Weather index construction method and system for low-temperature congelation disaster of power grid

Abstract

The invention discloses a power grid low-temperature congelation disaster weather index construction method and system, and relates to the technical field of power grid disaster assessment, and the method comprises the steps: obtaining low-temperature congelation disaster events based on power grid observation data, carrying out the preliminary classification, and determining the disaster types of which weather indexes need to be established; evaluating the matching relationship between the temperature and humidity in the high temporal-spatial resolution meteorological data and the power grid observation data, and establishing a multi-element database corresponding to the low-temperature congelation disaster event; extracting multi-element features and combinations corresponding to different disaster types in a multi-element database through a deep learning algorithm; and analyzing the multi-element characteristics and combination to obtain a corresponding relation between the power grid congelation disaster and the combined meteorological condition, and establishing a weather index. According to the method, high-temporal-spatial-resolution meteorological data, power grid observation data, a deep learning algorithm and the like are utilized, basic characteristics which effectively extract influence low-temperature congelation meteorological conditions are constructed for power grid high influence meteorological conditions, and power grid congelation weather indexes are established.

Description

Technical Field

[0001] This invention relates to the field of power grid disaster assessment technology, specifically to a method and system for constructing meteorological indicators for power grid low-temperature freezing disasters. Background Technology

[0002] The impact of low-temperature freezing on the power grid primarily affects exposed transmission lines and towers. First, when the weight of the ice accumulation is excessive, exceeding the mechanical arm's stress limit when the total weight of the ice and the line itself is beyond the limit, line breakage and tower collapse can occur. Second, ice accumulation between insulator discs can create bridging or insufficient spacing; when the voltage across the insulator exceeds the breakdown voltage, flashover and malfunctions can occur. Third, uneven ice accumulation on transmission lines, under wind influence, can cause vibrations, characterized by low frequency and large amplitude. This line galloping can lead to flashovers, tripping, and even damage to hardware, insulators, conductor breakage, and tower collapse—major electrical accidents. Fourth, uneven ice accumulation can cause mechanical damage to the lines. Uneven ice accumulation on transmission lines can create uneven stress, resulting in tension differences and line breakage. Predicting the potential impact of ice accumulation disasters on transmission line systems based on meteorological information provided by meteorological departments has become an urgent problem to solve. To address this issue, numerous studies on the meteorological conditions for power line icing have been conducted both domestically and internationally. However, due to differences in spatial and temporal resolution and coverage periods between meteorological and power grid data, as well as the lack of effective on-site monitoring data, research progress on low-temperature condensation meteorology has been slow.

[0003] Existing studies mostly employ statistical analysis of process data. For example, correlation analysis of meteorological elements reveals that the main meteorological conditions for low-temperature freezing are temperature, humidity, and wind speed. Analysis of the meteorological conditions for low-temperature freezing in January and February 2008 shows that freezing occurred under conditions of low temperature, low precipitation, low wind speed (or calm wind), and relatively humid air. Fuzzy information allocation methods and spectral clustering have been used to analyze the meteorological conditions for icing processes in power grids at different levels. However, the results of fuzzy information methods are relatively coarse, and spectral clustering requires pre-defined samples of reference categories. In recent years, with the development of deep learning algorithms, they have been increasingly applied in many fields.

[0004] Therefore, a method for constructing a meteorological index for low-temperature freezing disasters in power grids is needed. This method would comprehensively utilize high spatiotemporal resolution meteorological data, power grid observation data, and deep learning algorithms combined with traditional statistical methods to construct an index that effectively extracts the basic characteristics affecting low-temperature freezing meteorological conditions, such as the characteristics and combinations of multiple elements and the range of element changes, and their relationship with different types of icing. This would establish a meteorological index for power grid freezing, which is hoped to be applied to low-temperature freezing disaster zoning and monitoring, providing some basis and ideas for power grid disaster risk assessment. Summary of the Invention

[0005] In view of the above-mentioned problems, the present invention is proposed.

[0006] Therefore, the technical problem solved by this invention is: how to provide a method and system for constructing meteorological indicators for low-temperature freezing disasters in power grids.

[0007] To address the aforementioned technical problems, this invention provides the following technical solution: a method for constructing meteorological indicators for power grid low-temperature freezing disasters, comprising: acquiring low-temperature freezing disaster events within a target area based on power grid observation data, and preliminarily classifying them to determine the disaster types for which meteorological indicators need to be established; evaluating the correlation between temperature and humidity in high spatiotemporal resolution meteorological data and the power grid observation data, and establishing a multi-element database corresponding to the low-temperature freezing disaster events; extracting multi-element features and combinations corresponding to different disaster types in the multi-element database using a deep learning algorithm; analyzing the multi-element features and combinations to obtain the variation range of meteorological elements corresponding to power grid disasters, i.e., obtaining the correspondence between power grid freezing disasters and combined meteorological conditions, and establishing meteorological indicators.

[0008] As a preferred embodiment of the method for constructing meteorological indicators for low-temperature freezing disasters in power grids as described in this invention, the power grid observation data includes sensor observation data and manual observation data; wherein, the sensor observation data refers to the data generated by checking the ice accumulation of data during the observation process through power grid tension sensors; the elements of the manual observation data include observed ice thickness, air temperature, and relative humidity; the high spatiotemporal resolution meteorological data is acquired through CLDAS data, wherein CLDAS data refers to hourly data of CMA land surface data assimilation system products.

[0009] As a preferred embodiment of the method for constructing meteorological indicators for low-temperature freezing disasters in power grids according to the present invention, the following steps are taken: Based on power grid observation data, low-temperature freezing disaster events within the target area are acquired and preliminarily classified. Determining the disaster types for which meteorological indicators need to be established includes classifying the disaster types according to temperature and humidity. Specifically: when the temperature is less than or equal to a first preset threshold and the humidity is greater than or equal to a second preset threshold, the disaster type is determined to be low temperature and high humidity; when the temperature is greater than the first preset threshold and the humidity is less than the second preset threshold, the disaster type is determined to be high temperature and low humidity; when the temperature is less than or equal to the first preset threshold and the humidity is less than the second preset threshold, the disaster type is determined to be low temperature and low humidity; when the temperature is greater than the first preset threshold and the humidity is greater than or equal to the second preset threshold, the disaster type is determined to be high temperature and high humidity. Low temperature and high humidity, high temperature and low humidity, low temperature and low humidity, and high temperature and high humidity are identified as disaster types for which meteorological indicators need to be established.

[0010] As a preferred embodiment of the method for constructing meteorological indicators for low-temperature freezing disasters in power grids according to the present invention, the following steps are included in evaluating the correlation between temperature and humidity in high spatiotemporal resolution meteorological data and the power grid observation data, and establishing a multi-element database corresponding to low-temperature freezing disaster events:

[0011] The correlation coefficient r and root mean square error of temperature and humidity in the sensor observation data and manual observation data with the high spatiotemporal resolution meteorological data are calculated. According to the sample size n, the correlations all reach a reliability of 95%.

[0012] To test the differences in factors among different disaster types, a composite t-test was used for significance testing. Variables that achieved significant reliability were selected for focused analysis, i.e., |t|≥t. α The elements; among which, t α For a significant threshold

[0013] The root mean square error is used to measure the consistency between power grid observations and multiple elements of meteorological reanalysis on continuous data, and the specific formula is as follows:

[0014]

[0015] Where n is the number of samples, and the smaller the RMSE, the more consistent the results.

[0016] The t-test method refers to testing the significance of the difference between the mean of two groups of samples, namely, class 1 samples and class 2 samples, and the population mean, as well as the difference between the two. Specifically:

[0017] For the t-test of the difference between class 1 and the population, the following formula is used:

[0018]

[0019] in, s represents the sample mean and standard deviation, respectively; μ0 is the population mean;

[0020] To test class 2, then Replace with The significance threshold t is obtained by querying the statistical distribution table. α If |t|≥t α If , it indicates that there is significance;

[0021] The t-test is commonly used to evaluate the differences between variables corresponding to class 1 and class 2 samples.

[0022]

[0023] In the formula, n1 and n2 represent the mean and number of years of the corresponding elements x and y in the class 1 sample and the class 2 sample, respectively. and These represent the variances of the two factors, respectively.

[0024] The significance threshold t is obtained by querying the statistical distribution table. α If |t|≥t α This indicates that the differences between class 1 and class 2 samples are significant.

[0025] As a preferred embodiment of the method for constructing meteorological indicators for power grid low-temperature freezing disasters according to the present invention, the step of extracting multi-element features and combinations corresponding to different disaster types in the multi-element database using a deep learning algorithm is achieved by using the PDCDP algorithm. The PDCDP algorithm adds path and distribution attributes to the data points, based on the CDP algorithm which only considers data point density and distance attributes. Specifically, it includes the following steps:

[0026] Establish minimum distance connected path features for the dataset;

[0027] The path attributes and distribution attributes of data points are calculated based on the feature set of minimum distance connected paths, and an improved decision graph is generated.

[0028] Regression analysis is used to automatically determine the representative feature samples corresponding to the disaster type and to preliminarily classify the results.

[0029] The fuzzy weighted method is used to determine the characteristic disaster type samples to which the remaining data points should be classified and to separate the "transitional" class, thus completing the objective classification.

[0030] As a preferred embodiment of the method for constructing meteorological indicators for power grid low-temperature freezing disasters described in this invention, the step of establishing the minimum distance connected path features of the dataset includes the following steps.

[0031] A standardized multi-meteorological element dataset was established, using stations and icing disasters as identifiers.

[0032]

[0033] Where {1…n} represents the data point index, m represents the data dimension, and x represents the data position. ij Represents data value, id i Let ID = {id1, id2, ..., id3} represent the unique identifier of the i-th data point. n Calculate the distance matrix between its data points.

[0034]

[0035] Where, dij Let d represent the Euclidean distance between the i-th and j-th data points, and find the minimum value d in DS. mimj Given the row number mi and column number mj of the vector ES, generate a new vector ES = [mi, mj, d]. mimj ], and create a new n×n×4 three-dimensional matrix RFS = NaN to represent the path feature vector, let RFS mimj =RFS mjmi =[1,d mimj ,d 2 mimj ,d mimj Create a new set of indices di = {mi, mj}, where do is the complement of di in the set of indices {1, 2, ..., n};

[0036] Find the minimum value d in the subset DS(di,do) of DS. mdimdj , and its row number mdi and column number mdj in DS, generate ns = [mdi, mdj, d mdimdj The line is appended to ES as a newline. Then, the path feature vector from mdj to mdi is added (i.e., the RFS is updated). mdimdj =RFS mdjmdi =[1,d mdimdj ,d mdimdj 2 ,d mdimdj ]), and simultaneously add the path feature vectors of each element in the set odi = di - mdi to mdi (i.e., update RFS). odi(i)mdj =RFS mdjodi(i) =[RFS mdiodi(i)1 +1,RFS mdiodi(i)2 +d mdimdj RFS mdiodi(i)3 +d mdimdj 2 ,max(RFS mdiodi(i)4 ,d mdimd j Finally, remove mdj from set do and append it to the end of set di;

[0037] If do is not empty, return to step "Find the minimum value d in the subset DS(di,do) of DS". mdimdj Otherwise, end the step "Establish a standardized multi-meteorological element dataset based on the station and icing disaster" and complete the generation of MDCG[ID,ES] and RFS;

[0038] The process of calculating the path attributes and distribution attributes of data points based on the feature set of minimum distance connected paths and generating an improved decision graph includes the following steps:

[0039] With d cTo determine the cutoff distance parameter, the density attribute ρ for each data point is calculated using the DPC algorithm formula. i and distance attribute δ i Then we get

[0040] The path attribute of each data point is calculated as follows:

[0041]

[0042] In the formula, This indicates starting from data point i with δ i The other endpoint of the side of length;

[0043] The distribution attribute of each data point is calculated as follows:

[0044] θ i =∑(RFS) ij2 / d ij / RFS ij1 );

[0045] calculate and dgz i =dgx i ×dgy i This yields vectors dgx, dgy, and dgz.

[0046] As a preferred embodiment of the method for constructing meteorological indicators for low-temperature freezing disasters in power grids according to the present invention, the step of automatically determining the representative feature samples corresponding to the disaster type using regression analysis and initially classifying the results includes the following steps:

[0047] For vectors dgx, dgy, and dgz, regression analysis is performed on the constant fitting function dgz = C with confidence interval alpha, where C is an automatically calculated constant. This yields rint, an n×2 matrix, representing the upper and lower bounds of the residuals of n data points within the alpha confidence interval. Data points with the lower bound rint(:,1)>0 are taken as representative days of the characteristic weather type, i.e., cluster centers cc, with a number of ccn. The preset value alpha = 0.05 represents the 95% confidence interval.

[0048] Iterate through the cluster centers cc, and find the ci-th cluster center cc. ci Cutoff distance d c Data points within this cluster are assigned to this cluster as the cluster kernel, where a data point k belongs to cc. ci It also belongs to cc cj Then k will be arranged according to By assigning the ci-th or cj-th cluster core to the cluster core, the preliminary classification of the low-temperature freezing disaster event samples is completed, namely the division of the cluster cores co.

[0049] The process of using fuzzy weighting to determine the characteristic disaster type samples to which the remaining data points should be classified and separating the "transitional" class to complete the objective classification includes the following steps:

[0050] Generate weight matrix Calculate the membership matrix P, which is an oln×con matrix, where oln represents the number of unassigned data points and con represents the number of cluster cores.

[0051]

[0052] In the formula, co ci Let ol represent the ci-th cluster kernel in the cluster kernel set co. oi This represents the oi-th data point in the set of unassigned data points ol.

[0053] Pick like Then iterate through the data points ol oi d c Data points j within the range and update in Then set all rows oi of P to 0 and then set ol oi Delete from ol and move to co ci In the middle, it represents unassigned data points ol. oi Included in cluster nucleus co ci ;

[0054] Repeat the previous step until Then all data points in ol will be marked as noise.

[0055] To further address the aforementioned technical problems, this invention provides the following technical solution: a system for constructing meteorological indicators for power grid low-temperature freezing disasters, comprising: a data acquisition module for collecting power grid observation data, including power grid tension sensor observation data and manual observation data, to obtain low-temperature freezing disaster events within the target area; a data analysis module for evaluating the consistency between temperature, humidity, and power grid observation data in high spatiotemporal resolution meteorological data, calculating the correlation coefficient r and root mean square error, and establishing a multi-element database corresponding to power grid low-temperature freezing disaster events; a difference testing module for performing significant difference analysis on meteorological elements between different disaster types, and using a composite t-test method to screen out significant variables; a deep learning module for using the PDCDP algorithm combined with deep learning methods to extract multi-element features and combinations corresponding to different disaster types from the multi-element database, and generating an improved decision map; and an indicator construction module for analyzing multi-element features and combinations, obtaining the variation range of meteorological elements corresponding to power grid disasters, establishing the correspondence between power grid freezing disasters and combined meteorological conditions, and finally forming scientific meteorological indicators.

[0056] A computer device includes a memory and a processor, the memory storing a computer program, characterized in that the processor executes the computer program to implement the steps of the method for constructing meteorological indicators for low-temperature freezing disasters in power grids as described above.

[0057] A computer-readable storage medium having a computer program stored thereon, characterized in that, when the computer program is executed by a processor, it implements the steps of the method for constructing meteorological indicators for low-temperature freezing disasters in power grids as described above.

[0058] The beneficial effects of this invention are as follows: By comprehensively utilizing high spatiotemporal resolution meteorological data, power grid stress sensor observation data, and manual observation data, combined with deep learning algorithms and an improved density peak classification method, this invention can more accurately identify and classify low-temperature freezing disaster events and establish a scientific meteorological index model. This innovative method significantly improves the prediction accuracy and prevention capabilities of low-temperature freezing disasters, provides effective technical support for the stable operation of the power grid, and helps to take timely countermeasures to reduce the losses caused by the disaster. Attached Figure Description

[0059] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0060] Figure 1 The overall flowchart of a method for constructing meteorological indicators for low-temperature freezing disasters in power grids, provided in an embodiment of the present invention.

[0061] Figure 2 A flowchart is provided for establishing the meteorological indicators of the PDCDP algorithm of this invention.

[0062] Figure 3 This is a statistical chart of the number of stations where ice formation occurred, as presented in this invention.

[0063] Figure 4 This is a statistical chart of air humidity and temperature corresponding to a certain condensation station from December 16 to 21, 2023 (horizontal axis: temperature; vertical axis: humidity).

[0064] Figure 5 This is a schematic diagram showing the significance t-test values ​​of meteorological elements with and without freezing during low temperature and high humidity events.

[0065] Figure 6 This is a map showing the changes in ground features corresponding to freezing and condensation at low temperatures and high humidity. Detailed Implementation

[0066] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the protection scope of the present invention.

[0067] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0068] Example 1

[0069] Reference Figures 1-6 As an embodiment of the present invention, a method for constructing meteorological indicators for low-temperature freezing disasters in power grids is provided.

[0070] Low-temperature freezing is one of the high-impact weather conditions on power grids, primarily affecting exposed transmission lines and towers. Low-temperature freezing occurs when supercooled raindrops, fog droplets, or wet snow in cold, humid weather conditions condense and freeze on near-ground surfaces below 0°C, forming an ice layer. Freezing types include rime, hoarfrost, wet snow, and mixed rain / fog or rain / snow freezing. Hoarfrost and wet snow have relatively weak adhesion, while rime and mixed rime have strong adhesion and pose the greatest threat to power grids.

[0071] The local meteorological conditions associated with low-temperature freezing are mainly temperature, humidity, precipitation, and wind speed. The importance or threshold of these associated meteorological elements varies depending on the stage or type of low-temperature freezing. For example, in the early stages of freezing, the correlation coefficient between temperature and wind speed is relatively high, while the correlation coefficient between relative humidity is relatively low. In the later stages of freezing, the trend of these correlation coefficients is exactly the opposite. In the early stages of freezing, the correlation coefficients between temperature and wind speed remain high, while the correlation coefficient between relative humidity is relatively low. In the later stages of freezing, the correlation coefficient between relative humidity increases significantly compared to the early stages, while the correlation coefficients between temperature and wind speed show a decreasing trend, but still remain at a relatively high value. Furthermore, the meteorological conditions for the same disaster event may differ for different regions or terrains.

[0072] Therefore, based on these fundamental characteristics of low-temperature freezing, this invention provides a meteorological index for power grid freezing based on the multi-element characteristics and combination types, and the range of element changes in meteorological conditions for low-temperature freezing in power grids. It also aims to reflect the relationship between low-temperature freezing in power grids and multi-element meteorological conditions, so as to assess, monitor, and forecast power grid freezing disasters through changes in meteorological elements.

[0073] The present invention includes the following steps:

[0074] S1: Based on power grid observation data, acquire low-temperature freezing disaster events in the target area, classify them in advance, and determine the types of disasters for which meteorological indicators need to be established.

[0075] Specifically, power grid observation data includes sensor observation data and manual observation data. Sensor observation data refers to the data generated by using power grid tension sensors to check for icing during the observation process. The elements of manual observation data include ice thickness, air temperature, and relative humidity, and the observation frequency is twice a day.

[0076] Furthermore, disaster types are classified according to temperature and humidity, specifically: when the temperature is less than or equal to the first preset threshold and the humidity is greater than or equal to the second preset threshold, the disaster type is determined to be low temperature and high humidity; when the temperature is greater than the first preset threshold and the humidity is less than the second preset threshold, the disaster type is determined to be high temperature and low humidity; when the temperature is less than or equal to the first preset threshold and the humidity is less than the second preset threshold, the disaster type is determined to be low temperature and low humidity; when the temperature is greater than the first preset threshold and the humidity is greater than or equal to the second preset threshold, the disaster type is determined to be high temperature and high humidity; low temperature and high humidity, high temperature and low humidity, low temperature and low humidity, and high temperature and high humidity are considered as disaster types for which meteorological indicators need to be established.

[0077] In a preferred embodiment, if the first preset threshold is set to 0°C and the second preset threshold is set to 75%, then the freezing disaster can be preliminarily divided into four categories: 1. Low temperature and high humidity: temperature less than or equal to 0°C, humidity greater than or equal to 75%; 2. High temperature and low humidity: temperature greater than 0°C, humidity less than 75%; 3. Low temperature and low humidity: temperature less than or equal to 0°C, humidity less than 75%; 4. High temperature and high humidity: temperature greater than 0°C, humidity greater than or equal to 75%, as shown in Table 1.

[0078] Table 1. Number of stations with or without freezing / icing observed by artificial ice monitoring in Zhaotong power grid from December 16-22, 2023.

[0079]

[0080] S2: Evaluate the correlation between temperature and humidity in high spatiotemporal resolution meteorological data and power grid observation data, and establish a multi-element database corresponding to low-temperature freezing disaster events.

[0081] Specifically, high spatiotemporal resolution meteorological data is acquired through CLDAS data, which refers to hourly data from the CMA land surface data assimilation system product.

[0082] Specifically, the correlation coefficient *r* and root mean square error (RMSE) of temperature and humidity in sensor observation data, manual observation data, and high spatiotemporal resolution meteorological data are calculated, with a 95% reliability of the correlation based on the sample size *n*. The correlation coefficient *r* refers to the Pearson correlation coefficient, a value between -1 and 1. When the linear relationship between two variables strengthens, the correlation coefficient tends towards -1 or 1; when one variable increases and the other also increases, it indicates a positive correlation (correlation coefficient greater than 0); when one variable increases and the other decreases, it indicates a negative correlation (correlation coefficient less than 0); if the correlation coefficient equals 0, it indicates no correlation. In power grid observation data, a higher correlation between the two types of data indicates a closer relationship, i.e., better consistency. The following is the method for calculating the correlation:

[0083]

[0084] Among them, X i Y i These are the observed values ​​of two variables; This represents the average of the two variables. After determining the significance level α based on the sample size n, the significance threshold r is obtained by consulting the statistical distribution table. α If |r|≥r α This indicates that there is a significant correlation between the sequences.

[0085] The root mean square error (RMSE) is used to measure the consistency between power grid observations and multiple elements of meteorological reanalysis on continuous data. The specific formula is as follows:

[0086]

[0087] Where n is the number of samples, and the smaller the RMSE, the more consistent the results.

[0088] Table 2 Relationship and Deviation between CLDAS Data Elements and Power Grid Data

[0089]

[0090] As shown in Table 2, the data measured by the tension sensor is more consistent with the data measured by CLDAS.

[0091] To test the differences between different disaster types, a composite t-test was used. Although initially twenty-two variables were analyzed, including concurrent air temperature, 3-hour temperature variation, maximum and minimum temperatures within the previous 24 hours, hourly precipitation, 3-hour precipitation variation, 24-hour precipitation, concurrent relative humidity, 3-hour relative humidity variation, 24-hour maximum relative humidity, concurrent wind speed, 3-hour wind speed variation, 24-hour average wind speed, concurrent shortwave radiation, 3-hour shortwave radiation variation, 24-hour shortwave radiation variation, concurrent air pressure, 3-hour air pressure variation, 24-hour air pressure variation, concurrent specific humidity, 3-hour specific humidity variation, and 24-hour specific humidity variation, the variables that reached significant reliability were selected for focused analysis based on the significance test. That is, |t| ≥ t α The elements.

[0092] The t-test method is used to test the significance of the difference between the means of two groups of samples, namely, class 1 samples and class 2 samples, and the population mean, as well as the difference between the two groups. Specifically:

[0093] For the t-test of the difference between class 1 and the population, the following formula is used:

[0094]

[0095] in, s represents the sample mean and standard deviation, respectively; μ0 is the population mean; this formula follows a t-distribution with degrees of freedom v = n-2.

[0096] To test class 2, then Replace with The significance threshold t is obtained by querying the statistical distribution table. α If |t|≥t α If , it indicates that there is significance.

[0097] The t-test is commonly used to evaluate the differences between variables corresponding to class 1 and class 2 samples.

[0098]

[0099] In the formula, n1 and n2 represent the mean and number of years of the corresponding elements x and y in the class 1 sample and the class 2 sample, respectively. and These represent the variances of the two factors respectively; this formula follows a t-distribution with degrees of freedom ν = n1 + n2 - 2. The significance threshold t is obtained by consulting a statistical distribution table. α If |t|≥t α This indicates that the differences between class 1 and class 2 samples are significant.

[0100] S3: Extract multi-element features and combinations corresponding to different disaster types from a multi-element database using deep learning algorithms.

[0101] Specifically, in this embodiment, the PDCDP algorithm, which considers only the density and distance attributes of data points (such as weather phenomena composed of surface meteorological elements at a certain time and latitude and longitude) in the CDP algorithm, adds the path and distribution attributes of data points. While linearly increasing the time complexity, it achieves better performance than the CDP algorithm, making it more applicable to severe weather identification. It can control clustering with a single truncation distance parameter, automatically selecting cluster centers and dividing non-cluster center data points. Finally, it only needs to determine the disaster level criteria of the cluster centers and some small clusters of data points to complete the identification of multiple severe weather events in one go. The specific steps of this algorithm are as follows:

[0102] Establish minimum distance connected path features for the dataset:

[0103] A standardized multi-meteorological element dataset was established, using stations and icing disasters as identifiers.

[0104]

[0105] Where {1…n} represents the data point index, m represents the data dimension, and x represents the data position. ij Represents data value, id i Let ID = {id1, id2, ..., id3} represent the unique identifier of the i-th data point. n Calculate the distance matrix between its data points.

[0106]

[0107] Where, d ij Let d represent the Euclidean distance between the i-th and j-th data points, and find the minimum value d in DS. mimj Given the row number mi and column number mj of the vector ES, generate a new vector ES = [mi, mj, d]. mimj ], and create a new n×n×4 three-dimensional matrix RFS = NaN to represent the path feature vector, let RFS mimj =RFS mjmi =[1,d mimj ,d 2 mimj ,d mimj Create a new set of indices di = {mi, mj}, where do is the complement of di in the set of indices {1, 2, ..., n};

[0108] Find the minimum value d in the subset DS(di,do) of DS. mdimdj , and its row number mdi and column number mdj in DS, generate ns = [mdi, mdj, d mdimdjThe line is appended to ES as a newline; then the path feature vector from mdj to mdi is added (i.e., RFS is updated). mdimdj =RFS mdjmdi =[1,d mdimdj ,d mdimdj 2 ,d mdimdj ]), and simultaneously add the path feature vectors of each element in the set odi = di - mdi to mdi (i.e., update RFS). odi(i)mdj =RFS mdjodi(i) =[RFS mdiodi(i)1 +1,RFS mdiodi(i)2 +d mdimdj RFS mdiodi(i)3 +d mdimdj 2 ,max(RFS mdiodi(i)4 ,d mdimdj Finally, remove mdj from set do and append it to the end of set di;

[0109] If do is not empty, return to step "Find the minimum value d in the subset DS(di,do) of DS". mdimdj Otherwise, end the step "Establish a standardized multi-meteorological element dataset using the station and icing disaster as identifiers" and complete the generation of MDCG[ID,ES] and RFS.

[0110] Calculate the path attributes and distribution attributes of data points based on the feature set of minimum distance connected paths, and generate an improved decision graph: d c To determine the cutoff distance parameter, the density attribute ρ for each data point is calculated using the DPC algorithm formula. i and distance attribute δ i Then we get

[0111] The path attribute of each data point is calculated as follows:

[0112]

[0113] In the formula, This indicates starting from data point i with δ i The other endpoint of the side of length;

[0114] The distribution attribute of each data point is calculated as follows:

[0115] θ i =∑(RFS) ij2 / d ij / RFS ij1 );

[0116] calculate and dgz i=dgx i ×dgy i This yields vectors dgx, dgy, and dgz.

[0117] Regression analysis is used to automatically determine the representative feature samples corresponding to the disaster type and to preliminarily classify the results: the vectors dgx, dgy, and dgz are fitted with a constant function dgz = C with a confidence interval alpha, and C is an automatically calculated constant. Regression analysis is performed to obtain rint, an n×2 matrix, which represents the upper and lower limits of the residuals of n data points within the alpha confidence interval. The data points with the lower limit rint(:,1)>0 are taken as the representative days of the characteristic weather type, that is, the cluster centers cc, and the number of them is ccn. The preset value alpha = 0.05 represents the 95% confidence interval.

[0118] Iterate through the cluster centers cc, and find the ci-th cluster center cc. ci Cutoff distance d c Data points within this cluster are assigned to this cluster as the cluster kernel, where a data point k belongs to cc. ci It also belongs to cc cj Then k will be arranged according to By assigning the ci-th or cj-th cluster core to the sample, the preliminary classification of the objective classification of low-temperature freezing disaster events is completed, namely the division of the cluster cores co.

[0119] The fuzzy weighted method is used to determine the characteristic disaster type samples to which the remaining data points should be classified and to separate the "transitional" class, thus completing the objective classification:

[0120] Generate weight matrix Calculate the membership matrix P, which is an oln×con matrix, where oln represents the number of unassigned data points and con represents the number of cluster cores.

[0121]

[0122] In the formula, co ci Let ol represent the ci-th cluster kernel in the cluster kernel set co. oi This represents the oi-th data point in the set of unassigned data points ol.

[0123] Pick like Then iterate through the data points ol oi d c Data points j within the range and update in Then set all rows oi of P to 0 and then set ol oi Delete from ol and move to co ci In the middle, it represents unassigned data points ol. oiIncluded in cluster nucleus co ci ;

[0124] Repeat the previous step until Then all data points in ol will be marked as noise.

[0125] S4: Analyze the characteristics and combinations of multiple elements to obtain the range of variation of meteorological elements corresponding to power grid disasters, that is, to obtain the correspondence between power grid freezing disasters and combined meteorological conditions, and establish meteorological indicators.

[0126] Specifically, based on the classification identifiers, scatter plot analysis is performed on the key elements of the same type of multi-element to obtain their range of variation, and the threshold values ​​of elements occurring for different freezing types are combined.

[0127] In summary, this invention, by comprehensively utilizing high spatiotemporal resolution meteorological data, power grid stress sensor observations, and manual observation data, combined with deep learning algorithms and an improved density peak classification method, can more accurately identify and classify low-temperature freezing disaster events, and establish a scientific meteorological index model. This innovative method significantly improves the prediction accuracy and prevention capabilities of low-temperature freezing disasters, provides effective technical support for the stable operation of the power grid, and helps to take timely countermeasures to reduce the losses caused by disasters.

[0128] Example 2

[0129] Reference Figures 1-6 This invention provides a system for constructing meteorological indicators for power grid low-temperature freezing disasters, comprising: a data acquisition module for collecting power grid observation data, including power grid tension sensor observation data and manual observation data, to obtain low-temperature freezing disaster events in the target area; a data analysis module for evaluating the consistency between temperature, humidity and power grid observation data in high spatiotemporal resolution meteorological data, calculating the correlation coefficient r and root mean square error, and establishing a multi-element database corresponding to power grid low-temperature freezing disaster events; a difference testing module for performing significant difference analysis on meteorological elements between different disaster types, and using a composite t-test to screen out significant variables; a deep learning module for using the PDCDP algorithm combined with deep learning methods to extract multi-element features and combinations corresponding to different disaster types from the multi-element database, and generating an improved decision map; and an indicator construction module for analyzing multi-element features and combinations, obtaining the variation range of meteorological elements corresponding to power grid disasters, establishing the correspondence between power grid freezing disasters and combined meteorological conditions, and finally forming scientific meteorological indicators.

[0130] Example 3

[0131] This is one embodiment of the present invention, which differs from the previous embodiment in that:

[0132] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, essentially, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0133] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-included system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device.

[0134] More specific examples of computer-readable media (a non-exhaustive list) include: electrical connections (electronic devices) having one or more wires, portable computer disk drives (magnetic devices), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Furthermore, computer-readable media can even be paper or other suitable media on which the program can be printed, because the program can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in computer memory.

[0135] It should be understood that various parts of the present invention can be implemented in hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented in software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0136] Example 4

[0137] Reference Figures 1-6 As an embodiment of the present invention, a method for constructing meteorological indicators of low-temperature freezing disasters in power grids is provided. In order to verify the beneficial effects of the present invention, scientific demonstration is carried out through economic benefit calculation and simulation experiments.

[0138] The specific implementation method is illustrated using the process analysis from December 16-22, 2023 as an example.

[0139] like Figure 3 As shown, at 12:55 PM on December 16, 2023, meteorological departments in many parts of Yunnan Province issued a yellow cold wave warning signal. From December 16 to 22, 2023, affected by strong cold air, the daily average temperature in many parts of northeastern Yunnan dropped by more than 10℃ within 24 hours, resulting in a cold wave. Temperatures above 1500 meters in altitude remained below 0℃, and the minimum temperature above 2000 meters dropped below -3℃. This cold wave caused a significant temperature drop, leading to condensation disasters in the power grid in many parts of Yunnan Province, especially Zhaotong City, causing adverse effects. Low-temperature freezing mainly occurred in areas above 800 meters in altitude. More than 81% of the observation stations in the region experienced low-temperature freezing, with 2 stations experiencing freezing throughout the entire process and 5 stations experiencing no freezing. The total number of low-temperature freezing events in the region over time shows... Figure 3 Since the cold wave hit the area on December 16, 2023, low-temperature condensation events have occurred, with a high number of stations (nearly 20) experiencing condensation between 09:00 on the 17th and 09:00 on the 19th. By 14:00 on the 21st, the number of stations experiencing condensation events had decreased to 9, after which the number of stations experiencing condensation began to increase again, indicating the influence of new cold air. During the low-temperature freezing process from December 16th to 22nd, 2023, a total of 185 condensation events occurred at various stations.

[0140] For this process, the power grid observation data shows the following: 1. A total of 378 stations were manually observed during this process, with 186 stations experiencing freezing and 192 stations not experiencing freezing. If only Zhaotong City is considered, out of a total of 298 stations, 178 stations experienced freezing and 120 stations did not. As shown in Table 1, Zhaotong City experienced 107 freezing events under low temperature and high humidity, 2 under high temperature and low humidity, 56 under low temperature and low humidity, and 13 under high temperature and high humidity conditions. Freezing generally does not occur under high temperature and low humidity conditions above 0℃, and only two out of the 36 stations observed experienced this. Under high humidity conditions above 0℃, 13 out of the 30 stations observed experienced freezing. In this case, the proportion of stations not experiencing freezing under low temperature and low humidity conditions below 0℃ was 6 / 46. Although freezing mostly occurs under low temperature and high humidity conditions, 23 / 124 stations did not experience freezing. From the above analysis, it is clear that it is difficult to distinguish the occurrence of freezing based solely on meteorological conditions of temperature and humidity.

[0141] 2. Based on the data from the power grid tension sensors, the icing data (including observations from Zhaotong City at 10:00 on December 19, 2023 and 18:00 on December 21, 2023, twice a day) were examined. The freezing situation is shown in Table 2. Table 3 shows that Zhaotong City experienced 156 freezing events under low temperature and high humidity, 0 freezing events under high temperature and low humidity, 4 freezing events under low temperature and low humidity, and 0 freezing events under high temperature and high humidity. Of the 435 stations observed, no freezing occurred under low humidity conditions above 0℃; 790 stations under high humidity conditions above 0℃ also had no freezing events; in this example, the proportion of freezing events in low temperature and low humidity conditions below 0℃ was 4 / 26, and 878 / 1034 stations in low temperature and high humidity conditions did not experience freezing.

[0142] Table 3. Number of stations with or without freezing / icing detected by tension sensors in Zhaotong power grid from December 16-22, 2023.

[0143]

[0144] First, based on data observed by tensile sensors, at 09:00 on December 21, 2023, 148 out of a total of 473 stations in a certain area (excluding stations with abnormal observations and missing data) experienced low-temperature freezing and icing. The temperature at all freezing stations was below 0℃, with 145 stations having a relative humidity greater than 75%, and only 3 stations having a relative humidity less than 75% during the same period. Figure 4 The freezing of ice at low temperatures and high humidity is relatively easy to understand, but freezing under low humidity conditions requires further analysis of the preceding conditions. Further analysis using artificial ice observation revealed that freezing at low temperatures and low humidity occurs at the end of the process, while freezing occurs at temperatures above 0°C in the early stages. Furthermore, there are instances where freezing does not occur at temperatures below 0°C and humidity above 75%, indicating that freezing indices cannot be determined solely based on temperature and humidity conditions.

[0145] Then, the data from manual ice observation and sensor observation were compiled separately, and stations without latitude and longitude markings were removed. The Euclidean distance between the latitude and longitude of the power grid stations and the grid points of the CLDAS data was calculated, and the nearest point was selected and substituted into the corresponding time-related elements, including temperature, humidity, wind speed, sunshine, precipitation, and air pressure. Furthermore, considering that high-temperature and high-humidity freezing occurs in the early stages of the process, and low-temperature and low-humidity freezing occurs in the later stages, which may be accompanied by changes in these elements, the changes in precipitation, minimum and maximum temperatures, maximum humidity, and relative wind speed, sunshine, and air pressure values ​​for the same period and the previous time were calculated based on hourly CLDAS data. Comparative analysis yielded low-temperature freezing disaster events and their corresponding multi-element databases. In this example, out of 2312 samples, 324 were low-temperature and high-humidity events matching the CLDAS temperature, of which 112 were low-temperature and high-humidity freezing events and 212 were without freezing. Composite analysis was used to diagnose key meteorological elements showing significant differences in the presence or absence of freezing. (See attached instruction manual.) Figure 5 It is evident that there are significant differences in the meteorological elements and their changes under low temperature and high humidity conditions with and without freezing, with most differences except for specific humidity (see instruction manual). Figure 5 a) The significance level was reached at 95% (t0.05 = 1.65). Further analysis of cases with and without freezing during periods of no precipitation revealed significant differences in the following factors: minimum and maximum temperatures, changes in temperature and relative humidity, wind speed, shortwave radiation, and air pressure and its changes (see attached instructions). Figure 5 b).

[0146] Finally, using AI reinforcement learning + density peak classification method and the database obtained in the second step, we extracted different types of multi-element features and combinations corresponding to icing thickness. Through the analysis of different combinations of elements, we obtained the variation range of corresponding meteorological elements, that is, we obtained the correspondence between power grid freezing disasters and combined meteorological conditions, and established meteorological indicators. In this example, we classified 162 low temperature and high humidity events under no precipitation conditions, which can be divided into two categories. One category has all individual samples corresponding to freezing (see instruction manual). Figure 6 (ac), accounting for 51 / 162; most of the individual samples did not freeze (see instruction manual). Figure 6(df), accounting for 111 / 162. Classification experiment: First: Input all the above-mentioned elements corresponding to all samples (including concurrent air temperature, 3-hour air temperature change, highest and lowest air temperature in the previous 24 hours, hourly precipitation, 3-hour precipitation change, 24-hour precipitation, concurrent relative humidity, 3-hour relative humidity change, 24-hour maximum relative humidity, concurrent wind speed, 3-hour wind speed change, 24-hour average wind speed, concurrent shortwave radiation, 3-hour shortwave radiation change, 24-hour shortwave radiation change, concurrent air pressure, 3-hour air pressure change, 24-hour air pressure change, concurrent specific humidity, 3-hour specific humidity change, 24-hour specific humidity change); Second: Input only the elements with significant differences (including concurrent air temperature, highest and lowest air temperature in the previous 24 hours, concurrent relative humidity, 3-hour relative humidity change, 3-hour wind speed change, concurrent shortwave radiation, 3-hour air pressure change, 3-hour specific humidity change), and analyze using AI reinforcement learning + density peak classification method. The results were consistent. This indicates that as long as further analysis is performed on the elements with significant differences, the corresponding thresholds can be obtained. Included in the instruction manual Figure 6 It is evident that the maximum temperature corresponding to freezing is mostly below 0℃, shortwave radiation is greater than 200w / m2, wind speed is less than 1.2 m / s, and specific humidity increases within 3 hours (see instruction manual). Figure 6 (ac); Conversely, the maximum temperature corresponding to freezing is mostly above 0℃, shortwave radiation is less than 200w / m2, wind speed is greater than 0.5 m / s, and the specific humidity decreases within 3 hours (see instruction manual). Figure 6 (df). Based on the above analysis, when there is no precipitation in 24 hours, freezing occurs when shortwave radiation is greater than 200w / m2, corresponding to a maximum temperature ≤0℃, wind speed less than 1.2 m / s, and specific humidity >0g / kg in the past 3 hours; when shortwave radiation is less than 20w / m2, freezing is mostly absent, corresponding to a maximum temperature >0℃ and wind speed between 0.5 and 2.5 m / s, but freezing is more likely to occur when wind speed is less than 0.5 m / s and specific humidity <0g / kg in the past 3 hours.

[0147] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for constructing a weather index of power grid low-temperature freezing disaster, characterized in that, The method comprises the following steps: Based on the grid observation data, the low-temperature freezing disaster events in the target area are obtained, and the disaster types that need to establish weather index are determined by preliminary classification; The relationship between temperature, humidity in high spatiotemporal resolution meteorological data and the grid observation data is evaluated, and a multi-element database corresponding to the low-temperature freezing disaster events is established; The multi-element characteristics and combinations corresponding to different disaster types in the multi-element database are extracted by deep learning algorithm; The change range of meteorological elements corresponding to the power grid disaster is obtained by analyzing the multi-element characteristics and combinations, that is, the corresponding relationship between the power grid freezing disaster and the combined meteorological conditions is obtained, and the weather index is established.

2. The method of claim 1, wherein the method is characterized by: The grid observation data includes sensor observation data and manual observation data; wherein the sensor observation data refers to the data generated by the ice inspection of the observation process through the grid tension sensor; the elements of the manual observation data include ice thickness, air temperature and relative humidity; The high spatiotemporal resolution meteorological data is obtained by CLDAS data, and the CLDAS data refers to the CMA land surface data assimilation system product hourly data.

3. The method of claim 2, wherein the method is characterized by: Based on the grid observation data, the low-temperature freezing disaster events in the target area are obtained, and the disaster types that need to establish weather index are determined by preliminary classification, which comprises, The disaster types are divided according to the temperature and humidity, specifically: When the temperature is less than or equal to the first preset threshold and the humidity is greater than or equal to the second preset threshold, the disaster type is determined as low temperature and high humidity; When the temperature is greater than the first preset threshold and the humidity is less than the second preset threshold, the disaster type is determined as high temperature and low humidity; When the temperature is less than or equal to the first preset threshold and the humidity is less than the second preset threshold, the disaster type is determined as low temperature and low humidity; When the temperature is greater than the first preset threshold and the humidity is greater than or equal to the second preset threshold, the disaster type is determined as high temperature and high humidity; Low temperature and high humidity, high temperature and low humidity, low temperature and low humidity, and high temperature and high humidity are taken as the disaster types that need to establish weather index.

4. The method of claim 3, wherein the method is characterized by: The relationship between temperature, humidity in high spatiotemporal resolution meteorological data and the grid observation data is evaluated, and a multi-element database corresponding to the low-temperature freezing disaster events is established, which comprises the following steps: The correlation coefficient r and the root mean square error between the sensor observation data and the manual observation data and the temperature and humidity in the high spatiotemporal resolution meteorological data are calculated; For the difference test between different disaster types, the synthetic t test method is used for significance test, and the variables reaching significant reliability are screened out for key analysis, that is, |t|≥t α elements; wherein, t α is a significant threshold The root mean square error is used to measure the consistency between the grid observation on continuous data and the multi-element meteorological reanalysis, and the specific formula is as follows: Wherein, n is the sample number, and the smaller the RMSE, the more consistent; The t test method refers to the test of the average value of two types of samples, that is, class 1 sample and class 2 sample, and the difference between the total average and the two, specifically: For the difference t test between class 1 and the total, the following formula is used: wherein s represents the sample mean and standard deviation, respectively; and μ0is the population mean. If class 2 is to be tested, then is replaced by The threshold t is obtained from the statistical query distribution table α If |t| ≥ t α , then there is a significant For the variable difference t test corresponding to class 1 sample and class 2 sample: In the formula, n1, n2 represent the mean values of the elements x, y corresponding to the class 1 sample and the class 2 sample, respectively, and the number of years, and represent the variances of the two elements, respectively; From the statistical query distribution table, a significant threshold t is obtained α If |t|≥t α , it indicates that the difference between the class 1 sample and the class 2 sample is significant.

5. The method of claim 4, wherein the method is characterized by: The PDCDP algorithm increases the path attribute and distribution attribute of the data points on the basis of the CDP algorithm only considering the data point density attribute and distance attribute, and specifically includes the following steps: Establish the minimum distance connected path feature of the data set; Calculate the path attribute and distribution attribute of the data points based on the minimum distance connected path feature set and generate an improved decision graph; Use regression analysis to automatically determine the characteristic sample representative corresponding to the disaster type and preliminarily divide the classification result; Determine the characteristic disaster type sample to which the remaining data points belong by using the fuzzy weighted method and separate out the "transition" class to complete the objective classification.

6. The method of claim 5, wherein the method is characterized by: The establishment of the minimum distance connected path feature of the data set includes the following steps, A standardized multi-meteorological element dataset is established with sites and icing disasters as identifiers wherein {1…n} represents data point serial number, m represents data dimension, x ij represents data value, id i represents the i-th data point unique identification, let ID = {id1, id2…id n} calculates the distance matrix between each data point wherein d ij represents the Euclidean distance between the i, jth data points, find the minimum value d mimj in the DS, and its row number mi and column number mj, generate a new vector ES = [mi, mj, d mimj ], and a new n x n x 4 three-dimensional matrix RFS = NaN representing the path feature vector, let RFS mimj = RFS mjmi = [1, d mimj , d 2 mimj , d mimj ], and a new index set di = {mi, mj}, and do is the complement of di in the index set {1, 2, …, n}. Find the minimum value d in the DS subset DS(di, do) mdimdj , and its row index mdi and column index mdj in DS, generate ns = [mdi, mdj, d mdimdj ] as a new row appended to ES below; then add mdj to the path signature vector of mdi (i.e. update RFS mdimdj = RFS mdjmdi = [1, d mdimdj , d mdimdj 2 , d mdimdj ]) and simultaneously add mdj to the path signature vector of each element in the set odi = di - mdi (i.e. update RFS odi(i)mdj = RFS mdjodi(i) = [RFS mdiodi(i)1 + 1, RFS mdiodi(i)2 + d mdimdj , RFS mdiodi(i)3 + d mdimdj 2 , max(RFS mdiodi(i)4 , d mdimdj )]) and finally remove mdj from the set do and append it to the tail of di; If do is not empty, go to step "find the minimum value d in the DS subset DS(di, do)" mdimdj , otherwise end step "establish the standardized multi-meteorological element dataset set with station and icing disaster as identifier", complete the generation of MDCG[ID, ES] and RFS; The calculation of the path attribute and distribution attribute of the data points based on the minimum distance connected path feature set and the generation of the improved decision graph include the following steps: d c As the truncation distance parameter, the density attribute p i and the distance attribute d i of each data point are calculated according to the DPC algorithm formula, and then The path attribute of each data point is calculated as: wherein denotes the other end point of the edge of length δ i from data point i; The distribution attribute of each data point is calculated as: θ i = ∑(RFS ij2 / d ij / RFS ij1 ); Computing and dgz i = dgx i x dgy i , resulting in the vector dgx, dgy, dgz.

7. The method of claim 6, wherein the method is characterized by: The use of regression analysis to automatically determine the characteristic sample representative corresponding to the disaster type and preliminarily divide the classification result includes the following steps, The vector dgx, dgy, dgz is subjected to regression analysis according to the constant fitting function dgz = C with a confidence interval alpha, C is an automatically calculated constant, to obtain a n x 2 matrix rint, which represents the upper and lower bounds of the residual of the n data points in the alpha confidence interval, and the data points with a lower bound rint(:, 1) > 0 are taken as the characteristic weather type representative, that is, the cluster center cc, and the number of ccn, wherein the preset value alpha = 0.05 represents a 95% confidence interval; Traverse the cluster center cc, the first cluster center cc ci Cut-off distance d c Data points within the cluster are classified into the cluster as the cluster core part, wherein if a certain data point k belongs to cc ci Also belongs to cc cj k is classified according to The first cluster center or the second cluster center, complete the preliminary division of the objective classification of the low-temperature freezing disaster event sample, that is, the division of the cluster core co; The use of the fuzzy weighted method to determine the characteristic disaster type sample to which the remaining data points belong and to separate out the "transition" class to complete the objective classification includes the following steps, Generating a weight matrix and computing a membership matrix P, P is an oln x con matrix, where oln represents the number of unassigned data points and con represents the number of cluster cores where co ci denotes the ci-th cluster core in the cluster core set co, ol oi denotes the oi-th data point in the unassigned data point set ol, Take As Then traverse data points ol oi in the range of d c of data points j and update Where Then make the oi row of P all 0, and remove ol oi from ol and move to co ci , indicating that the unassigned data point ol oi is drawn into the cluster core co ci ; Repeat the previous step until Then all data points in ol are marked as noise.

8. A system for constructing a weather index of power grid low-temperature freezing disaster using the method according to any one of claims 1 to 7, characterized in that, It includes: A data acquisition module for acquiring power grid observation data, including power grid tension sensor observation data and manual observation data, to obtain low-temperature freezing disaster events in the target area; A data analysis module for evaluating the matching relationship between temperature, humidity and power grid observation data in high spatio-temporal resolution meteorological data, calculating the correlation coefficient r and the root mean square error, and establishing a multi-element database corresponding to the power grid low-temperature freezing disaster event; A difference test module for significant difference analysis of meteorological elements between different disaster types, using synthetic t-test method to screen out significant variables; A deep learning module for extracting multi-element features and combinations corresponding to different disaster types from the multi-element database using the PDCDP algorithm combined with deep learning method, and generating an improved decision graph; An index construction module for analyzing the multi-element features and combinations to obtain the change range of the corresponding meteorological elements of the power grid disaster, establishing the corresponding relationship between the power grid freezing disaster and the combined meteorological conditions, and finally forming a scientific weather index. 9.A computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the computer device is configured to perform the method according to any one of claims 1-8 when the computer program is executed by the processor. The processor implements the steps of the weather index construction method of the power grid low-temperature freezing disaster of any one of claims 1 to 7 when the computer program is executed.

10. A computer-readable storage medium having stored thereon a computer program, characterized in that, The computer program is executed by the processor to implement the steps of the weather index construction method of the power grid low-temperature freezing disaster of any one of claims 1 to 7.

Images