Pumped storage power station slope near-natural ecology monitoring method, system and equipment

Abstract

The invention discloses a pumped storage power station slope near-natural ecology monitoring method, system and equipment. Belongs to the technical field of natural ecological monitoring, water conservancy and hydropower engineering and water and soil conservation. S200, preprocessing the ecological environment point cloud atlas to generate a sample data set; s300, taking the sample data set as the input of a neural network, and obtaining a prediction network model by learning and training the neural network; s400, predicting the slope of the current pumped storage power station through the prediction network model, achieving the comprehensive monitoring and evaluation of the natural ecological environment of the pumped storage power station, and taking the natural state of the pumped storage power station before construction as an evaluation standard. And the natural ecological environment which is close to the natural ecological environment before the pumped storage power station is built as much as possible serves as a restoration target.

Description

Technical Field

[0001] The present invention relates to the technical field of near-natural ecological monitoring, and more specifically to a method, system and equipment for near-natural ecological monitoring of a slope of a pumped-storage power station. Background Art

[0002] The construction of pumped storage power stations involves large-scale earthwork excavation, backfilling and slope reshaping. These activities will disturb the original landforms on a large scale, destroy surface vegetation, and lead to soil erosion and other disturbances and damage to the slope ecosystem. Therefore, it is particularly important to conduct continuous, scientific and comprehensive monitoring and evaluation of the ecological status of the slopes of pumped storage power stations.

[0003] However, in the existing technology, the ecological status monitoring of the slope of the pumped storage power station is not comprehensive and accurate enough. For example, manual monitoring and ground fixed monitoring points have problems such as low accuracy, long time consumption and low efficiency. Manual monitoring is limited by the professional level and sense of responsibility of the monitoring personnel, and data errors are prone to occur; ground fixed monitoring points may not be able to fully cover the slope area due to the limited number of deployments.

[0004] The near-natural ecological monitoring method simulates natural ecological processes to conduct all-round and multi-level monitoring of slope ecosystems in order to assess their impact on the environment and formulate corresponding protection and restoration measures.

[0005] In summary, how to provide a method, system and equipment for monitoring the near-natural ecology of the slope of a pumped-storage power station is an urgent problem to be solved by those skilled in the art. Summary of the invention

[0006] In view of this, the present invention provides a method and system for monitoring the near-natural ecology of the slope of a pumped-storage power station, which can solve the above-mentioned problems, and is used to solve the technical problems existing in the above-mentioned prior art.

[0007] In order to achieve the above object, the present invention provides the following technical solutions:

[0008] A method for monitoring the near-natural ecology of a pumped storage power station slope comprises the following steps:

[0009] S100: Obtaining the ecological environment point cloud map of the slope of the pumped storage power station;

[0010] S200: preprocessing the ecological environment point cloud map to generate a sample data set;

[0011] S300: using the sample data set as an input of a neural network, and obtaining a prediction network model by learning and training the neural network;

[0012] S400: Predicting the slope of the current pumped-storage power station through the prediction network model and outputting the prediction result.

[0013] Preferably, the method further includes S500: adjusting the near-natural ecological control parameters of the slope of the pumped-storage power station according to the prediction results.

[0014] Preferably, the prediction network model includes a feature extraction layer, a regression prediction layer and a fusion output layer;

[0015] Wherein, the feature extraction layer is used to extract deep features to obtain a deep feature data set;

[0016] According to the near-natural ecological environment standard, the deep feature data is divided into: normal operation period, early degradation period and severe deterioration period. The regression prediction layer performs correlation prediction on each feature graph in the deep feature data set to obtain the correlation score of each feature graph in the deep feature, and outputs the result of normal operation period, early degradation period or severe deterioration period according to the correlation score;

[0017] The fusion output layer fuses the different prediction results output by the regression prediction layer, and uses a nonlinear function to enhance the fused features and then outputs the monitoring results.

[0018] Preferably, the feature extraction layer includes multiple convolution layers and pooling layers, each layer introduces nonlinearity through an activation function, and deep feature extraction is performed through a combination of convolution operation and nonlinear activation function to obtain a deep feature data set, which is expressed as:

[0019] H n =σ(W n *X+B n )

[0020] H m+1 =σ(W m *H m +B m )

[0021] Among them, H n H represents the output feature map of the nth layer; m+1 represents the output feature map of the mth layer; σ is the activation function; X represents the ecological environment point cloud map in the sample data set, W n and B n are the weight and bias of the nth layer respectively; W m and B m are the weight and bias of the mth layer respectively.

[0022] Preferably, the parameters of the feature extraction layer are adjusted using the stochastic gradient descent method, and the expression is:

[0023] g(W)=∑WHx ;

[0024]

[0025] In the formula, g(W) represents the feature extraction function, W represents the weight of the convolutional layer, h(W) represents the loss function, and y x represents the Xth sample vector, t represents the number of iterations, α represents the penalty term, and γ represents the learning rate in gradient descent. represents the gradient, y x represents the sample value of the Xth, H x Indicates the target output.

[0026] Preferably, the regression prediction layer performs correlation prediction on each feature graph in the deep feature data set to obtain a correlation score for each feature graph in the deep feature, and outputs the result of the normal operation period, the early degradation period or the severe deterioration period according to the correlation score, including:

[0027] Calculate the cosine distance between feature maps;

[0028] Calculate correlation based on cosine distance;

[0029] An adaptive neighborhood is initially constructed based on the correlation, and the results of the normal operation period, early degradation period or severe deterioration period are output.

[0030] Preferably, the correlation is calculated based on the cosine distance, and the expression is:

[0031]

[0032] Where, T x1 、T x2 It is represented as a vector of two feature maps, and L2 is the Euclidean length of the two feature maps.

[0033] Preferably, the fusion output layer fuses the features extracted by different convolutional layers, and uses a nonlinear function to enhance the fused features and then outputs the detection result, which is expressed as:

[0034]

[0035] Among them, F enh represents the enhanced fusion feature, L is the total number of layers in the fusion output layer, m∈[1,L], β m is the importance weight of the m-th layer feature, and γ1 and γ2 are learning parameters for adjusting the strength of nonlinear transformation.

[0036] A near-natural ecological monitoring system for a pumped storage power station slope, comprising:

[0037] Data acquisition module: obtain the ecological environment point cloud map of the slope of the pumped storage power station;

[0038] Data processing module: pre-process the ecological environment point cloud map to generate a sample data set;

[0039] Network construction module: taking the sample data set as the input of the neural network, and obtaining the prediction network model by learning and training the neural network;

[0040] Prediction module: predict the slope of the current pumped storage power station through the prediction network model and output the prediction result.

[0041] An electronic device comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor implements a near-natural ecological monitoring method for a slope of a pumped-storage power station when executing the computer program.

[0042] It can be seen from the above technical solution that, compared with the prior art, the present invention discloses a method and system for monitoring the near-natural ecology of the slope of a pumped-storage power station that can solve the above-mentioned problems. The acquired ecological environment point cloud map of the slope of the pumped-storage power station is used as the input of the neural network, and is classified based on the near-natural ecological environment standard as the output of the neural network. By learning and training the neural network, an accurate prediction network model is obtained. Through the feature extraction layer, regression prediction layer and fusion output layer of this prediction network model, the current pumped-storage power station slope can be predicted, and the prediction result can be output, thereby realizing continuous, scientific and comprehensive monitoring and evaluation of the ecological status of the slope of the pumped-storage power station. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.

[0044] Figure 1 It is a schematic diagram of the method flow of the present invention;

[0045] Figure 2 It is a schematic diagram of the system structure of the present invention. DETAILED DESCRIPTION

[0046] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0047] The purpose of the present invention is to provide a method, system and equipment for monitoring the near-natural ecology of the slope of a pumped-storage power station. The method comprises: S100: obtaining an ecological environment point cloud map of the slope of a pumped-storage power station; S200: preprocessing the ecological environment point cloud map to generate a sample data set; S300: using the sample data set as an input of a neural network, and obtaining a prediction network model by learning and training the neural network; S400: predicting the change trend of the near-natural ecological environment of the slope of a pumped-storage power station in the future through the prediction network model, and outputting the prediction results, thereby realizing all-round and multi-level monitoring of the slope ecosystem.

[0048] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0049] See also Figure 1 As shown, the embodiment of the present invention discloses a method for monitoring the near-natural ecology of a slope of a pumped storage power station, comprising the following steps:

[0050] S100: Obtaining the ecological environment point cloud map of the slope of the pumped storage power station;

[0051] S200: preprocessing the ecological environment point cloud map to generate a sample data set;

[0052] S300: taking the sample data set as the input of the neural network, and obtaining the prediction network model by learning and training the neural network;

[0053] S400: Predicting the slope of the current pumped-storage power station through the prediction network model and outputting the prediction result.

[0054] Specifically, in S100, a laser scanner or other three-dimensional scanning equipment is used to obtain an ecological environment point cloud map of the slope of the pumped-storage power station. The ecological environment point cloud map of the slope of the pumped-storage power station is an important data representation method, which can record the detailed information of the topography, vegetation, soil, etc. on the slope surface with high precision, and provide strong support for slope stability assessment and ecological environment monitoring.

[0055] Specifically, in the ecological environment point cloud map, a variety of parameters can be extracted and monitored. The following are some of the main monitoring parameters: topographic parameters, vegetation parameters, soil parameters, ecological environment quality parameters (biodiversity: the vegetation distribution and terrain characteristics reflected in the point cloud map can indirectly evaluate the biodiversity level of the region; ecosystem stability: combined with information on terrain, vegetation, soil and other aspects, the point cloud map can assist in evaluating the stability and anti-interference ability of the ecosystem), etc.

[0056] In a specific embodiment, the step further includes S500: adjusting the near-natural ecological control parameters of the slope of the pumped-storage power station according to the prediction results.

[0057] In a specific embodiment, the prediction network model includes a feature extraction layer, a regression prediction layer, and a fusion output layer;

[0058] Among them, the feature extraction layer is used to extract deep features and obtain a deep feature data set;

[0059] According to the near-natural ecological environment standard, the deep feature data is divided into: normal operation period, early degradation period and severe deterioration period. The regression prediction layer performs correlation prediction on each feature graph in the deep feature data set to obtain the correlation score of each feature graph in the deep feature, and outputs the result of normal operation period, early degradation period or severe deterioration period according to the correlation score;

[0060] The fusion output layer fuses the different prediction results output by the regression prediction layer, and uses a nonlinear function to enhance the fused features and then outputs the monitoring results.

[0061] In a specific embodiment, according to the near-natural ecological environment standard, the deep feature data is divided into: normal operation period, early degradation period and severe deterioration period. The normal operation period refers to the ecological environment system being in a stable and healthy state, various ecological indicators fluctuating within the normal range, and the ecosystem being able to maintain itself and perform normal ecological service functions. The early degradation period refers to the beginning of signs of degradation of the ecological environment system, and some ecological indicators deviate from the normal range, but have not yet reached the level of severe deterioration. At this point, the ecosystem still has a certain recovery capacity, but intervention measures need to be taken to prevent further degradation. The severe deterioration period refers to the ecological environment system that has suffered serious damage, a number of ecological indicators are seriously exceeded, the ecosystem function is basically lost, and recovery is extremely difficult.

[0062] Specifically, the topography, biodiversity, water quality, soil, vegetation coverage and ecosystem service functions have different characteristics in different periods. In order to ensure the accuracy and scientificity of the division, it is also necessary to comprehensively consider expert opinions and local actual conditions.

[0063] More specifically, according to the near-natural ecological environment standard, the depth feature data is divided into: normal operation period, early degradation period and severe deterioration period. Based on the above description, this embodiment first determines a series of monitoring indicators that can reflect the health status and dynamic changes of the slope ecosystem. These indicators should cover vegetation conditions, soil quality, hydrological processes, biodiversity, and possible environmental pollutants. For example:

[0064] Vegetation indicators: vegetation coverage rate, species diversity index, vegetation growth conditions, etc.

[0065] Soil indicators: soil erosion rate, soil fertility, soil microbial activity, etc.

[0066] Hydrological indicators: surface runoff, groundwater level, water quality parameters, etc.

[0067] Biodiversity indicators: species richness, species distribution, population structure, etc.

[0068] Environmental pollutant indicators: heavy metal content, organic pollutant concentration, etc.

[0069] According to the standards of near-natural ecological environment, the expert scoring method is used to divide the deep feature data into specific indicators of normal operation period, early degradation period and severe deterioration period to ensure the scientificity and rationality of the division.

[0070] During the division process, the near-natural ecological process is simulated, which can be achieved in a specific embodiment by constructing an ecosystem model, which can simulate the dynamic change process of the slope ecosystem under different environmental conditions.

[0071] The function of the regression prediction layer is to perform correlation prediction on each feature map in the deep feature data set, obtain the correlation score of each feature map in the deep feature, and output the results of the normal operation period, early degradation period or severe deterioration period according to the correlation score.

[0072] Specific classification method: Correlation analysis includes comparing the data characteristics of two adjacent times to observe the changing trends and amplitudes of various indicators. If all indicators remain within the normal range and there is no obvious downward trend, it can be judged as the normal operation period; if some indicators begin to decline but have not yet reached the warning line, it can be judged as the early degradation period; if multiple indicators are significantly lower than the normal range and the downward trend is obvious, it can be judged as the severe deterioration period.

[0073] The fusion output layer fuses the different prediction results output by the regression prediction layer, and uses nonlinear functions to enhance the fused features and output the monitoring results.

[0074] In a specific embodiment, the feature extraction layer includes multiple convolution layers and pooling layers. Each layer introduces nonlinearity through an activation function. Deep feature extraction is performed through a combination of convolution operations and nonlinear activation functions to obtain a deep feature data set, which is expressed as:

[0075] H n =σ(W n *X+B n )

[0076] H m+1 =σ(W m *H m +B m )

[0077] Among them, H n H represents the output feature map of the nth layer; m+1 represents the output feature map of the mth layer; σ is the activation function; X represents the ecological environment point cloud map in the sample data set, W n and B n are the weight and bias of the nth layer respectively; W m and B m are the weight and bias of the mth layer respectively.

[0078] According to the above technical solution, the feature representation of the data is extracted and transformed layer by layer through the feature extraction layer to capture the complex patterns and deep relationships in the data.

[0079] In a specific embodiment, the parameters of the feature extraction layer are adjusted using the stochastic gradient descent method, and the expression is:

[0080] g(W)=∑W Hx ;

[0081]

[0082] In the formula, g(W) represents the feature extraction function, W represents the weight of the convolutional layer, h(W) represents the loss function, and y x represents the Xth sample vector, t represents the number of iterations, α represents the penalty term, and γ represents the learning rate in gradient descent. represents the gradient, y x represents the sample value of the Xth, H x Indicates the target output.

[0083] Through the above technical solution, the accuracy of the feature extraction layer can be effectively improved.

[0084] In a specific embodiment, the regression prediction layer performs correlation prediction on each feature map in the deep feature data set to obtain a correlation score for each feature map in the deep feature, and outputs the results of the normal operation period, the early degradation period, or the severe deterioration period according to the correlation score, including:

[0085] Calculate the cosine distance between feature maps;

[0086] The correlation is calculated based on the cosine distance, and the expression is:

[0087]

[0088] Where, T x1 、T x2 It is represented as a vector of two feature maps, and L2 is the Euclidean length of the two feature maps;

[0089] An adaptive neighborhood is initially constructed based on the correlation, and the results of the normal operation period, early degradation period or severe deterioration period are output.

[0090] In a specific embodiment, the fusion output layer fuses the features extracted by different convolutional layers, and uses a nonlinear function to enhance the fused features and then outputs the detection result, which is expressed as:

[0091]

[0092] Among them, F enh represents the enhanced fusion feature, L is the total number of layers in the fusion output layer, m∈[1,L], β m is the importance weight of the m-th layer feature, and γ1 and γ2 are learning parameters for adjusting the strength of nonlinear transformation.

[0093] Specifically, by fusing the features extracted by different convolutional layers and enhancing the fused features using nonlinear functions, the prediction network model's ability to understand and express point cloud features can be effectively improved.

[0094] See also Figure 2 As shown, a pumped storage power station slope near-natural ecological monitoring system comprises:

[0095] Data acquisition module: obtain the ecological environment point cloud map of the slope of the pumped storage power station;

[0096] Data processing module: pre-process the ecological environment point cloud map to generate a sample data set;

[0097] Network construction module: Take the sample data set as the input of the neural network, and obtain the prediction network model by learning and training the neural network;

[0098] Prediction module: predict the slope of the current pumped storage power station through the prediction network model and output the prediction result.

[0099] An electronic device comprises a memory, a processor and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, a near-natural ecological monitoring method for a slope of a pumped-storage power station is implemented.

[0100] In this specification, each embodiment is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part.

[0101] The above description of the disclosed embodiments enables one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for monitoring the near-natural ecology of a pumped storage power station slope, characterized in that: The following steps are involved: S100: Obtaining the ecological environment point cloud map of the slope of the pumped storage power station; S200: preprocessing the ecological environment point cloud map to generate a sample data set; S300: using the sample data set as an input of a neural network, and obtaining a prediction network model by learning and training the neural network; S400: Predicting the slope of the current pumped-storage power station through the prediction network model and outputting the prediction result.

2. A method for monitoring the near-natural ecology of a pumped storage power station slope according to claim 1, characterized in that: It also includes S500: adjusting the near-natural ecological control parameters of the slope of the pumped storage power station according to the prediction results.

3. A method for monitoring the near-natural ecology of a pumped storage power station slope according to claim 1, characterized in that: The prediction network model includes a feature extraction layer, a regression prediction layer and a fusion output layer; Wherein, the feature extraction layer is used to extract deep features to obtain a deep feature data set; According to the near-natural ecological environment standard, the deep feature data is divided into: normal operation period, early degradation period and severe deterioration period. The regression prediction layer performs correlation prediction on each feature graph in the deep feature data set to obtain the correlation score of each feature graph in the deep feature, and outputs the result of normal operation period, early degradation period or severe deterioration period according to the correlation score; The fusion output layer fuses the different prediction results output by the regression prediction layer, and uses a nonlinear function to enhance the fused features and then outputs the monitoring results.

4. A method for monitoring the near-natural ecology of a pumped storage power station slope according to claim 3, characterized in that: The feature extraction layer includes multiple convolutional layers and pooling layers. Each layer introduces nonlinearity through an activation function. Deep feature extraction is performed through a combination of convolution operations and nonlinear activation functions to obtain a deep feature data set, which is expressed as: H n =σ(W n *X+B n ) H m+1 =σ(W m *H m +B m ) Among them, H n H represents the output feature map of the nth layer; m+1 represents the output feature map of the mth layer; σ is the activation function; X represents the ecological environment point cloud map in the sample data set, W n and B n are the weight and bias of the nth layer respectively; W m and B m are the weight and bias of the mth layer respectively.

5. A method for monitoring the near-natural ecology of a pumped storage power station slope according to claim 4, characterized in that: The parameters of the feature extraction layer are adjusted using the stochastic gradient descent method, and the expression is: g(W)=∑W Hx ; In the formula, g(W) represents the feature extraction function, W represents the weight of the convolutional layer, h(W) represents the loss function, and y x represents the Xth sample vector, t represents the number of iterations, α represents the penalty term, and γ represents the learning rate in gradient descent. represents the gradient, y x represents the sample value of the Xth, H x Indicates the target output.

6. A method for monitoring the near-natural ecology of a pumped storage power station slope according to claim 3, characterized in that: The regression prediction layer performs correlation prediction on each feature map in the deep feature data set to obtain a correlation score for each feature map in the deep feature, and outputs the results of the normal operation period, early degradation period, or severe deterioration period according to the correlation score, including: Calculate the cosine distance between feature maps; Calculate correlation based on cosine distance; An adaptive neighborhood is initially constructed based on the correlation, and the results of the normal operation period, early degradation period or severe deterioration period are output.

7. A method for monitoring the near-natural ecology of a pumped storage power station slope according to claim 6, characterized in that: The correlation is calculated based on the cosine distance, and the expression is: Where, T x1 , T x2 It is represented as a vector of two feature maps, and L2 is the Euclidean length of the two feature maps.

8. A method for monitoring the near-natural ecology of a pumped storage power station slope according to claim 3, characterized in that: The fusion output layer fuses the features extracted by different convolutional layers, and uses a nonlinear function to enhance the fused features and then outputs the detection result. The expression is: Among them, F enh represents the enhanced fusion feature, L is the total number of layers in the fusion output layer, m∈[1,L], β m is the importance weight of the m-th layer feature, and γ1 and γ2 are learning parameters for adjusting the strength of nonlinear transformation.

9. A pumped storage power station slope near-natural ecology monitoring system using the pumped storage power station slope near-natural ecology monitoring method according to any one of claims 1 to 8, characterized in that: include: Data acquisition module: obtain the ecological environment point cloud map of the slope of the pumped storage power station; Data processing module: pre-process the ecological environment point cloud map to generate a sample data set; Network construction module: taking the sample data set as the input of the neural network, and obtaining the prediction network model by learning and training the neural network; Prediction module: predict the slope of the current pumped storage power station through the prediction network model and output the prediction result.

10. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the near-natural ecological monitoring method for the slope of a pumped-storage power station as described in any one of claims 1 to 8 is implemented.

Images