Method and device for assessing geological hazard level

Abstract

The present invention provides a method and device for assessing geological hazard level, which determines a target assessment area from an assessment area library according to a preset screening rule; determines the disaster type and target assessment factor corresponding to the target assessment area according to the regional description information of the target assessment area; determines the edge computing device of the target assessment area based on the disaster type and target assessment factor corresponding to the target assessment area; the edge computing device is used to calculate and store monitoring data corresponding to the disaster type and target assessment factor; establishes a federated learning framework and performs model training according to the target assessment factor, a preset training data set, and the edge computing device and its monitoring data to obtain a geological hazard level assessment model; inputs the target assessment area into the geological hazard level assessment model, and outputs the assessment result of the target assessment area. This method combines the federated learning framework and conducts assessment with reference to the current assessment standards, so that the assessment result is more authoritative.

Description

Technical Field

[0001] The present invention relates to the technical field of geological disaster risk level assessment, and in particular to a geological disaster risk level assessment method and device. Background Art

[0002] Geological hazard assessment is a systematic analysis of geological environmental factors that affect the occurrence and development of landslide disasters. It is divided into zones according to the possible danger level of the disaster, and then prioritizes and treats them differently. It has become one of the important means to actively and effectively prevent and reduce the losses caused by landslide disasters. Geological hazard assessment methods can be roughly divided into heuristic methods, deterministic methods, statistical analysis methods and artificial intelligence methods.

[0003] However, the above-mentioned deterministic methods, statistical analysis methods and artificial intelligence methods usually require a large amount of complete and reliable historical data on disasters and related environmental conditions survey data to produce reliable results. Limited by existing technical means and theoretical basis, existing technologies usually use heuristic methods to select certain areas for geological hazard risk assessment as the final risk assessment result. Generally, they rely mainly on the personal experience and knowledge of experts, resulting in the incomparability of the results evaluated by different experts. This leads to a great deal of subjectivity in the selection of data and evaluation rules, lack of authority, and insufficient user recognition. Summary of the invention

[0004] The purpose of the present invention is to provide a method and device for assessing geological hazard levels to solve the problems of insufficient authority and insufficient user recognition.

[0005] In a first aspect, the present invention provides a method for assessing geological hazard levels, which is applied to an electronic device, wherein an assessment area library is stored in the electronic device, wherein the assessment area library contains multiple areas to be assessed, and area description information corresponding to each area to be assessed; the method comprises: determining a target assessment area from the assessment area library according to a preset screening rule; determining a disaster type and a target evaluation factor corresponding to the target assessment area according to the area description information of the target assessment area; determining an edge computing device for the target assessment area based on the disaster type and the target evaluation factor corresponding to the target assessment area; the edge computing device is used to calculate and store monitoring data corresponding to the disaster type and the target evaluation factor; establishing a federated learning framework and performing model training according to the target evaluation factor, a preset training data set, the edge computing device and its monitoring data, to obtain a geological hazard level assessment model; inputting the target assessment area into the geological hazard level assessment model, and outputting an assessment result of the target assessment area.

[0006] In an optional embodiment, the above-mentioned step of determining the target evaluation area from the evaluation area library according to preset screening rules includes: evaluating each area to be evaluated contained in the evaluation area library to obtain a first scoring result corresponding to each area to be evaluated; comparing the first scoring result corresponding to each area to be evaluated with a first preset scoring threshold; and determining the area to be evaluated corresponding to a first scoring result higher than the first preset scoring threshold as the target evaluation area.

[0007] In an optional embodiment, the above-mentioned first scoring result includes multiple first sub-results; the first scoring result corresponding to each area to be evaluated is compared with a first preset scoring threshold; the step of determining the area to be evaluated corresponding to a first scoring result higher than the first preset scoring threshold as a target evaluation area includes: determining the Kendall harmony coefficient of the current first scoring result according to the multiple first sub-results corresponding to each first scoring result; for each Kendall harmony coefficient, performing the following steps: comparing the Kendall harmony coefficient corresponding to each first scoring result with the first preset scoring threshold; determining the chi-square test result of the Kendall harmony coefficient higher than the first preset scoring threshold; determining the target Kendall harmony coefficient corresponding to the chi-square test result higher than the first preset test threshold; and determining the area to be evaluated corresponding to the maximum value of the chi-square test result as the target evaluation area.

[0008] In an optional embodiment, the regional description information of the target assessment area includes multiple preset disaster types and multiple preset evaluation factors belonging to the target assessment area; the step of determining the disaster type and target evaluation factor corresponding to the target assessment area according to the regional description information of the target assessment area includes: determining the multiple preset disaster types contained in the regional description information as the disaster type corresponding to the target assessment area; evaluating the multiple preset evaluation factors belonging to the target assessment area to obtain a second scoring result corresponding to each preset evaluation factor; for the second scoring result corresponding to each preset evaluation factor, performing the following steps: comparing the second scoring result corresponding to each preset evaluation factor with a second preset scoring threshold; determining the preset evaluation factor corresponding to the second scoring result higher than the second preset scoring threshold as the target evaluation factor.

[0009] In an optional embodiment, the above-mentioned second scoring result includes multiple second sub-results; the second sub-results are used to indicate the correlation level of the impact of the current preset evaluation factor and the hazard level; the second scoring result corresponding to each preset evaluation factor is compared with the second preset scoring threshold; the step of determining the preset evaluation factor corresponding to the second scoring result higher than the second preset scoring threshold as the target evaluation factor includes: determining the weighted mean of the current preset evaluation factor based on the multiple second sub-results corresponding to the second scoring result; and determining the preset evaluation factor corresponding to the weighted mean greater than the second threshold as the target evaluation factor.

[0010] In an optional embodiment, the above-mentioned step of determining the edge computing device of the target assessment area based on the disaster type and the target evaluation factor corresponding to the target assessment area includes: determining the required number of edge computing devices based on the number of disaster types corresponding to the target assessment area and the number of established monitoring system types corresponding to the types of disasters; determining the edge computing device and its deployment location of the target assessment area based on the location of the target assessment area and the required number of edge computing devices.

[0011] In an optional embodiment, the electronic device stores a hazard level database, which includes hazard level indicators corresponding to the monitoring data; there are multiple target evaluation factors, which include developmental level evaluation factors, hazard level evaluation factors and inducing factor evaluation factors; the training data set is determined in the following manner: comparing the monitoring data of each target evaluation factor with a preset monitoring data threshold; determining target monitoring data that is greater than the preset monitoring data threshold; obtaining a first decision value indicated by the target monitoring data based on the target monitoring data and the hazard level database; obtaining the developmental level evaluation factor corresponding to the developmental level evaluation factor based on the developmental level evaluation factor of each target evaluation factor and the target monitoring data corresponding to the current developmental level evaluation factor. degree target variable; based on the hazard degree evaluation factor of each target evaluation factor and the target monitoring data corresponding to the current hazard degree evaluation factor, obtain the hazard degree target variable corresponding to the hazard degree evaluation factor; according to the development degree target variable and the hazard degree target variable, as well as the stored hazard level database, obtain the second decision values ​​corresponding to the development degree target variable and the hazard degree target variable respectively; according to the second decision value and the stored hazard level database, determine the label value corresponding to the target evaluation factor; the label value is used to indicate the hazard level of the target evaluation factor; based on the target evaluation factor, the target monitoring data, the development degree target variable and the hazard degree target variable, as well as the first decision value, the second decision value and the label value, obtain a preset training data set.

[0012] In an optional embodiment, the regional description information of the target assessment area includes multiple preset disaster types and multiple preset evaluation factors to which the target assessment area belongs; there are multiple preset evaluation factors, and a monitoring data source characteristic table is stored in the electronic device, and the monitoring data source characteristic table includes the acquisition method and acquisition frequency of the monitoring data corresponding to each preset evaluation factor; the target monitoring data is determined by the following steps: according to the monitoring data source characteristic table and each target evaluation factor, the monitoring data corresponding to the current target evaluation factor is determined; for the monitoring data corresponding to each target evaluation factor, the following steps are performed: determine whether the value of the current monitoring data is less than a preset first monitoring data threshold; if yes, delete the current monitoring data; if no, determine whether the value of the current monitoring data is greater than a preset second monitoring data threshold; if yes, replace the value of the current monitoring data with the mean of the monitoring data corresponding to the current target evaluation factor, and determine the current monitoring data as the target monitoring data; if no, directly determine the current monitoring data as the target monitoring data.

[0013] In an optional embodiment, the above-mentioned electronic device stores preset data security coordination rules, which are used to indicate the encryption or decryption method of monitoring data; according to the target evaluation factor, the preset training data set and the edge computing device and its monitoring data, a federated learning framework is established and model training is performed to obtain a geological hazard hazard level assessment model, including: establishing a federated learning framework according to the preset training data set; determining the training method of the model according to the target evaluation factor, the preset training data set and the preset data security coordination rules; performing model training based on the model training method, the edge computing device and the federated learning framework to update the parameters of the model; and obtaining a geological hazard hazard level assessment model according to the updated model parameters.

[0014] In a second aspect, the present invention provides a geological hazard risk level assessment device, which is applied to an electronic device and the above-mentioned method, wherein an assessment area library is stored in the electronic device, and the assessment area library contains multiple areas to be assessed, and area description information corresponding to each area to be assessed; the device includes: an area determination module, which is used to determine the target assessment area from the assessment area library according to a preset screening rule; an information determination module, which is used to determine the disaster type and target evaluation factor corresponding to the target assessment area according to the area description information of the target assessment area; a computing device determination module, which is used to determine the edge computing device of the target assessment area based on the disaster type and target evaluation factor corresponding to the target assessment area; the edge computing device is used to train the assessment model; the model determination module, which is used to establish a federated learning framework and perform model training according to the evaluation factor, the preset training data set and the edge computing device, to obtain a geological hazard risk level assessment model; an assessment result determination module, which is used to input the target assessment area into the geological hazard risk level assessment model and output the assessment result of the target assessment area.

[0015] The embodiments of the present invention bring the following beneficial effects:

[0016] The present invention provides a method and device for assessing the risk level of geological disasters. First, a target assessment area to be ultimately assessed is selected from a plurality of assessment areas, and then a target assessment factor to be ultimately assessed is selected from a plurality of assessment factors of the target assessment area. Monitoring data is then acquired according to the target assessment factor and the type of disaster corresponding to the target assessment area. After that, the acquired monitoring data is processed to obtain a training data set, and a federated learning framework is established. The edge computing device participates in model training. The method not only evaluates and obtains the assessment result based on the experience of experts, but also combines the federated learning technology, and takes the current assessment standards as a reference basis, so as to solve the problems of insufficient authority of the current technical assessment results and insufficient user recognition.

[0017] Other features and advantages of the present invention will be set forth in the following description, or some features and advantages may be inferred or unambiguously determined from the description, or may be learned by implementing the above-mentioned technology of the present invention.

[0018] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, preferred embodiments are specifically listed below and described in detail with reference to the attached drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the specific implementation methods of the present invention or the technical solutions in the prior art, the drawings required for use in the specific implementation methods or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some implementation methods of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0020] Figure 1 A flow chart of a method for assessing geological hazard levels provided by an embodiment of the present invention;

[0021] Figure 2 A flow chart of another method for assessing geological disaster risk levels provided by an embodiment of the present invention;

[0022] Figure 3 A flow chart of another method for assessing geological disaster risk levels provided by an embodiment of the present invention;

[0023] Figure 4 A schematic diagram of the structure of a geological disaster risk level assessment device provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0024] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Generally, the components of the embodiments of the present invention described and shown in the drawings here can be arranged and designed in various different configurations.

[0025] Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the invention claimed for protection, but merely represents selected embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0026] Geological hazard assessment is a systematic analysis of geological environmental factors that affect the occurrence and development of landslide disasters. It is divided according to the possible risk level of the disaster, and then the severity and urgency are differentiated. It has become one of the important means to actively and effectively prevent and reduce the losses of landslide disasters. Geological hazard assessment methods can be roughly divided into deterministic methods, statistical analysis methods, artificial intelligence methods and heuristic methods. Deterministic assessment methods are based on various physical models of geological disasters for disaster assessment. Statistical analysis methods can evaluate the possibility and degree of danger of geological disasters in the current area to be assessed based on the distribution information of geological disasters that have occurred in the past and the main environmental conditions. Artificial intelligence assessment methods are mainly based on machine learning algorithms to build a geological hazard early warning model for the area to be assessed. Heuristic methods are mainly determined by comprehensive analysis by disaster assessment experts to determine the hazard of the disaster.

[0027] However, the deterministic assessment method theoretically requires the establishment of a physical model that is consistent with the environment of the area to be assessed. Limited by existing technical means and theoretical foundations, it is impossible to establish a completely consistent physical model; the statistical analysis assessment method requires a large amount of complete and reliable historical data on disasters and related environmental condition survey data to produce reliable results. Geological disasters are affected by many factors, such as topography, geology, and hydrological conditions. In actual operations, it is often unrealistic to obtain a large amount of complete and reliable data. Due to the complex geological structure, there may be obvious correlations between disaster influencing factors, or the control factors and influencing factors of the screening factors are insufficient, which may lead to reduced model accuracy and make it difficult to select and analyze the evaluation factors of the artificial intelligence assessment method. In addition, due to the different construction and management units of the existing system, there are no standards for sensor selection and deployment, which leads to great difficulties in obtaining sample data in a comprehensive and stable manner. The use of machine learning algorithms to classify or regress the geological disaster risk in the assessment area requires a large amount of sample data, so it is impossible to accurately obtain the assessment results through the artificial intelligence assessment method.

[0028] Heuristic assessment methods generally rely mainly on the personal experience and knowledge of experts. Since deterministic methods, statistical analysis methods and artificial intelligence methods usually require a large amount of complete and reliable historical data on disasters and related environmental conditions survey data to produce reliable results, and are limited by existing technical means and theoretical foundations, existing technologies usually use heuristic methods to select certain areas for geological hazard risk assessment as the final risk assessment results. They generally rely mainly on the personal experience and knowledge of experts, resulting in the incomparability of the results evaluated by different experts. This leads to a high degree of subjectivity in the selection of data and evaluation rules, lack of authority, and insufficient user recognition.

[0029] Based on the above problems, an embodiment of the present invention provides a method and device for assessing the risk level of geological disasters. This technology can be applied in the scenario of assessing the risk level of geological disasters. To facilitate the understanding of this embodiment, a method for assessing the risk level of geological disasters disclosed in an embodiment of the present invention is first introduced in detail. The method is applied to an electronic device, and an assessment area library is stored in the electronic device. The assessment area library contains multiple areas to be assessed, and area description information corresponding to each area to be assessed. Figure 1 As shown, the method comprises the following steps:

[0030] Step S102: determining a target evaluation area from an evaluation area library according to a preset screening rule.

[0031] Specifically, the areas to be assessed included in the assessment area library are assessment areas that are usually referenced when conducting geological hazard risk level assessments, such as mountain roads, rift valleys, bridges, and other common or uncommon assessment areas. In specific implementation, it is necessary to select the target assessment area to be ultimately used for assessment from the assessment area library. Furthermore, the target assessment area is determined through evaluations by multiple experts, wherein a part of the experts are used to propose a delineation plan for the assessment area, and another part of the experts are used to evaluate the delineation plan, and the target assessment plan is determined based on the evaluation results.

[0032] Step S104: determining the disaster type and target evaluation factor corresponding to the target assessment area according to the area description information of the target assessment area.

[0033] Specifically, after determining the above-mentioned target assessment area, the target assessment area contains multiple corresponding disaster types and multiple evaluation factors corresponding to each disaster type. In specific implementation, it is also necessary to select the target evaluation factor ultimately used for the evaluation from the above-mentioned multiple evaluation factors. Furthermore, the above-mentioned target evaluation factor is determined through multiple evaluations, and the multiple experts respectively evaluate the above-mentioned preset evaluation factors and determine the target evaluation factor based on the evaluation results.

[0034] Step S106, based on the disaster type and target evaluation factor corresponding to the target evaluation area, determine the edge computing device of the target evaluation area; the edge computing device is used to calculate and store monitoring data corresponding to the disaster type and target evaluation factor.

[0035] Specifically, the above-mentioned edge computing device is a hardware and software integrated system with functions such as computing, storage, communication, and security protection. Its main function is to realize local model training and updating. When the target assessment area for the final assessment, the type of disaster in the target assessment area and the evaluation factors used for the assessment are determined, the number of the above-mentioned edge computing devices will be determined according to the number of disaster types in the assessment area.

[0036] Step S108, based on the target evaluation factors, the preset training data set, the edge computing device and its monitoring data, a federated learning framework is established and model training is performed to obtain a geological hazard level assessment model.

[0037] After determining the evaluation factors and the number of edge computing devices, the target evaluation area will be monitored by sensors or established monitoring systems to obtain the monitoring data corresponding to the target evaluation factors. Then, the monitoring data will be normalized, standardized and discretized according to the current "GB / T 40112-2021 Specification for Geological Hazard Hazard Assessment" to determine the preset training data set. The preset training data set uses the processed monitoring data as the model training sample, and is trained according to the federated learning framework. The edge computing devices participate in the model training process, and finally obtain the geological hazard level assessment model. The concept and principle of federated learning are: there are several participants and collaborators who jointly perform the federated learning task. The participants (i.e., the data owners) generate intermediate data similar to gradients through the preset federated learning algorithm, which is handed over to the coordinator for further processing and then returned to each participant to prepare for the next round of training.

[0038] Step S110, input the target assessment area into the geological hazard risk level assessment model, and output the assessment result of the target assessment area.

[0039] A method for assessing geological hazard levels provided by an embodiment of the present invention first selects a target assessment area to be ultimately assessed from a plurality of assessment areas, then selects a target assessment factor to be ultimately assessed from a plurality of assessment factors of the target assessment area, and then acquires monitoring data according to the target assessment factor and the type of disaster corresponding to the target assessment area. After that, the acquired monitoring data is processed to obtain a training data set, and a federated learning framework is established, in which edge computing devices participate in model training. The assessment results are obtained not only through expert evaluation, but also in combination with federated learning technology, and the current assessment standards are used as a reference basis, so as to solve the problems of insufficient authority of current technical assessment results and insufficient user recognition.

[0040] In view of the above embodiment, the present invention also provides another method for assessing the geological hazard level, which is implemented on the basis of the above method for assessing the geological hazard level. The method focuses on describing the specific process of determining the target assessment area from the assessment area library according to the preset screening rules (implemented by the following steps S202-S206), and the specific process of determining the disaster type and target evaluation factor corresponding to the target assessment area according to the regional description information of the target assessment area (implemented by the following steps S208-S214), such as Figure 2 As shown, the method includes:

[0041] Step S202: Evaluate each to-be-evaluated area contained in the evaluation area library to obtain a first scoring result corresponding to each to-be-evaluated area.

[0042] The electronic device stores a plurality of objects for determining the area to be selected. In a specific implementation, an assessment area determination group is formed according to the stored plurality of objects for determining the area to be selected. The assessment area determination group includes a plurality of first objects, and the first objects may be experts who are qualified to serve in the "Technical Expert Database for Geological Hazard Prevention and Control Work" at the municipal level or above in the area to be assessed. In addition, the assessment area determination group is divided into a scheme group and a scoring group, wherein the number of first objects in the scheme group is N, and 3≤N≤7; the number of first objects in the scoring group is M, and 10≤M≤20.

[0043] Afterwards, each first object of the scheme group proposes a delineation scheme for the area to be assessed based on the assessment area library according to the natural environment, population and society of the area to be assessed, and the delineation scheme is used to determine the assessment area for geological hazard risk level assessment; then each first object of the scoring group scores the above N schemes respectively to obtain the first sub-result corresponding to each area to be assessed; wherein the above first sub-result follows the following rules: the first word result of each area to be assessed by the same first object is different, and the scoring value C of the above first word result is an integer, and 1≤C≤N, the higher the score, the more recognized the scheme, wherein the value of N is consistent with the number N of the first objects of the scheme group, indicating that the N schemes can be sorted by comparing the size of the scoring value C.

[0044] Step S204: compare the first scoring result corresponding to each to-be-evaluated area with a first preset scoring threshold.

[0045] Step S206: The to-be-evaluated area corresponding to the first scoring result being higher than the first preset scoring threshold is determined as the target evaluation area.

[0046] In a specific implementation, the Kendall harmony coefficient of the current first scoring result is determined according to the multiple first sub-results corresponding to each first scoring result.

[0047] Kendall's harmony coefficient is a correlation quantity that calculates the degree of correlation of multiple grade variables. It is applicable to data with multiple columns of related grade data, that is, k raters evaluate N objects, or the same person evaluates N objects k times in succession. By obtaining the Kendall's harmony coefficient, good works or good raters can be selected more objectively. In specific implementation, the first word result of the first object in the scoring group is analyzed, and the Kendall's harmony coefficient of the first scoring result corresponding to each area to be evaluated (i.e., the above-mentioned demarcation scheme) is calculated according to the following formula:

[0048]

[0049] in, S is the variance of the total scheme score and the sum of the ratings of each evaluated object (i.e. the above-mentioned area to be evaluated) R i The average of all these sums For each Kendall coefficient of harmony, perform the following steps 20-23:

[0050] Step 20: compare the Kendall harmony coefficient corresponding to each first scoring result with a first preset scoring threshold.

[0051] Step 21, determining a chi-square test result of the Kendall coefficient of harmony that is higher than a first preset score threshold.

[0052] Specifically, the above-mentioned first preset scoring threshold is 0.85. When the above-mentioned Kendall harmony coefficient W<0.85, the M first objects of the scoring group are reorganized to score and judge again. If it still fails to meet the standard after 3 times, the N first objects of the plan group are organized to re-propose the above-mentioned demarcation plan; when the above-mentioned Kendall harmony coefficient W>0.85, a chi-square test is performed on the current Kendall harmony coefficient to obtain the chi-square test result corresponding to the current Kendall harmony coefficient.

[0053] The chi-square test is used to measure the deviation between the actual observed value and the theoretical inference value of the statistical sample. The deviation between the actual observed value and the theoretical inference value determines the value of the chi-square test. It is a very widely used hypothesis testing method. Its application in the statistical inference of categorical data includes the chi-square test for comparing two rates or two component ratios, the chi-square test for comparing multiple rates or multiple component ratios, and the correlation analysis of categorical data. The formula of the chi-square test is: 2 =M(P-1)W, where the degree of freedom df=P-1, that is, the current Kendall harmony coefficient adopts a suitability test method with only one constraint condition.

[0054] Step 22, determining whether the chi-square test result is higher than the target Kendall coefficient of harmony corresponding to the first preset test threshold.

[0055] Step 23, the to-be-evaluated area corresponding to the maximum value of the chi-square test result is determined as the target evaluation area.

[0056] When the chi-square test result reaches a confidence level of 95%, it means that the Kendall concordance coefficient W reaches a significance level. The delineation scheme corresponding to the Kendall concordance coefficient with the maximum value in the chi-square test result (i.e., the above-mentioned area to be evaluated) is selected and determined as the target evaluation area. Otherwise, the M first subjects of the scoring group are reorganized to perform scoring and make the above-mentioned judgment. If the standard is still not met after three times, the N first subjects of the scheme group are organized to re-propose the above-mentioned delineation scheme.

[0057] After the target assessment area is determined, the area description information of the target assessment area can be obtained, wherein the area description information of the target assessment area includes a plurality of preset disaster types and a plurality of preset evaluation factors to which the target assessment area belongs.

[0058] The above-mentioned preset evaluation factors include two parts: stable factors and dynamic factors. The above-mentioned stable factors refer to evaluation factors with fixed or slowly changing reference indicators after the scope of the above-mentioned target evaluation area is determined. Such evaluation factors only need to be obtained once or updated regularly to meet the risk assessment requirements.

[0059] The above dynamic factors refer to the factors whose index data are affected by natural or human factors, change rapidly, and need to be monitored in real time or updated quickly by technical means such as sensors. Among them, the dynamic factor index items, index acquisition methods and update frequencies of various types of geological disasters in the above regional description information are determined by the following Table 1:

[0060] Table 1

[0061]

[0062]

[0063] Step S208: Determine the multiple preset disaster types included in the above-mentioned area description information as the disaster types corresponding to the target assessment area.

[0064] Step S210: Evaluate the multiple preset evaluation factors to which the target evaluation area belongs, and obtain a second scoring result corresponding to each preset evaluation factor.

[0065] In a specific implementation, the electronic device also stores a plurality of evaluation factor determination objects, and an evaluation factor determination group is formed according to the stored plurality of evaluation factor determination objects. The evaluation factor determination group includes a plurality of second objects, and the second objects may be experts who are qualified to serve in the "Technical Expert Database for Geological Hazard Prevention and Control Work" at the district or city level or above to be evaluated. The number of second objects in the evaluation factor determination group is A, and 15≤A≤30.

[0066] A second object performs correlation scoring on a plurality of preset evaluation factors in the target assessment area according to the number and distribution of disaster types within the target assessment area, combined with the geological, meteorological, hydrological and other historical data of the target assessment area. Let the correlation score of the second object j on the influence of the i-th preset evaluation factor and the geological disaster risk level of the target assessment area be recorded as x ij (i=1,2,...,B;j=1,2,...,A), then a second scoring result corresponding to each of the above-mentioned preset evaluation factors is obtained, and the second scoring result includes multiple second sub-results, and the above-mentioned second sub-results are used to indicate the correlation level of the impact of the current preset evaluation factor on the above-mentioned hazard level.

[0067] For each second scoring result corresponding to the preset evaluation factor, perform the following steps:

[0068] Step S212: compare the second scoring result corresponding to each preset evaluation factor with a second preset scoring threshold.

[0069] Step S214: The preset evaluation factor corresponding to the second scoring result being higher than the second preset scoring threshold is determined as the target evaluation factor.

[0070] Specifically, the above target evaluation factor is determined through the following steps 30-31:

[0071] Step 30: Determine a weighted average of the current preset evaluation factors based on the multiple second sub-results corresponding to the second scoring result.

[0072] After the second scoring result is obtained, the multiple second sub-results of the second scoring result of each preset evaluation factor are converted into a grade number (i.e., a ranking value), and the ranking value of the second object j for the second sub-result of the i-th preset evaluation factor is recorded as y ij , where 1≤y ij ≤B, calculate the rank correlation coefficient ρ of the second object j and all second objects j , the above rank correlation coefficient is calculated by the following formula:

[0073]

[0074] Among them, y kj Represents the ranking value of the second object j for the second sub-result of the kth preset evaluation factor.

[0075] The Coefficient of Rank Correlation, also known as the rank correlation coefficient, is a statistic obtained by arranging the sample values ​​of two factors in order of data size and replacing the actual data with the rank of each factor sample value. It is a statistical analysis indicator that reflects the degree of rank correlation.

[0076] When the above rank correlation coefficient ρ is obtained j Then, let ρ j The number of second subjects with a value >0 is S. Among the S second subjects, the score of the second subject r on the i-th preset evaluation factor is z ir (i=1,2,…,B;r=1,2,…,S), the rank correlation coefficient ρ between the second object r and all second objects r is the calculated value obtained by the formula of the rank correlation coefficient, and the weighted mean of the scores of the S second objects for the preset evaluation factor i is calculated. The weighted mean is calculated by the following steps:

[0077]

[0078] in, ρ t Represents the rank correlation coefficient of the tth second object among the above S second objects.

[0079] Step 31: determine the preset evaluation factor corresponding to the weighted mean value greater than the second threshold as the target evaluation factor.

[0080] Get the above removed weighted mean E i After that, remove the weighted mean E i The preset evaluation factor corresponding to the average value is significantly lower than that of the above average value, which is selected according to the actual usage. i The specific data that is significantly lower than the average value is the above-mentioned second threshold; then the remaining preset evaluation factors are selected as the target evaluation factors of the above-mentioned target evaluation area, and the target evaluation factors are the geological disaster risk level assessment indicators.

[0081] Step S216, determining the required number of edge computing devices based on the number of disaster types corresponding to the target assessment area and the number of established monitoring system types corresponding to the disaster types.

[0082] Specifically, when multiple disaster types corresponding to the target assessment area are obtained, the number U of established monitoring system types corresponding to the above disaster types is obtained according to the number K of the above disaster types, and the number Q of deployed edge computing devices is determined according to the number K of the above disaster types and the number N of established monitoring system types, that is: when U=0, Q=K; when U≠0, and the number of disaster types corresponding to the above established monitoring systems is D, Q=K+UD.

[0083] Step S218, determining the edge computing devices in the target evaluation area and their deployment locations based on the location of the target evaluation area and the required number of edge computing devices.

[0084] After determining the required number Q of the above-mentioned edge computing devices, the above-mentioned edge computing devices are deployed according to the location of the above-mentioned target evaluation area and the required number Q of the above-mentioned edge computing devices, and the above-mentioned edge computing devices follow the principle of nearby deployment, that is, they are deployed on-site in the above-mentioned target evaluation area or in the above-mentioned built monitoring system room.

[0085] Step S220, based on the target evaluation factors, the preset training data set, the edge computing device and its monitoring data, a federated learning framework is established and model training is performed to obtain a geological hazard level assessment model.

[0086] Specifically, the above preset training data set is determined by the following steps 40-47:

[0087] Step 40, comparing the monitoring data of each target evaluation factor with a preset monitoring data threshold.

[0088] Step 41, determining target monitoring data that is greater than a preset monitoring data threshold.

[0089] Specifically, the electronic device stores a hazard level database, which contains hazard level indicators corresponding to the monitoring data of the above-mentioned multiple preset evaluation factors; specifically, the hazard level database is the "GB / T40112-2021 Geological Hazard Hazard Assessment Specification". In specific implementation, the electronic device stores a monitoring data source feature table. The specific content refers to the description of the stability factor part in the above-mentioned regional description information and Table 1 of the dynamic factor, which will not be repeated here. The above-mentioned monitoring data source feature table includes the acquisition method and acquisition frequency of the monitoring data corresponding to each preset evaluation factor; the above-mentioned target monitoring data is determined by the following steps 50-53:

[0090] Step 50, determining the monitoring data corresponding to the current target evaluation factor according to the monitoring data source characteristic table and each target evaluation factor.

[0091] The monitoring data corresponding to the above-mentioned target evaluation factor is acquired according to the acquisition method and acquisition frequency of the corresponding monitoring data in the monitoring data source characteristic table. In addition, the above-mentioned acquired monitoring data is classified according to the above-mentioned target evaluation factor.

[0092] In some optional embodiments, it is also determined whether the types and quantities of the above-mentioned built monitoring systems can meet the requirements for obtaining the required monitoring data based on the types of disasters in the above-mentioned target assessment area and the corresponding target evaluation factors, as well as the number of types U of the above-mentioned built monitoring systems and the required number Q of the above-mentioned edge computing devices. When the types and numbers of monitoring sensors deployed by the above-mentioned built monitoring systems are insufficient to meet the needs, additional construction is carried out based on the types of disasters in the above-mentioned target assessment area and the corresponding target evaluation factors.

[0093] Perform the following steps for the monitoring data corresponding to each target evaluation factor:

[0094] Step 51, determining whether the value of the current monitoring data is less than a preset first monitoring data threshold.

[0095] Step 52, if yes, delete the current monitoring data; if no, determine whether the value of the current monitoring data is greater than a preset second monitoring data threshold.

[0096] Step 53, if yes, replace the value of the current monitoring data with the mean value of the monitoring data corresponding to the current target evaluation factor, and determine the current monitoring data as the target monitoring data; if no, directly determine the current monitoring data as the target monitoring data.

[0097] Specifically, after obtaining the monitoring data corresponding to the above-mentioned target evaluation factor, determine whether the monitoring data is null value data (such as a small proportion). When the monitoring data is null value data, directly delete the monitoring data. Whether the above-mentioned null value data is the above-mentioned first monitoring data threshold, and the specific value of the above-mentioned null value data is selected according to the actual usage; if the monitoring data is not null value data, determine whether the monitoring data has a large proportion. If the monitoring data has a large proportion, use the mean substitution method to complete the data, and then determine the above-mentioned monitoring data that has not been deleted and the monitoring data completed by the mean substitution method as the above-mentioned target monitoring data. The above-mentioned specific value used to determine whether the monitoring data has a large proportion is the above-mentioned second monitoring data threshold, and the specific value of the above-mentioned second monitoring data threshold is selected according to the actual usage.

[0098] Step 42, obtaining a first decision value indicated by the target monitoring data according to the target monitoring data and the danger level database.

[0099] Specifically, after obtaining the target monitoring data, the monitoring data of the target evaluation factors are distinguished according to the types of the disasters, and are numbered respectively, and the monitoring data of different target evaluation factors are distinguished by the disaster potential point numbers. Then, the target monitoring data of the target evaluation factors are normalized, standardized and discretized according to the hazard level database to obtain the first determination value; the first determination value is determined by the following Table 2:

[0100] Table 2

[0101]

[0102]

[0103]

[0104]

[0105]

[0106]

[0107]

[0108]

[0109] The corresponding numerical value assigned in the above Table 2 is the above first determination numerical value.

[0110] Step 43, based on the developmental degree evaluation factor of each target evaluation factor and the target monitoring data corresponding to the current developmental degree evaluation factor, obtain the developmental degree target variable corresponding to the developmental degree evaluation factor.

[0111] Step 44, based on the hazard level evaluation factor of each target evaluation factor and the target monitoring data corresponding to the current hazard level evaluation factor, obtain the hazard level target variable corresponding to the hazard level evaluation factor.

[0112] Step 45, according to the development level target variable and the hazard level target variable, and the stored danger level database, obtain the second determination values ​​corresponding to the development level target variable and the hazard level target variable respectively.

[0113] Specifically, the above-mentioned target evaluation factors include developmental level evaluation factors, hazard level evaluation factors and inducing factor evaluation factors. After obtaining the above-mentioned target monitoring data, the developmental level target variables and hazard level target variables corresponding to the above-mentioned target evaluation factors can be obtained. Specifically, the developmental level target variable is recorded as F and the hazard level target variable is recorded as G. Then, the developmental level target variables and hazard level target variables corresponding to the above-mentioned target evaluation factors are normalized, standardized and discretized according to the above-mentioned hazard level database to obtain the above-mentioned second determination value. When the above-mentioned developmental level target variable F is determined to be strongly developed according to the monitoring data, the second determination value of the above-mentioned developmental level target variable F is determined to be 3; when the above-mentioned developmental level target variable F is determined to be moderately developed according to the monitoring data, the second determination value of the above-mentioned developmental level target variable F is determined to be 2; when the above-mentioned developmental level target variable F is determined to be weakly developed according to the monitoring data, the second determination value of the above-mentioned developmental level target variable F is determined to be 1. When the hazard level target variable G is determined to be strongly developed according to the monitoring data, the second determination value of the hazard level target variable G is determined to be 3; when the hazard level target variable G is determined to be moderately developed according to the monitoring data, the second determination value of the hazard level target variable G is determined to be 2; when the hazard level target variable G is determined to be weakly developed according to the monitoring data, the second determination value of the hazard level target variable G is determined to be 1. The above-mentioned development level determination standard is determined by the value change of the above-mentioned development level target variable F or the above-mentioned hazard level target variable G.

[0114] Step 46, determining a label value corresponding to the target evaluation factor according to the second decision value and the stored danger level database; the label value is used to indicate the danger level of the target evaluation factor.

[0115] The above-mentioned label value refers to the result of the geological disaster risk determination, which is recorded as label value C. In specific implementation, the value of the above-mentioned label value C is determined based on the above-mentioned second decision value and the above-mentioned hazard level database.

[0116] Specifically, when the second determined value of the above-mentioned developmental level target variable F is 3, and the second determined value of the above-mentioned harm level target variable G is 3, the above-mentioned label value C is 3; when the second determined value of the above-mentioned developmental level target variable F is 3, and the second determined value of the above-mentioned harm level target variable G is 2, the above-mentioned label value C is 3; when the second determined value of the above-mentioned developmental level target variable F is 3, and the second determined value of the above-mentioned harm level target variable G is 1, the above-mentioned label value C is 2; when the second determined value of the above-mentioned developmental level target variable F is 2, and the second determined value of the above-mentioned harm level target variable G is 3, the above-mentioned label value C is 3; when the second determined value of the above-mentioned developmental level target variable F is 2, and the second determined value of the above-mentioned harm level target variable G is 3 When the second determined value of the harm degree target variable G is 2, the above-mentioned label value C is 2; when the second determined value of the above-mentioned development level target variable F is 2 and the second determined value of the above-mentioned harm degree target variable G is 1, the above-mentioned label value C is 1; when the second determined value of the above-mentioned development level target variable F is 1 and the second determined value of the above-mentioned harm degree target variable G is 3, the above-mentioned label value C is 2; when the second determined value of the above-mentioned development level target variable F is 1 and the second determined value of the above-mentioned harm degree target variable G is 2, the above-mentioned label value C is 2; when the second determined value of the above-mentioned development level target variable F is 1 and the second determined value of the above-mentioned harm degree target variable G is 1, the above-mentioned label value C is 1.

[0117] Among them, when the value of the above-mentioned label value C is 3, it means that the geological hazard risk determination result corresponding to the above-mentioned target evaluation factor is highly dangerous; when the value of the above-mentioned label value C is 2, it means that the geological hazard risk determination result corresponding to the above-mentioned target evaluation factor is medium dangerous; when the value of the above-mentioned label value C is 1, it means that the geological hazard risk determination result corresponding to the above-mentioned target evaluation factor is low dangerous.

[0118] Step 47, obtaining a preset training data set based on the target evaluation factor, the target monitoring data, the development level target variable and the hazard level target variable, as well as the first decision value, the second decision value and the label value.

[0119] Specifically, the monitoring data of the above developmental evaluation factor is recorded as {X i}, the above monitoring data of the above hazard level evaluation factor is {Y j}, the above monitoring data of the above inducing factor evaluation factor is {Z k}, then the target evaluation factor of the target evaluation area is {X i , Y j , Z k}, wherein the above-mentioned development level target variable F is obtained by monitoring data {X iThe above-mentioned hazard level target variable G is determined by the monitoring data {Y j According to the first decision value, the second decision value and the label value obtained through the stored danger level database, a preset training data set is obtained.

[0120] Among them, the geological hazard risk level of the above-mentioned target assessment area can be preliminarily determined by the above-mentioned development degree evaluation factor and the above-mentioned hazard degree evaluation factor. In addition, the above-mentioned development degree evaluation factor and the above-mentioned hazard degree evaluation factor mainly correspond to the stability factor in the preset evaluation factors.

[0121] Step S222, input the target assessment area into the geological hazard risk level assessment model, and output the assessment result of the target assessment area.

[0122] Another method for assessing geological hazard level provided by an embodiment of the present invention is to form an assessment area determination group, and divide the assessment area determination group into a scheme group and a scoring group. The scheme group proposes a delineation scheme for the target assessment area, and the scoring group evaluates the above delineation scheme. Then, it is judged whether the scoring result of the above scoring group meets the first scoring threshold to determine the final target assessment area. Whether the standard of meeting the first scoring threshold is obtained by calculating the Kendall concordance coefficient and the chi-square test result of the scoring result, and the method has engineering feasibility; an evaluation factor determination group is formed to evaluate multiple preset evaluation factors in the above target assessment area, and it is judged whether the second scoring result of each of the above evaluation factors meets the second preset scoring threshold to determine the final target evaluation factor used for evaluation. Whether the second scoring result meets the second preset scoring threshold is determined by calculating the rank correlation grade coefficient of each second object in the evaluation factor determination group and the weighted mean of the evaluation factor. The selection of the above target evaluation factor is quantified, and the method has engineering feasibility.

[0123] In addition, the number of edge computing devices designed in the embodiment of the present invention is determined according to the type of disaster and the number of established monitoring systems. Edge computing devices are only deployed for disaster types for which no monitoring system has been built. This allows for efficient use of the monitoring data of established monitoring systems while avoiding duplication of construction and waste of resources.

[0124] The embodiment of the present invention also obtains monitoring data based on the target evaluation factor and the type of disaster corresponding to the target assessment area, and then converts the acquired monitoring data into a hazard level (i.e., the first decision value, the second decision value and the label value) according to the hazard level database, and quantifies the monitoring data of the above target evaluation factor. On the basis of the current national standard "GB / T 40112-2021 Specification for Geological Hazard Hazard Assessment", technical innovation is carried out to obtain an engineered and executable method for constructing a geological hazard risk assessment model; when training the model, it only needs to be calculated according to the corresponding hazard level, which is convenient for calculation and subsequent processing, and can also avoid network delays and calculation waste caused by calculation.

[0125] With respect to the above embodiment, the present invention also provides another method for assessing geological hazard level, which is implemented on the basis of the above method for assessing geological hazard level. The method focuses on describing the specific process of establishing a federated learning framework and performing model training based on the target evaluation factor, the preset training data set, the edge computing device and its monitoring data, and obtaining a geological hazard level assessment model (implemented by the following steps S308-S314), as shown in FIG. Figure 3 As shown, the method includes:

[0126] Step S302: determining a target evaluation area from an evaluation area library according to a preset screening rule.

[0127] Step S304: Determine the disaster type and target evaluation factor corresponding to the target assessment area according to the area description information of the target assessment area.

[0128] Step S306: Determine the edge computing device of the target assessment area based on the disaster type and target evaluation factor corresponding to the target assessment area.

[0129] Step S308: Establish a federated learning framework based on a preset training data set.

[0130] After the preset training data set is obtained, a federated learning framework is established according to the source characteristics of the monitoring data. Specifically, in this embodiment, the federated learning framework adopts a vertical federated learning framework.

[0131] Step S310, determining the training method of the model according to the target evaluation factor, the preset training data set and the preset data security coordination rules.

[0132] Step S312: Perform model training based on the model training method, edge computing device and federated learning framework to update the parameters of the model.

[0133] Step S314, obtaining a geological disaster risk level assessment model based on the updated model parameters.

[0134] Specifically, the hierarchical architecture of the vertical federated learning framework includes: data preprocessing layer, sample alignment layer, federated learning layer, and data application layer. This embodiment includes two participants, one of which is a cloud computing center, which provides the above-mentioned monitoring data and the above-mentioned label values ​​as the main participant, and plays the role of a collaborator in the training process, participates in the encryption and decryption of sensitive data in the above-mentioned monitoring data and coordinates the training process. The sensitive data of the above-mentioned monitoring data is determined according to actual usage requirements. The other participant is the above-mentioned Q edge computing devices, which, as a slave participant, only provides encrypted monitoring data and is only a data provider. The source of its data is determined according to the above-mentioned method, and comes from the real-time monitoring data of various sensors and the historical data of the above-mentioned established monitoring system.

[0135] The electronic device stores preset data security coordination rules, which are used to indicate the encryption or decryption method of the monitoring data. In specific implementation, the data preprocessing layer refers to integrating the monitoring data (i.e., using the disaster potential point number to distinguish the monitoring data), cleaning (i.e., if the value of the current monitoring data is less than the preset first monitoring data threshold, the monitoring data is deleted), transforming (i.e., determining the first decision value, the second decision value and the label value according to the hazard level database), constructing the preset training data set and verifying the training data set.

[0136] The sample alignment layer refers to encrypting the above monitoring data using encryption ID matching technology according to the above stored data security coordination rules before modeling of the above-mentioned participants, so as to ensure the security and privacy of the above monitoring data. In addition, the common monitoring data of each participant is also aligned, which can be achieved by encryption algorithms such as RSA asymmetric encryption algorithm.

[0137] The above-mentioned federated learning layer refers to model training through encrypted parameter exchange. To ensure privacy, a trusted collaborator should be introduced to use privacy protection technology (such as homomorphic encryption) to encrypt and decrypt the above sample monitoring data and coordinate the training process. The steps can generally be summarized as follows: the collaborator generates a key pair and sends the public key to each participant, and each participant encrypts the sample data; each participant calculates the intermediate results in an encrypted parameter exchange manner, and the intermediate results are used to calculate the gradient and loss value; each participant calculates the encrypted gradient and loss value and uploads it to the collaborator; the collaborator uses the private key to decrypt the gradient and loss value, and sends this gradient information back to each participant, and each participant updates the current model parameters based on this gradient information.

[0138] The above-mentioned data application layer refers to the process in which each participant updates the parameters of the local training model after training at the federated learning layer and outputs the prediction results, thereby obtaining the evaluation results of the above-mentioned target evaluation area.

[0139] Step S316, input the target assessment area into the geological hazard risk level assessment model, and output the assessment result of the target assessment area.

[0140] Another method for assessing the geological hazard level provided by an embodiment of the present invention adopts a vertical federated learning framework, which is characterized in that the sample space of the monitoring data between the participants has a large overlap, while the feature space is different, that is, the range of the monitoring data used has a large overlap (the target evaluation factor and the disaster type may be the same), and its specific content is determined according to the corresponding target evaluation factor and the disaster type, which is convenient for calculating the evaluation results when the various monitoring data are correlated. In the embodiment of the present invention, a data security coordination rule for indicating the encryption or decryption method of the monitoring data is also designed, which can ensure the use of the monitoring data and the data security of the above monitoring data during model training.

[0141] Based on the above system embodiment, the embodiment of the present invention further provides a geological disaster risk level assessment device, which is applied to an electronic device, wherein an assessment area library is stored in the electronic device, and the assessment area library contains multiple areas to be assessed, and area description information corresponding to each area to be assessed; Figure 4 As shown, the device comprises:

[0142] The region determination module 401 is used to determine a target evaluation region from an evaluation region library according to a preset screening rule.

[0143] The information determination module 402 is used to determine the disaster type and target evaluation factor corresponding to the target assessment area according to the area description information of the target assessment area.

[0144] The computing device determination module 403 is used to determine the edge computing device of the target assessment area based on the disaster type and target evaluation factor corresponding to the target assessment area; the edge computing device is used to calculate and store monitoring data corresponding to the disaster type and target evaluation factor.

[0145] The model determination module 404 is used to establish a federated learning framework and perform model training based on the target evaluation factors, the preset training data set, the edge computing device and its monitoring data to obtain a geological hazard level assessment model.

[0146] The assessment result determination module 405 is used to input the target assessment area into the geological hazard risk level assessment model and output the assessment result of the target assessment area.

[0147] A geological disaster risk level assessment device provided by an embodiment of the present invention first selects a target assessment area to be ultimately assessed from a plurality of assessment areas, then selects a target assessment factor to be ultimately assessed from a plurality of assessment factors of the target assessment area, then acquires monitoring data according to the target assessment factor and the type of disaster corresponding to the target assessment area, then processes the acquired monitoring data to obtain a training data set, establishes a federated learning framework, and uses edge computing devices to participate in model training. The assessment results are not only obtained through expert evaluation, but also combined with federated learning technology, and are based on current assessment standards as a reference, thereby solving the problems of insufficient authority of current technical assessment results and insufficient user recognition.

[0148] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for assessing geological hazard level, characterized in that: The method is applied to an electronic device, wherein an evaluation area library is stored in the electronic device, wherein the evaluation area library contains a plurality of areas to be evaluated and area description information corresponding to each of the areas to be evaluated; the method comprises: According to a preset screening rule, determining a target evaluation area from the evaluation area library; Determine the disaster type and target evaluation factor corresponding to the target evaluation area according to the area description information of the target evaluation area, wherein the target evaluation factor includes a development degree evaluation factor, a hazard degree evaluation factor and an inducing factor evaluation factor; The regional description information of the target assessment area includes a plurality of preset disaster types and a plurality of preset evaluation factors to which the target assessment area belongs; The step of determining the disaster type and target evaluation factor corresponding to the target assessment area according to the area description information of the target assessment area comprises: Determine the multiple preset disaster types included in the area description information as the disaster types corresponding to the target assessment area; evaluate the multiple preset evaluation factors belonging to the target assessment area to obtain a second scoring result corresponding to each of the preset evaluation factors; For each second scoring result corresponding to the preset evaluation factor, perform the following steps: Compare the second scoring result corresponding to each of the preset evaluation factors with a second preset scoring threshold; The preset evaluation factor corresponding to the second scoring result being higher than the second preset scoring threshold is determined as the target evaluation factor; The second scoring result includes a plurality of second sub-results; the second sub-results are used to indicate the correlation level of the impact of the current preset evaluation factor and the risk level; Comparing the second scoring result corresponding to each of the preset evaluation factors with a second preset scoring threshold; The step of determining the preset evaluation factor corresponding to the second scoring result being higher than the second preset scoring threshold as the target evaluation factor comprises: Determining a weighted average of the current preset evaluation factors based on a plurality of second sub-results corresponding to the second scoring result; The preset evaluation factor corresponding to the weighted mean being greater than the second threshold is determined as the target evaluation factor; Based on the disaster type and target evaluation factor corresponding to the target evaluation area, determining the edge computing device of the target evaluation area; the edge computing device is used to calculate and store the monitoring data corresponding to the disaster type and the target evaluation factor; The step of determining the edge computing device of the target assessment area based on the disaster type and target evaluation factor corresponding to the target assessment area includes: Determining the required number of edge computing devices based on the number of disaster types corresponding to the target assessment area and the number of established monitoring systems corresponding to the disaster types; Determine the edge computing device of the target evaluation area and its deployment location based on the location of the target evaluation area and the required number of the edge computing devices; According to the target evaluation factor, the preset training data set, the edge computing device and its monitoring data, a federated learning framework is established and model training is performed to obtain a geological hazard level assessment model; The electronic device stores a preset data security coordination rule, wherein the data security coordination rule is used to indicate an encryption or decryption method of the monitoring data; The step of establishing a federated learning framework and performing model training according to the target evaluation factor, the preset training data set, the edge computing device and its monitoring data to obtain a geological hazard level assessment model includes: Establishing a federated learning framework according to the preset training data set; Determining a training method for the model according to the target evaluation factor, the preset training data set and the preset data security coordination rule; Based on the training method of the model, the edge computing device and the federated learning framework, perform model training to update parameters of the model; According to the updated parameters of the model, the geological disaster risk level assessment model is obtained; The target assessment area is input into the geological disaster risk level assessment model, and the assessment result of the target assessment area is output.

2. The method according to claim 1, characterized in that The step of determining the target evaluation area from the evaluation area library according to the preset screening rules comprises: Evaluate each of the to-be-evaluated areas contained in the evaluation area library to obtain a first scoring result corresponding to each of the to-be-evaluated areas; Compare the first scoring result corresponding to each of the to-be-evaluated areas with a first preset scoring threshold; Determine the to-be-evaluated area corresponding to the first scoring result being higher than the first preset scoring threshold as the target evaluation area; The first scoring result includes a plurality of first sub-results; wherein each first sub-result is obtained by evaluating the area to be evaluated by different experts in the "geological disaster prevention and control technical expert database"; The first scoring result corresponding to each of the to-be-evaluated areas is compared with a first preset scoring threshold; The step of determining the to-be-evaluated area corresponding to the first scoring result being higher than the first preset scoring threshold as the target evaluation area comprises: Determining the Kendall harmony coefficient of the current first scoring result according to the multiple first sub-results corresponding to each of the first scoring results; For each of the Kendall concordance coefficients, perform the following steps: Comparing the Kendall harmony coefficient corresponding to each of the first scoring results with a first preset scoring threshold; Determining a chi-square test result of the Kendall coefficient of concordance that is above the first preset score threshold; Determine that the chi-square test result is higher than a target Kendall coefficient of harmony corresponding to a first preset test threshold; The to-be-evaluated area corresponding to the maximum value of the chi-square test result is determined as the target evaluation area.

3. The method according to claim 1, characterized in that The electronic device stores a danger level database, which contains danger level indicators corresponding to the monitoring data; there are multiple target evaluation factors, which include development degree evaluation factors, hazard degree evaluation factors and inducing factor evaluation factors; The training data set is determined in the following manner: Comparing the monitoring data of each target evaluation factor with a preset monitoring data threshold; Determining target monitoring data that is greater than the preset monitoring data threshold; Obtaining a first decision value indicated by the target monitoring data according to the target monitoring data and the danger level database; Based on the developmental degree evaluation factor of each target evaluation factor and the target monitoring data corresponding to the current developmental degree evaluation factor, obtaining the developmental degree target variable corresponding to the developmental degree evaluation factor; Based on the hazard level evaluation factor of each target evaluation factor and the target monitoring data corresponding to the current hazard level evaluation factor, obtaining a hazard level target variable corresponding to the hazard level evaluation factor; According to the development level target variable and the hazard level target variable, and the stored hazard level database, obtaining second determination values ​​corresponding to the development level target variable and the hazard level target variable respectively; Determining a label value corresponding to the target evaluation factor according to the second decision value and the stored danger level database; the label value is used to indicate the danger level of the target evaluation factor; The preset training data set is obtained based on the target evaluation factor, the target monitoring data, the development level target variable and the hazard level target variable, as well as the first decision value, the second decision value and the label value.

4. The method according to claim 3, characterized in that The regional description information of the target assessment area includes multiple preset disaster types and multiple preset evaluation factors to which the target assessment area belongs; there are multiple preset evaluation factors, and the electronic device stores a monitoring data source feature table, and the monitoring data source feature table includes the acquisition method and acquisition frequency of the monitoring data corresponding to each preset evaluation factor; the target monitoring data is determined by the following steps: Determine the monitoring data corresponding to the current target evaluation factor according to the monitoring data source characteristic table and each target evaluation factor; The following steps are performed for the monitoring data corresponding to each target evaluation factor: Determine whether the current value of the monitoring data is less than the preset first monitoring data threshold; If yes, delete the current monitoring data; if no, determine whether the value of the current monitoring data is greater than the preset second monitoring data threshold; If yes, the value of the current monitoring data is replaced with the mean value of the monitoring data corresponding to the current target evaluation factor, and the current monitoring data is determined as the target monitoring data; If not, the current monitoring data is directly determined as the target monitoring data.

5. A geological hazard risk level assessment device, the device being applied to an electronic device and the method described in any one of claims 1 to 4, wherein an assessment area library is stored in the electronic device, the assessment area library contains a plurality of areas to be assessed, and area description information corresponding to each of the areas to be assessed; characterized in that: The device comprises: An area determination module, used to determine a target evaluation area from the evaluation area library according to a preset screening rule; An information determination module, used to determine the disaster type and target evaluation factor corresponding to the target assessment area according to the area description information of the target assessment area; A computing device determination module, configured to determine an edge computing device in the target assessment area based on the disaster type and target assessment factor corresponding to the target assessment area; the edge computing device is configured to calculate and store monitoring data corresponding to the disaster type and the target assessment factor; A model determination module is used to establish a federated learning framework and perform model training according to the target evaluation factor, the preset training data set, the edge computing device and its monitoring data, so as to obtain a geological hazard level assessment model; The assessment result determination module is used to input the target assessment area into the geological disaster risk level assessment model and output the assessment result of the target assessment area.

Images