A method, device and equipment for power transmission line forest fire monitoring feature optimization

Abstract

This invention relates to the field of power transmission line wildfire monitoring technology, specifically to a method, apparatus, and equipment for feature optimization in power transmission line wildfire monitoring. The method includes: acquiring a set of candidate features and multiple feature selectors; ranking the features by importance using each selector; calculating the difference between selectors by summing the absolute values ​​of the differences in their rankings across different lists, and selecting the selectors with the highest difference to form an optimal subset; finally, voting on the candidate features nominated by the optimal subset, and selecting features receiving more than half the votes to form the final feature set. This application improves the quality and robustness of feature selection by constructing diverse selector combinations, laying the foundation for building a high-precision wildfire monitoring model.

Description

Technical Field

[0001] This invention relates to the field of power transmission line wildfire monitoring technology, specifically to a method, apparatus, and equipment for optimizing the characteristics of power transmission line wildfire monitoring. Background Technology

[0002] Power transmission lines often traverse complex terrain, and wildfires pose a significant threat to their safe operation. To provide early warnings of wildfire risks, existing technologies typically establish risk assessment models based on multi-dimensional monitoring data. This monitoring data constitutes the input features of the model, such as meteorological data (e.g., temperature, humidity, wind speed), topographic data (e.g., slope, aspect), and vegetation data (e.g., type, coverage).

[0003] The selection of the input feature set is crucial when constructing a risk assessment model. Too many features not only increase the model's computational burden and reduce response efficiency, but may also introduce irrelevant or redundant noise, interfering with the model's judgment and leading to overfitting, thus reducing the model's generalization ability and accuracy. Conversely, too few features, or failure to include key influencing factors, may result in underfitting, making it unable to accurately capture the complex patterns of wildfire occurrence.

[0004] To address this issue, existing technologies have proposed ensemble feature selection methods, which simultaneously use multiple different feature selection algorithms and attempt to fuse their results to improve the robustness of the selection. For example, some technical solutions disclose a framework for selecting models by evaluating the differences between different models and then using a voting method for fusion. However, these ensemble methods themselves have drawbacks. If the selected feature selectors are based on similar principles or have similar evaluation perspectives, their evaluation results will tend to be consistent. This "redundant" combination of selectors does not provide effective information supplementation and may instead reinforce the bias of certain single perspectives, resulting in a suboptimal feature set in the final selection. Therefore, existing technologies generally lack a systematic and quantitative method for evaluating and optimizing feature selectors themselves to construct a selector combination with good diversity, leading to a poor quality final feature set. Summary of the Invention

[0005] In view of this, the purpose of the present invention is to provide a method, apparatus and equipment for optimizing the characteristics of wildfire monitoring of power transmission lines, so as to overcome the problems existing in the current prior art.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: Firstly, this application provides a method for optimizing the characteristics of wildfire monitoring on power transmission lines, including: Step S1: Obtain a set of candidate features and n feature selectors; wherein, the set of candidate features includes N wildfire monitoring features and their historical monitoring data, and n is an integer greater than or equal to 2; Step S2: The importance of the N wildfire monitoring features in the candidate feature set is evaluated by the n feature selectors and sorted according to the evaluation results to obtain a list of n sorted features. Step S3: Based on the n sorted feature lists, calculate the difference between any two feature selectors; calculate the sum or average difference between each feature selector and all other feature selectors, and sort them in descending order according to the sum or average difference, selecting the top n / 2 feature selectors as the preferred feature selector subset, where if n / 2 is not an integer, round up; Step S4: Sort the wildfire monitoring features corresponding to each feature selector in the preferred feature selector subset by importance, and extract the top m features by importance in the sorted feature list as candidate features, where m is a preset integer greater than or equal to 5; vote on all candidate features, and select the candidate features that receive more than half the number of feature selectors in the preferred feature selector subset to form the feature set of important wildfire impact factors.

[0007] Furthermore, in the method described above, calculating the difference between any two feature selectors includes: The difference between any two feature selectors is calculated using the difference calculation formula; wherein the difference calculation formula is:

[0008] Among them, ST f|i ST represents the ordinal position of the f-th feature after sorting by the i-th feature selector. f|j This indicates the order of the f-th feature after sorting by the j-th feature selector.

[0009] Furthermore, in the method described above, in step S4, the voting method for all candidate features is weighted voting, wherein each feature selector in the subset of preferred feature selectors is assigned a voting weight, and the final number of votes for a candidate feature is the sum of the weights of all the votes it receives. The candidate features that rank in the top K by weighted total score, or the candidate features whose weighted total score exceeds a preset threshold, are selected to form the feature set of important factors affecting wildfires, where K is a positive integer.

[0010] Furthermore, in the method described above, the voting weight is positively correlated with the magnitude of the difference of the feature selector.

[0011] Furthermore, in the above-described method, the wildfire monitoring features include at least one of meteorological features, topographic features, vegetation features, and line body features; the feature selector includes at least two of the following: correlation coefficient method, entropy weight method, Gini index method, information gain method, chi-square test method, ReliefF algorithm, and LASSO regression algorithm.

[0012] Secondly, this application provides a device for optimizing the characteristics of wildfire monitoring on power transmission lines, comprising: The data acquisition module is used to acquire a set of candidate features and n feature selectors; wherein, the set of candidate features includes N wildfire monitoring features and their historical monitoring data, and n is an integer greater than or equal to 2; The sorting module is used to evaluate the importance of N wildfire monitoring features in the candidate feature set through n feature selectors and sort them according to the evaluation results to obtain a list of n sorted features. The selector optimization module is used to calculate the difference between any two feature selectors based on the n sorted feature lists; calculate the sum or average difference between each feature selector and all other feature selectors, and sort them in descending order according to the sum or average difference, selecting the top n / 2 feature selectors as the preferred feature selector subset, where if n / 2 is not an integer, it is rounded up; The feature set generation module is used to sort the wildfire monitoring features corresponding to each feature selector in the preferred feature selector subset by importance, extract the top m features in the sorted feature list as candidate features, where m is a preset integer greater than or equal to 5; vote on all candidate features, and select the candidate features that receive more than half the number of feature selectors in the preferred feature selector subset to form the feature set of important wildfire impact factors.

[0013] Furthermore, in the apparatus described above, calculating the difference between any two feature selectors includes: The difference between any two feature selectors is calculated using the difference calculation formula; wherein the difference calculation formula is:

[0014] Among them, ST f|i ST represents the ordinal position of the f-th feature after sorting by the i-th feature selector. f|j This indicates the order of the f-th feature after sorting by the j-th feature selector.

[0015] Furthermore, in the apparatus described above, the feature set generation module is further configured to: All candidate features are voted on using a weighted voting method, wherein each feature selector in the subset of preferred feature selectors is assigned a voting weight that is positively correlated with its degree of difference, and the final number of votes for a candidate feature is the sum of the weights of all the votes it receives; The candidate features that rank in the top K by weighted total score, or the candidate features whose weighted total score exceeds a preset threshold, are selected to form the feature set of important factors affecting wildfires, where K is a positive integer.

[0016] Furthermore, in the apparatus described above, the voting weight is positively correlated with the magnitude of the difference in the feature selector.

[0017] Thirdly, this application provides a device for optimizing the characteristics of wildfire monitoring on power transmission lines, including a processor and a memory, wherein the processor is connected to the memory: The processor is used to call and execute the program stored in the memory; The memory is used to store the program, which is used to execute at least the above-described method for optimizing the characteristics of wildfire monitoring on transmission lines.

[0018] The beneficial effects of this invention are as follows: 1. Improve the quality of the feature set and the accuracy of the model: This application constructs a combination of selectors with diverse perspectives and complementary information by calculating and screening the feature selectors with the greatest differences. This avoids redundancy and bias caused by similar selector principles and can screen out features that have a key impact on the occurrence of wildfires with a higher probability, thereby improving the quality of the final feature set and laying the foundation for building a wildfire monitoring model with higher accuracy.

[0019] 2. Enhanced robustness and objectivity of feature selection: The method in this application does not rely on the performance of any single algorithm. By systematically evaluating and screening selectors, the randomness of the selection process is reduced, making the feature selection process more objective and stable.

[0020] 3. Balancing efficiency and effectiveness: This application adopts a two-step strategy of first filtering the selector and then filtering the features, which ensures the filtering effect while controlling the computational complexity, thus achieving an effective balance between accuracy and computational efficiency. Attached Figure Description

[0021] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. 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.

[0022] Figure 1 This is a flowchart of an embodiment of the method for optimizing the monitoring features of wildfires on power transmission lines according to the present invention; Figure 2 This is a schematic diagram of the structure of an embodiment of the wildfire monitoring feature optimization device for power transmission lines according to the present invention; Figure 3 This is a schematic diagram of the structure of an embodiment of a preferred device for monitoring wildfires on power transmission lines according to the present invention. Detailed Implementation

[0023] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be described in detail below. Obviously, the described embodiments are merely some embodiments of this invention, and not all embodiments. Based on the embodiments of this invention, all other implementation methods obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0024] Example 1 Figure 1 This is a flowchart illustrating an embodiment of a method for optimizing the characteristics of wildfire monitoring on power transmission lines according to the present invention. Please refer to [link / reference]. Figure 1 This embodiment may include the following steps: Step S1: Obtain the candidate feature set and n feature selectors; wherein, the candidate feature set contains N wildfire monitoring features and their historical monitoring data, and n is an integer greater than or equal to 2; Step S2: Use n feature selectors to evaluate the importance of N wildfire monitoring features in the feature set to be selected and sort them according to the evaluation results to obtain a list of n sorted features. Step S3: Based on the n sorted feature lists, calculate the difference between any two feature selectors; calculate the sum or average difference between each feature selector and all other feature selectors, and sort them in descending order according to the sum or average difference, selecting the top n / 2 feature selectors as the preferred feature selector subset, where if n / 2 is not an integer, round up; Step S4: Sort the wildfire monitoring features corresponding to each feature selector in the preferred feature selector subset by importance, and extract the top m features in the sorted feature list as candidate features, where m is a preset integer greater than or equal to 5; vote on all candidate features, and select the candidate features that receive more than half the number of feature selectors in the preferred feature selector subset to form the feature set of important wildfire impact factors.

[0025] Preferably, calculating the difference between any two feature selectors includes: The difference between any two feature selectors is calculated using the following formula:

[0026] Among them, ST f|i ST represents the ordinal position of the f-th feature after sorting by the i-th feature selector. f|j This indicates the order of the f-th feature after sorting by the j-th feature selector.

[0027] Preferably, in step S4, the voting method for all candidate features is weighted voting, wherein each feature selector in the subset of preferred feature selectors is assigned a voting weight, and the final number of votes for a candidate feature is the sum of the weights of all the votes it receives. The top K candidate features with weighted total scores, or candidate features with weighted total scores exceeding a preset threshold, will be selected to form a feature set of important factors affecting wildfires, where K is a positive integer.

[0028] Preferably, the voting weight is positively correlated with the magnitude of the difference of the feature selector.

[0029] Preferably, the wildfire monitoring features include at least one of the following: meteorological features, topographic features, vegetation features, and line body features; the feature selector includes at least two of the following: correlation coefficient method, entropy weight method, Gini index method, information gain method, chi-square test method, ReliefF algorithm, and LASSO regression algorithm.

[0030] In practice, preprocessed historical monitoring data is first acquired to form a candidate feature set S. In this embodiment, it is assumed that the candidate feature set S contains N=20 wildfire monitoring features. Specifically, these features can be divided into several categories: 1. Meteorological characteristics, such as daily maximum temperature, daily minimum temperature, average relative humidity, minimum relative humidity, maximum wind speed, average wind speed, prevailing wind direction, and rainfall in the past 24 hours; 2. Topographic features, such as the elevation, slope, and aspect of the location where the power transmission line runs; 3. Vegetation characteristics, such as vegetation coverage, vegetation type, normalized difference vegetation index, thickness of litter layer, and moisture content of combustibles; 4. Line characteristics, such as the width of the line corridor, the distance between the conductor and the ground, and the operating current of the line. These characteristics together constitute an N-dimensional feature space describing the environment in which wildfires occur.

[0031] Simultaneously, a predefined feature selector set {FS} is loaded, which contains n different feature selection algorithms. In this embodiment, n=6, and the six feature selectors are as follows: FS1: Correlation coefficient method, such as Pearson correlation coefficient, which measures the linear correlation between features.

[0032] FS2: Gini index method, commonly used in decision tree models, to measure the contribution of features to reducing sample impurity.

[0033] FS3: Information Gain Method, based on information entropy theory, measures the amount of information brought by the system.

[0034] FS4: Chi-square test, used to test the independence of two categorical variables, often applied to the selection of categorical features.

[0035] FS5: ReliefF algorithm, a feature selection algorithm based on the nearest neighbor concept, capable of handling multi-class problems and sensitive to feature interactions.

[0036] FS6: The LASSO regression algorithm, which achieves feature selection by adding an L1 regularization term to the loss function, which shrinks the coefficients of some features to zero. It should be noted that these six selectors are based on different mathematical principles and evaluation perspectives, providing a good foundation for subsequent differentiated evaluation.

[0037] The feature set S is sorted by feature importance using each feature selector set {ST}. In this step, the candidate feature set S and the feature selector set {FS} are received as input, and the feature set S is input to the six selectors FS1 to FS6 in parallel or serially. Each selector FSi evaluates the importance of the 20 features in S and outputs a sorted list STi from most important to least important. For example, FS1 (correlation coefficient method) calculates the correlation coefficient between each feature and the wildfire occurrence label and sorts them in descending order of their absolute values; while FS2 (Gini index method) sorts the features based on the average reduction in Gini impurity in the random forest model. After this process, the sorting module outputs n=6 sorted feature lists, denoted as {ST1, ST2, ST3, ST4, ST5, ST6}. Each list contains all 20 features, but their order is different.

[0038] From the initial 6 selectors, select the subset with the greatest internal differences and the strongest information complementarity.

[0039] Calculate the difference DIF between any two selectors. The system calculates the difference between any two feature selectors FSi and FSj using the following formula:

[0040] Among them, ST f|iST represents the ordinal position of the f-th feature after sorting by the i-th feature selector. f|j This indicates the order of the f-th feature after sorting by the j-th feature selector.

[0041] The physical meaning of this formula is that it obtains a total difference score by summing the differences in the ranking of all features in the two ranking lists. It can be understood that if the two selectors have highly similar assessments of the importance of all features, the rankings of most features will be very close, resulting in a smaller DIF value; conversely, if the two selectors have very different evaluation perspectives, the rankings of many features will differ significantly, resulting in a larger DIF value.

[0042] The selectors are sorted according to DIF, and the top n / 2 are selected to form FSe. Next, the selector optimization module calculates a total dissimilarity score for each selector. This score can be the sum or average of the dissimilarity scores of the selector with the other (n-1) selectors. After calculating the total dissimilarity scores of all 6 selectors, these 6 selectors are sorted in descending order of score. The selector with the highest score means that its difference from the "mainstream" opinion is the greatest and its evaluation perspective is the most unique. Subsequently, the top n / 2 feature selectors are selected according to preset rules. In this embodiment, n=6, so n / 2=3. Therefore, the top three feature selectors with the highest total dissimilarity scores are selected to form the preferred feature selector subset FSe. Assuming that after calculation and sorting, the three selectors with the highest scores are FS5 (ReliefF algorithm), FS2 (Gini index method), and FS1 (correlation coefficient method), then the preferred subset FSe = {FS1, FS2, FS5}. As an alternative implementation, if the initial number of selectors n is odd (e.g., n=7), then n / 2=3.5, which can be rounded up to 4, meaning the top 4 are selected.

[0043] The system extracts the top m features from the sorted set corresponding to FSe. First, a parameter m is set, representing the number of candidate features nominated by each selector. This value needs to be large enough to include all possible key features, but should not be too large to avoid introducing excessive noise. In this application, m is preset to an integer greater than or equal to 5; in this embodiment, m=5. Then, the system extracts the top 5 features from the sorted lists ST1, ST2, and ST5 corresponding to the preferred subset FSe = {FS1, FS2, FS5}, respectively. Assume the extraction results are as follows: Candidate features from ST1: {Feature A, Feature B, Feature C, Feature D, Feature E} Candidate features from ST2: {Feature A, Feature C, Feature F, Feature G, Feature H} Candidate features from ST5: {Feature B, Feature C, Feature F, Feature I, Feature J} These features are merged and deduplicated to obtain a candidate feature pool of 71: {A, B, C, D, E, F, G, H, I, J}.

[0044] Sub-step S420: Count the votes and select features that receive more than half the votes to form Se. The feature set generation module performs voting statistics on each feature in the candidate feature pool 72. If a feature appears in the top m list of a certain preferred selector, it receives one vote. The statistical results are as follows: Feature A: Nominated by ST1 and ST2, receiving 2 votes.

[0045] Feature B: Nominated by ST1 and ST5, receiving 2 votes.

[0046] Feature C: Nominated by ST1, ST2, and ST5, and received 3 votes.

[0047] Feature D: Nominated by ST1 and received 1 vote.

[0048] Feature E: Nominated by ST1 and received 1 vote.

[0049] Feature F: Nominated by ST2 and ST5, receiving 2 votes.

[0050] Feature G: Nominated by ST2 and received 1 vote.

[0051] Feature H: Nominated by ST2 and received 1 vote.

[0052] Feature I: Nominated by ST5 and received 1 vote.

[0053] Feature J: Nominated by ST5 and received 1 vote.

[0054] Finally, the final selected features are determined according to the screening rules. The rule is: the number of votes received must be more than half the number of feature selectors in the preferred feature selector subset. In this example, the preferred subset FSe has 3 selectors, and half of its number is 1.5. Therefore, any feature that receives more than 1.5 votes (i.e., equal to or greater than 2 votes) will be selected. According to the above statistics, the features that receive 2 or 3 votes are {A, B, C, F}. Correspondingly, the final generated feature set Se(73) of important wildfire impact factors is {feature A, feature B, feature C, feature F}. This concise set of 4 features is the result of the combined recognition of three different evaluation dimensions and perspectives of selectors. Using this feature set Se to train wildfire risk assessment models (such as logistic regression, support vector machine or neural network models) can achieve higher prediction accuracy, stronger generalization ability and faster model inference speed compared to feature sets selected by a single selector or feature sets that simply integrate the results of all 6 selectors.

[0055] Example 2 This embodiment is an improved version of Embodiment 1, providing a variant scheme for adaptively determining the number of preferred selector subsets. In Embodiment 1, the number of preferred subsets is fixed at n / 2, which is a simple and effective rule, but may be somewhat rigid in some cases. For example, when the initial n selectors have a highly concentrated distribution of differences, only the first two selectors may have significant differences, while the third and fourth have similar and low differences. In this case, forcibly selecting n / 2 may introduce redundancy. Conversely, if the difference distribution is relatively flat, more than n / 2 selectors may have high differences, and selecting only n / 2 may result in missing valuable evaluation perspectives. This embodiment aims to solve this problem by dynamically determining the number of preferred subsets in a data-driven manner.

[0056] In this embodiment, after calculating the total difference score and sorting all selectors in descending order, the first n / 2 selectors are not directly chosen. Instead, an "inflection point" detection algorithm is executed to automatically determine the optimal cutoff point. The specific process is as follows: First, calculate the total difference score for each selector and obtain a descending score sequence D = {d1, d2, ..., dn}, where d1 is the highest score and dn is the lowest score.

[0057] Next, the difference between adjacent scores in this score sequence is calculated, i.e., the gradient. The gradient sequence is G = {g1, g2,..., g_{n-1}}, where g_i = d_i - d_{i+1}. This gradient value reflects the rate at which the score decreases.

[0058] Next, we search for the maximum value g_max in the gradient sequence G. The location where this maximum gradient occurs signifies the most drastic drop in score and can be considered a natural dividing point, or "inflection point." We denote the index position corresponding to this maximum gradient g_max as k.

[0059] Finally, the top k feature selectors are selected as the preferred feature selector subset FSe.

[0060] To better understand this technical solution, a specific example is given below: Assume there are initially n=10 feature selectors. After dissimilarity calculation and sorting, the resulting total dissimilarity score sequence D is: [100, 98, 95, 70, 68, 67, 65, 64, 63, 60]. The corresponding gradient sequence G is: g1 = 100 - 98 = 2 g² = 98 - 95 = 3 g3 = 95 - 70 = 25 g4 = 70 - 68 = 2 g5 = 68 - 67 = 1 g6 = 67 - 65 = 2 g7 = 65 - 64 = 1 g8 = 64 - 63 = 1 g9 = 63 - 60 = 3 The gradient sequence G is [2, 3, 25, 2, 1, 2, 1, 1, 3]. The maximum gradient g_max is 25, occurring at the 3rd position (g3), thus the inflection point k=3. This indicates that the score shows the most significant decrease between the 3rd and 4th selectors, suggesting that the first 3 selectors constitute a highly differentiated "first tier." Therefore, the selector optimization module determines the size of the preferred subset to be 3 and selects the top 3 selectors in terms of total difference score to form FSe.

[0061] In contrast, if the fixed ratio method of Example 1 is used, n / 2 = 5 selectors would be selected for n=10. However, according to the adaptive method of this example, only 3 are selected, thereby avoiding the introduction of the 4th and 5th selectors (scores 70 and 68) with lower differences into the preferred subset, making the subset more refined and efficient.

[0062] This embodiment introduces an inflection point detection algorithm to achieve intelligent and automated determination of the optimal number of selectors. This method eliminates the reliance on manually preset fixed ratios (such as n / 2) and can better adapt to the inherent differences in the distribution characteristics of different initial selector sets, making the selection process more objective and reasonable.

[0063] Example 3 The present invention also provides a device for optimizing the characteristics of wildfire monitoring on power transmission lines, for implementing the above-described method embodiments. Figure 2 This is a schematic diagram of the structure of an embodiment of a power transmission line wildfire monitoring feature optimization device according to the present invention. Figure 2 As shown, it includes: Data acquisition module 1 is used to acquire a set of candidate features and n feature selectors; wherein, the set of candidate features contains N wildfire monitoring features and their historical monitoring data, and n is an integer greater than or equal to 2; Sorting module 2 is used to evaluate the importance of N wildfire monitoring features in the feature set to be selected by n feature selectors and sort them according to the evaluation results to obtain a list of n sorted features. Selector optimization module 3 is used to calculate the difference between any two feature selectors based on n sorted feature lists; calculate the sum or average difference between each feature selector and all other feature selectors, sort them in descending order according to the sum or average difference, and select the top n / 2 feature selectors as the preferred feature selector subset, where if n / 2 is not an integer, it is rounded up; The feature set generation module 4 is used to sort the importance of the wildfire monitoring features corresponding to each feature selector in the preferred feature selector subset, extract the top m features in the sorted feature list as candidate features, where m is a preset integer greater than or equal to 5; vote on all candidate features, and select the candidate features that receive more than half the number of feature selectors in the preferred feature selector subset to form the feature set of important wildfire impact factors.

[0064] Preferably, calculating the difference between any two feature selectors includes: The difference between any two feature selectors is calculated using the following formula:

[0065] Among them, ST f|i ST represents the ordinal position of the f-th feature after sorting by the i-th feature selector. f|j This indicates the order of the f-th feature after sorting by the j-th feature selector.

[0066] Preferably, the feature set generation module 4 is further configured as follows: A weighted voting method is used to vote on all candidate features. Each feature selector in the subset of preferred feature selectors is assigned a voting weight that is positively correlated with its degree of difference. The final number of votes for a candidate feature is the sum of the weights of all the votes it receives. The top K candidate features with weighted total scores, or candidate features with weighted total scores exceeding a preset threshold, will be selected to form a feature set of important factors affecting wildfires, where K is a positive integer.

[0067] Preferably, the voting weight is positively correlated with the magnitude of the difference of the feature selector.

[0068] Regarding the apparatus in the above embodiments, the specific manner in which each module performs its operation has been described in detail in the embodiments related to the method, and will not be elaborated upon here.

[0069] Example 4 The present invention also provides a device for optimizing the characteristics of wildfire monitoring on power transmission lines, for implementing the above-described method embodiments. Figure 3This is a schematic diagram of the structure of an embodiment of a power transmission line wildfire monitoring feature optimization device according to the present invention. Figure 3 As shown, the transmission line wildfire monitoring feature optimization device of this embodiment includes a processor 21 and a memory 22, with the processor 21 connected to the memory 22. The processor 21 is used to call and execute the program stored in the memory 22; the memory 22 is used to store the program, which is at least used to execute the transmission line wildfire monitoring feature optimization method in the above embodiments.

[0070] The specific implementation scheme of the transmission line wildfire monitoring feature optimization device provided in this application embodiment can refer to the implementation scheme of the transmission line wildfire monitoring feature optimization method in any of the above embodiments, and will not be repeated here.

[0071] It is understood that the same or similar parts in the above embodiments can be referred to each other, and the contents not described in detail in some embodiments can be referred to the same or similar contents in other embodiments.

[0072] It should be noted that in the description of this invention, the terms "first," "second," etc., are used for descriptive purposes only and should not be construed as indicating or implying relative importance. Furthermore, in the description of this invention, unless otherwise stated, "a plurality of" means at least two.

[0073] Any process or method description in the flowchart or otherwise herein can be understood as representing a module, segment, or portion of code comprising one or more executable instructions for implementing a particular logical function or process, and the scope of the preferred embodiments of the invention includes additional implementations in which functions may be performed not in the order shown or discussed, including substantially simultaneously or in reverse order depending on the functions involved, as will be understood by those skilled in the art to which embodiments of the invention pertain.

[0074] 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.

[0075] Those skilled in the art will understand that all or part of the steps of the methods in the above embodiments can be implemented by a program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, the program includes one or a combination of the steps of the method embodiments.

[0076] Furthermore, the functional units in the various embodiments of the present invention can be integrated into a processing module, or each unit can exist physically separately, or two or more units can be integrated into a module. The integrated module can be implemented in hardware or as a software functional module. If the integrated module is implemented as a software functional module and sold or used as an independent product, it can also be stored in a computer-readable storage medium.

[0077] The storage media mentioned above can be read-only memory, disk, or optical disk, etc.

[0078] In the description of this specification, references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0079] Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present invention.

Claims

1. A method for optimizing the monitoring features of wildfires along power transmission lines, characterized in that, include: Step S1: Obtain a set of candidate features and n feature selectors; wherein, the set of candidate features includes N wildfire monitoring features and their historical monitoring data, and n is an integer greater than or equal to 2; Step S2: The importance of the N wildfire monitoring features in the candidate feature set is evaluated by the n feature selectors and sorted according to the evaluation results to obtain a list of n sorted features. Step S3: Based on the n sorted feature lists, calculate the difference between any two feature selectors; calculate the sum or average difference between each feature selector and all other feature selectors, and sort them in descending order according to the sum or average difference, selecting the top n / 2 feature selectors as the preferred feature selector subset, where if n / 2 is not an integer, round up; Step S4: Sort the wildfire monitoring features corresponding to each feature selector in the preferred feature selector subset by importance, and extract the top m features by importance in the sorted feature list as candidate features, where m is a preset integer greater than or equal to 5; vote on all candidate features, and select the candidate features that receive more than half the number of feature selectors in the preferred feature selector subset to form the feature set of important wildfire impact factors.

2. The method according to claim 1, characterized in that, The calculation of the difference between any two feature selectors includes: The difference between any two feature selectors is calculated using the difference calculation formula; wherein the difference calculation formula is: Among them, ST f|i ST represents the ordinal position of the f-th feature after sorting by the i-th feature selector. f|j This indicates the order of the f-th feature after sorting by the j-th feature selector.

3. The method according to claim 1, characterized in that, In step S4, the voting method for all candidate features is weighted voting, wherein each feature selector in the subset of preferred feature selectors is assigned a voting weight, and the final number of votes for a candidate feature is the sum of the weights of all the votes it receives. The candidate features that rank in the top K by weighted total score, or the candidate features whose weighted total score exceeds a preset threshold, are selected to form the feature set of important factors affecting wildfires, where K is a positive integer.

4. The method according to claim 3, characterized in that, The voting weight is positively correlated with the degree of difference of the feature selector.

5. The method according to claim 1, characterized in that, The wildfire monitoring features include at least one of the following: meteorological features, topographic features, vegetation features, and line body features; the feature selector includes at least two of the following: correlation coefficient method, entropy weight method, Gini index method, information gain method, chi-square test method, ReliefF algorithm, and LASSO regression algorithm.

6. A device for optimizing the characteristics of wildfire monitoring on power transmission lines, characterized in that, include: The data acquisition module is used to acquire a set of candidate features and n feature selectors; wherein, the set of candidate features includes N wildfire monitoring features and their historical monitoring data, and n is an integer greater than or equal to 2; The sorting module is used to evaluate the importance of N wildfire monitoring features in the candidate feature set through n feature selectors and sort them according to the evaluation results to obtain a list of n sorted features. The selector optimization module is used to calculate the difference between any two feature selectors based on the n sorted feature lists; calculate the sum or average difference between each feature selector and all other feature selectors, and sort them in descending order according to the sum or average difference, selecting the top n / 2 feature selectors as the preferred feature selector subset, where if n / 2 is not an integer, it is rounded up; The feature set generation module is used to sort the wildfire monitoring features corresponding to each feature selector in the preferred feature selector subset by importance, extract the top m features in the sorted feature list as candidate features, where m is a preset integer greater than or equal to 5; vote on all candidate features, and select the candidate features that receive more than half the number of feature selectors in the preferred feature selector subset to form the feature set of important wildfire impact factors.

7. The apparatus according to claim 6, characterized in that, The calculation of the difference between any two feature selectors includes: The difference between any two feature selectors is calculated using the difference calculation formula; wherein the difference calculation formula is: Among them, ST f|i ST represents the ordinal position of the f-th feature after sorting by the i-th feature selector. f|j This indicates the order of the f-th feature after sorting by the j-th feature selector.

8. The apparatus according to claim 6, characterized in that, The feature set generation module is further configured to: All candidate features are voted on using a weighted voting method, wherein each feature selector in the subset of preferred feature selectors is assigned a voting weight that is positively correlated with its degree of difference, and the final number of votes for a candidate feature is the sum of the weights of all the votes it receives; The candidate features that rank in the top K by weighted total score, or the candidate features whose weighted total score exceeds a preset threshold, are selected to form the feature set of important factors affecting wildfires, where K is a positive integer.

9. The apparatus according to claim 8, characterized in that, The voting weight is positively correlated with the degree of difference of the feature selector.

10. A device for optimizing the characteristics of wildfire monitoring on power transmission lines, characterized in that, It includes a processor and a memory, wherein the processor is connected to the memory: The processor is used to call and execute the program stored in the memory; The memory is used to store the program, which is at least used to execute the method for optimizing the characteristics of wildfire monitoring of transmission lines as described in any one of claims 1-5.

Images