Reservoir site selection method, device, equipment and medium

Abstract

The invention relates to a reservoir site selection method, device and equipment and a medium, and the method comprises the steps: obtaining topographic feature information corresponding to target digital elevation data in a research region based on a GIS system; based on a trained embedded physical constraint neural network model, obtaining multiple pieces of preliminary site selection information corresponding to the target digital elevation data and the topographic feature information; on the basis of the constructed parameter space, obtaining feasibility evaluation information corresponding to each piece of preliminary site selection information; and on the basis of the multiple pieces of feasibility evaluation information, obtaining the optimal preliminary site selection information, and taking the optimal preliminary site selection information as the target site selection of the corresponding reservoir in the research area. The method has the advantages that the site selection efficiency of the reservoir is improved, it is ensured that the output preliminary site selection information conforms to physical laws of terrains, hydrology and the like, and multiple factors influencing site selection of the reservoir are comprehensively considered.

Description

Technical Field

[0001] The present application relates to the technical field of reservoir site selection, and in particular to a reservoir site selection method, device, equipment and medium. Background Art

[0002] With the advancement of water transfer technology and the reduction of water transfer project costs, some domestic institutions have been conducting research on water transfer in the western region in recent years to optimize water resource allocation and solve problems such as uneven water resource distribution. The water transfer system is an important energy load and peak-shaving system after the large-scale development of new energy in the west. By arranging reservoirs along the water transfer line as energy storage devices to adjust the energy output of the entire system, the dynamic balance, safety and stability of the energy system can be achieved.

[0003] However, traditional reservoir site selection methods often rely on manual experience and simple terrain analysis, which makes it difficult to fully consider all factors. They are subjective, have a large workload and a long construction period, making the site selection process inefficient. Summary of the invention

[0004] In order to improve the efficiency of reservoir site selection, various factors affecting site selection are comprehensively considered. This application provides a reservoir site selection method, device, equipment and medium.

[0005] In a first aspect, the present application provides a reservoir site selection method, comprising:

[0006] Based on the GIS system, obtain the terrain feature information corresponding to the target digital elevation data in the study area;

[0007] Based on the trained embedded physical constraint neural network model, a plurality of preliminary site selection information corresponding to the target digital elevation data and the terrain feature information is obtained, wherein the embedded physical constraint neural network model limits and optimizes the output of the output layer through the embedded physical constraint conditions;

[0008] Based on the constructed parameter space, feasibility assessment information corresponding to each of the preliminary site selection information is obtained, wherein the parameter space includes a plurality of parameter combinations, each of which includes a multidimensional parameter affecting the site selection of the corresponding reservoir in the study area;

[0009] Based on the plurality of feasibility assessment information, the optimal preliminary site selection information is obtained, and the optimal preliminary site selection information is used as the target site selection for the corresponding reservoir in the study area.

[0010] The beneficial effects of this application are: automatically conducting the site selection of reservoirs in the study area, improving the efficiency of reservoir site selection. The embedded physical constraint neural network model can automatically consider physical constraints during the training process, while minimizing the prediction error, ensuring that the output preliminary site selection information conforms to the physical laws of topography and hydrology, etc., which not only improves the accuracy and robustness of the embedded physical constraint neural network model, but also enhances the actual feasibility and engineering safety of the output preliminary site selection information. The optimal preliminary site selection information is screened out based on multiple feasibility assessment information, realizing the comprehensive consideration of factors that affect the site selection of reservoirs in many aspects.

[0011] Furthermore, before obtaining the terrain feature information corresponding to the target digital elevation data in the study area based on the GIS system, the method further includes:

[0012] The original digital elevation data in the study area is acquired, and the original digital elevation data is preprocessed to screen out error data to obtain the target digital elevation data.

[0013] The beneficial effect of adopting the above further scheme is that through the preprocessing step, the quality of the target digital elevation data can be significantly improved, ensuring the accuracy and reliability of the subsequent extraction and analysis of terrain feature information, etc. It provides a solid data foundation for reservoir site selection, enabling intelligent site selection analysis to be carried out with the support of high-quality terrain data, effectively improving the scientificity and accuracy of site selection.

[0014] Furthermore, the terrain feature information includes valley lines, ridge lines and water ranges, wherein the valley lines are lines connecting multiple local surface lowest points in the study area, the ridge lines are lines connecting multiple local surface highest points in the study area, and the water ranges are parameters characterizing the boundaries and shapes of the simulated water areas in the study area;

[0015] The obtaining of terrain feature information corresponding to the target digital elevation data in the study area includes:

[0016] Performing depression-filling processing on the target digital elevation data to obtain depression-free data, and performing confluence accumulation analysis based on the depression-free data to obtain a first analysis result;

[0017] Based on the first analysis result, obtaining the ridgeline;

[0018] Performing inversion processing on the target digital elevation data to obtain inverse terrain data, and performing confluence accumulation analysis based on the inverse terrain data to obtain a second analysis result;

[0019] Based on the second analysis result, obtaining the valley line;

[0020] The water area range is determined based on the flood topology algorithm and the valley line.

[0021] The beneficial effect of adopting the above further scheme is: improving the convenience and accuracy of extracting valley lines, ridge lines and water areas.

[0022] Further, the embedded physical constraint neural network model includes a comprehensive loss function, and the comprehensive loss function includes an initial loss function and a physical constraint function;

[0023] The physical constraint function includes a first constraint function, a second constraint function, a third constraint function and a fourth constraint function. The first constraint function is used to ensure that the elevation of the boundary of the reservoir is consistent, the second constraint function is used to ensure that the elevation of the center of the reservoir is lower than the elevation of the boundary, the third constraint function is used to limit the dam height of the reservoir to within a preset dam height range, and the fourth constraint function is used to ensure that the terrain height and the dam height are both within their respective corresponding preset ranges.

[0024] The beneficial effect of adopting the above further scheme is that the comprehensive loss function can ensure that the model output meets certain physical conditions, thereby improving the stability and credibility of the embedded physical constraint neural network model, and ensuring that the preliminary site selection information output by the embedded physical constraint neural network model meets the actual geographical and engineering needs. The neural network architecture with embedded physical constraints solves the limitations of traditional machine learning models in dealing with problems related to physical laws, thereby providing a more reliable and effective model.

[0025] Further, based on the constructed parameter space, obtaining feasibility assessment information corresponding to each of the preliminary site selection information includes:

[0026] Obtaining an objective function corresponding to the parameter space, wherein the objective function is a weighted combination of multiple sub-objective functions, each of which is used to evaluate the feasibility of each of the preliminary site selection information;

[0027] For each piece of preliminary site selection information, calculating the first function value of each sub-objective function corresponding to the preliminary site selection information;

[0028] For each piece of preliminary site selection information, based on each of the first function values, the second function value of the objective function corresponding to the preliminary site selection information is calculated, and based on the second function value, the feasibility assessment information is generated.

[0029] The beneficial effect of adopting the above further scheme is: finding the best balance point among multiple evaluation indicators, and based on the various feasibility assessment information, taking the best preliminary site selection information in the feasibility assessment information as the target site selection for the corresponding reservoir in the study area.

[0030] Furthermore, the multiple sub-objective functions include a first sub-function, a second sub-function and a third sub-function. The first sub-function is used to evaluate the terrain suitability corresponding to each of the preliminary site selection information, the second sub-function is used to evaluate the engineering feasibility corresponding to each of the preliminary site selection information, and the third sub-function is used to evaluate the environmental protection corresponding to each of the preliminary site selection information.

[0031] The beneficial effect of adopting the above further scheme is that the parameter space can balance the requirements of terrain suitability, engineering feasibility and ecological protection through multi-objective optimization technology.

[0032] Further, the determining of the water area based on the flood topology algorithm and the valley line includes:

[0033] Based on the valley line, a seed point is obtained, where the seed point is the starting point for simulating a flooding process based on the flooding topology algorithm;

[0034] Determining an initial range based on the flood topology algorithm and the seed point;

[0035] The initial range is post-processed and the post-processed initial range is used as the water area range, wherein the post-processing includes noise removal processing and boundary line smoothing processing.

[0036] The beneficial effects of adopting the above further scheme are: improving the accuracy and stability of the water area extraction results, eliminating the uncertainty caused by data noise or algorithm errors, and ensuring the accuracy and reliability of the final water area.

[0037] In a second aspect, the present application provides a reservoir site selection device, comprising:

[0038] The first acquisition module is used to acquire terrain feature information corresponding to the target digital elevation data in the study area based on the GIS system;

[0039] A second acquisition module is used to acquire a plurality of preliminary site selection information corresponding to the target digital elevation data and the terrain feature information based on a trained embedded physical constraint neural network model, wherein the embedded physical constraint neural network model limits and optimizes the output of the output layer through embedded physical constraint conditions;

[0040] A third acquisition module is used to acquire feasibility assessment information corresponding to each of the preliminary site selection information based on the constructed parameter space, wherein the parameter space includes multiple parameter combinations, each of which includes multidimensional parameters affecting the site selection of the corresponding reservoir in the study area;

[0041] The fourth acquisition module is used to acquire the optimal preliminary site selection information based on the multiple feasibility assessment information, and use the optimal preliminary site selection information as the target site selection of the corresponding reservoir in the study area.

[0042] In a third aspect, the present application provides an electronic device, including a processor and a memory, wherein the processor is coupled to the memory;

[0043] The processor is used to execute the computer program stored in the memory so that the electronic device performs the method as described in any one of the first aspects.

[0044] In a fourth aspect, the present application provides a computer-readable storage medium, comprising a computer program or instructions, which, when executed on a computer, enables the computer to execute the method as described in any one of the first aspects. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 A schematic diagram of the process of the reservoir site selection method according to an embodiment of the present application;

[0046] Figure 2 A schematic diagram of the process of extracting valley lines and ridge lines in an embodiment of the present application;

[0047] Figure 3 A schematic diagram of a neural network model with embedded physical constraints in an embodiment of the present application;

[0048] Figure 4 This is a structural block diagram of the reservoir site selection device according to an embodiment of the present application;

[0049] Figure 5 This is a structural block diagram of an electronic device according to an embodiment of the present application. DETAILED DESCRIPTION

[0050] The present application is further described in detail below in conjunction with the accompanying drawings.

[0051] The embodiment of the present application provides a reservoir site selection method, which can be executed by a device, which can be a server or a terminal device, wherein the server can be an independent physical server, a server cluster or a distributed system composed of multiple physical servers, or a cloud server providing cloud computing services. The terminal device can be a smart phone, a tablet computer, a desktop computer, etc., but is not limited thereto.

[0052] like Figure 1 As shown, a reservoir site selection method is performed by an electronic device, and the main process of the method is described as follows (steps S101 to S104):

[0053] Step S101: Based on the GIS system, obtain terrain feature information corresponding to the target digital elevation data in the study area.

[0054] In this embodiment, the GIS system is a geographic information system. Digital elevation data (DEM) is an important basic data for reservoir site selection analysis in the study area. Digital elevation data provides terrain elevation information of the study area. Analysis of terrain elevation information is a key step to ensure the accuracy of reservoir site selection analysis.

[0055] Step S102: Based on the trained embedded physical constraint neural network model, a plurality of preliminary site selection information corresponding to the target digital elevation data and the terrain feature information is obtained, and the embedded physical constraint neural network model limits and optimizes the output of the output layer through the embedded physical constraint conditions.

[0056] It is easy to understand that the embedded physical constraint neural network model includes an input layer, a hidden layer and an output layer. The target digital elevation data and terrain feature information can be data corresponding to the input layer of the embedded physical constraint neural network model, and the output layer can output a plurality of preliminary site selection information corresponding to the target digital elevation data and terrain feature information, and the preliminary site selection information can include key parameters such as the location coordinates of the reservoir and the dam height.

[0057] Step S103: Based on the constructed parameter space, feasibility assessment information corresponding to each of the preliminary site selection information is obtained, the parameter space includes multiple parameter combinations, and each of the parameter combinations includes multidimensional parameters that affect the site selection of the corresponding reservoir in the study area.

[0058] The parameter space is a parameter space model constructed based on multidimensional parameters such as known reservoir construction parameters and terrain feature information. Each preliminary site selection information corresponds to a parameter combination, and each parameter combination is a reservoir site selection plan.

[0059] Step S104: based on the plurality of feasibility assessment information, obtaining the optimal preliminary site selection information, and using the optimal preliminary site selection information as the target site selection for the corresponding reservoir in the study area.

[0060] In this embodiment, the site selection of reservoirs in the study area can be automatically carried out, which improves the efficiency of reservoir site selection. The embedded physical constraint neural network model can automatically consider physical constraints during the training process, while minimizing the prediction error, ensuring that the output preliminary site selection information conforms to the physical laws of topography and hydrology, etc., which not only improves the accuracy and robustness of the embedded physical constraint neural network model, but also enhances the actual feasibility and engineering safety of the output preliminary site selection information. The optimal preliminary site selection information is screened out based on multiple feasibility assessment information, realizing the comprehensive consideration of factors that affect the site selection of reservoirs in many aspects.

[0061] In this embodiment, before step S101, the method further includes: acquiring original digital elevation data in the study area, performing preprocessing of error data screening on the original digital elevation data, and obtaining the target digital elevation data.

[0062] In practical applications, the original digital elevation data may contain data errors such as noise and outliers, which may come from measurement errors, data transmission errors or natural factors (such as vegetation coverage, water reflection, etc.). If preprocessing is not performed, these noises and outliers will affect the subsequent extraction and analysis of terrain feature information, resulting in inaccurate final reservoir site selection results. Therefore, preprocessing includes denoising the original digital elevation data and removing outliers.

[0063] In this embodiment, denoising is the first step of preprocessing, and its purpose is to eliminate random noise in the original digital elevation data. The denoising method of denoising can include any one of smoothing filtering, median filtering, etc. Exemplarily, the denoising method of denoising can be smoothing filtering, which can reduce high-frequency noise by the average value of neighboring pixels. Smoothing filtering can be expressed as:

[0064]

[0065] Among them, H(x, y) is the original elevation value corresponding to each pixel in the original digital elevation data, H(x, y) is the elevation value after smoothing the original elevation value corresponding to each pixel, k is the radius of the filter window, and N is the number of pixels in the filter window. It is easy to understand that the filter window is an area or window used to calculate the new value of each pixel during image processing and smoothing filtering.

[0066] Outliers usually refer to pixels that significantly deviate from the normal elevation range. Outliers may be caused by equipment errors or data processing errors. Methods for eliminating outliers may include statistical analysis and neighborhood analysis. Statistical analysis eliminates abnormal elevation values ​​that exceed a certain threshold by calculating the mean and standard deviation of elevation values. Neighborhood analysis eliminates obviously abnormal values ​​by comparing the elevation values ​​of a pixel with its neighboring pixels.

[0067] Through the preprocessing steps of denoising and removing outliers, the quality of the target digital elevation data can be significantly improved, ensuring the accuracy and reliability of subsequent extraction and analysis of terrain feature information. It provides a solid data foundation for reservoir site selection, enabling intelligent site selection analysis to be carried out with the support of high-quality terrain data, effectively improving the scientificity and accuracy of site selection.

[0068] In this embodiment, the terrain feature information may include valley lines, ridge lines and water ranges. The valley lines are lines connecting multiple local lowest surface points in the study area, the ridge lines are lines connecting multiple local highest surface points in the study area, and the water ranges are parameters that characterize the boundaries and shapes of the simulated water areas in the study area.

[0069] like Figure 2 As shown, step S101 specifically includes the following processing:

[0070] Performing depression-filling processing on the target digital elevation data to obtain depression-free data, and performing confluence accumulation analysis based on the depression-free data to obtain a first analysis result;

[0071] Based on the first analysis result, obtaining the ridgeline;

[0072] Performing inversion processing on the target digital elevation data to obtain inverse terrain data, and performing confluence accumulation analysis based on the inverse terrain data to obtain a second analysis result;

[0073] Based on the second analysis result, obtaining the valley line;

[0074] The water area range is determined based on the flood topology algorithm and the valley line.

[0075] In this embodiment, obtaining the ridge line based on the first analysis result may include: acquiring positive terrain data corresponding to the target digital elevation data, and obtaining the ridge line based on the first analysis result and the positive terrain data, wherein the positive terrain data is terrain data that characterizes the original terrain corresponding to the target digital elevation data and is higher than a preset surrounding area.

[0076] Based on the second analysis result, obtaining the valley line may include: acquiring negative terrain data corresponding to the target digital elevation data, and obtaining the valley line based on the second analysis result and the negative terrain data, wherein the negative terrain data is terrain data that characterizes the original terrain corresponding to the target digital elevation data and is lower than a preset surrounding area.

[0077] Figure 2 The original DEM data is the target digital elevation data, the inverse terrain DEM is the inverse terrain data, the no-depression DEM is the no-depression data, the negative terrain is the negative terrain data, and the positive terrain is the positive terrain data.

[0078] The extraction of valley lines and ridge lines is a key step in extracting terrain feature information. Valley lines usually correspond to the locations of rivers and streams, and ridge lines correspond to the locations of ridges and highlands. By performing a flow accumulation analysis on the non-depression data and extracting the grid with a flow accumulation value of zero, the watershed line can be obtained, thereby identifying the first analysis result.

[0079] For the extraction of valley lines, the characteristics of inverse terrain can be used. By subtracting the target digital elevation data from a larger value, terrain data that is completely opposite to the original terrain corresponding to the target digital elevation data, namely inverse terrain data, is obtained, so that the ridges in the original terrain corresponding to the target digital elevation data become valleys in the inverse terrain data, and the valleys in the original terrain corresponding to the target digital elevation data become ridges in the inverse terrain data.

[0080] The inverse terrain data can be expressed as:

[0081] H 反 (x, y) = H max -H(x,y)

[0082]

[0083] Among them, H 反 (x, y) is the elevation value in the inverse terrain data, H(x, y) is the elevation value in the original terrain corresponding to the target digital elevation data, and H max It is the highest point elevation value in the original terrain corresponding to the target digital elevation data, and θ reflects the inclination characteristics of the terrain surface.

[0084] In the inverse terrain data, the same method as that for extracting ridge lines is used, that is, extracting the grid with zero confluence accumulation value after confluence accumulation analysis, and the second analysis result can be obtained by combining the first analysis result with the positive terrain data, and combining the second analysis result with the negative terrain data to obtain the valley line. The positive terrain data and the negative terrain data can respectively correct the first analysis result and the second analysis result, and can ensure the accuracy of the ridge line and valley line extraction results.

[0085] The flood topology algorithm simulates the natural flow of water on the terrain, starting from a preset seed point and gradually spreading to the surrounding areas. It uses the slope and elevation information of the terrain to accurately determine the boundary and shape of the water area.

[0086] In the present embodiment, the water area range is determined based on the flood topology algorithm and the valley line, specifically including: obtaining a seed point based on the valley line, the seed point being the starting point for simulating the flood process based on the flood topology algorithm; determining an initial range based on the flood topology algorithm and the seed point; post-processing the initial range, and using the post-processed initial range as the water area, the post-processing including noise removal processing and boundary line smoothing processing.

[0087] The seed point is a point selected from the valley line. Starting from the seed point, the flood topology algorithm can simulate the process of water spreading to the surrounding area based on the slope and elevation information of the terrain. In this simulation process, water will expand from lowlands to high places until it covers the set preset area, which can be an area within the study area.

[0088] In order to improve the accuracy and stability of the water area extraction results, the initial range can also be post-processed, such as noise removal and boundary line smoothing, to eliminate the uncertainty caused by data noise or algorithm errors and ensure the accuracy and reliability of the final water area.

[0089] In the process of extracting the initial range, some small isolated areas or noise points may be introduced. These small areas that do not conform to the actual terrain characteristics need to be removed, so the noise removal process may include using dilation and erosion operations to remove small isolated pixels or noise areas. In order to make the boundary of the extracted water range smoother and more continuous, the smooth boundary line process may include using a moving average method to take the average value of its neighboring points for each point on the boundary line as the new coordinate of the point.

[0090] In this embodiment, the method for selecting seed points on the valley line may include:

[0091] For each valley line, divide the valley line into multiple valley line segments, and obtain the length information and complexity information corresponding to each valley line segment;

[0092] For each valley segment, obtain a complexity weight corresponding to the complexity information of the valley segment, multiply the length information of the valley segment by the corresponding complexity weight, and obtain a weighted length corresponding to the valley segment;

[0093] For each valley line, the weighted lengths of multiple valley line segments corresponding to the valley line are added together to obtain a total weighted length corresponding to the valley line;

[0094] For each valley line, the average density value corresponding to the valley line is calculated based on the number of target seed points and the total weighted length, and the seed points corresponding to the valley line are generated based on the average density value and various complexity information.

[0095] In this embodiment, the complexity information can be evaluated based on factors such as the curvature of the valley segment, the number of branches, and the rate of change of the terrain. According to the complexity evaluation result of the valley segment, a complexity weight is defined for each valley segment. The complexity weight can be a function based on the curvature, the number of branches, or the rate of change of the terrain. The higher the complexity information, the greater the complexity weight. The weighted length can be obtained by multiplying the length information of each valley segment with its corresponding complexity weight, so that the valley segment with a long and complex length will have a greater impact on the final seed point density.

[0096] According to the analysis requirements and data accuracy requirements, a target number of seed points can be set. The set target number of seed points should be reasonable, which can reflect the details of the valley line without causing excessive computational burden. By dividing the target number of seed points by the total weighted length of the valley line, an average density value can be obtained. The average density value indicates the number of seed points that should be distributed per unit length on the valley line after the weighted average of the complexity information and length information.

[0097] In this embodiment, the average density needs to be further adjusted according to the complexity information, and the correction amount of the average density corresponding to each complexity information is stored in the electronic device. For each complexity information, the average density is corrected according to the correction amount corresponding to the complexity information, and the corrected density is used as the density of the seed points of the valley line segment corresponding to the complexity information. Valley line segments with high complexity information (such as many bends and branches) require denser seed points to capture their subtle features; while valley line segments with low complexity information (such as relatively straight and few branches) can appropriately reduce the seed point density, thereby realizing dynamic adjustment of the seed point density and improving the rationality of valley line seed point collection.

[0098] After determining the density of seed points on the valley line, you can use tools in the GIS system (such as point generation tools, random point generation tools combined with certain rules) to generate multiple seed points on the valley line to ensure that multiple seed points are not too densely clustered together or completely missing in certain areas.

[0099] On the extracted valley line, multiple valley line segments can be divided according to the natural breakpoints and complexity information on the valley line. The multiple valley line segments after division are verified and adjusted to ensure their rationality and accuracy. In this embodiment, the rationality of the valley line segment division can be verified by superimposing auxiliary data such as topographic maps and elevation profiles.

[0100] The valley line is divided according to the natural breakpoints or significant change points. The natural breakpoints or significant change points can be the sharp changes in terrain slope, curvature or elevation, such as the intersection, branch point or turning point of the valley. The natural breakpoints or significant change points are used as the division points of the valley line segment. Adjacent parts with similar complexity information can also be divided into the same valley line segment, thus avoiding simply dividing the line segments with the same complexity information into two different valley line segments.

[0101] In this embodiment, the electronic device also stores a terrain feature recognition model for identifying valley lines and ridge lines. The terrain feature recognition model can use a machine learning algorithm (such as a support vector machine, a random forest, a deep learning, etc.) to train a training data set to learn the relationship between valley lines, ridge lines and target digital elevation data. Through the trained terrain feature recognition model, the valley lines and ridge lines corresponding to the target digital elevation data can be automatically identified.

[0102] The valley lines and ridge lines output by the terrain feature recognition model are compared with the calculated valley lines and ridge lines to determine whether the ridge lines and valley lines obtained by the two methods are consistent. If they are inconsistent, the valley lines or ridge lines output by the terrain feature recognition model and the calculated valley lines or ridge lines are respectively superimposed with the target digital elevation data, and the valley lines or ridge lines with higher consistency are taken as the correct results, and prompt information corresponding to the wrong method is generated and fed back to the user end. In this way, the automatic extraction and optimization of ridge lines and valley lines can be realized, the accuracy and efficiency of ridge line and valley line extraction are improved, and strong support is provided for subsequent site selection analysis.

[0103] like Figure 3 As shown, in this embodiment, the embedded physical constraint neural network model includes a comprehensive loss function, and the comprehensive loss function includes an initial loss function and a physical constraint function. The physical constraint function may include a first constraint function, a second constraint function, a third constraint function, and a fourth constraint function. The first constraint function is used to ensure that the elevation of the boundary of the reservoir is consistent, the second constraint function is used to ensure that the elevation of the center of the reservoir is lower than the elevation of the boundary, the third constraint function is used to limit the dam height of the reservoir to a preset dam height range, and the fourth constraint function is used to ensure that the terrain height and the dam height are both within their respective corresponding preset ranges. Figure 3 The loss function is the comprehensive loss function, and the physical constraint is the physical constraint function.

[0104] The embedded physical constraint neural network model can use a multi-layer feedforward neural network as the basic structure, including an input layer, multiple hidden layers and an output layer. The neurons in each layer are connected through weights and biases, and nonlinear characteristics are introduced through activation functions.

[0105] The first constraint function can be expressed as:

[0106] |P 90 (mask boundary )-P 10 (mask boundary )|=0

[0107] Among them, P 90 (mask boundary ), P 10 (mask boundary ) represent the 90th percentile elevation and the 10th percentile elevation of the boundary mask, respectively. These two elevation values ​​are close, which means that the elevation distribution in the boundary area is relatively flat, avoiding abrupt changes at the boundary and ensuring that the boundary elevation difference is minimized.

[0108] The second constraint function can be expressed as:

[0109] mask center <P 90 (mask boundary )

[0110] Among them, mask center Represents the elevation of the center of the reservoir, ensuring that it is below the 90th percentile elevation of the boundaries.

[0111] The third constraint function can be expressed as:

[0112] P 90 (mask boundary )-min(P 90-100 (mask boundary ))<200m

[0113] Among them, P 90-100 (mask boundary ) represents the elevation of 10% of the boundary mask, P 90 (mask boundary ) can be considered as the higher position of the boundary, because the elevation of the reservoir center is guaranteed to be lower than the 90th percentile elevation of the boundary. 90-100 (mask boundary )) represents the lowest height of the first 10% in the boundary area. The dam height is defined by this difference to ensure that the height difference between the highest and lowest points of the dam does not exceed 200 meters. The third constraint function mainly limits the maximum value of the dam height to ensure that the dam height does not exceed 200 meters.

[0114] The fourth constraint function can be expressed as:

[0115]

[0116] Where H(x, y) represents the terrain height at the location (x, y). represents the target water level, D(x, y) represents the difference in terrain elevation from the position (x, y) to the nearest dam location, B represents the dam height, S represents the deviation measure between the terrain and the dam height, and the fourth constraint function contains two error terms: H(x, y) and The deviation between α and β and the deviation between D(x, y) and B, λ represents the weight factor, which is used to balance the importance of these two errors.

[0117] The fourth constraint function is to ensure that both terrain height and dam height meet the set target by comprehensively evaluating the terrain height and dam height. Although the fourth constraint function itself does not directly constrain the S value, its influence is reflected in the total loss function. The limitation of the S value is achieved through the overall optimization process of Physics Loss. As the training progresses, the weight factor λ will be adjusted according to the actual deviation of terrain height and dam height to ensure that the physical constraint has enough influence on the model output to keep the terrain height and dam height within the set expected range.

[0118] Based on the initial loss function, the physical constraint function is introduced to construct a comprehensive loss function, which can be expressed as:

[0119] Loss=Data Loss+α·Physics Loss

[0120] Among them, Data Loss represents the initial loss function, Ph ysics Loss represents the physical constraint function, and α represents the weight factor, which is used to balance the impact of data error and physical constraint error.

[0121] The physical constraint function can be expressed as:

[0122]

[0123] In traditional neural networks, the model mainly relies on the data-driven training process, which is easy to ignore or violate the basic laws of physics. This defect may cause the model to output unreasonable results when facing nonlinear, complex and changeable terrain. The comprehensive loss function can ensure that the model output meets certain physical conditions, thereby improving the stability and credibility of the embedded physical constraint neural network model, and ensuring that the preliminary site selection information output by the embedded physical constraint neural network model meets the actual geographical and engineering needs. The neural network architecture with embedded physical constraints solves the limitations of traditional machine learning models in dealing with problems related to physical laws, thereby providing a more reliable and effective model.

[0124] In this embodiment, the physical constraint function in the embedded physical constraint neural network model can also be set according to actual conditions. The physical constraint function is designed as a pluggable modular form, and each module corresponds to a physical constraint function. When constructing the embedded physical constraint neural network model, different physical constraint modules can be flexibly selected and combined according to the specific site selection situation. The modular design of the physical constraint function not only improves the reusability of the code, but also facilitates subsequent maintenance and updating.

[0125] First, the specific site selection requirements and physical environment of different reservoirs are analyzed, including but not limited to the reservoir's topography, hydrological conditions, geological structure, climatic conditions, ecological environment, and socioeconomic factors. Based on the analysis results of these analysis contents, the physical constraints corresponding to the reservoir to be sited are identified, such as the reservoir boundary elevation, the relative height between the reservoir center and the boundary, the dam height limit, and the terrain stability requirements. According to the analysis results corresponding to the reservoir to be sited, the corresponding physical constraint module is automatically selected and configured. At the same time, the parameters of the physical constraint function can be automatically adjusted according to the specific physical conditions and data characteristics to ensure that the physical constraint function can accurately reflect the actual situation of the reservoir to be sited.

[0126] In this embodiment, step S103 specifically includes the following processing:

[0127] Obtaining an objective function corresponding to the parameter space, wherein the objective function is a weighted combination of multiple sub-objective functions, each of which is used to evaluate the feasibility of each of the preliminary site selection information;

[0128] For each piece of preliminary site selection information, calculating the first function value of each sub-objective function corresponding to the preliminary site selection information;

[0129] For each piece of preliminary site selection information, based on each of the first function values, the second function value of the objective function corresponding to the preliminary site selection information is calculated, and based on the second function value, the feasibility assessment information is generated.

[0130] Each sub-objective function corresponds to an evaluation index of the preliminary site selection information, and the feasibility assessment information may include the evaluation results of each evaluation index.

[0131] In this embodiment, the plurality of sub-objective functions include a first sub-function, a second sub-function and a third sub-function, wherein the first sub-function is used to evaluate the terrain suitability corresponding to each of the preliminary site selection information, the second sub-function is used to evaluate the engineering feasibility corresponding to each of the preliminary site selection information, and the third sub-function is used to evaluate the environmental protection corresponding to each of the preliminary site selection information. The first sub-function, the second sub-function and the third sub-function can all be evaluated by judging whether they meet the respective corresponding preset parameter ranges. The first sub-function can be judged by checking whether the terrain height is within the set range suitable for reservoir construction, the second sub-function can be used to see whether the engineering parameters such as the dam length meet the construction requirements, and the third sub-function can be determined by evaluating whether the parameters of the preliminary site selection information affecting the surrounding ecology are within an acceptable range.

[0132] The objective function can be expressed as:

[0133] f(θ)=ω1·Terrain Suitability(θ)+ω2·Engineering Feasibility(θ)+ω3·Environmental Impact(θ)

[0134] Among them, ω1, ω2 and ω3 represent weight factors, which are used to balance the importance of different optimization objectives, TerrainSuitability(θ) represents the first sub-function, Engineering Feasibility(θ) represents the second sub-function, and Environmental Impact(θ) represents the third sub-function. Specific sub-objective functions can be defined and adjusted according to actual needs.

[0135] The parameter space can balance the requirements of terrain suitability, engineering feasibility and ecological protection through multi-objective optimization technology. Find the best balance point among multiple evaluation indicators to achieve comprehensive optimization, and make a trade-off between minimizing construction costs and maximizing ecological protection. Based on the feasibility assessment information, the preliminary site selection information that is both economical and environmentally friendly in the feasibility assessment information is used as the target site selection for the corresponding reservoir in the study area.

[0136] Based on the same technical concept, the present application also provides a reservoir site selection device, such as Figure 4 As shown, the reservoir site selection device 200 mainly includes:

[0137] The first acquisition module 201 is used to acquire terrain feature information corresponding to target digital elevation data in the study area based on the GIS system;

[0138] A second acquisition module 202 is used to acquire a plurality of preliminary site selection information corresponding to the target digital elevation data and the terrain feature information based on a trained embedded physical constraint neural network model, wherein the embedded physical constraint neural network model limits and optimizes the output of the output layer through embedded physical constraints;

[0139] The third acquisition module 203 is used to acquire feasibility assessment information corresponding to each of the preliminary site selection information based on the constructed parameter space, wherein the parameter space includes multiple parameter combinations, each of which includes multidimensional parameters affecting the site selection of the corresponding reservoir in the study area;

[0140] The fourth acquisition module 204 is used to acquire the optimal preliminary site selection information based on the multiple feasibility assessment information, and use the optimal preliminary site selection information as the target site selection of the corresponding reservoir in the study area.

[0141] Optionally, before the first acquisition module 201, the method further includes:

[0142] The preprocessing module is used to obtain the original digital elevation data in the study area, perform preprocessing of error data screening on the original digital elevation data, and obtain the target digital elevation data.

[0143] Optionally, the terrain feature information includes valley lines, ridge lines and water ranges, the valley lines are lines connecting multiple local lowest surface points in the study area, the ridge lines are lines connecting multiple local highest surface points in the study area, and the water ranges are parameters that characterize the boundaries and shapes of the waters simulated in the study area.

[0144] The first acquisition module 201 includes:

[0145] A depression-filling processing submodule is used to perform depression-filling processing on the target digital elevation data to obtain depression-free data, and perform confluence accumulation analysis based on the depression-free data to obtain a first analysis result;

[0146] A first obtaining submodule, configured to obtain the ridgeline based on the first analysis result;

[0147] A negation processing submodule, used for performing negation processing on the target digital elevation data to obtain inverse terrain data, and performing confluence accumulation analysis based on the inverse terrain data to obtain a second analysis result;

[0148] A second obtaining submodule, configured to obtain the valley line based on the second analysis result;

[0149] The determination submodule is used to determine the water area range based on the flood topology algorithm and the valley line.

[0150] Optionally, the embedded physical constraint neural network model includes a comprehensive loss function, and the comprehensive loss function includes an initial loss function and a physical constraint function;

[0151] The physical constraint function includes a first constraint function, a second constraint function, a third constraint function and a fourth constraint function. The first constraint function is used to ensure that the elevation of the boundary of the reservoir is consistent, the second constraint function is used to ensure that the elevation of the center of the reservoir is lower than the elevation of the boundary, the third constraint function is used to limit the dam height of the reservoir to within a preset dam height range, and the fourth constraint function is used to ensure that the terrain height and the dam height are both within their respective corresponding preset ranges.

[0152] Optionally, the third acquisition module 203 includes:

[0153] A first acquisition submodule is used to acquire an objective function corresponding to the parameter space, wherein the objective function is a weighted combination of multiple sub-objective functions, and each of the sub-objective functions is used to evaluate the feasibility of each of the preliminary site selection information;

[0154] A first calculation submodule, configured to calculate, for each piece of the preliminary site selection information, a first function value of each of the sub-objective functions corresponding to the preliminary site selection information;

[0155] The second calculation submodule is used to calculate, for each piece of the preliminary site selection information, the second function value of the objective function corresponding to the preliminary site selection information based on each of the first function values, and generate the feasibility assessment information based on the second function value.

[0156] Optionally, the multiple sub-objective functions include a first sub-function, a second sub-function and a third sub-function, the first sub-function is used to evaluate the terrain suitability corresponding to each of the preliminary site selection information, the second sub-function is used to evaluate the engineering feasibility corresponding to each of the preliminary site selection information, and the third sub-function is used to evaluate the environmental protection corresponding to each of the preliminary site selection information.

[0157] Optionally, determine that the submodules include:

[0158] A second acquisition submodule is used to acquire a seed point based on the valley line, wherein the seed point is a starting point for simulating a flooding process based on the flooding topology algorithm;

[0159] An initial range determination module, configured to determine an initial range based on the flood topology algorithm and the seed point;

[0160] The post-processing submodule is used to perform post-processing on the initial range and use the post-processed initial range as the water area range. The post-processing includes noise removal processing and boundary line smoothing processing.

[0161] In one example, the module in any of the above devices can be one or more integrated circuits configured to implement the above methods, such as: one or more application specific integrated circuits (ASICs), or, one or more digital signal processors (DSPs), or, one or more field programmable gate arrays (FPGAs), or a combination of at least two of these integrated circuit forms.

[0162] For another example, when the modules in the device can be implemented in the form of a processing element scheduling program, the processing element can be a general-purpose processor, such as a central processing unit (CPU) or other processor that can call a program. For another example, these modules can be integrated together and implemented in the form of a system-on-a-chip (SOC).

[0163] Various objects such as various messages / information / equipment / network elements / systems / devices / actions / operations / processes / concepts that may appear in this application are named. It can be understood that these specific names do not constitute a limitation on the relevant objects. The names assigned may change with factors such as scenarios, contexts or usage habits. The understanding of the technical meaning of the technical terms in this application should be mainly determined from the functions and technical effects embodied / executed in the technical scheme.

[0164] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and modules described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0165] Those of ordinary skill in the art will appreciate that the modules and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0166] Based on the same technical concept, the present application also provides an electronic device, such as Figure 5 As shown, the electronic device 300 includes a processor 301 and a memory 302 , and may further include an information input / information output (I / O) interface 303 , one or more of a communication component 304 , and a communication bus 305 .

[0167] The processor 301 is used to control the overall operation of the electronic device 300 to complete all or part of the steps in the reservoir site selection method described above; the memory 302 is used to store various types of data to support the operation of the electronic device 300, and these data may include, for example, instructions for any application or method used to operate on the electronic device 300, and data related to the application. The memory 302 can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk.

[0168] The I / O interface 303 provides an interface between the processor 301 and other interface modules, and the above-mentioned other interface modules can be keyboards, mice, buttons, etc. These buttons can be virtual buttons or physical buttons. The communication component 304 is used to test the wired or wireless communication between the electronic device 300 and other devices. Wireless communication, such as Wi-Fi, Bluetooth, Near Field Communication (NFC), 2G, 3G or 4G, or a combination of one or more of them, so the corresponding communication component 304 can include: Wi-Fi components, Bluetooth components, NFC components.

[0169] The communication bus 305 may include a path to transmit information between the above components. The communication bus 305 may be a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus, etc. The communication bus 305 may be divided into an address bus, a data bus, a control bus, etc.

[0170] The electronic device 300 can be implemented by one or more application specific integrated circuits (ASIC), digital signal processors (DSP), digital signal processing devices (DSPD), programmable logic devices (PLD), field programmable gate arrays (FPGA), controllers, microcontrollers, microprocessors or other electronic components to execute the reservoir site selection method given in the above embodiment.

[0171] The electronic device 300 may include but is not limited to mobile terminals such as digital broadcast receivers, PDAs (personal digital assistants), PMPs (portable multimedia players), etc., and fixed terminals such as digital TVs, desktop computers, etc., and may also be servers, etc.

[0172] Based on the same technical concept, the present application also provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of the above-mentioned reservoir site selection method are implemented.

[0173] The computer-readable storage medium may include: a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, and other media that can store program codes.

[0174] The terms "comprises," "comprising," or any other variations thereof are intended to cover non-exclusive inclusion such that a process, method, article, or apparatus that includes a list of elements includes not only those elements but also other elements not expressly listed, or also includes elements inherent to such process, method, article, or apparatus.

[0175] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include at least one of the features. In the description of this application, the meaning of "plurality" is at least two, such as two, three, etc., unless otherwise clearly and specifically defined.

[0176] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present application. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine the different embodiments or examples described in this specification and the features of the different embodiments or examples, without contradiction.

[0177] Although the embodiments of the present application have been shown and described above, it can be understood that the above embodiments are exemplary and cannot be understood as limitations on the present application. Ordinary technicians in this field can change, modify, replace and modify the above embodiments within the scope of the present application.

Claims

1. A reservoir site selection method, characterized in that: include: Based on the GIS system, obtain the terrain feature information corresponding to the target digital elevation data in the study area; Based on the trained embedded physical constraint neural network model, a plurality of preliminary site selection information corresponding to the target digital elevation data and the terrain feature information is obtained, wherein the embedded physical constraint neural network model limits and optimizes the output of the output layer through the embedded physical constraint conditions; Based on the constructed parameter space, feasibility assessment information corresponding to each of the preliminary site selection information is obtained, wherein the parameter space includes a plurality of parameter combinations, each of which includes a multidimensional parameter affecting the site selection of the corresponding reservoir in the study area; Based on the plurality of feasibility assessment information, the optimal preliminary site selection information is obtained, and the optimal preliminary site selection information is used as the target site selection for the corresponding reservoir in the study area.

2. A reservoir site selection method according to claim 1, characterized in that: include: Before obtaining the terrain feature information corresponding to the target digital elevation data in the study area based on the GIS system, the method further includes: The original digital elevation data in the study area is acquired, and the original digital elevation data is preprocessed to screen out error data to obtain the target digital elevation data.

3. A reservoir site selection method according to claim 1, characterized in that: The terrain feature information includes valley lines, ridge lines and water ranges, wherein the valley lines are lines connecting multiple local surface lowest points in the study area, the ridge lines are lines connecting multiple local surface highest points in the study area, and the water ranges are parameters characterizing the boundaries and shapes of the simulated water areas in the study area; The obtaining of terrain feature information corresponding to the target digital elevation data in the study area includes: Performing depression-filling processing on the target digital elevation data to obtain depression-free data, and performing confluence accumulation analysis based on the depression-free data to obtain a first analysis result; Based on the first analysis result, obtaining the ridgeline; Performing inversion processing on the target digital elevation data to obtain inverse terrain data, and performing confluence accumulation analysis based on the inverse terrain data to obtain a second analysis result; Based on the second analysis result, obtaining the valley line; The water area range is determined based on the flood topology algorithm and the valley line.

4. A reservoir site selection method according to claim 1, characterized in that: The embedded physical constraint neural network model includes a comprehensive loss function, and the comprehensive loss function includes an initial loss function and a physical constraint function; The physical constraint function includes a first constraint function, a second constraint function, a third constraint function and a fourth constraint function. The first constraint function is used to ensure that the elevation of the boundary of the reservoir is consistent, the second constraint function is used to ensure that the elevation of the center of the reservoir is lower than the elevation of the boundary, the third constraint function is used to limit the dam height of the reservoir to within a preset dam height range, and the fourth constraint function is used to ensure that the terrain height and the dam height are both within their respective corresponding preset ranges.

5. A reservoir site selection method according to claim 1, characterized in that: The obtaining of feasibility assessment information corresponding to each of the preliminary site selection information based on the constructed parameter space includes: Obtaining an objective function corresponding to the parameter space, wherein the objective function is a weighted combination of multiple sub-objective functions, each of which is used to evaluate the feasibility of each of the preliminary site selection information; For each piece of preliminary site selection information, calculating the first function value of each sub-objective function corresponding to the preliminary site selection information; For each piece of preliminary site selection information, based on each of the first function values, the second function value of the objective function corresponding to the preliminary site selection information is calculated, and based on the second function value, the feasibility assessment information is generated.

6. A reservoir site selection method according to claim 5, characterized in that: The multiple sub-objective functions include a first sub-function, a second sub-function and a third sub-function. The first sub-function is used to evaluate the terrain suitability corresponding to each of the preliminary site selection information, the second sub-function is used to evaluate the engineering feasibility corresponding to each of the preliminary site selection information, and the third sub-function is used to evaluate the environmental protection corresponding to each of the preliminary site selection information.

7. A reservoir site selection method according to claim 3, characterized in that: The determining of the water area based on the flood topology algorithm and the valley line includes: Based on the valley line, a seed point is obtained, where the seed point is the starting point for simulating a flooding process based on the flooding topology algorithm; Determining an initial range based on the flood topology algorithm and the seed point; The initial range is post-processed and the post-processed initial range is used as the water area range, wherein the post-processing includes noise removal processing and boundary line smoothing processing.

8. A reservoir site selection device, characterized in that: include: The first acquisition module is used to acquire terrain feature information corresponding to the target digital elevation data in the study area based on the GIS system; A second acquisition module is used to acquire a plurality of preliminary site selection information corresponding to the target digital elevation data and the terrain feature information based on a trained embedded physical constraint neural network model, wherein the embedded physical constraint neural network model limits and optimizes the output of the output layer through embedded physical constraint conditions; A third acquisition module is used to acquire feasibility assessment information corresponding to each of the preliminary site selection information based on the constructed parameter space, wherein the parameter space includes multiple parameter combinations, each of which includes multidimensional parameters affecting the site selection of the corresponding reservoir in the study area; The fourth acquisition module is used to acquire the optimal preliminary site selection information based on the multiple feasibility assessment information, and use the optimal preliminary site selection information as the target site selection of the corresponding reservoir in the study area.

9. An electronic device, characterized in that: comprising a processor and a memory, wherein the processor is coupled to the memory; The processor is configured to execute a computer program stored in the memory, so that the electronic device executes the method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: The method comprises a computer program or an instruction, which, when executed on a computer, causes the computer to execute the method according to any one of claims 1 to 7.

Images