A neural network-based power transmission and transformation project slope stability evaluation method and system

Abstract

A kind of neural network-based power transmission project slope stability evaluation method, comprising: obtaining the original three-dimensional point cloud data of power transmission project slope, collecting the original geological meteorological data of power transmission project slope;According to the original three-dimensional point cloud data, generate the three-dimensional model of slope, the original geological meteorological data is standardized, and the geological meteorological data is obtained;From the three-dimensional model of slope, the geometric feature of slope is extracted, surface morphology feature, combined with geometric feature, surface morphology feature, geological meteorological data constructs the feature vector of slope;Based on recurrent neural network, construct the slope stability prediction evaluation model, the feature vector of slope is identified by the slope stability prediction evaluation model, and the slope stability evaluation result is obtained, the stability grade of slope is judged according to the evaluation result, corresponding preventive measures are taken based on stability grade.The present application improves the accuracy of slope stability evaluation, reduces the risk of accident, and promotes the intelligentization of slope monitoring and management.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of slope stability evaluation, and particularly relates to a method and system for evaluating slope stability of power transmission and transformation projects based on a neural network. BACKGROUND

[0002] In modern power transmission and transformation projects, slope stability evaluation is an important link to ensure the safety and reliable operation of the project. Although existing technologies have introduced some advanced tools and methods, such as monocular vision-based image processing technology and other automated equipment, there are still some significant defects.

[0003] Firstly, although these technologies improve the recognition efficiency, due to the lack of depth information, it is difficult to conduct accurate three-dimensional spatial quantitative analysis, which limits the accuracy of the evaluation results and makes them susceptible to light changes and occlusions in complex environments, leading to instability of the recognition results. Existing technologies still rely on limited field data, which cannot fully reflect the actual situation of the slope, thereby affecting the reliability of the evaluation results. SUMMARY

[0004] The present application aims to overcome the problem of over-reliance on limited field data in the evaluation of slope stability in power transmission and transformation projects, which cannot fully reflect the actual situation of the slope, and provides a method and system for evaluating slope stability in power transmission and transformation projects based on a neural network.

[0005] To achieve the above objectives, the technical solution of the present application is as follows:

[0006] In a first aspect, the present application provides a method for evaluating slope stability in power transmission and transformation projects based on a neural network, comprising the following steps:

[0007] S1, using a three-dimensional laser scanner to scan the slope of the power transmission and transformation project, obtaining the original three-dimensional point cloud data of the slope of the power transmission and transformation project, and collecting the original geological and meteorological data of the slope of the power transmission and transformation project;

[0008] S2, denoising, filtering and reconstructing the original three-dimensional point cloud data to obtain a three-dimensional model of the slope, and standardizing the original geological and meteorological data to obtain geological and meteorological data;

[0009] S3, extracting the geometric features and surface morphology features of the slope of the power transmission and transformation project from the three-dimensional model of the slope, and constructing a slope feature vector combining the geometric features, the surface morphology features and the geological and meteorological data;

[0010] S4, constructing a slope stability prediction and evaluation model based on a recurrent neural network, identifying the slope feature vector through the slope stability prediction and evaluation model to obtain a slope stability evaluation result, judging the stability level of the power transmission and transformation project slope according to the slope stability evaluation result, and taking corresponding preventive measures based on the stability level.

[0011] Preferably, based on the above technical solutions, step S1 comprises:

[0012] S11, selecting a plurality of optimal scanning positions according to the topographic features of the power transmission and transformation project slope, setting up three-dimensional laser scanners at each optimal scanning position, performing horizontal calibration and angle adjustment on each three-dimensional laser scanner, and setting scanning parameters of the three-dimensional laser scanners, wherein the scanning parameters include scanning resolution, scanning range and scanning frequency, so as to obtain original three-dimensional point cloud data of the power transmission and transformation project slope;

[0013] S12, determining the rock-soil type and stratum structure of the power transmission and transformation project slope through drilling sampling, installing a groundwater level monitoring device, regularly recording groundwater level data of the power transmission and transformation project slope through the groundwater level monitoring device, and monitoring meteorological data of the power transmission and transformation project slope in real time through an automatic weather station, and combining the rock-soil type, the stratum structure, the groundwater level data and the meteorological data to obtain original geological and meteorological data of the power transmission and transformation project slope.

[0014] Preferably, based on the above technical solutions, step S2 comprises:

[0015] S21, performing denoising processing on the original three-dimensional point cloud data through a statistical outlier detection method, eliminating abnormal points and noise points of the original three-dimensional point cloud data to obtain denoised three-dimensional point cloud data, performing smoothing processing on the denoised three-dimensional point cloud data by using an adaptive bilateral filtering algorithm to obtain smoothed three-dimensional point cloud data, and performing three-dimensional reconstruction on the smoothed three-dimensional point cloud data based on a Poisson reconstruction algorithm to obtain a slope three-dimensional model;

[0016] S22, performing normalization processing on the original geological and meteorological data to obtain normalized geological and meteorological data under a unified scale, identifying abnormal values in the normalized geological and meteorological data through a quartile method, performing interpolation on the abnormal values to obtain interpolated geological and meteorological data, and performing feature extraction and dimension reduction on the interpolated geological and meteorological data by using a hybrid principal component analysis method to obtain standardized processed geological and meteorological data.

[0017] Preferably, based on the above technical solutions, the denoising processing on the original three-dimensional point cloud data through the statistical outlier detection method comprises:

[0018] Calculate a statistical outlier of each data point in the original three-dimensional point cloud data, and remove an outlier in the original three-dimensional point cloud data according to the statistical outlier;

[0019] The calculation formula of the statistical outlier of each data point in the original three-dimensional point cloud data is:

[0020]

[0021] Wherein, Z i is the statistical outlier of the i th data point in the original three-dimensional point cloud data, X i is the i th data point in the original three-dimensional point cloud data, μ is the mean value of the original three-dimensional point cloud data, σ is the standard deviation of the original three-dimensional point cloud data, and n is the number of data points of the original three-dimensional point cloud data.

[0022] The mixed principal component analysis method is used for feature extraction and dimension reduction of the interpolated geological meteorological data, including:

[0023] The principal component analysis method is used for dimension reduction of the interpolated geological meteorological data, and the principal component features of the interpolated geological meteorological data are obtained.

[0024] The independent component analysis method is used for processing the principal component features, and the standardized processed geological meteorological data is obtained.

[0025] The calculation formula of the geological meteorological data is:

[0026] F=w1PC j +w2IC k ;

[0027] Wherein, F is the geological meteorological data, w1 and w2 are the weight of the principal component analysis method and the weight of the independent component analysis method, PC j and IC k are the j th principal component feature of the principal component analysis method and the k th independent component feature of the independent component analysis method, respectively.

[0028] On the basis of the above technical scheme, preferably, the step S3 comprises:

[0029] S31, the three-dimensional grid division algorithm is used to divide the slope three-dimensional model into a plurality of grid units, the local slope, the local slope direction and the local elevation value of each grid unit are calculated, the curvature feature of the surface of the slope three-dimensional model is calculated based on the Gaussian curvature and the average curvature, the roughness index of the slope surface is calculated by using the multi-scale analysis method, and the distribution map of the geometric feature and the surface morphology feature of the whole slope of the power transmission project is generated by statistical analysis and spatial interpolation method.

[0030] S32, normalize the geometric features and the surface morphology features to obtain first geometric features and first surface morphology features, perform dimension reduction on the first geometric features and the first surface morphology features by using a principal component analysis method, extract principal component features of the first geometric features and the first surface morphology features to obtain second geometric features and second surface morphology features, splice the second geometric features and the second surface morphology features with the geologic and meteorological data, introduce an attention mechanism, give different weights to different features, and construct a multi-level slope feature vector, the slope feature vector including global features and local features.

[0031] Based on the above technical solutions, preferably, step S4 comprises:

[0032] S41, constructing an initial slope stability prediction and evaluation model based on a recurrent neural network, initializing model parameters including weight matrices and bias vectors, setting training strategies and hyperparameters including learning rates, batch sizes, and iteration numbers;

[0033] S42, obtaining historical slope stability data, and training the initial slope stability prediction and evaluation model by using the historical slope stability data through a gradient descent algorithm to obtain a slope stability prediction and evaluation model;

[0034] S43, preprocessing the slope feature vector to obtain a unified slope feature vector consistent with the data format of the historical slope stability data, inputting the unified slope feature vector into the slope stability prediction and evaluation model to perform stability prediction, obtaining a prediction result, and smoothing the prediction result by using a post-processing algorithm to obtain a slope stability evaluation result;

[0035] S44, mapping the slope stability evaluation result to a corresponding stability level according to a preset stability level division standard, formulating corresponding prevention and control measures for different stability levels, notifying relevant management personnel, implementing the corresponding prevention and control measures, establishing a slope monitoring and feedback mechanism, and monitoring the stability level change of the power transmission and transformation engineering slope in real time.

[0036] Based on the above technical solutions, preferably, step S4 further comprises:

[0037] The state update calculation formula of the initial slope stability prediction and evaluation model is:

[0038] h t =tanh(W x x t +W h h t-1 +b);

[0039] wherein, h tis a hidden state vector at time t, x t is an input feature vector at time t, h t-1 is a hidden state vector at time t-1, W x and W h are a weight of the input feature vector at time t and a weight of the hidden state vector at time t-1 respectively, b is a bias vector, and tanh(·) is a hyperbolic tangent activation function;

[0040] The loss function calculation formula of the initial slope stability prediction and evaluation model is:

[0041]

[0042] Wherein, L is a loss function value, N is a sample quantity of the initial slope stability prediction and evaluation model, y l is a true value of the lth sample of the initial slope stability prediction and evaluation model, is a predicted value of the lth sample of the initial slope stability prediction and evaluation model, w l is a weight of the lth sample of the initial slope stability prediction and evaluation model, λ is a regularization coefficient, and θ is a model parameter vector, is an L2 norm square of θ;

[0043] The calculation formula of the gradient descent algorithm is:

[0044]

[0045] θ (m+1) =θ (m) +v (m+1) ;

[0046] Wherein, v (m+1) and v (m) are a velocity vector of the m+1th iteration and a velocity vector of the mth iteration respectively, η is a momentum coefficient, τ is a learning rate, is a gradient of the loss function to θ (m) , θ (m+1) and θ (m) are a parameter vector of the m+1th iteration and a parameter vector of the mth iteration respectively.

[0047] In the second aspect, the application provides a neural network-based slope stability evaluation system for power transmission and transformation projects, which comprises:

[0048] A data acquisition module is configured to scan a slope of a power transmission and transformation project by using a three-dimensional laser scanner, acquire original three-dimensional point cloud data of the slope of the power transmission and transformation project, and collect original geological and meteorological data of the slope of the power transmission and transformation project.

[0049] a three-dimensional model module, configured to denoise, filter and reconstruct the original three-dimensional point cloud data to obtain a three-dimensional model of the slope, and to standardize the original geological and meteorological data to obtain geological and meteorological data;

[0050] a feature vector module, configured to extract geometric features and surface morphology features of the slope of the power transmission and transformation project from the three-dimensional model of the slope, and to construct a slope feature vector in combination with the geometric features, the surface morphology features and the geological and meteorological data;

[0051] a stability evaluation module, configured to construct a slope stability prediction and evaluation model based on a recurrent neural network, to identify the slope feature vector through the slope stability prediction and evaluation model to obtain a slope stability evaluation result, to determine a stability level of the slope of the power transmission and transformation project according to the slope stability evaluation result, and to take corresponding preventive measures based on the stability level.

[0052] In a third aspect, the present application further provides an electronic device, comprising at least one processor, at least one memory, a communication interface and a bus;

[0053] The processor, the memory and the communication interface can communicate with each other through the bus, the memory stores program instructions executable by the processor, and the processor invokes the program instructions to implement the steps of the method.

[0054] In a fourth aspect, the present application further provides a computer readable storage medium storing computer instructions, which enable a computer to implement the steps of the method.

[0055] Compared with the prior art, the present application has the following advantages:

[0056] 1. In the method, high-precision three-dimensional point cloud data is obtained by using a three-dimensional laser scanner, and geometric features and surface morphology features are extracted in combination with geological and meteorological data to construct a slope feature vector, thereby enhancing comprehensive understanding of slope features, fully reflecting the actual situation of the slope, denoising, filtering and reconstructing the data to ensure the accuracy of the data, constructing a three-dimensional model of the slope, standardizing the geological and meteorological data to improve the consistency of the data, predicting the stability based on a stability prediction and evaluation model to improve the accuracy of the evaluation, and taking preventive measures in a timely manner through a real-time monitoring and feedback mechanism to reduce safety risks, improve the accuracy of slope stability evaluation, reduce accident risks and promote the intelligentization of slope monitoring and management.

[0057] 2、The kind of neural network based on power transmission and transformation project slope stability evaluation method in the application, through the statistical outlier detection method is used to the original three-dimensional point cloud data is handled to the noise, and the noise point is removed, and the adaptive bilateral filtering algorithm is used to the three-dimensional point cloud data after denoising is handled to the smooth, the three-dimensional point cloud data after smoothing is reconstructed based on Poisson reconstruction algorithm, the three-dimensional model of slope is obtained, the original geological and meteorological data is handled to the normalization, the abnormal value identification and interpolation, the feature extraction and dimension reduction are carried out to the mixed principal component analysis method, and the standardized geological and meteorological data is obtained, the accuracy of slope feature extraction and stability evaluation is improved.

[0058] 3、The kind of neural network based on power transmission and transformation project slope stability evaluation method in the application, the gradient descent algorithm is used to train and optimize the initial model, and the slope stability prediction evaluation model is obtained, so as to identify the slope feature vector, and the prediction result of the slope stability is given, the post-processing algorithm is used to smooth the prediction result, and the final slope stability evaluation result is obtained, the slope stability is divided into different grades according to the evaluation result, corresponding prevention and control measures are formulated for different grades, and the monitoring and feedback mechanism is established, the slope stability change is monitored in real time, and the slope stability is accurately predicted. BRIEF DESCRIPTION OF DRAWINGS

[0059] Figure 1 It is a flowchart of the neural network based on power transmission and transformation project slope stability evaluation method provided by the embodiment of the application.

[0060] Figure 2 It is a structure schematic view of the neural network based on power transmission and transformation project slope stability evaluation system provided by the embodiment of the application.

[0061] Figure 3 It is a structure schematic view of the electronic equipment provided by the embodiment of the application. DETAILED DESCRIPTION

[0062] The technical scheme in the embodiment of the application will be clearly and completely described below in combination with the embodiment of the application. Obviously, the described embodiment is only a part of the embodiment of the application, not all the embodiments. Based on the embodiment of the application, all other embodiments obtained by those skilled in the art without creative labor belong to the protection scope of the application.

[0063] Please refer to Figure 1 The application provides a neural network based on power transmission and transformation project slope stability evaluation method, which comprises the following steps:

[0064] S1, scanning the power transmission and transformation project slope by using a three-dimensional laser scanner, obtaining original three-dimensional point cloud data of the power transmission and transformation project slope, and collecting original geological and meteorological data of the power transmission and transformation project slope;

[0065] S2, denoising, filtering and reconstructing the original three-dimensional point cloud data to obtain a slope three-dimensional model, and standardizing the original geological and meteorological data to obtain geological and meteorological data;

[0066] S3, extracting geometric features and surface morphology features of the power transmission and transformation project slope from the slope three-dimensional model, and constructing a slope feature vector combining the geometric features, the surface morphology features and the geological and meteorological data;

[0067] S4, constructing a slope stability prediction and evaluation model based on a recurrent neural network, identifying the slope feature vector through the slope stability prediction and evaluation model to obtain a slope stability evaluation result, judging the stability grade of the power transmission and transformation project slope according to the slope stability evaluation result, and taking corresponding preventive measures based on the stability grade.

[0068] Specifically, the embodiment obtains high-precision three-dimensional point cloud data by using a three-dimensional laser scanner, extracts geometric features and surface morphology features in combination with geological and meteorological data, constructs a slope feature vector, enhances comprehensive understanding of slope features, comprehensively reflects the actual situation of the slope, denoises, filters and reconstructs the data to ensure the accuracy of the data, constructs a slope three-dimensional model, standardizes the geological and meteorological data to improve the consistency of the data, predicts the stability based on a stability prediction and evaluation model to improve the accuracy of the evaluation, and takes preventive measures in a timely manner through a real-time monitoring and feedback mechanism to reduce safety risks, improve the accuracy of slope stability evaluation, reduce accident risks, and promote the intelligentization of slope monitoring and management.

[0069] Step S1 includes:

[0070] S11, selecting a plurality of optimal scanning positions according to the topographic features of the power transmission and transformation project slope, setting a three-dimensional laser scanner at each optimal scanning position, horizontally calibrating and angle adjusting each three-dimensional laser scanner, and setting scanning parameters of the three-dimensional laser scanner, the scanning parameters including scanning resolution, scanning range and scanning frequency, to obtain the original three-dimensional point cloud data of the power transmission and transformation project slope;

[0071] S12, determine the rock-soil type and stratum structure of the power transmission and transformation project slope through drilling sampling, install groundwater level monitoring equipment, record the groundwater level data of the power transmission and transformation project slope regularly through the groundwater level monitoring equipment, monitor the weather data of the power transmission and transformation project slope in real time through an automatic weather station, and obtain the original geological and meteorological data of the power transmission and transformation project slope in combination with the rock-soil type, the stratum structure, the groundwater level data and the weather data.

[0072] Specifically, the embodiment can make the three-dimensional laser scanner obtain more comprehensive and accurate original three-dimensional point cloud data of the power transmission and transformation project slope by selecting multiple optimal scanning positions and performing horizontal calibration and angle adjustment, and can set scanning parameters including resolution, range and frequency reasonably to ensure that the obtained three-dimensional point cloud data has sufficient detail information and time resolution.

[0073] The rock-soil type and stratum structure of the slope are determined through drilling sampling, the geological conditions of the slope are better understood, the groundwater level and weather data of the slope are obtained in real time by installing the groundwater level monitoring equipment and the automatic weather station, and key environmental factor data for slope stability evaluation are provided. Ultimately, the rock-soil type, the stratum structure, the groundwater level and the weather data are comprehensively utilized to construct more comprehensive and accurate original geological and meteorological data.

[0074] Step S2 includes:

[0075] S21, the original three-dimensional point cloud data is denoised by a statistical outlier detection method, abnormal points and noise points in the original three-dimensional point cloud data are removed, denoised three-dimensional point cloud data is obtained, the denoised three-dimensional point cloud data is smoothed by using an adaptive bilateral filtering algorithm, smoothed three-dimensional point cloud data is obtained, and a three-dimensional model of the slope is obtained by three-dimensional reconstruction of the smoothed three-dimensional point cloud data based on a Poisson reconstruction algorithm.

[0076] S22, the original geological and meteorological data is normalized to obtain normalized geological and meteorological data under a unified scale, the abnormal values in the normalized geological and meteorological data are identified by using a quartile method, the abnormal values are interpolated to obtain interpolated geological and meteorological data, and the interpolated geological and meteorological data is subjected to feature extraction and dimension reduction by using a hybrid principal component analysis method to obtain standardized geological and meteorological data.

[0077] Specifically, the original three-dimensional point cloud data is denoised by a statistical outlier detection method, abnormal points and noise points are removed, the denoised three-dimensional point cloud data is smoothed by using an adaptive bilateral filtering algorithm, noise in the data is reduced, and a more real and complete three-dimensional model of the slope is generated by three-dimensional reconstruction of the smoothed three-dimensional point cloud data based on a Poisson reconstruction algorithm.

[0078] The original geological meteorological data is mapped to a unified scale range through normalization processing, the influence of dimensional difference is eliminated, the abnormal values in the normalized geological meteorological data are identified and interpolated by using the quartile method, the influence of noise on the analysis result is reduced, and the interpolated geological meteorological data is subjected to feature extraction and dimension reduction by using the mixed principal component analysis method, so that the potential features of the data are captured.

[0079] The step S2 further includes:

[0080] The calculation formula of the statistical outlier detection method is:

[0081]

[0082] Wherein, Z i is the statistical outlier value of the i th data point in the original three-dimensional point cloud data, X i is the i th data point in the original three-dimensional point cloud data, μ is the mean value of the original three-dimensional point cloud data, σ is the standard deviation of the original three-dimensional point cloud data, and n is the number of data points of the original three-dimensional point cloud data.

[0083] The calculation formula of the normalization processing is:

[0084]

[0085] Wherein, Y norm is the data point of the normalized geological meteorological data, Y i is the i th data point of the normalized geological meteorological data, Y min is the minimum data point of the normalized geological meteorological data, Y max is the maximum data point of the normalized geological meteorological data, and α is a non-zero positive number.

[0086] The mixed principal component analysis method includes:

[0087] The interpolated geological meteorological data is subjected to dimension reduction by using the principal component analysis method, the principal component features of the interpolated geological meteorological data are obtained, the principal component features are processed by using the independent component analysis method, and the normalized geological meteorological data is obtained, and the calculation formula of the geological meteorological data is:

[0088] F=w1PC j +w2IC k ;

[0089] Wherein, F is the geological meteorological data, w1 and w2 are the weight of the principal component analysis method and the weight of the independent component analysis method respectively, PC j and IC k are the j th principal component feature of the principal component analysis method and the k th independent component feature of the independent component analysis method respectively.

[0090] Specifically, the embodiment identifies outliers by calculating the relationship of each data point with the mean and standard deviation of the data set, uses statistical methods to improve the accuracy of outlier detection, and avoids misjudging normal data as abnormal. For example, the embodiment can introduce more complex statistical models (such as Z-score or IQR method) to enhance the ability to identify outliers.

[0091] The normalization processing formula ensures that different features have the same influence in model training by scaling the data to a unified scale. A more flexible normalization method, such as Z-score standardization or Min-Max scaling, is used to adapt to the characteristics of different data distributions and reduce the instability of model training caused by differences in feature scales.

[0092] By combining principal component analysis and independent component analysis, the principal component features and independent component features of the original three-dimensional point cloud data are extracted. After dimensionality reduction using principal component analysis, the correlation between features is removed using independent component analysis, thereby obtaining more independent feature representations, improving the effectiveness of feature extraction, and enhancing the generalization ability of the model.

[0093] The embodiment improves the quality of input data through outlier detection and processing, reduces the impact of noise on model training, and ensures that all features are on the same scale through normalization processing, reducing the instability of training caused by differences in feature scales. The application of mixed principal component analysis makes the feature vector more representative and better captures the internal structure of the data. Through accurate slope stability evaluation results, more reliable decision support is provided for the safety management of power transmission and transformation projects, and potential risks are reduced.

[0094] Step S3 includes:

[0095] S31, divide the slope three-dimensional model into a plurality of grid cells by a three-dimensional grid division algorithm, calculate the local slope, local slope direction and local elevation value of each grid cell, calculate the curvature feature of the surface of the slope three-dimensional model based on Gaussian curvature and average curvature, calculate the roughness index of the slope surface by a multi-scale analysis method, and generate the distribution map of the overall geometric features and surface morphology features of the power transmission and transformation project slope by statistical analysis and spatial interpolation method;

[0096] S32, normalize the geometric features and the surface morphology features to obtain first geometric features and first surface morphology features, reduce dimensions of the first geometric features and the first surface morphology features by using a principal component analysis method, extract principal component features of the first geometric features and the first surface morphology features to obtain second geometric features and second surface morphology features, splice the second geometric features and the second surface morphology features with the geologic and meteorological data, introduce an attention mechanism, give different weights to different features, and construct a multi-level slope feature vector, wherein the slope feature vector comprises global features and local features.

[0097] Specifically, the slope three-dimensional model is divided into a plurality of grid units by a three-dimensional grid division algorithm, the local slope, local slope direction and local elevation value of each grid unit are calculated, the geometric features of the slope are comprehensively captured, the curvature features of the surface of the slope three-dimensional model are calculated based on the Gaussian curvature and the average curvature, the roughness index of the surface of the slope is calculated by using a multi-scale analysis method, the surface morphology features of the slope can be extracted, and the distribution maps of the geometric features and the surface morphology features of the slope are generated by statistical analysis and spatial interpolation methods.

[0098] The extracted geometric features and surface morphology features are normalized, and the principal component analysis method is used for dimension reduction to obtain more effective feature vectors. The geometric features, surface morphology features and geologic and meteorological data are spliced to construct a multi-level slope feature vector, and an attention mechanism is introduced to give different weights to different features, thereby enhancing the expression ability of the features.

[0099] Step S4 comprises:

[0100] S41, constructing an initial slope stability prediction and evaluation model based on a recurrent neural network, initializing model parameters, the model parameters comprising a weight matrix and a bias vector, setting a training strategy and hyperparameters, the hyperparameters comprising a learning rate, a batch size and an iteration number;

[0101] S42, obtaining historical slope stability data, and training the initial slope stability prediction and evaluation model by using the historical slope stability data by a gradient descent algorithm to obtain a slope stability prediction and evaluation model;

[0102] S43, preprocessing the slope feature vector to obtain a unified slope feature vector consistent with the data format of the historical slope stability data, inputting the unified slope feature vector into the slope stability prediction and evaluation model, performing stability prediction to obtain a prediction result, and smoothing the prediction result by a post-processing algorithm to obtain a slope stability evaluation result;

[0103] S44, according to the preset stability grade division standard, the slope stability evaluation result is mapped to the corresponding stability grade, the corresponding prevention and control measures are formulated for different stability grades, the relevant management personnel are informed, the corresponding prevention and control measures are implemented, the slope monitoring and feedback mechanism is established, and the stability grade change of the slope of the power transmission and transformation project is monitored in real time.

[0104] Specifically, the parameters of the recurrent neural network model, including the weight matrix and the bias vector, are initialized, and reasonable training strategies and hyperparameters, including the learning rate, the batch size and the iteration number, are set, to ensure the effectiveness of the initial slope stability prediction and evaluation model. The gradient descent algorithm is used to train the initial slope stability prediction and evaluation model combined with historical slope stability data. The momentum term is introduced into the gradient descent algorithm to accelerate the convergence process of the model and improve the training efficiency. The input feature vector is preprocessed to make it consistent with the format of the historical slope stability data. The post-processing algorithm is used to smooth the model prediction results, which can effectively reduce the volatility of the prediction results.

[0105] According to the slope stability evaluation result mapped to different stability grades, the corresponding prevention and control measures are formulated, the slope monitoring and feedback mechanism is established, the slope stability grade change is monitored in real time, the effective prevention measures are ensured in time, and the safety of the project is improved.

[0106] In a specific embodiment, the corresponding prevention and control measures are formulated for different stability grades.

[0107] According to the slope stability evaluation result, the slope stability is divided into different grades such as high, medium and low.

[0108] For the slope with high stability grade, regular patrol and vegetation maintenance are taken as the conventional prevention measures, for the slope with medium stability grade, slope support and drainage system maintenance are taken as the enhanced prevention measures, and for the slope with low stability grade, slope reinforcement and slope reshaping are taken as the control prevention measures.

[0109] And the relevant management personnel are informed to implement the corresponding prevention and control measures.

[0110] The slope stability evaluation result and the corresponding prevention and control measures are timely informed to the management department and maintenance personnel of the power transmission and transformation project, the relevant personnel are organized to formulate specific prevention and control plans, and the corresponding prevention and control measures are implemented.

[0111] A slope monitoring and feedback mechanism is established to monitor the slope stability grade change in real time.

[0112] Monitoring equipment is installed at key positions of the slope, including inclination monitors, displacement monitors and the like, to monitor indicators such as deformation and displacement of the slope in real time, to establish an automatic collection, analysis and early warning mechanism for slope monitoring data, to timely discover changes in the stability level of the slope, and to immediately notify relevant management personnel once the stability level of the slope changes, and to adjust preventive control measures according to the changes.

[0113] Step S4 further includes:

[0114] The state update calculation formula of the initial slope stability prediction and evaluation model is:

[0115] h t =tanh(W x x t +W h h t-1 +b);

[0116] where h t is a hidden state vector at time t, x t is an input feature vector at time t, h t-1 is a hidden state vector at time t-1, W x and W h are a weight of the input feature vector at time t and a weight of the hidden state vector at time t-1, respectively, b is a bias vector, and tanh(·) is a hyperbolic tangent activation function.

[0117] The loss function calculation formula of the initial slope stability prediction and evaluation model is:

[0118]

[0119] where L is a loss function value, N is a sample quantity of the initial slope stability prediction and evaluation model, y l is a true value of the lth sample of the initial slope stability prediction and evaluation model, is a predicted value of the lth sample of the initial slope stability prediction and evaluation model, w l is a weight of the lth sample of the initial slope stability prediction and evaluation model, λ is a regularization coefficient, θ is a model parameter vector, is an L2 norm square of θ;

[0120] The calculation formula of the gradient descent algorithm is:

[0121]

[0122] θ (m+1) =θ (m) +v (m+1) ;

[0123] where v (m+1) and v(m) are the velocity vectors of the m+1th iteration and the mth iteration, respectively, η is the momentum coefficient, τ is the learning rate, is the loss function for θ (m) The gradient of θ (m+1) and θ (m) are the parameter vectors for the m+1th iteration and the parameter vector for the mth iteration respectively;

[0124] The calculation formula of the post-processing algorithm is:

[0125]

[0126] in, is the slope stability assessment result at time T, is the original predicted value at time T, is the smoothed prediction value at time T-1, and β is the smoothing coefficient.

[0127] Specifically, the state update calculation formula of this embodiment introduces the input feature vector at the current moment and the hidden state vector at the previous moment, combines weights and biases, updates the hidden state of the model, and captures the dynamic characteristics in the time series data.

[0128] In this embodiment, the hyperbolic tangent activation function tanh(·) is used. Its nonlinear characteristics can effectively handle the nonlinear relationship of the input data and enhance the expressiveness of the model. The loss function is used to measure the difference between the model prediction value and the true value. By introducing a regularization term, the model is prevented from overfitting and the generalization ability of the model is improved. The L2 norm squared regularization is used to control the model complexity and ensure that the model performs balanced on the training set and the test set.

[0129] The gradient descent algorithm of this embodiment accelerates convergence and reduces oscillations by introducing a momentum term. The momentum coefficient is used to allow the model to take into account the previous update direction when updating parameters, thereby improving the convergence speed. Combined with the learning rate and gradient information, the step size of the parameter update is dynamically adjusted, making the optimization process more efficient.

[0130] The post-processing algorithm smoothes the prediction results through the smoothing coefficient, reduces the fluctuation of the prediction results, makes the final slope stability assessment results more stable, effectively reduces the impact of noise on the assessment results, and improves the credibility of the assessment results.

[0131] See also Figure 2 The present invention also provides a neural network-based slope stability assessment system for power transmission and transformation projects, the system comprising:

[0132] The data acquisition module is configured to scan the power transmission and transformation project slope using a three-dimensional laser scanner, acquire original three-dimensional point cloud data of the power transmission and transformation project slope, and collect original geological and meteorological data of the power transmission and transformation project slope.

[0133] The three-dimensional model module is configured to denoise, filter, and reconstruct the original three-dimensional point cloud data to obtain a slope three-dimensional model, and standardize the original geological and meteorological data to obtain geological and meteorological data.

[0134] The feature vector module is configured to extract geometric features and surface morphology features of the power transmission and transformation project slope from the slope three-dimensional model, and construct a slope feature vector based on the geometric features, the surface morphology features, and the geological and meteorological data.

[0135] The stability evaluation module is configured to construct a slope stability prediction and evaluation model based on a recurrent neural network, identify the slope feature vector through the slope stability prediction and evaluation model to obtain a slope stability evaluation result, determine a stability level of the power transmission and transformation project slope according to the slope stability evaluation result, and take corresponding preventive measures based on the stability level.

[0136] Specifically, the data acquisition module of the embodiment acquires original three-dimensional point cloud data and geological and meteorological data of a slope by high-precision scanning of the slope using a three-dimensional laser scanner. The three-dimensional model module denoises, filters, and reconstructs the original three-dimensional point cloud data to generate a high-quality slope three-dimensional model. The feature vector module extracts geometric features and surface morphology features from the slope three-dimensional model and constructs a slope feature vector based on the geological and meteorological data. The stability evaluation module constructs a slope stability prediction and evaluation model based on a recurrent neural network, intelligently identifies and analyzes the slope feature vector, improves the efficiency and accuracy of slope stability determination, and thus identifies potential risks in a timely manner.

[0137] The neural network-based power transmission and transformation project slope stability evaluation system of the embodiment can monitor the stability change of a slope in real time and take corresponding preventive measures according to the evaluation result. The real-time monitoring and feedback mechanism ensures the safety of the project and can respond to possible slope instability risks in a timely manner.

[0138] Please refer to Figure 3The application further discloses an electronic device, comprising at least one processor, at least one memory, a communication interface and a bus; wherein the processor, the memory and the communication interface communicate with each other through the bus; the memory stores program instructions executable by the processor; and the processor invokes the program instructions to realize the neural network-based power transmission and transformation project slope stability evaluation method.

[0139] The application further discloses a computer readable storage medium storing computer instructions, wherein the computer instructions make the computer realize all or part of steps of the neural network-based power transmission and transformation project slope stability evaluation method.

[0140] The above merely describes preferred embodiments of the application and is not intended to limit the application, and any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the application shall be included in the protection scope of the application.

Claims

1. A neural network-based slope stability assessment method for power transmission and transformation projects, characterized by: S1. Scanning a slope of a power transmission and transformation project using a three-dimensional laser scanner to obtain original three-dimensional point cloud data of the slope of the power transmission and transformation project, and collecting original geological and meteorological data of the slope of the power transmission and transformation project; S2. Denoising, filtering, and reconstructing the original three-dimensional point cloud data to obtain a three-dimensional slope model, and standardizing the original geological and meteorological data to obtain geological and meteorological data; S3. Extracting geometric features and surface morphological features of the power transmission and transformation project slope from the three-dimensional slope model, and constructing a slope feature vector by combining the geometric features, the surface morphological features, and the geological and meteorological data; S4. Constructing a slope stability prediction and assessment model based on a recurrent neural network, identifying the slope characteristic vector using the slope stability prediction and assessment model to obtain a slope stability assessment result, determining the stability level of the power transmission and transformation project slope according to the slope stability assessment result, and taking corresponding preventive measures based on the stability level; The S3 includes: S31. Dividing the three-dimensional slope model into a plurality of grid cells using a three-dimensional grid division algorithm, calculating the local slope, local slope aspect, and local elevation of each grid cell, calculating the curvature characteristics of the surface of the three-dimensional slope model based on Gaussian curvature and mean curvature, calculating the roughness index of the slope surface using a multi-scale analysis method, and generating a distribution map of the overall geometric characteristics and surface morphological characteristics of the power transmission and transformation project slope through statistical analysis and spatial interpolation methods; S32. Normalize the geometric features and the surface morphological features to obtain first geometric features and first surface morphological features, use principal component analysis to reduce the dimensionality of the first geometric features and the first surface morphological features, extract the principal component features of the first geometric features and the first surface morphological features, obtain second geometric features and second surface morphological features, splice the second geometric features and the second surface morphological features with the geological and meteorological data, introduce an attention mechanism, assign different weights to different features, and construct a multi-level slope feature vector, wherein the slope feature vector includes global features and local features.

2. The method for evaluating slope stability of power transmission and transformation projects based on neural networks according to claim 1, characterized in that: S1 includes, S11. Select multiple scanning locations based on the topographical characteristics of the power transmission and transformation project slope, set a three-dimensional laser scanner at each scanning location, perform horizontal calibration and angle adjustment on each three-dimensional laser scanner, and set scanning parameters of the three-dimensional laser scanner, including scanning resolution, scanning range, and scanning frequency, to obtain original three-dimensional point cloud data of the power transmission and transformation project slope; S12. Determine the rock and soil type and stratum structure of the power transmission and transformation project slope by drilling sampling, install groundwater level monitoring equipment, regularly record groundwater level data of the power transmission and transformation project slope by the groundwater level monitoring equipment, monitor the meteorological data of the power transmission and transformation project slope in real time by an automatic weather station, and obtain original geological and meteorological data of the power transmission and transformation project slope by combining the rock and soil type, stratum structure, groundwater level data, and meteorological data.

3. The method for evaluating slope stability of power transmission and transformation projects based on neural networks according to claim 2, characterized in that: S2 includes, S21, denoising the original three-dimensional point cloud data using a statistical outlier detection method to remove abnormal points and noise points in the original three-dimensional point cloud data to obtain denoised three-dimensional point cloud data, smoothing the denoised three-dimensional point cloud data using an adaptive bilateral filtering algorithm to obtain smoothed three-dimensional point cloud data, and performing three-dimensional reconstruction on the smoothed three-dimensional point cloud data based on a Poisson reconstruction algorithm to obtain a three-dimensional slope model; S22. Normalize the original geological and meteorological data to obtain normalized geological and meteorological data under a unified scale, identify outliers in the normalized geological and meteorological data by using the quartile method, interpolate the outliers to obtain interpolated geological and meteorological data, and use a mixed principal component analysis method to extract features and reduce the dimension of the interpolated geological and meteorological data to obtain standardized geological and meteorological data.

4. The neural network-based slope stability assessment method for power transmission and transformation projects according to claim 3, characterized in that: In S21, the calculation formula of the statistical outlier detection method is: ； in, is the statistical outlier value of the i-th data point in the original 3D point cloud data, is the i-th data point in the original 3D point cloud data, is the mean of the original 3D point cloud data, is the standard deviation of the original 3D point cloud data, and n is the number of data points in the original 3D point cloud data; In S22, the method of using a mixed principal component analysis method to perform feature extraction and dimension reduction on the interpolated geological and meteorological data includes: The interpolated geological and meteorological data are reduced in dimension by principal component analysis to obtain the principal component characteristics of the interpolated geological and meteorological data. The principal component characteristics are processed by independent component analysis to obtain the standardized geological and meteorological data. The calculation formula of the geological and meteorological data is: ； Among them, F is geological and meteorological data, and are the weights of principal component analysis and independent component analysis, and They are the jth principal component feature of the principal component analysis method and the kth independent component feature of the independent component analysis method, respectively.

5. The method for evaluating slope stability of power transmission and transformation projects based on neural networks according to claim 1, wherein: S4 includes, S41. Constructing an initial slope stability prediction and assessment model based on a recurrent neural network, initializing model parameters, including a weight matrix and a bias vector, setting a training strategy and hyperparameters, including a learning rate, a batch size, and a number of iterations; S42, obtaining historical slope stability data, and using the historical slope stability data to train the initial slope stability prediction and evaluation model through a gradient descent algorithm to obtain a slope stability prediction and evaluation model; S43, pre-processing the slope characteristic vector to obtain a unified slope characteristic vector consistent with the format of historical slope stability data, inputting the unified slope characteristic vector into the slope stability prediction and assessment model to perform stability prediction to obtain a prediction result, and smoothing the prediction result through a post-processing algorithm to obtain a slope stability assessment result; S44. According to the preset stability grade classification standard, the slope stability assessment results are mapped to the corresponding stability grade. For different stability grades, corresponding preventive and control measures are formulated, and relevant management personnel are notified to implement the corresponding preventive and control measures. A slope monitoring and feedback mechanism is established to monitor the stability grade changes of the slope of the power transmission and transformation project in real time.

6. A method for evaluating slope stability of power transmission and transformation projects based on a neural network according to claim 5, characterized in that: S4 also includes, The state update calculation formula of the initial slope stability prediction and evaluation model is: ； in, is the hidden state vector at time t, is the input feature vector at time t, is the hidden state vector at time t-1, and are the weights of the input feature vector at time t and the hidden state vector at time t-1, b is the bias vector, is the hyperbolic tangent activation function; The loss function calculation formula of the initial slope stability prediction and evaluation model is: ； Among them, L is the loss function value, N is the number of samples of the initial slope stability prediction and evaluation model, The initial slope stability prediction and evaluation model The true value of the sample, The initial slope stability prediction and evaluation model The predicted value of the sample, The initial slope stability prediction and evaluation model The weight of the samples, is the regularization coefficient, is the model parameter vector, for The L2 norm squared; The calculation formula of the gradient descent algorithm is: ； ； in, and are the velocity vector of the m+1th iteration and the velocity vector of the mth iteration, is the momentum coefficient, is the learning rate, is the loss function pair The gradient, and are the parameter vectors for the m+1th iteration and the parameter vector for the mth iteration respectively.

7. A neural network-based slope stability assessment system for power transmission and transformation projects, characterized in that: The system comprises: A data acquisition module is used to scan the slope of the power transmission and transformation project using a three-dimensional laser scanner to obtain original three-dimensional point cloud data of the slope of the power transmission and transformation project, and to collect original geological and meteorological data of the slope of the power transmission and transformation project; A three-dimensional model module is used to perform denoising, filtering and reconstruction on the original three-dimensional point cloud data to obtain a three-dimensional slope model, and to perform standardization processing on the original geological and meteorological data to obtain geological and meteorological data; The feature vector module is used to extract the geometric features and surface morphological features of the power transmission and transformation project slope from the three-dimensional slope model, and construct a slope feature vector by combining the geometric features, the surface morphological features and the geological and meteorological data, including: The three-dimensional slope model is divided into a plurality of grid cells using a three-dimensional grid division algorithm, and the local slope, local slope aspect, and local elevation value of each grid cell are calculated. The curvature characteristics of the surface of the three-dimensional slope model are calculated based on Gaussian curvature and mean curvature. The roughness index of the slope surface is calculated using a multi-scale analysis method. Through statistical analysis and spatial interpolation methods, a distribution map of the overall geometric characteristics and surface morphological characteristics of the power transmission and transformation project slope is generated; Normalizing the geometric features and the surface morphological features to obtain first geometric features and first surface morphological features, performing dimensionality reduction on the first geometric features and the first surface morphological features using principal component analysis, extracting principal component features of the first geometric features and the first surface morphological features to obtain second geometric features and second surface morphological features, concatenating the second geometric features and the second surface morphological features with the geological and meteorological data, introducing an attention mechanism to assign different weights to different features, and constructing a multi-level slope feature vector, wherein the slope feature vector includes global features and local features; The stability assessment module is used to construct a slope stability prediction and assessment model based on a recurrent neural network, identify the slope characteristic vector through the slope stability prediction and assessment model, obtain a slope stability assessment result, judge the stability level of the slope of the power transmission and transformation project according to the slope stability assessment result, and take corresponding preventive measures based on the stability level.

8. An electronic device, characterized in that: include: at least one processor, at least one memory, a communication interface, and a bus; The processor, memory, and communication interface communicate with each other via the bus, the memory stores program instructions that can be executed by the processor, and the processor calls the program instructions to implement the method according to any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that The computer-readable storage medium stores computer instructions, and when the computer program is executed by a processor, the method according to any one of claims 1 to 6 is implemented.

Images