Method and system for predicting inundation loss of beach area based on artificial intelligence

Abstract

The invention discloses an artificial intelligence-based beach area inundation loss prediction method and system, and relates to data processing and artificial intelligence technologies, and the method comprises the steps: obtaining the topographic data and satellite image data of a target drainage basin, and determining the main river channel range of the target drainage basin; identifying a possible beach area coverage range; performing grid division on the coverage area of any beach area, and determining a loss cardinal number in each grid; determining a plurality of submerging elevations based on the any beach area and the divided grids; acquiring current meteorological data and upstream water level monitoring data of the target drainage basin; utilizing a pre-trained LSTM model to predict the river flow of the future target drainage basin based on the meteorological data and the upstream water level monitoring data; and according to the predicted river flow and the reservoir capacity of the beach area under each submerging elevation, estimating submerging duration and submerging loss. The method and the device are used for realizing graded inundation loss estimation and improving the effect of pre-prediction of flood disasters.

Description

Technical Field

[0001] The present application relates to the technical fields of data processing and artificial intelligence, and particularly relates to a method and system for predicting beach area inundation losses based on artificial intelligence. Background Art

[0002] Rivers provide a continuous source of water for humans and are of great significance to human production and life. However, once a flood occurs in a river, it will bring heavy disasters to humans. Therefore, identifying the areas along the river that are prone to inundation is of great significance for flood prevention and mitigation, flood risk analysis, construction site selection, etc.

[0003] Currently, the identification of flood inundation areas mainly determines the inundation range by comparing remote sensing images before and after the flood to evaluate the affected losses. For example, by combining multi-period optical and radar satellites before and after the flood, analyzing the water body identification capabilities of different sensor images using the threshold method and the machine model migration method, and analyzing the flood disaster situation in the study area.

[0004] There is also a method of constructing a rapid extraction model for flood inundation range by combining mountain shadow mask data to analyze the flood inundation range.

[0005] However, the existing methods are all for the post-disaster analysis and evaluation of flood inundation areas and cannot be applied to the pre-disaster prediction of flood disasters. Summary of the Invention

[0006] The embodiments of the present application provide a method and system for predicting beach area inundation losses based on artificial intelligence, which can realize the prediction of inundation water levels for beach area inundation and achieve a hierarchical estimation of inundation losses, improving the effect of pre-disaster prediction of flood disasters.

[0007] The embodiments of the present application propose a method for predicting beach area inundation losses based on artificial intelligence, including: Obtaining topographic data and satellite image data of the target river basin, and determining the main river channel range of the target river basin; Identifying the possible beach area coverage range according to the main river channel range, the topographic data, and the satellite image data; Dividing the coverage range of any beach area into grids, and determining the loss base in each grid; Based on any beach area and the divided grids, determining multiple inundation elevations, and configuring corresponding loss factors for each inundation elevation; Obtaining the current meteorological data and upstream water level monitoring data of the target river basin; Using a pre-trained LSTM model to predict the future river channel flow of the target river basin based on the meteorological data and the upstream water level monitoring data; Based on the predicted river channel flow and the reservoir capacity of the floodplain at various inundation elevations, the inundation duration and inundation losses are estimated.

[0008] Optionally, based on the main river channel range, the terrain data, and the satellite image data, identifying the possible floodplain coverage range includes: Selecting typical riverbank pixels on both sides of the main river channel range from the satellite image data; Removing the typical riverbank pixels from the satellite image data; Determining multiple continuous pixel paths in the remaining satellite image data; Taking the paths with the shortest pixel distance within a preset range from the main river channel range among the multiple continuous pixel paths as the inundation intrusion paths.

[0009] Optionally, based on the main river channel range, the terrain data, and the satellite image data, identifying the possible floodplain coverage range further includes: Determining partial terrain data around any inundation intrusion path from the terrain data; Determining a first area with undulation less than a first preset threshold in the partial terrain data; and, Based on the terrain data, determining a second area with a height difference less than a second preset threshold from the main river channel range; Taking the union of the first area and the second area as the floodplain coverage range.

[0010] Optionally, performing grid division on the coverage range of any floodplain and determining the loss base in each grid includes: Dividing the coverage range of any floodplain into multiple adaptive grids, and after division, making the inundation flow accommodated by each grid similar, where the grid containing the inundation intrusion path is determined based on the boundary of the inundation intrusion path; For any grid, determining the inundated object data therein and setting an inundation base for each inundated object; Summing up the inundation bases to obtain the loss base within the grid.

[0011] Optionally, based on any floodplain and the divided grids, determining multiple inundation elevations includes: Extracting the inundated object data in each grid and determining the types of inundated objects included in all grids of any floodplain; and, For any inundated object, judging according to its reference inundation height and the terrain data whether the proportion of inundating all grids of the any inundated object reaches a preset threshold; Based on the corresponding reference inundation height and terrain data, determining the inundation elevation of the any inundated object.

[0012] Optionally, it further includes training the LSTM model in the following manner in advance: Sampling historical meteorological data and local meteorological observation data, and fusing the sampled data; Based on the fused sampled data and historical upstream water level monitoring data, constructing a training sequence; Using transfer learning, retaining the underlying spatio-temporal features in training the LSTM model, and generating a flow prediction interval for the output of the LSTM model using the Monte Carlo method; Estimating the inundation flow according to the satellite image data of the floodplain area under the historical upstream water level monitoring data, and using the deviation between the estimated inundation flow and the generated flow interval as the loss to train the LSTM model.

[0013] Optionally, it further includes obtaining the topographic cross-section data within the river channel range of any one of the floodplain areas; After predicting the river channel flow of the future target basin based on the meteorological data and the upstream water level monitoring data using the pre-trained LSTM model, it further includes: Estimating the possible temporary inundation peak according to the predicted river channel flow and the topographic cross-section data.

[0014] Optionally, according to the predicted river channel flow and the reservoir capacity of the floodplain area at each inundation elevation, estimating the inundation duration and inundation loss includes: Determining the inflow into the floodplain area according to part of the topographic data of the floodplain area and the predicted river channel flow, and estimating the inundation height according to the inflow and the reservoir capacity of the floodplain area at each inundation elevation; Determining the inundation loss according to the estimated inundation height and the inundation elevation of any inundated object; Starting from the inundation height reaching the preset height threshold, determining the inundation duration according to the duration when the continuously predicted river channel flow exceeds the river channel capacity threshold characterized by part of the topographic data of the floodplain area.

[0015] An embodiment of the present application also proposes an artificial intelligence-based floodplain inundation loss prediction system, including a processor and a memory, and a computer program is stored on the memory. When the computer program is executed by the processor, the steps of the artificial intelligence-based floodplain inundation loss prediction method as described above are implemented.

[0016] The method of the present application is used to realize the prediction of the inundation water level of the floodplain inundation, and realize the hierarchical estimation of the inundation loss, improving the effect of the pre-disaster prediction of the flood disaster situation.

[0017] The above description is only an overview of the technical solution of this application. In order to understand the technical means of this application more clearly, it can be implemented according to the content of the specification. And in order to make the above and other purposes, features and advantages of this application more obvious and understandable, the following specific embodiments of the present invention are specifically given. Brief Description of the Drawings

[0018] By reading the following detailed description of the preferred embodiments, various other advantages and benefits will become clear to those of ordinary skill in the art. The drawings are only for the purpose of showing the preferred embodiments and are not considered to be a limitation of the present invention. Moreover, throughout the drawings, the same reference numerals are used to represent the same components. In the drawings: Figure 1 It is a schematic diagram of the basic process of the floodplain inundation loss prediction method based on artificial intelligence in this embodiment. Detailed Embodiments

[0019] The exemplary embodiments of the present disclosure will be described in more detail below with reference to the drawings. Although the exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided so that the present disclosure can be more thoroughly understood and the scope of the present disclosure can be fully conveyed to those skilled in the art.

[0020] An embodiment of this application proposes a floodplain inundation loss prediction method based on artificial intelligence, as Figure 1 shown, including the following steps: In step S101, topographic data and satellite image data of the target basin are acquired, and the main channel range of the target basin is determined. In a specific example, the elevation model DEM of the area can be trimmed to retain the data of the basin part as the topographic data of the target basin, and the satellite image data can be the image data in the dry season or the normal water season. According to the acquired topography and satellite images, the main channel range is determined respectively therein.

[0021] In step S102, according to the main channel range, the topographic data and the satellite image data, the possible floodplain coverage range is identified. The possible floodplain areas can be judged according to the identified main channel range and topographic data, and then the floodplain range is further identified according to the satellite image data.

[0022] In step S103, the coverage range of any floodplain is divided into grids, and the loss base in each grid is determined. In the case where there are multiple floodplains in the target basin, the coverage ranges of each floodplain are divided into grids, and the loss base refers to the loss caused, for example, after a certain building is inundated.

[0023] In step S104, based on any of the floodplains and the divided grids, multiple inundation elevations are determined, and corresponding loss factors are configured for each inundation elevation. According to the aforementioned inundation of a building, different inundation loss weights can be set according to the building height as loss factors.

[0024] In step S105, the current meteorological data and upstream water level monitoring data of the target basin are obtained. In the embodiments of the present application, there is no need to add any data collection equipment to the floodplain, and the downstream floodplain flow is directly predicted based on the meteorological data and upstream water level monitoring data.

[0025] In step S106, using a pre-trained LSTM model, the channel flow of the future target basin is predicted based on the meteorological data and upstream water level monitoring data. In some examples, according to the historical upstream water level monitoring data and downstream flow, in a specific example, the historical downstream flow can be calculated based on historical data, and meteorological labels are added to the upstream water level monitoring data to train the LSTM model, so as to perform prediction after training.

[0026] In step S107, according to the predicted channel flow and the storage capacity of the floodplain at each inundation elevation, the inundation duration and inundation loss are estimated. The inundation height is estimated based on the storage capacity of the floodplain at each inundation elevation and the predicted channel flow, and then the inundation loss is determined according to the aforementioned loss base and loss factor.

[0027] The method of the present application is used to realize the prediction of the inundation water level of the floodplain and realize the hierarchical estimation of the inundation loss, so as to improve the effect of the pre-disaster prediction of the flood disaster situation.

[0028] In some embodiments, according to the main channel range, the terrain data, and the satellite image data, the possible floodplain coverage range is identified, including: From the satellite image data, typical riverbank pixels on both sides of the main channel range are selected, the representative pixel range of the river channel can be determined, the representative pixels of the river channel are removed, and combined with the farmland, buildings, etc. on both sides of the riverbank to determine the typical riverbank pixels.

[0029] Remove the typical riverbank pixels from the satellite image data.

[0030] Determine multiple continuous pixel paths in the remaining satellite image data. In some examples, after removing the above pixels, by identifying the continuous pixel paths, the possible tributaries in the floodplain can be determined. In some examples, the performance of the tributaries in the satellite image may be different from that of the main channel. By removing the typical pixels and performing path identification, the identification effect of the tributaries can be improved.

[0031] The path with the shortest pixel distance between multiple consecutive pixel paths and the main river channel range within a preset range is used as the inundation intrusion path. In a specific example, to determine whether a tributary introduces into the main river channel, by judging the shortest pixel distance between the consecutive pixel path and the main river channel range, the tributaries introducing into the main river channel can be determined. For the gullies that do not introduce into the main river channel, adaptive grid division is performed in subsequent examples.

[0032] In some embodiments, based on the main river channel range, the terrain data, and the satellite image data, identifying the possible floodplain coverage range further includes: Determining a part of the terrain data around any inundation intrusion path from the terrain data; Determining a first area in the part of the terrain data with undulation less than a first preset threshold, and through this step, the extended identification of the floodplain near the tributary introducing into the main river channel is realized.

[0033] According to the terrain data, determining a second area with a height difference less than a second preset threshold from the main river channel range, and further combining the terrain data to determine an area with a height difference less than a set value from the main river channel bank.

[0034] Taking the union of the first area and the second area as the floodplain coverage range.

[0035] In some embodiments, grid division is performed on the coverage range of any floodplain, and determining the loss base in each grid includes: Dividing the coverage range of any floodplain into multiple adaptive grids, and after division, making the inundation flow rates accommodated by each grid similar. The grid containing the inundation intrusion path is determined based on the boundary of the inundation intrusion path. Particularly, for the inundation situation of the floodplain, it generally cannot reach the state of complete inundation. By dividing the adaptive grids, the flow rates accommodated by the grids are made close, so that in the subsequent prediction process, the number of inundated grids can be directly judged according to the predicted inflow rate. In some examples, a serial number representing the inundation order can also be added to any adaptive grid according to the height of the terrain data and the distances from the main river channel and the identified tributaries.

[0036] For any grid, determining the inundated object data therein, and setting an inundation base for each inundated object. The inundation base can be set according to, for example, crops, buildings, etc. in the grid.

[0037] Summing up the inundation bases to obtain the loss base in the grid.

[0038] In some embodiments, based on any floodplain and the divided grids, determining multiple inundation elevations includes: Extracting the inundated object data in each grid, and determining the types of inundated objects included in all grids of any floodplain.

[0039] For any submerged object, based on its reference submerged height and the terrain data, determine whether the proportion of the grid cells in which the submerged object is submerged reaches a preset threshold. For example, for crops, different submerged proportions can be determined according to the aforementioned numbers, and multiple submerged proportion thresholds can be set accordingly.

[0040] Based on the corresponding reference submerged height and terrain data, determine the submerged elevation of the submerged object. In the case where multiple submerged proportion thresholds are set for the same type of submerged object, multiple submerged elevations can be determined respectively. In particular, more accurate submerged height parameters can be set for important things in the floodplain area, so as to focus on this submerged object in subsequent predictions.

[0041] In some embodiments, it further includes pre-training the LSTM model in the following manner: Sample the historical meteorological data and local meteorological observation data, and fuse the sampled data. For example, the mean value can be taken after sampling the historical meteorological data and local meteorological observation data to form the fused data.

[0042] Based on the fused sampled data and historical upstream water level monitoring data, construct training sequences. For example, set a time window for sequential sliding to obtain multiple training sequences.

[0043] Using transfer learning, retain the underlying spatio-temporal features in training the LSTM model, and, for the output of the LSTM model, use the Monte Carlo method to generate a flow prediction interval. For example, freeze the first 3 convolutional layers during training and adjust the parameters of other parts of the model. Use the Monte Carlo method to generate a flow prediction interval to describe the uncertain propagation law of the flow under submerged conditions.

[0044] Estimate the submerged flow based on the satellite image data of the floodplain area under the historical upstream water level monitoring data, and use the deviation between the estimated submerged flow and the generated flow interval as the loss to train the LSTM model.

[0045] In some embodiments, it further includes obtaining the terrain cross-section data within the range of any floodplain river channel. In some examples, the terrain cross-section data can be determined based on the monitoring data of the river surface during the dry season. For example, for the floodplain location, it can be determined according to the river area and slope exposed during the dry season. In practice, only the floodplain area can be monitored, and for the river channels that are not exposed, the terrain cross-section data can be determined by empirical calculation.

[0046] After predicting the channel flow of the future target basin based on meteorological data and upstream water level monitoring data using a pre-trained LSTM model, it further includes: estimating the possible peak of temporary inundation according to the predicted channel flow and topographic cross-section data. In some specific examples, the peak of temporary inundation in the short term can be estimated according to the change of the predicted channel flow, as well as the channel area and channel topographic cross-section data under the current flow.

[0047] In some embodiments, estimating the inundation duration and inundation loss according to the predicted channel flow and the storage capacity of the floodplain at each inundation elevation includes: According to part of the topographic data of the floodplain and the predicted channel flow, determine the inflow of the floodplain, and estimate the inundation height according to the inflow and the storage capacity of the floodplain at each inundation elevation. In a specific example, according to the number of the aforementioned inundation sequence, for the predicted flow exceeding the main channel flow threshold, take the value before exceeding the main channel flow threshold as the main channel flow threshold, and take the subsequent flow as the inflow for subtraction, and fill the numbered grid according to the difference to estimate the inundation height.

[0048] Determine the inundation loss according to the estimated inundation height and the inundation elevation of any inundated object, that is, calculate according to the weight and base value at different inundation heights to obtain the inundation loss.

[0049] Starting from the inundation height reaching the preset height threshold, determine the inundation duration according to the duration that the continuously predicted channel flow exceeds the channel capacity threshold characterized by part of the topographic data of the floodplain.

[0050] The method of the embodiment of the present application predicts and analyzes the future flow through adaptive grid division, combined with meteorological data and water level data, so as to predict the inundated area and inundation loss of the floodplain, and realize hierarchical estimation of inundation loss, improving the effect of pre-disaster prediction of flood disasters.

[0051] The embodiment of the present application also proposes a floodplain inundation loss prediction system based on artificial intelligence, including a processor and a memory, and a computer program is stored on the memory. When the computer program is executed by the processor, it realizes the steps of the aforementioned floodplain inundation loss prediction method based on artificial intelligence.

[0052] In addition, although exemplary embodiments have been described herein, the scope includes any and all embodiments based on the present disclosure having equivalent elements, modifications, omissions, combinations (e.g., schemes of cross-combination of various embodiments), adaptations or changes. It is not limited to the examples described in this specification or during the implementation of this application, and the examples will be construed as non-exclusive.

[0053] The above description is intended to be illustrative and not restrictive. For example, the above examples (or one or more aspects thereof) can be used in combination with each other. For example, those of ordinary skill in the art can use other embodiments when reading the above description.

[0054] The above embodiments are only exemplary embodiments of the present disclosure. Those skilled in the art can make various modifications or equivalent replacements to the present invention within the essence and scope of protection of the present disclosure, and such modifications or equivalent replacements should also be regarded as falling within the scope of protection of the present invention.

Claims

1. An artificial intelligence-based prediction method for floodplain inundation losses, characterized in that, Including: Obtain the topographic data and satellite image data of the target basin, and determine the main river channel range of the target basin; Identify the possible floodplain coverage range according to the main river channel range, the topographic data and the satellite image data; Perform grid division on the coverage range of any floodplain, and determine the loss base in each grid; Based on any floodplain and the divided grids, determine multiple inundation elevations, and configure corresponding loss factors for each inundation elevation; Obtain the current meteorological data and upstream water level monitoring data of the target basin; Use the pre-trained LSTM model to predict the future river channel flow of the target basin based on the meteorological data and the upstream water level monitoring data; According to the predicted river channel flow and the storage capacity of the floodplain at each inundation elevation, estimate the inundation duration and inundation loss.

2. The method for predicting flood loss in floodplain areas based on artificial intelligence according to claim 1, characterized in that, Identifying the possible floodplain coverage range according to the main river channel range, the topographic data and the satellite image data includes: Select typical riverbank pixels on both sides of the main river channel range from the satellite image data; Eliminate the typical riverbank pixels in the satellite image data; Determine multiple continuous pixel paths in the remaining satellite image data; Take the path with the shortest pixel distance within a preset range between the multiple continuous pixel paths and the main river channel range as the inundation intrusion path.

3. The method for predicting flood inundation losses in floodplains based on artificial intelligence according to claim 2, wherein Identifying the possible floodplain coverage range according to the main river channel range, the topographic data and the satellite image data further includes: Determine part of the topographic data around any inundation intrusion path from the topographic data; Determine the first area with undulation less than the first preset threshold in the part of the topographic data; and, Determine the second area with a height difference less than the second preset threshold from the main river channel range according to the topographic data; Take the union of the first area and the second area as the floodplain coverage range.

4. The method for predicting flood loss in floodplain areas based on artificial intelligence according to claim 1, wherein Performing grid division on the coverage range of any floodplain and determining the loss base in each grid includes: Divide the coverage range of any floodplain into multiple adaptive grids, and after division, make the inundation flow accommodated by each grid similar, where the grid containing the inundation intrusion path is determined based on the boundary of the inundation intrusion path; For any grid, determine the inundated object data therein, and set the inundation base for each inundated object; Sum up each inundation base to obtain the loss base in the grid.

5. The method for predicting flood inundation losses in floodplains based on artificial intelligence according to claim 4, characterized in that, Determining multiple inundation elevations based on any floodplain and the divided grids includes: Extract the inundated object data in each grid, and determine the types of inundated objects included in all grids of any floodplain; and, For any inundated object, judge whether the proportion of inundating all grids of the any inundated object reaches the preset threshold according to its reference inundation height and the topographic data; Determine the inundation elevation of the any inundated object according to the corresponding reference inundation height and the topographic data.

6. The method for predicting flood inundation losses in floodplains based on artificial intelligence according to claim 5, wherein It also includes pre-training the LSTM model in the following way: Sample the historical meteorological data and the local meteorological observation data, and fuse the sampled data; Construct a training sequence based on the fused sampled data and the historical upstream water level monitoring data; Using transfer learning, retain the underlying spatio-temporal features during the training of the LSTM model, and, for the output of the LSTM model, adopt the Monte Carlo method to generate a flow prediction interval; Estimate the inundation flow based on the historical upstream water level monitoring data and the satellite image data of the floodplain area, and use the deviation between the estimated inundation flow and the generated flow interval as the loss to train the LSTM model.

7. The method for predicting flood inundation losses in floodplains based on artificial intelligence according to claim 6, wherein It also includes obtaining the topographic cross-section data within the river channel range of any of the floodplain areas; After predicting the river channel flow of the future target basin based on the meteorological data and the upstream water level monitoring data by using the pre-trained LSTM model, it further includes: Estimate the possible temporary inundation peak according to the predicted river channel flow and the topographic cross-section data.

8. The method for predicting flood inundation losses in floodplains based on artificial intelligence according to claim 7, wherein According to the predicted river channel flow and the storage capacity of the floodplain area at each inundation elevation, estimate the inundation duration and the inundation loss, including: Determine the flow rate flowing into the floodplain area according to the partial topographic data of the floodplain area and the predicted river channel flow, and estimate the inundation height according to the inflow rate and the storage capacity of the floodplain area at each inundation elevation; Determine the inundation loss according to the estimated inundation height and the inundation elevation of any inundated object; Taking the inundation height reaching the preset height threshold as the starting point, determine the inundation duration according to the duration of the continuously predicted river channel flow exceeding the river channel capacity threshold characterized by the partial topographic data of the floodplain area.

9. An artificial intelligence-based prediction system for flood inundation losses in floodplains, characterized in that, It includes a processor and a memory, and a computer program is stored on the memory. When the computer program is executed by the processor, it realizes the steps of the method for predicting the inundation loss of the floodplain area based on artificial intelligence according to any one of claims 1 to 8.

Images