative position data and the encoded feature vector of atmospheric electric field intensity time series fluctuation, and inputting them into the classifier for analysis, the influence of various factors on the lightning warning level can be comprehensively considered. The warning level obtained in this way can more accurately reflect the actual threat level of lightning to the tank area, and reduce false alarms and missed alarms. For example, when the atmospheric electric field intensity fluctuates abnormally, and the predicted value of lightning relative position shows that lightning may approach the tank area, the classifier can give a more appropriate warning level based on this comprehensive information. The lightning warning level results here can be level 1 warning, level 2 warning, and level 3 warning. Specifically, the first-level warning indicates that there may be lightning activity, the atmospheric electric field in the covered area is increasing, the electric field is fluctuating, and the ground lightning backstroke point is located in the adjacent area 10 kilometers away from the port's major hazardous source tank area for dangerous goods, which may cause lightning accidents; the second-level warning indicates that there is a high possibility of lightning, the atmospheric electric field in the covered area is rapidly increasing, and the electric field changes and fluctuations are intensifying. The ground lightning backstroke point is 5-10 kilometers away from the port's major hazardous source tank area for dangerous goods, which increases the possibility of lightning accidents; the third-level warning indicates that lightning is about to occur, the atmospheric electric field in the covered area is fluctuating violently, and the ground lightning backstroke point is 0-5 kilometers away from the port's major hazardous source tank area for dangerous goods, which may cause lightning accidents.

[0077] In summary, the lightning warning system 100 for the major hazard source tank area of ​​dangerous goods in the port based on the embodiment of the present application is explained, which uses artificial intelligence-based data processing technology to split the time series data set of lightning characteristic parameters, and then performs relative position prediction based on the time series characteristics on the split time series data set of the lightning position, and at the same time, performs message constraint transmission based on the time series characteristics on the time series data set of the atmospheric electric field strength value, so as to automatically determine the lightning warning level according to the predicted value of the lightning relative position data and the atmospheric electric field strength time series fluctuation coding characteristics after the time series transmission. In this way, by real-time monitoring of the changes in the atmospheric electric field strength and accurately capturing the location and movement path of lightning, the occurrence time and intensity of lightning can be more accurately predicted, thereby providing reliable safety protection for the port dangerous goods tank area.

Claims

1. A lightning warning system for a major hazard source tank area of ​​dangerous goods in a port, characterized by: include: A lightning parameter acquisition module, used to obtain a time series data set of lightning characteristic parameters acquired by the lightning detection module, wherein the lightning characteristic parameters include an atmospheric electric field intensity value acquired by an atmospheric electric field instrument and a lightning position acquired by a lightning location system; A data splitting module, used for splitting the time series data set of the lightning characteristic parameters to obtain a time series data set of the atmospheric electric field intensity value and a time series data set of the lightning position; An analysis module is used to determine the lightning warning level based on the time series characteristics of the time series data set of the atmospheric electric field strength values ​​and the predicted relative position of the time series data set of the lightning positions, wherein the analysis module includes: a coding unit, used to perform electric field strength spatial constraint time series propagation and relative position prediction on the time series data set of the atmospheric electric field strength values ​​and the time series data set of the lightning positions, respectively, to obtain the atmospheric electric field strength time series fluctuation coding characteristics and the predicted value of the lightning relative position data; a level determination unit, used to determine the lightning warning level based on the predicted value of the lightning relative position data and the atmospheric electric field strength time series fluctuation coding characteristics.

2. The lightning warning system for the major hazard source tank area of ​​dangerous goods in the port according to claim 1 is characterized in that: The encoding unit comprises: A relative position time series data set construction subunit is used to obtain the position data of the hazardous source irrigation area, and to construct a time series data set of lightning relative position data based on the time series data set of the lightning position and the position data of the hazardous source irrigation area; An atmospheric electric field intensity coding subunit, used for performing message dynamic propagation based on local time series coding on the time series data set of the atmospheric electric field intensity value to obtain a coding feature of the atmospheric electric field intensity time series fluctuation; The relative position prediction subunit is used to perform time series encoding and decoding on the time series data set of the lightning relative position data to obtain the predicted value of the lightning relative position data.

3. The lightning warning system for the major hazard source tank area of ​​dangerous goods in the port according to claim 2 is characterized in that: The atmospheric electric field intensity encoding subunit comprises: The atmospheric electric field intensity time series coding secondary subunit is used to perform one-dimensional convolution coding on the time series data set of the atmospheric electric field intensity value to obtain a sequence distribution of the atmospheric electric field intensity local time series fluctuation coding feature vector; The atmospheric electric field intensity time series fluctuation coding secondary subunit is used to perform time dimension message transmission of atmospheric electric field intensity local constraint on the sequence distribution of the atmospheric electric field intensity local time series fluctuation coding feature vector to obtain the atmospheric electric field intensity time series fluctuation coding feature vector as the atmospheric electric field intensity time series fluctuation coding feature.

4. The lightning warning system for the major hazard source tank area of ​​dangerous goods in the port according to claim 3 is characterized in that: The atmospheric electric field intensity temporal fluctuation encoding secondary subunit comprises: The atmospheric electric field strength anchoring three-level subunit is used to perform tail constraint anchoring and axis constraint anchoring on the sequence distribution of the atmospheric electric field strength local time series fluctuation coding feature vector to obtain the atmospheric electric field strength local time series message transmission space tail constraint anchoring coding vector and the atmospheric electric field strength time series message transmission space axis constraint anchoring coding vector; The constraint factor calculation tertiary subunit is used to calculate the atmospheric electric field strength message transmission constraint factor of each atmospheric electric field strength local time series fluctuation coding feature vector in the sequence distribution of the atmospheric electric field strength local time series fluctuation coding feature vector based on the atmospheric electric field strength local time series message transmission space tail constraint anchor coding vector and the atmospheric electric field strength time series message transmission space axis constraint anchor coding vector; The atmospheric electric field intensity time series fluctuation coding feature generates a three-level sub-unit, which is used to perform message dynamic constraint transmission encoding on the sequence distribution of the atmospheric electric field intensity local time series fluctuation coding feature vectors based on the atmospheric electric field intensity message transmission constraint factor of each atmospheric electric field intensity local time series fluctuation coding feature vector to obtain the atmospheric electric field intensity time series fluctuation coding feature vector.

5. The lightning warning system for the major hazard source tank area of ​​dangerous goods in the port according to claim 4 is characterized in that: The atmospheric electric field intensity anchoring three-level subunit is used for: Extracting the last atmospheric electric field intensity local time series fluctuation coding feature vector from the sequence distribution of the atmospheric electric field intensity local time series fluctuation coding feature vector as the atmospheric electric field intensity local time series message transmission space tail end constraint anchor coding vector; A cluster analysis is performed on the sequence distribution of the characteristic vector of the local temporal fluctuation coding of the atmospheric electric field intensity to obtain the spatial axis constraint anchor coding vector of the atmospheric electric field intensity temporal message transmission.

6. The lightning warning system for the major hazard source tank area of ​​dangerous goods in the port according to claim 5 is characterized in that: The constraint factor calculation three-level sub-unit includes: The fourth-level sub-unit for calculating the tail end message transmission constraint factor of the atmospheric electric field intensity is used to calculate the tail end message transmission constraint factor of each atmospheric electric field intensity local time series fluctuation coding feature vector in the sequence distribution of the atmospheric electric field intensity local time series fluctuation coding feature vector relative to the tail end constraint anchor coding vector of the atmospheric electric field intensity local time series message transmission space; The fourth-level sub-unit for calculating the atmospheric electric field strength axis message transmission constraint factor is used to calculate the atmospheric electric field strength axis message transmission constraint factor of each atmospheric electric field strength local time series fluctuation coding feature vector in the sequence distribution of the atmospheric electric field strength local time series fluctuation coding feature vector relative to the atmospheric electric field strength time series message transmission space axis constraint anchor coding vector; The fourth-level sub-unit for calculating the atmospheric electric field strength message transmission constraint factor is used to calculate the atmospheric electric field strength message transmission constraint factor of each atmospheric electric field strength local time series fluctuation coding feature vector based on the atmospheric electric field strength tail message transmission constraint factor and the atmospheric electric field strength axis message transmission constraint factor of each atmospheric electric field strength local time series fluctuation coding feature vector in the sequence distribution of the atmospheric electric field strength local time series fluctuation coding feature vector.

7. The lightning warning system for the major hazard source tank area of ​​dangerous goods in the port according to claim 6 is characterized in that: The atmospheric electric field intensity message transmission constraint factor calculation four-level subunit is used for: Use a standard space constraint matrix to perform linear transverse covariance on the atmospheric electric field intensity tail end message transmission constraint factor and the atmospheric electric field intensity axis line message transmission constraint factor corresponding to the atmospheric electric field intensity local time series fluctuation encoding feature vector to obtain the atmospheric electric field intensity tail end message transmission covariance constraint factor and the atmospheric electric field intensity axis line message transmission covariance constraint factor; Based on the atmospheric electric field strength tail end message transmission covariance constraint factor and the atmospheric electric field strength axis line message transmission covariance constraint factor, the atmospheric electric field strength tail end message transmission constraint factor and the atmospheric electric field strength axis line message transmission constraint factor are globally consistent with each other to obtain an optimized atmospheric electric field strength tail end message transmission constraint factor and an optimized atmospheric electric field strength axis line message transmission constraint factor; The optimized atmospheric electric field strength tail end message transmission constraint factor and the optimized atmospheric electric field strength axis message transmission constraint factor are weighted based on the sigmoid function to obtain the atmospheric electric field strength message transmission constraint factor corresponding to the atmospheric electric field strength local time series fluctuation encoding feature vector.

8. The lightning warning system for the major hazard source tank area of ​​dangerous goods in the port according to claim 7 is characterized in that: The relative position prediction subunit is used to: Performing time series coding based on a bidirectional recurrent neural network on the time series data set of the lightning relative position data to obtain a time series coding feature vector of the lightning relative position data; The lightning relative position data time series encoding feature vector is input into a decoder-based relative position predictor to obtain a predicted value of the lightning relative position data.

9. The lightning warning system for the major hazard source tank area of ​​dangerous goods in the port according to claim 8 is characterized in that: The level determination unit is used to: add the predicted value of the lightning relative position data to the tail of the atmospheric electric field intensity time series fluctuation encoding feature vector and then input it into a classifier-based level analyzer to obtain the lightning warning level.

Images