Power transmission line geological disaster risk assessment method based on prototype

Abstract

The invention belongs to the technical field of artificial intelligence, and discloses a prototype-based power transmission line geological disaster risk assessment method, which comprises the steps of improving geological disaster risk assessment precision and robustness, providing a power transmission line geological disaster risk assessment method, and obtaining a geological disaster positive sample prototype based on a frequency acquisition method; according to different data types, geographical environment factors are divided into discrete environment factors and continuous environment factors; calculating the credibility of the positive sample and the negative sample by using the positive sample prototype; constructing a positive sample set and a negative sample set with credibility by utilizing the credibility of the positive samples and the negative samples; the positive sample set and the negative sample set with credibility are utilized to construct a geological disaster risk assessment model, so that the geological disaster risk assessment precision is improved, and the line safety is guaranteed.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of artificial intelligence, and particularly relates to a power transmission line geological disaster risk assessment method based on a prototype. BACKGROUND

[0002] Geological disasters are one of the important challenges faced by regional power systems, especially in low-temperature and high-humidity weather conditions, power lines, towers and other power equipment are often affected by icing. Geological disasters refer to the phenomenon of rock-soil sliding in mountainous areas when encountering extreme weather processes. With the occurrence of geological disasters, the external force borne by the lines and equipment increases dramatically, and in severe cases, it may lead to accidents such as power line breakage, tower collapse, equipment failure, etc., which seriously threatens the stable operation and reliable power supply of the power system. In order to effectively respond to this risk, timely and accurate monitoring and assessment of the risk of geological disasters and their potential impact on power equipment have become an urgent demand of the power industry.

[0003] The traditional geological disaster risk assessment method mainly evaluates the geological disaster risk degree to obtain the spatial distribution information of the geological disaster risk degree. By evaluating the geological disaster risk degree, the possibility of geological disaster occurrence is predicted, which is the basis and premise of geological disaster emergency and management work. Geological disaster risk degree evaluation is a method of using geographic information system technology to predict the possibility of geological disaster occurrence in a certain area. One of the theoretical foundations of the geological disaster risk degree evaluation method is the "geological disaster risk degree-environmental factor relationship" model, which assumes that there is a coordinated change relationship between the geological disaster risk degree and the environmental variables. Once this relationship is obtained, the geological disaster risk degree of the study area can be predicted.

[0004] According to the basic idea of the theory, the core content of the geological disaster risk assessment is to obtain the relationship between the geological disaster risk and the environmental factors. The data sources for obtaining the relationship between the geological disaster and the environmental variables mainly include two categories: expert knowledge and training samples. Expert knowledge is usually provided by local experts in the field of geological disasters. Expert knowledge reflects the experience and cognition of geological disaster experts on the relationship between geological disaster risk and environmental factors, which can be obtained by knowledge engineers through interviews with experts in the field of geological disasters. The understanding of the relationship between geological disasters and environmental variables by experts in the field of geological disasters is usually from the perspective of mechanism and personal experience. Therefore, expert knowledge has strong mechanism and interpretability. However, the quality of expert knowledge is crucial for the extraction and expression of the relationship between geological disaster risk and environmental factors. However, the subjectivity of expert knowledge is one of the important problems faced by this method. For the relationship between geological disaster risk and environmental factors, different experts in the field of geological disasters have different expert knowledge, and the universality of the obtained relationship between geological disaster risk and environmental factors is difficult to guarantee. Moreover, many areas do not have experts in the field of geological disasters, making it difficult to implement this method. Therefore, the acquisition of expert knowledge as a data source for obtaining the relationship between geological disaster risk and environmental factors is limited. Training samples are another data source for obtaining the relationship between geological disaster risk and environmental factors. The basic idea is to use machine learning models to obtain the relationship from training samples and environmental factor data. Machine learning models are widely used and are currently a research hotspot for obtaining the relationship between geological disaster risk and environmental factors, as they overcome the subjectivity of expert knowledge and are easy to model. Training samples are composed of geological disasters (positive samples) and non-geological disasters (negative samples). Positive samples refer to points (pixels) that have experienced geological disasters, which are usually collected from areas where geological disasters have occurred based on historical geological disaster inventory data. However, the collected positive samples have different representativeness, and each positive sample has different representative ability for the relationship between geological disasters and environmental conditions. Therefore, the reliability of the obtained positive samples is not the same. Negative samples refer to points (pixels) that have not experienced geological disasters. Negative samples cannot be directly obtained and are usually collected from areas where geological disasters have not occurred. It is possible that false negative samples are included in the collected negative samples, which have similar environmental conditions to positive samples but have not experienced geological disasters due to the lack of triggering factors. However, these negative samples are potential positive samples. Therefore, the reliability of negative samples is also difficult to guarantee. Existing research has shown that the quality of positive and negative sample collection has a significant impact on the reliability of machine learning modeling, which in turn affects the accuracy of geological disaster risk assessment. Therefore, how to measure the reliability of positive and negative samples to improve the quality of sample collection and improve the reliability of machine learning modeling is a key problem in the field of geological disaster risk assessment.

[0005] In summary, the research on the reliability measurement method of geological disaster training sample is an important research direction and research content in the field of geological disaster risk assessment. Solving the problem of geological disaster sample reliability measurement helps to perfect the existing geological disaster risk assessment theory and method system, has important scientific significance, and improves the accuracy of geological disaster risk assessment. It also has practical application value in geological disaster prevention, and provides more accurate and real-time technical support for the safe operation of power systems. SUMMARY

[0006] In view of the deficiencies of the prior art, the present application provides a prototype-based power transmission line geological disaster risk assessment method to solve the technical problems existing in the prior art.

[0007] The above technical purpose of the present application is realized by the following technical scheme:

[0008] A prototype-based power transmission line geological disaster risk assessment method, comprising the following steps:

[0009] S1, acquiring geological disaster positive sample prototypes based on a frequency acquisition method;

[0010] S2, according to the different data types, the geographical environment factors are divided into discrete environment factors and continuous environment factors; the reliability of positive samples and negative samples is calculated by using the positive sample prototypes;

[0011] S3, using the reliability of positive samples and negative samples, a positive sample set and a negative sample set with reliability are constructed;

[0012] S4, using the positive sample set and the negative sample set with reliability, a geological disaster risk assessment model is constructed.

[0013] Preferably, the geological disaster positive sample prototypes are acquired, comprising the following steps:

[0014] S1.1, acquiring geological disaster positive sample prototypes based on a frequency acquisition method, the key assumption of the frequency curve acquisition method is that if the point with the maximum frequency in each environment factor can best represent the typical position of the geological disaster polygon in the environment factor combination, then the point with high reliability can be acquired by using the frequency acquisition method;

[0015] The basic flow of the geological disaster positive sample prototype extraction and prototype library construction is based on the main steps of the frequency curve acquisition as follows:

[0016] S1.2: constructing a frequency histogram for each environment factor in the environment factor combination possessed by the geological disaster polygon;

[0017] S1.3: Since the value range, distribution form and type of each environmental factor of geological disasters are inconsistent, a proper histogram bin width should be established for each environmental factor according to its characteristics, and the establishment of the bin width uses the following empirical formula

[0018]

[0019] wherein, bin e is the histogram bin width of the environmental factor e, n and IQR e are the number of points and the interquartile range of the environmental factor e, respectively;

[0020] S1.4: The pixel points with the value of the environmental factor located in the maximum frequency interval are considered as the positive sample prototypes of the environmental factor. If the number of points of an environmental factor in two or more bin widths is consistent and is the maximum, then these points can be selected as the positive sample prototypes;

[0021] S1.5: The positive sample prototypes of geological disasters are constructed by summarizing the positive sample prototypes of each environmental factor. Due to the influence of geographical environmental characteristics and the contingency of geological disasters, geological disasters exhibit different geometric morphological characteristics in the geographical feature space. For different geometric morphological polygons of geological disasters, the above method can obtain the positive sample prototypes of each geological disaster, and finally construct the positive sample prototype library of geological disasters.

[0022] Preferably, according to the different types of data, the geographical environmental factors are divided into discrete environmental factors, continuous environmental factors, and the credibility of calculating positive samples and negative samples.

[0023] The selection method of the environmental factors mainly selects the geographical environmental characteristics and the existing research results. In order to facilitate the calculation of the geographical environmental similarity, the selected geographical environmental factors are divided into discrete environmental factors and continuous environmental factors according to the different types of data.

[0024] Preferably, the frequency ratio method is used to calculate the functional relationship between the frequency of the discrete environmental factor and the positive sample prototype of geological disasters, and the calculation method is as follows:

[0025]

[0026] wherein, p i,j is the frequency of the positive sample prototype in the category j of the environmental factor i; A i,j is the area in the category j of the environmental factor i; m is the number of categories of the environmental factor i; S i,j ' is the frequency of the positive sample prototype in the category j of the environmental factor i; A represents the total area of the study area;

[0027] to Si,j ' Normalization yields the similarity between category j in environmental factor i and the prototype of the positive sample in environmental factor i, as shown in the following formula:

[0028]

[0029] Among them, S i,j S represents the similarity between category j of environmental factor i and the typical value of the positive sample prototype in environmental factor i. i,j The range of its value is [0, 1].

[0030] Ideally, for kernel density estimation methods, it is assumed that there are n positive sample prototype points, whose environmental factor x has values ​​x1, x2, x3, ..., x n The kernel density estimate for environmental factor x is then given by the formula:

[0031]

[0032] Where f(x) is the probability density function relating environmental factor x to the frequency of positive prototypes; n is the number of positive prototype occurrences; k(·) is the kernel function; h is the bandwidth, the size of which constrains the smoothness and shape of the kernel density function; xx i Let x be the environmental factor and x be the environmental factor value of the positive sample prototype. i The difference between them;

[0033] The kernel density curve is inferred using the Gaussian kernel function, and the bandwidth h is calculated using the "rule of thumb," as shown in the following formula:

[0034]

[0035] Where σ is the standard deviation of the environmental factor x values ​​of the n positive sample prototypes.

[0036] Ideally, normalizing the estimated probability density function can yield the similarity between a single environmental factor and its positive sample prototype. The normalization is shown in the formula:

[0037]

[0038] Where f(x) is the probability density function between environmental factor x and the positive sample prototype frequency; f max (x) is the maximum value of f(x); S x This represents the similarity between environmental factor x and its positive sample prototype.

[0039] The similarity of the single environmental factor and the positive sample prototype of the factor is integrated to deduce the similarity of the point and the positive sample prototype of the comprehensive environmental factor, and the calculation formula is as follows:

[0040] S = f (s1, s2, …, s k ,…,s n )

[0041] Wherein, s k is the similarity of the kth environmental factor and the positive sample prototype of the factor, and S is the similarity of the comprehensive environmental factor, and the value range is [0, 1]; f represents a comprehensive method function;

[0042] The following formula is used to calculate the positive sample reliability and the negative sample reliability, the value range of the positive sample reliability is [0, 1], and the greater the value, the higher the positive sample reliability, and similarly, the value range of the negative sample reliability is [0, 1], and the greater the value, the higher the negative sample reliability.

[0043] Reliability i,j = S i,j

[0044] Reliability' i,j = 1-S i,j

[0045] Wherein, S i,j is the value of the geographical environment similarity of the (i, j) point and the positive sample prototype; Reliability i,j is the value of the positive sample reliability of the (i, j) point; Reliability' i,j is the value of the negative sample reliability of the (i, j) point.

[0046] Preferably, the method for constructing the positive sample set and the negative sample set with reliability by using the positive sample reliability and the negative sample reliability comprises:

[0047] According to the method for setting the sample reliability threshold, the higher the overall reliability of the sample set constructed by the more dissimilar positive samples and negative samples, the threshold is set to collect the positive samples and the negative samples, and the positive sample set and the negative sample set with high sample reliability are constructed.

[0048] By comprehensively considering the positive sample reliability, negative sample reliability and overall sample reliability, and in order to solve the problem that the too large collection interval cannot reflect the difference between the geological disaster samples under different reliability thresholds, and the problem that the too small collection interval causes redundant results and hinders the inductive analysis, a mutual exclusion reliability threshold setting method sets the reliability threshold of the sample collection interval and the collection interval, so that the positive sample set and the negative sample set are separated in the environmental feature space, and then a geological disaster training sample set with high reliability can be obtained, thereby improving the quality of the geological disaster risk evaluation;

[0049] The specific steps of the mutual exclusion reliability threshold setting method are as follows: the reliability value range collection interval of the positive sample is set to 0.1, and the reliability thresholds of the positive sample are >0.5 (the reliability threshold collection interval is 0.5 to 1), >0.6 (the reliability threshold collection interval is 0.6 to 1), >0.7 (the reliability threshold collection interval is 0.7 to 1), >0.8 (the reliability threshold collection interval is 0.8 to 1), and >0.9 (the reliability threshold collection interval is 0.9 to 1).

[0050] Similarly, the reliability value range of the negative sample is also set to >0.5 (the reliability threshold collection interval is 0.5 to 1), >0.6 (the reliability threshold collection interval is 0.6 to 1), >0.7 (the reliability threshold collection interval is 0.7 to 1), >0.8 (the reliability threshold collection interval is 0.8 to 1), and >0.9 (the reliability threshold collection interval is 0.9 to 1).

[0051] Preferably, the specific steps of the geological disaster sample collection method are as follows:

[0052] After calculating the reliability of the geological disaster positive sample and the negative sample, the geological disaster positive sample reliability and the negative sample reliability spatial distribution diagram can be obtained.

[0053] According to the threshold of the geological disaster positive sample reliability and the negative sample reliability, the collectable areas of the positive sample and the negative sample under different reliability value ranges are obtained respectively.

[0054] The same number of positive samples and negative samples are randomly collected in the respective collection areas to form the positive sample set and the negative sample set under different reliability value ranges, and then a geological disaster training sample set with reliability is formed.

[0055] Preferably, the method for constructing a geological disaster risk assessment model using the positive sample set and the negative sample set with reliability comprises the following steps:

[0056] The geological disaster risk evaluation based on sample reliability is to use a machine learning model to determine a geological disaster risk-environment factor relationship from a training sample set with reliability and environment factor data, and express the relationship in a linear or nonlinear form to establish a prediction model for evaluating the geological disaster risk.

[0057] The most prominent feature of the geological disaster risk evaluation based on sample reliability is that the training samples can be selected by setting the threshold values of the positive sample and negative sample reliability to achieve the goal of improving the accuracy of the geological disaster risk evaluation.

[0058] The machine learning model is another important module of the geological disaster risk evaluation based on sample reliability, and a machine learning model commonly used in the field of geological disaster risk mapping is selected to establish a prediction model.

[0059] In summary, the present application mainly has the following advantages:

[0060] The present application proposes a method of extracting prototypes using frequency collection based on geological disaster polygons and geographic environment factor data, designs a way of expressing prototypes with environment factor value vectors, and constructs a method of managing a prototype library based on environment factor membership curves, thereby providing a data basis for measuring the reliability of positive samples and negative samples.

[0061] The present application proposes a method of measuring the reliability of positive samples based on the similarity of the geographic environment of the prototype, and a method of measuring the reliability of negative samples based on the dissimilarity of the positive sample prototype, realizes intuitive and quantitative understanding of the reliability of positive samples and negative samples, proposes a method of setting mutually exclusive reliability threshold values, and designs a method of collecting positive samples and negative samples based on reliability, thereby providing data for the geological disaster risk evaluation method based on sample reliability.

[0062] The present application proposes a geological disaster risk evaluation method based on sample reliability, and conducts case studies in two research areas with different areas and geographic environment characteristics. The results show that the method can improve the quality of geological disaster sample collection by setting the reliability threshold value, improve the effect of geological disaster risk evaluation, and has better effectiveness and stability than existing methods, thereby perfecting the existing geological disaster risk evaluation method system. BRIEF DESCRIPTION OF DRAWINGS

[0063] Fig. 1 is a flowchart of the method of the present application;

[0064] Fig. 2 is a flowchart of the prototype extraction and mapping method of the present application;

[0065] Fig. 3 is a flowchart of the training sample collection method of the present application. DETAILED DESCRIPTION

[0066] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments in the present application, all the other embodiments obtained by a person of ordinary skill in the art without creative effort belong to the scope of the present application.

[0067] Reference Figs. 1-3 A prototype-based power transmission line geological disaster risk assessment method includes:

[0068] The method includes the following steps:

[0069] S1, acquiring a geological disaster positive sample prototype based on a frequency collection method;

[0070] S2, according to the different data types, the geographical environment factors are divided into discrete environment factors and continuous environment factors; the credibility of the positive samples and the negative samples is calculated by using the positive sample prototype;

[0071] S3, using the credibility of the positive samples and the negative samples, a positive sample set and a negative sample set with credibility are constructed;

[0072] S4, using the positive sample set and the negative sample set with credibility, a geological disaster risk assessment model is constructed.

[0073] The acquisition of the geological disaster positive sample prototype includes the following steps:

[0074] S1.1, a geological disaster positive sample prototype is acquired by using a frequency collection method, and the key assumption of the frequency curve collection method is that if the point with the maximum frequency in each environment factor can best represent the typical position of the geological disaster polygon in the environment factor combination, then the point with high credibility can be acquired by using the frequency collection method;

[0075] The basic flow of the geological disaster positive sample prototype extraction and the prototype library construction is based on the main steps of the frequency curve collection as follows:

[0076] S1.2: a frequency histogram is constructed for each environment factor in the environment factor combination possessed by the geological disaster polygon;

[0077] S1.3: since the value range, distribution form and type of each environment factor of the geological disaster are inconsistent, a proper histogram interval should be established for each environment factor according to the characteristics of each environment factor, and the establishment of the interval uses the following empirical formula

[0078]

[0079] bine is the histogram bin width of the environmental factor e, n and IQR e are the number of points and the interquartile range of the environmental factor e, respectively;

[0080] S1.4: The pixel points with the environmental factor value in the maximum frequency interval are considered as positive sample prototypes of the environmental factor. If the number of points of an environmental factor in two or more bins is consistent and is the maximum, these points can be selected as positive sample prototypes;

[0081] S1.5: The positive sample prototypes of each environmental factor are summarized to construct a geological disaster positive sample prototype library. Due to the influence of geographical environmental characteristics and the randomness of geological disasters, geological disasters exhibit different geometric morphological characteristics in geographical feature space. For different geometric morphological polygons of geological disasters, the above method can obtain the positive sample prototypes of each geological disaster, and finally construct a geological disaster positive sample prototype library. According to the different types of data, the geographical environmental factors are divided into discrete environmental factors and continuous environmental factors. The credibility of calculating positive samples and negative samples;

[0082] The selection method of environmental factors mainly depends on geographical environmental characteristics and existing research results. In order to facilitate the calculation of geographical environmental similarity, we divide the selected geographical environmental factors into discrete environmental factors and continuous environmental factors according to the different types of data.

[0083] The frequency ratio method is used to calculate the functional relationship between the occurrence frequency of discrete environmental factors and geological disaster positive sample prototypes. The calculation method is as follows:

[0084]

[0085] Among them, p i,j is the frequency of the occurrence of positive sample prototypes in class j of environmental factor i; A i,j is the area of class j of environmental factor i; m is the number of classes of environmental factor i; S i,j ' is the frequency of the occurrence of positive sample prototypes in class j of environmental factor i; A represents the total area of the study area;

[0086] The normalization of S i,j ' can obtain the similarity between class j of environmental factor i and the occurrence of positive sample prototypes in environmental factor i, and the formula is as follows:

[0087]

[0088] Among them, S i,j is the similarity between class j of environmental factor i and the typical value of the occurrence of positive sample prototypes in environmental factor i. Si,j The value range of S is [0, 1]; for the kernel density estimation method, assuming that there are n positive sample prototype points, the values of the environmental factors x of which are x1, x2, x3, …, xn, then the kernel density estimation of the environmental factor x is shown in the following formula: n

[0089]

[0090] Wherein, f(x) is the probability density function between the environmental factor x and the positive sample prototype frequency; n is the number of positive sample prototype occurrences, k(·) is the kernel function; h is the bandwidth, and the size of the bandwidth will restrict the smoothing and shape of the kernel density function; x-x i is the difference between the environmental factor x and the value x i of the environmental factor of the positive sample prototype;

[0091] The Gaussian kernel function is used to infer the kernel density curve, and the bandwidth h is calculated by using the empirical rule “rule of thumb”, and the formula is as follows:

[0092]

[0093] Wherein, σ is the standard deviation of the values of the environmental factor x of the n positive sample prototypes.

[0094] The estimated probability density function is normalized, and the similarity between a single environmental factor and the positive sample prototype of the factor can be obtained, and the normalization is shown in the following formula:

[0095]

[0096] Wherein, f(x) is the probability density function between the environmental factor x and the positive sample prototype frequency; f max (x) is the maximum value of f(x); S x is the similarity of the environmental factor x and the positive sample prototype of the factor;

[0097] After synthesizing the similarity of a single environmental factor and the positive sample prototype of the factor, the comprehensive environmental factor similarity of the point and the positive sample prototype can be inferred, and the calculation formula is as follows:

[0098] S = f(s1, s2, …, s k ,…,s n )

[0099] Wherein, s k is the similarity of the kth environmental factor and the positive sample prototype of the factor, and S is the comprehensive environmental factor similarity, and the value range is [0, 1]; f represents the comprehensive method function;

[0100] ​The reliability of the positive sample and the reliability of the negative sample are calculated by the following formula, the value range of the reliability of the positive sample is [0, 1], and the greater the value, the higher the reliability of the positive sample; similarly, the value range of the reliability of the negative sample is [0, 1], and the greater the value, the higher the reliability of the negative sample.

[0101] Reliability i,j =S i,j

[0102] Reliability' i,j =1-S i,j

[0103] wherein, S i,j is the value of the geographical environment similarity of the (i, j) point to the prototype of the positive sample; Reliability i,j is the value of the reliability of the (i, j) point as a positive sample; Reliability' i,j is the value of the reliability of the (i, j) point as a negative sample.

[0104] The method for constructing the positive sample set and the negative sample set with reliability by using the reliability of the positive sample and the reliability of the negative sample comprises the following steps:

[0105] According to the higher overall reliability of the sample set constructed by the more dissimilar positive samples and negative samples, a method for setting a sample reliability threshold is proposed, so as to set the threshold to collect positive samples and negative samples, and to construct a positive sample set and a negative sample set with high sample reliability.

[0106] By comprehensively considering the reliability of the positive sample, the reliability of the negative sample and the overall sample reliability, and in order to solve the problems that the difference between the geological disaster samples under different reliability thresholds cannot be reflected due to too large collection interval, and the result is redundant and the analysis is hindered due to too small collection interval, a mutual exclusion reliability threshold setting method sets the reliability threshold for the collection interval and the collection interval of the sample reliability, so that the positive sample set and the negative sample set are separated in the environmental feature space, and then a geological disaster training sample set with high reliability can be obtained, thereby improving the quality of the geological disaster risk evaluation;

[0107] The specific steps of the mutual exclusion reliability threshold setting method are as follows: the reliability value range collection interval of the positive sample is set to 0.1, and the reliability threshold of the positive sample is >0.5 (the reliability threshold collection interval is 0.5 to 1), >0.6 (the reliability threshold collection interval is 0.6 to 1), >0.7 (the reliability threshold collection interval is 0.7 to 1), >0.8 (the reliability threshold collection interval is 0.8 to 1), and >0.9 (the reliability threshold collection interval is 0.9 to 1).

[0108] Similarly, the negative sample credibility value range is also set as: >0.5 (credibility threshold collection interval is 0.5 to 1), >0.6 (credibility threshold collection interval is 0.6 to 1), >0.7 (credibility threshold collection interval is 0.7 to 1), >0.8 (credibility threshold collection interval is 0.8 to 1), >0.9 (credibility threshold collection interval is 0.9 to 1).

[0109] The specific steps of the geological disaster sample collection method are as follows:

[0110] After calculating the credibility of the positive and negative samples of the geological disaster, the positive and negative sample credibility spatial distribution maps of the geological disaster can be obtained.

[0111] According to the threshold values of the positive and negative sample credibility of the geological disaster, the collectable areas of the positive and negative samples under different credibility value ranges are obtained respectively.

[0112] The same number of positive and negative samples are randomly collected in the respective collection areas to form the positive and negative sample sets under different credibility value ranges, and then the geological disaster training sample set with credibility is formed.

[0113] The method for constructing a geological disaster risk assessment model using the positive and negative sample sets with credibility comprises the following steps:

[0114] The geological disaster hazard degree evaluation based on sample credibility uses a machine learning model to determine the relationship between the geological disaster hazard degree and environmental factors from the training sample set with credibility and environmental factor data, and expresses this relationship in a linear or nonlinear form to establish a prediction model for evaluating the geological disaster hazard degree.

[0115] The most significant feature of the geological disaster hazard degree evaluation based on sample credibility is that the training samples can be selected by setting the threshold values of the positive and negative sample credibility to achieve the goal of improving the accuracy of the geological disaster hazard degree evaluation.

[0116] The machine learning model is another important module of the geological disaster hazard degree evaluation based on sample credibility, and a commonly used machine learning model in the field of geological disaster hazard degree mapping is selected to establish the prediction model.

[0117] Although embodiments of the present application have been shown and described, it will be understood by those of ordinary skill in the art that various changes, modifications, substitutions and alterations can be made thereto without departing from the principles and spirit of the present application, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A prototype-based method for assessing geological hazard risks of power transmission lines, characterized in that, Includes the following steps: S1. Obtain positive sample prototypes of geological disasters based on frequency acquisition methods; S2. Based on the different data types, geographical environmental factors are divided into discrete environmental factors and continuous environmental factors; using positive sample prototypes, the credibility of positive and negative samples is calculated. S3. Construct a set of positive and negative samples with credibility using the credibility of positive and negative samples; S4. Construct a geological hazard risk assessment model using a set of credible positive and negative samples.

2. The method for geological hazard risk assessment of transmission lines based on a prototype, as described in claim 1, is characterized in that... The process of obtaining positive prototype samples of geological hazards includes the following steps: S1.

1. The frequency-based acquisition method is used to obtain positive sample prototypes of geological hazards. The key assumption of the frequency curve acquisition method is that if the point with the highest frequency in each environmental factor combination can best represent the typical location of the geological hazard polygon in the environmental factor combination, then the frequency-based acquisition method can be used to obtain points with high reliability. The basic process for extracting positive samples of geological hazards and constructing the prototype library, based on frequency curve acquisition, includes the following main steps: S1.2: Construct a frequency histogram for each environmental factor in the combination of environmental factors possessed by the geological hazard polygon; S1.3: Given that the value range, distribution form, and type of various environmental factors of geological hazards are inconsistent, appropriate histogram group intervals should be established for each environmental factor according to its characteristics. The group intervals are established using the following empirical formula. Among them, bin e The histogram group interval of environmental factor e, n, and IQR e These are the number of sites and the quartile difference of environmental factor e, respectively. S1.4: Pixels whose environmental factor values ​​are in the maximum frequency range are considered as positive sample prototypes of the environmental factor. If the number of points of a certain environmental factor is the same in two or more class intervals and they are all the maximum points, then these points can be selected as positive sample prototypes. S1.5: Construct a positive sample prototype library for geological hazards by summarizing the positive sample prototypes of each environmental factor. Due to the influence of geographical environmental characteristics and the randomness of geological hazard occurrence, geological hazards exhibit different geometric morphological characteristics in geographical feature space. For geological hazard polygons with different geometric shapes, the above method can be used to obtain the positive sample prototypes of each geological hazard, and finally construct a positive sample prototype library for geological hazards.

3. The method for geological hazard risk assessment of transmission lines based on a prototype, as described in claim 2, is characterized in that... Based on the different data types, geographical environmental factors are divided into discrete environmental factors and continuous environmental factors, and the reliability of positive and negative samples is calculated. The environmental factor selection method adopted is mainly based on geographical environmental characteristics and existing research results. In order to facilitate the calculation of geographical environmental similarity, we divide the selected geographical environmental factors into discrete environmental factors and continuous environmental factors according to different data types.

4. The method for geological hazard risk assessment of transmission lines based on a prototype according to claim 3, characterized in that, The frequency ratio method was used to calculate the functional relationship between discrete environmental factors and the occurrence frequency of the prototype geological disaster in positive samples. The calculation method is as follows: Where, p i,j A represents the frequency of positive prototype occurrences in category j of environmental factor i; i,j S represents the area within category j of environmental factor i; m represents the number of categories of environmental factor i; S i,j ′ represents the frequency of positive sample prototypes occurring in category j of environmental factor i; A represents the total area of ​​the study area; For S i,j Normalization yields the similarity between category j in environmental factor i and the prototype of the positive sample in environmental factor i, as shown in the following formula: Among them, S i,j S represents the similarity between category j of environmental factor i and the typical value of the positive sample prototype in environmental factor i. i,j The range of its value is [0, 1].

5. The method for geological hazard risk assessment of transmission lines based on a prototype according to claim 4, characterized in that, For the kernel density estimation method, it is assumed that there are n positive sample prototype points, whose environmental factor x has values ​​x1, x2, x3, ..., x n The kernel density estimate for environmental factor x is then given by the formula: Where f(x) is the probability density function relating environmental factor x to the frequency of positive prototypes; n is the number of positive prototype occurrences; k(·) is the kernel function; h is the bandwidth, the size of which constrains the smoothness and shape of the kernel density function; xx i Let x be the environmental factor and x be the environmental factor value of the positive sample prototype. i The difference between them; The kernel density curve is inferred using the Gaussian kernel function, and the bandwidth h is calculated using the "rule of thumb," as shown in the following formula: Where σ is the standard deviation of the environmental factor x values ​​of the n positive sample prototypes.

6. The method for geological hazard risk assessment of transmission lines based on a prototype, as described in claim 5, is characterized in that... Normalizing the estimated probability density function yields the similarity between a single environmental factor and its positive prototype. The normalization is shown in the formula: Where f(x) is the probability density function between environmental factor x and the positive sample prototype frequency; f max (x) is the maximum value of f(x); S x This represents the similarity between environmental factor x and its positive sample prototype. By combining the similarity of a single environmental factor with its positive sample prototype, the overall environmental factor similarity between that point and the positive sample prototype can be inferred. The calculation formula is as follows: S=f(s1,s2,…,s k ,…,s n ) Among them, s k Let S be the similarity between the k-th environmental factor and its positive sample prototype; S is the comprehensive environmental factor similarity, with a range of [0,1]; f represents the comprehensive method function. The confidence scores of positive and negative samples are calculated using the following formulas. The confidence score of positive samples is in the range of [0,1]. The larger the value, the higher the confidence score of the positive sample. Similarly, the confidence score of negative samples is in the range of [0,1]. The larger the value, the higher the confidence score of the negative sample. Reliability i,j =S i,j Reliability′ i,j =1-S i,j Among them, S i,j It is the value of the geographical similarity between point (i,j) and the positive sample prototype; Reliability i,j The point (i,j) represents the reliability value of a positive sample; Reliability′ i,j The point (i,j) represents the confidence value of the negative sample.

7. The method for geological hazard risk assessment of transmission lines based on a prototype, as described in claim 6, is characterized in that... The method for constructing a set of positive and negative samples with credibility based on the credibility of positive and negative samples includes: Based on the principle that the more dissimilar the positive and negative samples are, the higher the overall credibility of the constructed sample set, a method for setting a sample credibility threshold is proposed. This method allows for the collection of positive and negative samples by setting a threshold, thereby constructing positive and negative sample sets with high sample credibility. By comprehensively considering the credibility of positive samples, negative samples, and overall samples, and in order to solve the problems of excessively large collection intervals making it impossible to represent the differences between geological disaster samples under different credibility thresholds, and excessively small collection intervals causing redundancy and hindering inductive analysis, a mutually exclusive credibility threshold setting method is proposed. This method sets credibility thresholds for the collection interval and collection range of samples, so that the positive sample set and the negative sample set are separated in the environmental feature space, thereby obtaining a geological disaster training sample set with high credibility and improving the quality of geological disaster risk assessment. The specific steps for setting the mutual exclusion confidence threshold are as follows: the confidence value range collection interval for positive samples is set to 0.1, and the confidence thresholds for positive samples are: >0.5 (confidence threshold collection interval is 0.5 to 1), >0.6 (confidence threshold collection interval is 0.6 to 1), >0.7 (confidence threshold collection interval is 0.7 to 1), >0.8 (confidence threshold collection interval is 0.8 to 1), >0.9 (confidence threshold collection interval is 0.9 to 1); Similarly, the confidence threshold range for negative samples is also set as follows: >0.5 (confidence threshold range is 0.5 to 1), >0.6 (confidence threshold range is 0.6 to 1), >0.7 (confidence threshold range is 0.7 to 1), >0.8 (confidence threshold range is 0.8 to 1), >0.9 (confidence threshold range is 0.9 to 1).

8. The method for geological hazard risk assessment of transmission lines based on a prototype according to claim 7, characterized in that, The specific steps for collecting geological disaster samples are as follows: By calculating the credibility of positive and negative samples of geological disasters, spatial distribution maps of the credibility of positive and negative samples of geological disasters can be obtained. Based on the confidence thresholds for positive and negative samples of geological disasters, the areas where positive and negative samples can be collected under different confidence value ranges are obtained respectively. The same number of positive and negative samples are randomly collected within their respective collection areas to form positive and negative sample sets under different confidence value ranges, thereby forming a geological disaster training sample set with confidence.

9. The method for geological hazard risk assessment of transmission lines based on a prototype as described in claim 8, characterized in that, The method for constructing a geological hazard risk assessment model using a reliable set of positive and negative samples includes the following steps: Geological hazard risk assessment based on sample credibility uses machine learning models to determine the relationship between geological hazard risk and environmental factors from a credible training sample set and environmental factor data, and expresses this relationship in a linear or nonlinear form to establish a prediction model for evaluating geological hazard risk. The most significant feature of geological hazard risk assessment based on sample credibility is that training samples can be selected by setting thresholds for the credibility of positive and negative samples, thereby improving the accuracy of geological hazard risk assessment. Machine learning models are another important module in geological hazard risk assessment based on sample reliability. We selected a machine learning model commonly used in the field of geological hazard risk mapping to build the inference model.

Images