Reservoir landslide environmental parameter importance evaluation method and system

Abstract

The invention discloses a reservoir landslide environmental parameter importance assessment method and system, and the method comprises the steps: collecting the environmental monitoring data of a reservoir landslide region through a plurality of sensors, carrying out the preprocessing of the environmental monitoring data, and obtaining the multi-source features and feature matrixes of all environments; constructing a dynamic weight genetic algorithm, inputting the feature matrix into the dynamic weight genetic algorithm, calculating a fitness value of an individual corresponding to the multi-source features, performing crossover variation iteration on the fitness value of the individual, and when the improvement amplitude of the fitness value of continuous iteration for a plurality of times is smaller than a preset threshold value, executing crossover variation iteration on the fitness value of the individual. And ending iteration, outputting an optimal weight vector, carrying out importance calculation and normalization on the multi-source features of each environment to obtain a sorted feature importance result, and carrying out landslide risk response analysis on the collected environmental monitoring data of the reservoir landslide area. Compared with the prior art, the problem that traditional feature importance analysis is caught in local optimum is solved through the dynamic weight genetic algorithm.

Description

Technical Field

[0001] This invention relates to the field of natural disaster monitoring technology, and more specifically, to a method and system for assessing the importance of environmental parameters in reservoir landslides. Background Technology

[0002] Currently, reservoir landslide monitoring systems typically utilize various types of sensors to collect environmental parameters. This data is multi-dimensional, spans long periods, and exhibits strong nonlinearity, with complex coupling relationships between parameters. Traditional feature importance assessment methods struggle to accurately identify the comprehensive contribution of each environmental parameter to landslide deformation behavior when processing high-dimensional, nonlinear, and multi-source heterogeneous data. Furthermore, existing feature selection methods based on optimization algorithms suffer from problems such as being prone to getting trapped in local optima, slow convergence speed, and limited global search capabilities.

[0003] Existing technology discloses a landslide hazard identification method and system based on multi-source data fusion. This method employs multi-source data fusion, acquiring soil moisture, groundwater level, horizontal displacement, and vertical displacement data by deploying sensors in the landslide area. Combined with meteorological data, it uses principal component analysis and deep learning models to construct a landslide risk model, enabling real-time monitoring and assessment of landslide risk and issuing timely warnings. However, this approach has a drawback: when dealing with high-dimensional, nonlinear, and multi-source heterogeneous data, it often struggles to accurately identify the comprehensive contribution of various environmental parameters to landslide deformation behavior.

[0004] Therefore, in light of the above requirements and the shortcomings of existing technologies, this application proposes a method and system for assessing the importance of environmental parameters in reservoir landslides. Summary of the Invention

[0005] This invention provides a method and system for assessing the importance of environmental parameters in reservoir landslides, which can realize a quantitative assessment of the impact of multi-sensor environmental parameters on landslide evolution.

[0006] The primary objective of this invention is to solve the aforementioned technical problems. The technical solution of this invention is as follows: The first aspect of this invention provides a method for assessing the importance of environmental parameters in reservoir landslides, the method comprising the following steps: S1. Use several types of sensors to collect environmental monitoring data of the reservoir landslide area, preprocess the environmental monitoring data to obtain multi-source features of each environment, and obtain a feature matrix by splicing.

[0007] S2. Construct a dynamic weighted genetic algorithm, which is used to globally optimize the weights of the multi-source features. The global optimization method is to construct a fitness function based on the feature matrix and the variables used to assess landslide risk.

[0008] S3. Input the feature matrix into the dynamic weighted genetic algorithm, use the fitness function to calculate the fitness value of the individual corresponding to the multi-source features, and perform crossover mutation iteration on the fitness value of the individual in combination with the landslide risk correlation. When the fitness value improvement of several consecutive iterations is less than the preset threshold, the iteration ends and the optimal weight vector is output.

[0009] S4. Based on the optimal weight vector, the importance of the multi-source features of each environment is calculated and normalized to obtain the ranked feature importance results.

[0010] S5. Based on the aforementioned feature importance results, conduct landslide risk response analysis on the collected environmental monitoring data of the reservoir landslide area.

[0011] The second aspect of the present invention provides a reservoir landslide environmental parameter importance assessment system, which is used in the aforementioned reservoir landslide environmental parameter importance assessment method, and includes: a multi-source data acquisition and preprocessing module, a dynamic weighted genetic algorithm module, a feature fitness calculation module, a feature importance calculation and ranking module, and a landslide risk assessment module.

[0012] The multi-source data acquisition and preprocessing module collects environmental monitoring data of the reservoir landslide area using several sensors, preprocesses the environmental monitoring data to obtain multi-source features of each environment, and obtains a feature matrix by concatenation. The dynamic weighted genetic algorithm module is equipped with a dynamic weighted genetic algorithm to globally optimize the weights of the multi-source features. The feature fitness calculation module receives the feature matrix and calls the dynamic weighted genetic algorithm module to calculate the fitness value of individuals corresponding to the multi-source features. It then performs cross-mutation iteration on the fitness values ​​of individuals in combination with the landslide risk correlation. When the fitness value improvement of several consecutive iterations is less than a preset threshold, the iteration ends and the optimal weight vector is output. The feature importance calculation and ranking module calculates and normalizes the importance of the multi-source features of each environment based on the optimal weight vector to obtain the ranked feature importance results. The landslide risk assessment module outputs the landslide risk response analysis results of the environmental monitoring data of the reservoir landslide area based on the feature importance results.

[0013] Compared with the prior art, the beneficial effects of the technical solution of the present invention are: This invention provides a method and system for assessing the importance of environmental parameters in reservoir landslides. By employing a dynamic weighted genetic algorithm, it overcomes the limitation of traditional feature importance analysis, which is prone to getting trapped in local optima, and achieves quantitative contribution analysis of multi-source sensor parameters. It provides explanatory evidence for landslide mechanism research; dynamically adjusting weights and crossover mutation probabilities enhances the model's responsiveness to environmental changes; this method can be extended to multi-parameter feature importance analysis in other geological disaster scenarios; and based on the weighting results, sensor layout and landslide early warning threshold settings can be optimized, thereby improving the efficiency and reliability of the overall monitoring system. Attached Figure Description

[0014] Figure 1 This is a flowchart of a method for assessing the importance of environmental parameters in reservoir landslides according to the present invention.

[0015] Figure 2 This is a flowchart of a genetic algorithm in one embodiment of the present invention.

[0016] Figure 3 This is a schematic diagram of a reservoir landslide environmental parameter importance assessment system according to the present invention. Detailed Implementation

[0017] To better understand the above-mentioned objectives, features, and advantages of the present invention, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be noted that, unless otherwise specified, the embodiments and features described in these embodiments can be combined with each other.

[0018] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and therefore the scope of protection of the invention is not limited to the specific embodiments disclosed below.

[0019] Example 1 like Figure 1 As shown, this invention provides a method for assessing the importance of environmental parameters in reservoir landslides. This method includes the following steps: S1. Use several types of sensors to collect environmental monitoring data of the reservoir landslide area, preprocess the environmental monitoring data to obtain multi-source features of each environment, and obtain a feature matrix by splicing.

[0020] S2. Construct a dynamic weighted genetic algorithm, which is used to globally optimize the weights of the multi-source features. The global optimization method is to construct a fitness function based on the feature matrix and the variables used to assess landslide risk.

[0021] S3. Input the feature matrix into the dynamic weighted genetic algorithm, use the fitness function to calculate the fitness value of the individual corresponding to the multi-source features, and perform crossover mutation iteration on the fitness value of the individual in combination with the landslide risk correlation. When the fitness value improvement of several consecutive iterations is less than the preset threshold, the iteration ends and the optimal weight vector is output.

[0022] S4. Based on the optimal weight vector, the importance of the multi-source features of each environment is calculated and normalized to obtain the ranked feature importance results.

[0023] S5. Based on the aforementioned feature importance results, conduct landslide risk response analysis on the collected environmental monitoring data of the reservoir landslide area.

[0024] In this embodiment, several types of sensors are used to collect environmental monitoring data of the reservoir landslide area. These sensors include a rainfall monitoring sensor, a soil temperature and humidity sensor, an infrasound and ground acoustic wave sensor, a tilt sensor, a water level and pore water pressure sensor, a GNSS monitoring device, and an air humidity, temperature, and phreatic line sensor. Specifically, the rainfall monitoring sensor acquires rainfall sequences; the soil temperature and humidity sensor acquires soil moisture content and temperature changes; the infrasound and ground acoustic wave sensor acquires acoustic emission signals from underground fissures; the tilt sensor acquires changes in the slope's tilt angle; the water level and pore water pressure sensor acquires hydraulic changes in the bank slope; the GNSS monitoring device acquires bank slope surface displacement data; and the air humidity, temperature, and phreatic line sensor reflects changes in climate and seepage boundary conditions.

[0025] In step S1, the various sensors collect environmental monitoring data from the reservoir landslide area to obtain the original sequence. The original sequence is then subjected to time alignment, missing value imputation, standardization, and anomaly detection. The aligned hourly data is then used as a unified time step to convert the data into numerical features and concatenate them into a fixed-dimensional feature vector. :

[0026] The raw sequences from various sensors were converted into numerical features that could be used for model calculations, including rainfall sequences: hourly cumulative rainfall. Rainfall variation 24-hour cumulative rainfall Soil moisture content and temperature: mean moisture content Rate of change in water content Average temperature Acoustic emission signal from underground fissures: 1-hour RMS energy High-frequency energy Slope angle: mean slope angle Inclination rate of change Hydraulic parameters: pore pressure Water level , immersion line Hydraulic gradient GNSS displacement: cumulative displacement .

[0027] Then, a feature matrix is ​​constructed using time steps as the unit time scale. :

[0028] The feature matrix Each feature is a preprocessed feature vector. , Representing the real number field means that all elements in the matrix are real numbers. This represents the length of the time dimension, i.e., the number of time steps. This indicates the number of feature dimensions, i.e., the total number of types of environmental monitoring data.

[0029] like Figure 2 As shown, in step S2, the fitness function is specifically:

[0030]

[0031]

[0032]

[0033] in, This represents the feature weight of the i-th eigenvector in the feature matrix input to the dynamic weight genetic algorithm. This represents the cumulative displacement data of the bank slope surface obtained using GNSS monitoring equipment. Indicates the predicted displacement. This represents the average value of the target variable. Indicates the goodness of fit. Represents the root mean square error, regularization term To prevent overfitting, the feature weights are constrained to be too large. , , All represent weight balancing coefficients. In this embodiment, the weight balancing coefficients are... , , Used to balance accuracy and sparsity, the fitness function aims to maximize... That is, to find the optimal combination of feature weights while ensuring prediction accuracy. Let represent the feature sequence corresponding to the i-th eigenvector at time t.

[0034] In step S3, before inputting the feature matrix into the dynamic weight genetic algorithm, the feature matrix needs to be encoded and initialized. Specifically, this involves assigning weights to each feature. Perform real-number encoding to form chromosome individuals of length i, and randomly generate a population. , where N is the population size, N=50~200, each individual in the population represents a set of candidate feature weights, and the chromosome is represented as a binary vector.

[0035] The process of encoding the real number is as follows: ,in , .

[0036] In the aforementioned encoding and initialization, each column of the feature matrix corresponds to a selectable environmental feature. Therefore, the chromosome is designed as a binary vector of length d for gene selection and combined with real-valued weighted encoding to form a unified chromosome.

[0037] in, This indicates that the i-th environmental feature was selected into the model. This indicates that the characteristic is not used; such as chromosomes. This indicates that features such as rainfall, water content, and tilt angle are selected from the feature matrix X for fitness calculation. During initialization, N chromosomes are randomly generated, each corresponding to a feature combination scheme, and used for subsequent fitness calculation. For the corresponding optimizable weights of this feature, if Then the corresponding It is not included in fitness calculation.

[0038] The aforementioned real-number encoding of each feature involves further assigning weights to the features selected by binary encoding using real-number encoding. In one embodiment, the selected rainfall weight , pore pressure weight The two work together to achieve the technical effect of selecting features first and then optimizing weights, avoiding invalid features from participating in weight calculation and improving algorithm efficiency.

[0039] In step S3, when calculating the fitness value of an individual, the corresponding feature weights only participate in model prediction when the binary vector of the chromosome is 1.

[0040] in, The predicted value is a landslide risk prediction calculated based on the selected features and their weights, used to compare with the actual GNSS displacement. contrast; This is the binary vector of the chromosome; Let be the i-th feature value at time step t. After traversing the fitness values ​​of all individuals, select individuals with fitness values ​​greater than a preset threshold and perform genetic operations on them. Specifically, in each evolutionary process, a certain number of individuals are randomly selected from the parent generation, and then selected from these individuals... Genetic manipulation is performed on individuals close to 1, and this process is repeated until the offspring population is the same size as the parent population.

[0041] The genetic operations include crossover and mutation operations, and further include dynamically adjusting the crossover probability during the genetic process using a population diversity metric D. With the probability of mutation The crossover operation is based on the crossover probability. Randomly select parent pairs from the population, and then randomly select chromosomes from the selected parent pairs. and feature weights Gene exchange is performed to generate offspring individuals; the mutation operation is based on the mutation probability. Randomly select individuals from the population, and for each selected individual, randomly flip their chromosomes. The binary value or the feature weights are fine-tuned within a preset range of [-1, 1]. The process generates mutated offspring; the generation of offspring or mutated individuals is considered the completion of one iteration, and the newly generated individuals join the population and participate in the next iteration; the crossover probability during the genetic process is dynamically adjusted. With the probability of mutation Specifically:

[0042]

[0043]

[0044] in As a measure of population diversity, it represents the average dissimilarity of the current population and is a measure of the average weight of the i-th feature among all individuals. Related functions, The initial crossover probability, The initial mutation probability, This represents the upper limit of population diversity, which is dynamically set based on the historical upper limit of diversity.

[0045] In this embodiment, the gene transformation of the crossover operation can be the exchange of the rainfall characteristic weight of parent A and the pore pressure characteristic weight of parent B, and the mutation operation can be... Become Remove a certain feature; or Become Add a certain feature.

[0046] In each iteration, the feature weights also need to be updated. The specific process includes dynamically updating the feature weights by combining the correlation between landslide risk and information entropy.

[0047] in, Let be the change in the correlation between the i-th feature and the GNSS displacement; The entropy change of the i-th feature information; and In this embodiment, parameters are dynamically adjusted. , Updated feature weights Limiting the values ​​to a preset range [-1, 1], feature weights exceeding the preset range are set as boundary values ​​of the preset range, such as... When the time is right, force it to be set to 1; When the value is set to -1, it is forced to be set to -1.

[0048] The updated feature weights Directly used for fitness calculations in the next generation population:

[0049] in, This represents the predicted value for the t-th sample in the (t+1)-th generation, which is the performance output of the next generation for historical samples during the dynamic optimization process. Let represent the value of the i-th chromosome in generation t+1, i.e., the binary vector value of the i-th environmental monitoring data feature; if the optimal fitness value in generation t is... The increase would be:

[0050] in, This indicates the rate of improvement in fitness in generation t; This represents the optimal fitness value in generation t, if for k consecutive generations it satisfies If the algorithm converges, then the algorithm is considered to have converged. This represents a preset threshold. The feature weights calculated for this fitness value are then the optimal output weights. The optimal weight vector is formed by iterating through the optimal weights of all multi-source features. .

[0051] In this embodiment, the change in correlation It is calculated based on the Pearson correlation coefficient, and the formula is:

[0052] in, It is the correlation coefficient between the i-th characteristic of generation t and the GNSS displacement. This indicates that the influence of this feature on displacement is enhanced, and the weight needs to be increased.

[0053] Information entropy change It measures feature uncertainty based on information entropy, and the formula is:

[0054]

[0055] in, It is the probability of the k-th value of the i-th feature. This indicates increased characteristic fluctuations. In a specific embodiment, this could be a sudden change in rainfall, requiring an increase in weights to capture the abnormal signal.

[0056] In step S4, the optimal weight vector of the output is... Perform a normalization operation to obtain the importance index of the multi-source features of the environmental monitoring data collected by each sensor:

[0057] in, Let be the importance coefficient of the i-th sensor feature. ;according to The sensor features are sorted in descending order to obtain the relative importance ranking of different sensors in landslide risk prediction. In one embodiment, this could be... .

[0058] In this embodiment, in the converged algorithm, the chromosome with the highest fitness is the current best individual. The best individual selects which features, i.e., the chromosome locus is 1. The weights corresponding to these features are calculated by the fitness function regression model:

[0059] in, This is a predicted value; The optimal weight; For the selected feature set, the binary gene bits of the optimal individual are used. It is determined that it contains characteristics that significantly contribute to landslide risk.

[0060] In step S5, the landslide risk response analysis includes risk sensitivity analysis, multi-source feature visualization, feature weight feedback correction, and decision optimization; the risk sensitivity analysis includes: constructing risk sensitivity curves for different multi-source features, specifically:

[0061] in This represents the importance coefficient of the i-th feature. This represents the standard deviation of the characteristic; Expressed as the standard deviation of GNSS displacement; based on sensitivity The statistical distribution is divided into high contribution factors, medium contribution factors, and low contribution factors based on quartiles. If the sensitivity of any parameter in the high contribution factor increases by more than a preset value within a preset time, the warning threshold is dynamically adjusted, and the warning level is raised until the sensitivity of the parameter is lower than the preset safety value, at which point the warning threshold is restored to the initial value.

[0062] In one specific embodiment, the contribution factor interval is divided according to the following rules: As a high contribution factor, It is a medium contribution factor; It has a low contribution factor. , For this feature sensitivity sequence The median and 75th percentile of the statistical distribution over all historical periods are used to reflect the boundary between typical and high-risk levels.

[0063] The process of dynamically adjusting the warning threshold is as follows: if the sensitivity of any one of the high-contribution factors increases significantly over L consecutive time periods, the warning level is increased according to the intensity of contribution; for example, rainfall sensitivity. With pore pressure sensitivity Simultaneously entering a high contribution range can lower the triggering condition for the GNSS displacement threshold, making the early warning more sensitive. If one high contribution factor rainfall... Continuous hours to satisfy If there are two or more high-contribution factors, such as rainfall, the GNSS displacement warning threshold will be lowered by 10%; and pore pressure Enter at the same time If the threshold is lowered by 20%, an orange alert will be triggered; if the sensitivity of high contribution factors falls back to [a certain level], [the alert will be triggered]. If the threshold is not adjusted, it will be restored to its initial value to avoid over-warning. It should be noted that the threshold adjustment range is an empirically optimized value and can be adjusted based on on-site landslide sensitivity, historical displacement threshold systems, and industry standards; it is not a fixed constant.

[0064] Based on the aforementioned technical features, the risk sensitivity curve proposed in this invention is used to identify the main control factors. These main control factors influence the early warning threshold and monitoring strategy. Changes in the threshold guide a new distribution of monitoring data, which is then used as input to re-enter the previous step. The algorithm automatically updates the weights and feature importance. Through this iterative process, a continuous adaptive landslide risk assessment process is achieved.

[0065] Example 2 like Figure 3 As shown, the present invention also provides a reservoir landslide environmental parameter importance assessment system. The system is used in the aforementioned reservoir landslide environmental parameter importance assessment method and includes: a multi-source data acquisition and preprocessing module, a dynamic weighted genetic algorithm module, a feature fitness calculation module, a feature importance calculation and ranking module, and a landslide risk assessment module.

[0066] The multi-source data acquisition and preprocessing module collects environmental monitoring data of the reservoir landslide area using several sensors, preprocesses the environmental monitoring data to obtain multi-source features of each environment, and obtains a feature matrix by concatenation. The dynamic weighted genetic algorithm module is equipped with a dynamic weighted genetic algorithm to globally optimize the weights of the multi-source features. The feature fitness calculation module receives the feature matrix and calls the dynamic weighted genetic algorithm module to calculate the fitness value of individuals corresponding to the multi-source features. It then performs cross-mutation iteration on the fitness values ​​of individuals in combination with the landslide risk correlation. When the fitness value improvement of several consecutive iterations is less than a preset threshold, the iteration ends and the optimal weight vector is output. The feature importance calculation and ranking module calculates and normalizes the importance of the multi-source features of each environment based on the optimal weight vector to obtain the ranked feature importance results. The landslide risk assessment module outputs the landslide risk response analysis results of the environmental monitoring data of the reservoir landslide area based on the feature importance results.

[0067] In the embodiments provided in this application, it should be understood that the disclosed systems and methods can be implemented in other ways. Those skilled in the art will understand that all or part of the steps of the above method embodiments can be implemented by hardware related to program instructions, and the aforementioned program can be stored in a computer-readable storage medium. When the program is executed, it performs the steps of the above method embodiments; and the aforementioned storage medium includes various media capable of storing program code, such as mobile storage devices, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0068] Alternatively, if the above embodiments of the present invention are implemented as software functional modules and sold or used as independent products, they can also be stored in a computer-readable storage medium. Based on this understanding, the technical solutions of the embodiments of the present invention, or the parts that contribute to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device to execute all or part of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as mobile storage devices, ROM, RAM, magnetic disks, or optical disks.

[0069] Obviously, the above embodiments of the present invention are merely examples for clearly illustrating the present invention, and are not intended to limit the implementation of the present invention. The icons depicting structural positional relationships in the accompanying drawings are for illustrative purposes only and should not be construed as limiting the present invention. Those skilled in the art can make other variations or modifications based on the above description. It is neither necessary nor possible to exhaustively describe all embodiments here. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the scope of protection of the claims of the present invention.

Claims

1. A method for assessing the importance of environmental parameters in reservoir landslides, characterized in that, Includes the following steps: S1. Collect environmental monitoring data of the reservoir landslide area using several types of sensors, preprocess the environmental monitoring data to obtain multi-source features of each environment, and obtain a feature matrix by splicing them together. S2. Construct a dynamic weighted genetic algorithm, which is used to globally optimize the weights of the multi-source features. The global optimization method is to construct a fitness function based on the feature matrix and the variables used to assess landslide risk. S3. Input the feature matrix into the dynamic weight genetic algorithm, use the fitness function to calculate the fitness value of the individual corresponding to the multi-source features, and perform cross-mutation iteration on the fitness value of the individual in combination with the landslide risk correlation. When the fitness value improvement of several consecutive iterations is less than the preset threshold, the iteration ends and the optimal weight vector is output. S4. Based on the optimal weight vector, calculate and normalize the importance of the multi-source features of each environment to obtain the ranked feature importance results. S5. Based on the aforementioned feature importance results, conduct landslide risk response analysis on the collected environmental monitoring data of the reservoir landslide area.

2. The method for assessing the importance of environmental parameters in reservoir landslides according to claim 1, characterized in that, In step S1, the various sensors collect environmental monitoring data from the reservoir landslide area to obtain the original sequence. The original sequence is then subjected to time alignment, missing value imputation, standardization, and anomaly detection. The aligned hourly data is then used as a unified time step to convert the data into numerical features and concatenate them into a fixed-dimensional feature vector. Then, a feature matrix is ​​constructed using time steps as the unit time scale. : The feature matrix Each feature is a preprocessed feature vector. , Represents the real number field. This represents the length of the time dimension, i.e., the number of time steps. This indicates the number of feature dimensions, i.e., the total number of types of environmental monitoring data.

3. The method for assessing the importance of environmental parameters in reservoir landslides according to claim 2, characterized in that, In step S2, the fitness function is specifically: in, This represents the feature weight of the i-th eigenvector in the feature matrix input to the dynamic weight genetic algorithm. This represents the cumulative displacement data of the bank slope surface obtained using GNSS monitoring equipment. Indicates the predicted displacement. This represents the average value of the target variable. Indicates the goodness of fit. This represents the root mean square error. , , All represent weight balancing coefficients, and the objective of the fitness function is to maximize... That is, to find the optimal combination of feature weights while ensuring prediction accuracy. Let represent the feature sequence corresponding to the i-th eigenvector at time t.

4. The method for assessing the importance of environmental parameters in reservoir landslides according to claim 3, characterized in that, In step S3, before inputting the feature matrix into the dynamic weight genetic algorithm, the feature matrix needs to be encoded and initialized. Specifically, this involves assigning weights to each feature. Perform real-number encoding to form chromosome individuals of length i, and randomly generate a population. , where N is the population size, each individual in the population represents a set of candidate feature weights, and the chromosome is represented as a binary vector.

5. The method for assessing the importance of environmental parameters in reservoir landslides according to claim 4, characterized in that, In step S3, when calculating the fitness value of an individual, the corresponding feature weights only participate in model prediction when the binary vector of the chromosome is 1. in, The predicted value is a landslide risk prediction calculated based on the selected features and their weights, used to compare with the actual GNSS displacement. contrast; This is the binary vector of the chromosome; Let be the i-th feature value at time step t; after traversing the fitness values ​​of all individuals, select the individuals whose fitness values ​​are greater than the preset threshold and perform genetic operations.

6. The method for assessing the importance of environmental parameters in reservoir landslides according to claim 5, characterized in that, The genetic operations include crossover and mutation operations, and further include dynamically adjusting the crossover probability during the genetic process using a population diversity metric D. With the probability of mutation The crossover operation is based on the crossover probability. Randomly select parent pairs from the population, and then randomly select chromosomes from the selected parent pairs. and feature weights Gene exchange is performed to generate offspring individuals; the mutation operation is based on the mutation probability. Randomly select individuals from the population, and for each selected individual, randomly flip their chromosomes. The binary value or the feature weights are fine-tuned within a preset range. Generate mutated offspring; after generating offspring or mutated individuals, it is considered that one round of iteration is completed, and the newly generated individuals join the population and participate in the next round of iteration; This involves dynamically adjusting the crossover probability during the genetic process. With the probability of mutation Specifically: in As a measure of population diversity, it represents the average dissimilarity of the current population and is a measure of the average weight of the i-th feature among all individuals. Related functions, The initial crossover probability, The initial mutation probability, This represents the upper limit of population diversity, which is dynamically set based on the historical upper limit of diversity.

7. The method for assessing the importance of environmental parameters in reservoir landslides according to claim 6, characterized in that, In each iteration, the feature weights also need to be updated. The specific process includes dynamically updating the feature weights by combining the correlation between landslide risk and information entropy. in, Let be the change in the correlation between the i-th feature and the GNSS displacement; The entropy change of the i-th feature information; and For dynamically adjusted parameters; updated feature weights The updated feature weights are limited to a preset numerical range, and feature weights exceeding the preset numerical range are set as boundary values ​​of the preset numerical range. Directly used for fitness calculations in the next generation population: in, This represents the predicted value for the t-th sample in the (t+1)-th generation, which is the performance output of the next generation for historical samples during the dynamic optimization process. Let represent the value of the i-th chromosome in generation t+1, i.e., the binary vector value of the i-th environmental monitoring data feature; if the optimal fitness value in generation t is... The increase would be: in, This indicates the rate of improvement in fitness in generation t; This represents the optimal fitness value in generation t, if for k consecutive generations it satisfies If the algorithm converges, then the algorithm is considered to have converged. This represents a preset threshold. The feature weights calculated for this fitness value are then the optimal output weights. The optimal weight vector is formed by iterating through the optimal weights of all multi-source features. .

8. The method for assessing the importance of environmental parameters in reservoir landslides according to claim 7, characterized in that, In step S4, the optimal weight vector of the output is... Perform a normalization operation to obtain the importance index of the multi-source features of the environmental monitoring data collected by each sensor: in, Let be the importance coefficient of the i-th sensor feature. ;according to By sorting the sensor features in descending order, the relative importance of different sensors in landslide risk prediction is obtained.

9. The method for assessing the importance of environmental parameters in reservoir landslides according to claim 8, characterized in that, In step S5, the landslide risk response analysis includes risk sensitivity analysis, multi-source feature visualization, feature weight feedback correction, and decision optimization; the risk sensitivity analysis includes: constructing risk sensitivity curves for different multi-source features, specifically: in This represents the importance coefficient of the i-th feature. This represents the standard deviation of the characteristic; Expressed as the standard deviation of GNSS displacement; based on sensitivity The statistical distribution is divided into high contribution factors, medium contribution factors, and low contribution factors based on quartiles. If the sensitivity of any parameter in the high contribution factor increases by more than a preset value within a preset time, the warning threshold is dynamically adjusted, and the warning level is raised until the sensitivity of the parameter is lower than the preset safety value, at which point the warning threshold is restored to the initial value.

10. A system for assessing the importance of environmental parameters in reservoir landslides, the system being used in the method for assessing the importance of environmental parameters in reservoir landslides as described in any one of claims 1-9, characterized in that, It includes: a multi-source data acquisition and preprocessing module, a dynamic weighted genetic algorithm module, a feature fitness calculation module, a feature importance calculation and ranking module, and a landslide risk assessment module; The multi-source data acquisition and preprocessing module collects environmental monitoring data from the reservoir landslide area using several sensors, preprocesses the environmental monitoring data to obtain multi-source features of each environment, and concatenates them to obtain a feature matrix. The dynamic weight genetic algorithm module is equipped with a dynamic weight genetic algorithm, which is used to globally optimize the weights of the multi-source features. The feature fitness calculation module receives the feature matrix and calls the dynamic weight genetic algorithm module to calculate the fitness value of the individuals corresponding to the multi-source features. It then performs cross-mutation iteration on the fitness value of the individuals in combination with the landslide risk correlation. When the fitness value improvement of several consecutive iterations is less than a preset threshold, the iteration ends and the optimal weight vector is output. The feature importance calculation and ranking module calculates and normalizes the importance of the multi-source features of each environment based on the optimal weight vector, and obtains the ranked feature importance results. Based on the importance of the features, the landslide risk assessment module outputs the landslide risk response analysis results from the environmental monitoring data of the reservoir landslide area.

Images