A flood prediction and deduction system, method, medium and device based on CNN model

Abstract

The present invention discloses a flood prediction and deduction system, method, medium, and device based on a CNN model. The method includes: step S1: determining the watershed for flood prediction and deduction, and collecting rainfall in the watershed; calculating the distribution weight coefficients for the rainfall collected by the rainfall monitoring station to be diffused to the tributaries on both sides; calculating the contribution rainfall data contributed by the rainfall monitoring stations on both sides of the tributary to the tributary; collecting flood water level rise data to form a training data set; constructing and training a CNN convolutional neural network model, and outputting the trained CNN convolutional neural network model; and calculating the water level rise derived from the main stream flood forecast point. The system includes a rainfall monitoring system and a data analysis module. A computer-readable storage medium stores a computer program for the above method. A terminal device executes the above method. The present invention solves the problems of low accuracy, poor timeliness, and insufficient processing capabilities for multi-source heterogeneous data in existing flood prediction methods.

Description

Technical Field

[0001] The present invention relates to the field of flood prediction, and in particular to a flood prediction and deduction system, method, medium and equipment based on a CNN model. Background Art

[0002] Flood disasters seriously threaten the safety of human life and property and socio-economic development. Accurate flood prediction and deduction are crucial for taking flood prevention and disaster reduction measures in advance. Traditional flood prediction methods, such as those based on physical models, often require a large number of parameter settings and complex calculations, and have extremely high requirements for data integrity and accuracy. In practical applications, due to factors such as complex underlying surface conditions in the basin and difficulties in obtaining meteorological data, the prediction accuracy and timeliness of traditional physical models are greatly limited. With the development of deep learning technology, prediction methods based on data-driven and high-precision processing have gradually been applied to the field of flood prediction. However, when processing flood-related rainfall data, existing deep learning models have problems such as inaccurate extracted rainfall and poor model generalization ability, making it difficult to meet the needs of high-precision flood prediction. Summary of the Invention

[0003] In response to the above-mentioned deficiencies in the prior art, the present invention provides a flood prediction and deduction system, method, medium and equipment based on the CNN model. Through high-precision rainfall data analysis, the diffusion contribution of rainfall data in the basin to tributaries is accurately obtained, ensuring high-precision flood prediction and deduction.

[0004] In order to achieve the above-mentioned object of the invention, the technical solution adopted by the present invention is:

[0005] A flood prediction and deduction method based on a CNN model is provided, which includes:

[0006] Step S1: Determine the river basin for flood prediction and deduction, obtain the main stream and tributaries of the river basin, and collect the rainfall in the river basin based on the rainfall monitoring stations set up within the river basin on both sides of the tributaries, as well as the three-dimensional coordinates of the corresponding rainfall monitoring stations distributed within the river basin;

[0007] Step S2: Analyze the diffusion of rainfall around the rainfall monitoring station to the tributaries on both sides based on the terrain data and distance data between the rainfall monitoring station and the tributaries, and calculate the distribution weight coefficients of the rainfall collected by the rainfall monitoring station to the tributaries on both sides;

[0008] Step S3: Calculate the contribution rainfall to the tributaries on both sides according to the allocation weight coefficient and the rainfall, and obtain the contribution rainfall data contributed by the rainfall monitoring stations on both sides of the tributary to the tributary;

[0009] Step S4: Collect historical rainfall data from rainfall monitoring stations on both sides of each tributary during the flood disaster process, construct a tributary rainfall data matrix, and collect flood water level rise data that changes with time after the historical tributary floods flow into the main stream to form a training data set;

[0010] Step S5: construct a CNN convolutional neural network model, train the CNN convolutional neural network model using a training data set, determine whether the CNN convolutional neural network model has converged based on a loss function, correct the weight coefficients and bias of the CNN convolutional neural network model, and output the trained CNN convolutional neural network model;

[0011] Step S6: Collect real-time rainfall data from rain monitoring stations on both sides of the tributary, calculate the actual contribution rainfall data diffused to the tributary, input it into the trained CNN convolutional neural network model, and calculate the water level rise height deduced from the main stream flood forecast point.

[0012] Furthermore, step S2 includes:

[0013] Step S21: Evenly select several reference points on the tributaries on both sides of the rainfall monitoring station so that the river lengths between the reference points are equal, and obtain the three-dimensional coordinates of each reference point , i is the number of the reference point;

[0014] Draw a perpendicular line from the location of the rainfall monitoring station to the tributary, obtain the perpendicular point between the perpendicular line and the tributary, and traverse the reference points on both sides of the perpendicular point along the tributary based on the perpendicular point, and calculate the altitude difference between the traversed reference points and the rainfall monitoring station , ;

[0015] Step S22: Filter out all altitude differences The reference point u is connected to the rainfall monitoring station, and several path points are taken on the connection line to obtain the altitude z of each path point. u , if there is a path point with an altitude , then delete the reference point corresponding to the line where the path point is located, and e is the number of the path point;

[0016] At the same time, the distance threshold L0 between the rainfall monitoring station and the reference point is set;

[0017] like , then it is determined that the rainfall around the rainfall monitoring station cannot spread to the tributary section where the reference point u is located, and the reference point u is also deleted; is the three-dimensional coordinate of the reference point u, is the three-dimensional coordinate of rainfall monitoring station n;

[0018] Step S23: Obtain the number U of rainfall diffusion paths established between the remaining reference points and the rainfall monitoring stations, and the average length of the rainfall diffusion paths established between the remaining reference points and the rainfall monitoring stations. ;

[0019] Step S24: Repeat steps S21-S23 to obtain the distribution weight coefficients of the number of diffusion paths and the length of the diffusion paths when the rainfall collected by rainfall monitoring station n diffuses to the tributaries on both sides;

[0020] ;

[0021] in, are the distribution weight coefficients of the number of diffusion paths when diffusing to the two tributaries, are the distribution weight coefficients of the diffusion path length when diffusing to the two tributaries, are the number of diffusion paths corresponding to the tributaries on both sides, are the average values ​​of the diffusion path lengths corresponding to the tributaries on both sides;

[0022] Step S25: Based on the average value of the altitude differences between the remaining reference points and the rainfall monitoring stations , calculate the distribution weight coefficient of the terrain factor when the rainfall at the rainfall monitoring station n spreads to the tributaries on both sides ;

[0023] ;

[0024] in, are the altitudes of the remaining reference points on both sides of the tributaries.

[0025] Furthermore, step S3 includes:

[0026] Step S31: Calculate the rainfall according to the distribution weight coefficient The contribution of rainfall to the tributaries on both sides under the conditions ;

[0027] ;

[0028] Step S32: Obtain the rainfall collected by the rainfall monitoring stations distributed on both sides of the tributary Contributing rainfall data for tributaries ; j is the number of the tributary, The contribution rainfall corresponding to the Nth rainfall monitoring station distributed on both sides of the tributary, N is the number of rainfall monitoring stations distributed on both sides of the tributary;

[0029] ;

[0030] in, For rainfall The distribution weight coefficient for the number of diffusion paths when diffusing to the jth tributary, For rainfall The distribution weight coefficient of the diffusion path length when diffusing to the jth tributary, For rainfall The distribution weight coefficient of terrain factors when diffusing to the jth tributary, The contribution rainfall corresponding to the nth rainfall monitoring station distributed on both sides of the tributary.

[0031] Furthermore, step S4 includes:

[0032] Step S41: Collect historical rainfall data from rainfall monitoring stations on both sides of each tributary during the flood disaster, calculate the contribution of each rainfall data to the tributary, and construct the rainfall data matrix of the tributary. ;

[0033] ;

[0034] Among them, T is the time when historical rainfall data was collected, is the contribution rainfall corresponding to the Nth rainfall monitoring station on both sides of the tributary at time T;

[0035] Step S42: Collect the flood level rise data of the flood flowing into the main stream at each moment during the flood disaster in the historical period 1-T. , , is the average time that flood water flows in the tributary, is the height of the flood water level at time t;

[0036] Step S43: Rainfall data matrix As input data, flood water level rise height data As output data, a training dataset is formed.

[0037] Furthermore, step S5 includes:

[0038] Step S51: Construct a CNN convolutional neural network model. The convolution kernel size of the CNN convolutional neural network model is , the activation function of the CNN convolutional neural network model is:

[0039] ;

[0040] in, is the output function, is the input function, is the size of the convolution kernel, is the contribution rainfall corresponding to the nth rainfall monitoring station on both sides of the tributary at time t, is the weight coefficient, is bias;

[0041] Step S52: Initialize the weight coefficients of the CNN convolutional neural network model, and input the training data set into the CNN convolutional neural network model for training. The loss function L of the training process is:

[0042] ;

[0043] in, is the number of training samples, is the predicted value;

[0044] Step S53: Setting the loss function threshold ,but , the CNN convolutional neural network model has converged, and the trained CNN convolutional neural network model is output; otherwise, the weight coefficients and biases are updated, and the process returns to step S52 to train the CNN convolutional neural network model with the updated weight coefficients and biases until the CNN convolutional neural network model converges;

[0045] The update method of weight coefficient and bias is:

[0046] ;

[0047] Among them, g is the number of weight coefficient and bias update iterations, are the weight coefficients and biases after updating iterations, are the weight coefficient and bias before updating iteration respectively, Represents the weight coefficient Including weight coefficient and weight coefficient , Indicates bias Including bias and bias , is the learning rate.

[0048] Furthermore, step S6 includes:

[0049] Step S61: Collect the real-time rainfall data from the rain monitoring stations on both sides of the tributary, and execute steps S31 and S32 to calculate the actual contribution rainfall data to the tributary, and input it into the trained CNN convolutional neural network model to output the time The predicted rise in the main stream water level ;

[0050] Step S62: Calculate the flood forecast point at time M based on the number of tributaries M upstream of the flood forecast point on the main stream. The subsequent water level rise .

[0051] A flood prediction and deduction system is provided for executing the above-mentioned flood prediction and deduction method based on the CNN model, which comprises:

[0052] A rainfall monitoring system installed in the watershed includes a number of rainfall monitoring stations distributed throughout the watershed, and a wireless communication module installed in each monitoring station. The rainfall data collected by the monitoring stations is uploaded to the rainfall monitoring system via the wireless communication module.

[0053] The data analysis module is equipped with a CNN convolutional neural network model. The rainfall monitoring system transmits rainfall data to the data analysis module, which uses the CNN convolutional neural network model to predict the flood level and output the height of the flood level rise.

[0054] A computer-readable storage medium is provided, which stores a computer program. When the computer program is executed by a processor, the above-mentioned flood prediction and deduction method based on the CNN model is executed.

[0055] A terminal device is provided, which is used to call and run a computer program from a memory, so that the terminal device executes the above-mentioned flood prediction and deduction method based on the CNN model.

[0056] The beneficial effects of the present invention are as follows: By rationally analyzing topographic and distance data on both sides of a tributary, the present invention accurately obtains rainfall data contributed by the basin to the tributaries on both sides, achieving high-precision rainfall data analysis, and obtaining the diffusion contribution of rainfall data within the basin to the tributaries. This provides high-precision training data for training CNN models, ensuring high-precision flood prediction and deduction. This solves the problems of low accuracy, poor timeliness, and insufficient processing capacity for multi-source heterogeneous data in existing flood prediction methods, improves the accuracy and reliability of flood prediction, and provides strong support for flood prevention and disaster reduction decision-making. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] Figure 1 The flowchart of the flood prediction and deduction method based on the CNN model.

[0058] Figure 2 Schematic diagram of the principle of allocating weight coefficient analysis. DETAILED DESCRIPTION

[0059] The specific embodiments of the present invention are described below to facilitate understanding of the present invention by those skilled in the art. However, it should be clear that the present invention is not limited to the scope of the specific embodiments. For those skilled in the art, as long as various changes are within the spirit and scope of the present invention as defined and determined by the appended claims, these changes are obvious, and all inventions and creations utilizing the concepts of the present invention are protected.

[0060] like Figure 1 As shown in FIG, a flood prediction and deduction method based on a CNN model includes:

[0061] Step S1: Determine the river basin for flood prediction and deduction, obtain the main stream and tributaries of the river basin, and collect the rainfall in the river basin according to the rainfall monitoring stations set up within the river basin on both sides of the tributaries. , and the three-dimensional coordinates of the corresponding rainfall monitoring stations distributed in the basin , n is the number of the rainfall monitoring station, It represents the horizontal distribution coordinates of the rainfall monitoring station in the watershed. The three-dimensional coordinates can be obtained through the three-dimensional map of the watershed. The coordinates of the rainfall monitoring station can be marked on the three-dimensional map. Indicates the altitude of the rainfall monitoring station in the basin. It is expressed as precipitation per unit time (for example, rainfall is 50 mm / h).

[0062] Step S2: Based on the terrain data and distance data between the rainfall monitoring station and the tributaries, analyze the state of rainfall diffusion around the rainfall monitoring station to the tributaries on both sides, and calculate the rainfall collected by the rainfall monitoring station The distribution weight coefficients for diffusion to the tributaries on both sides are respectively included in the following steps:

[0063] Step S21: Figure 2 As shown, several reference points are evenly selected on the tributaries on both sides of the rainfall monitoring station so that the river lengths between the reference points are equal, and the three-dimensional coordinates of each reference point are obtained. , i is the number of the reference point, draw a perpendicular line from the location of the rainfall monitoring station to the tributary, get the perpendicular point between the perpendicular line and the tributary, traverse the reference points on both sides of the perpendicular point along the tributary based on the perpendicular point, and calculate the altitude difference between the traversed reference point and the rainfall monitoring station , ;

[0064] Step S22: Filter out all altitude differences The reference point u is connected to the rainfall monitoring station, and several path points are taken on the connection line to obtain the altitude z of each path point. u , if there is a path point with an altitude , then delete the reference point corresponding to the line where the path point is located, and e is the number of the path point;

[0065] At the same time, the distance threshold L0 between the rainfall monitoring station and the reference point is set;

[0066] like , it is determined that the distance between the rainfall monitoring station and the reference point is too far, and the rainfall cannot reach the tributary section where the reference point is located when flowing to the tributary. It is then determined that the rainfall around the rainfall monitoring station cannot spread to the tributary section where the reference point u is located, and the reference point u is also deleted; is the three-dimensional coordinate of the reference point u, is the three-dimensional coordinate of rainfall monitoring station n;

[0067] Step S23: Obtain the number U of rainfall diffusion paths established between the remaining reference points and the rainfall monitoring stations, and the average length of the rainfall diffusion paths established between the remaining reference points and the rainfall monitoring stations. ;

[0068] Step S24: Repeat steps S21-S23 to obtain the distribution weight coefficients of the number of diffusion paths and the length of the diffusion paths when the rainfall collected by rainfall monitoring station n diffuses to the tributaries on both sides;

[0069] ;

[0070] in, are the distribution weight coefficients of the number of diffusion paths when diffusing to the two tributaries, are the distribution weight coefficients of the diffusion path length when diffusing to the two tributaries, are the number of diffusion paths corresponding to the tributaries on both sides, are the average values ​​of the diffusion path lengths corresponding to the tributaries on both sides;

[0071] Step S25: Based on the average value of the altitude differences between the remaining reference points and the rainfall monitoring stations , calculate the distribution weight coefficient of the terrain factor when the rainfall at the rainfall monitoring station n spreads to the tributaries on both sides ;

[0072] ;

[0073] in, are the altitudes of the remaining reference points on both sides of the tributaries.

[0074] Step S3: According to the distribution weight coefficient and rainfall Calculate the contribution rainfall to the tributaries on both sides, and obtain the contribution rainfall data contributed by the rain monitoring stations on both sides of the tributary to the tributary. Step S3 specifically includes:

[0075] Step S31: Calculate the rainfall according to the distribution weight coefficient The contribution of rainfall to the tributaries on both sides under the conditions ;

[0076] ;

[0077] Step S32: Obtain the rainfall collected by the rainfall monitoring stations distributed on both sides of the tributary Contributing rainfall data for tributaries ; j is the number of the tributary, The contribution rainfall corresponding to the Nth rainfall monitoring station distributed on both sides of the tributary, N is the number of rainfall monitoring stations distributed on both sides of the tributary;

[0078] ;

[0079] in, For rainfall The distribution weight coefficient for the number of diffusion paths when diffusing to the jth tributary, For rainfall The distribution weight coefficient of the diffusion path length when diffusing to the jth tributary, For rainfall The distribution weight coefficient of terrain factors when diffusing to the jth tributary, The contribution rainfall corresponding to the nth rainfall monitoring station distributed on both sides of the tributary.

[0080] In this embodiment, the rainfall amount distributed to the two tributaries based on the rainfall collected by the same rainfall monitoring station is first calculated. According to different distribution weight coefficients, the rainfall amount distributed to the two tributaries is different, and the distributed rainfall amount is determined by at least three distribution weight coefficients, which effectively increases the rationality of the rainfall distribution calculation. For example, if the rainfall collected by one rainfall monitoring station is 50 mm / h, and the corresponding three distribution weights calculated for the two tributaries are respectively , then the contribution rainfall allocated to the tributaries on both sides based on the rainfall collected by the rainfall monitoring station is .

[0081] Step S4: Collect historical rainfall data from rainfall monitoring stations on both sides of each tributary during the flood disaster, construct a tributary rainfall data matrix, and collect historical flood water level rise data after the tributary floods flow into the main stream, forming a training data set. Step S4 specifically includes:

[0082] Step S41: Collect historical rainfall data from rainfall monitoring stations on both sides of each tributary during the flood disaster, calculate the contribution of each rainfall data to the tributary, and construct the rainfall data matrix of the tributary. ;

[0083] ;

[0084] Among them, T is the time when historical rainfall data was collected, is the contribution rainfall corresponding to the Nth rainfall monitoring station on both sides of the tributary at time T;

[0085] Step S42: Collect the flood level rise data of the flood flowing into the main stream at each moment during the flood disaster in the historical period 1-T. , , is the average time for flood to flow in the tributary. Since the time difference of flood to flow to the main stream in the basin of tributary is small, Take the average time it takes for tributary floods to move toward the main stream during historical flood disasters. is the height of the flood water level at time t;

[0086] Step S43: Rainfall data matrix As input data, flood water level rise height data As output data, a training dataset is formed.

[0087] Step S5: Construct a CNN convolutional neural network model, train the CNN convolutional neural network model using the training data set, determine whether the CNN convolutional neural network model has converged based on the loss function, correct the weight coefficients and bias of the CNN convolutional neural network model, and output the trained CNN convolutional neural network model. Step S5 specifically includes:

[0088] Step S51: Construct a CNN convolutional neural network model. The convolution kernel size of the CNN convolutional neural network model is , the activation function of the CNN convolutional neural network model is:

[0089] ;

[0090] in, is the output function, is the input function, is the size of the convolution kernel, is the contribution rainfall corresponding to the nth rainfall monitoring station on both sides of the tributary at time t, is the weight coefficient, is bias;

[0091] Step S52: Initialize the weight coefficients of the CNN convolutional neural network model, and input the training data set into the CNN convolutional neural network model for training. The loss function L of the training process is:

[0092] ;

[0093] in, is the number of training samples, is the predicted value;

[0094] Step S53: Setting the loss function threshold ,but , the CNN convolutional neural network model has converged, and the trained CNN convolutional neural network model is output; otherwise, the weight coefficients and biases are updated, and the process returns to step S52 to train the CNN convolutional neural network model with the updated weight coefficients and biases until the CNN convolutional neural network model converges;

[0095] The update method of weight coefficient and bias is:

[0096] ;

[0097] Among them, g is the number of weight coefficient and bias update iterations, are the weight coefficients and biases after updating iterations, are the weight coefficient and bias before updating iteration respectively, Represents the weight coefficient Including weight coefficient and weight coefficient , Indicates bias Including bias and bias , is the learning rate.

[0098] Step S6: Collect real-time rainfall data from rainfall monitoring stations on both sides of the tributary, calculate the actual contribution of rainfall to the tributary, input it into the trained CNN convolutional neural network model, and calculate the water level rise height deduced from the main stream flood forecast point. Step S6 specifically includes:

[0099] Step S61: Collect the real-time rainfall data from the rain monitoring stations on both sides of the tributary, and execute steps S31 and S32 to calculate the actual contribution rainfall data to the tributary, and input it into the trained CNN convolutional neural network model to output the time The predicted rise in the main stream water level ;

[0100] Step S62: Calculate the flood forecast point at time M based on the number of tributaries M upstream of the flood forecast point on the main stream. The subsequent water level rise .

[0101] A flood prediction and deduction system is used to implement the above-mentioned flood prediction and deduction method based on the CNN model, comprising:

[0102] A rainfall monitoring system installed in the watershed includes a number of rainfall monitoring stations distributed throughout the watershed, and a wireless communication module installed in each monitoring station. The rainfall data collected by the monitoring stations is uploaded to the rainfall monitoring system via the wireless communication module.

[0103] The data analysis module is equipped with a CNN convolutional neural network model. The rainfall monitoring system transmits rainfall data to the data analysis module, which uses the CNN convolutional neural network model to predict the flood level and output the height of the flood level rise.

[0104] A computer-readable storage medium stores a computer program, which, when executed by a processor, executes the flood prediction and deduction method based on the CNN model.

[0105] A terminal device is used to call and run a computer program from a memory, so that the terminal device executes the above-mentioned flood prediction and deduction method based on the CNN model.

[0106] This method uses terrain and distance data from both sides of a tributary to analyze accurately, obtaining rainfall data from the basin to the tributaries on both sides. This allows for high-precision rainfall data analysis, capturing the diffusion contribution of rainfall data within the basin to the tributaries, and generating high-precision training data for training CNN models, ensuring high-precision flood prediction and deduction. This method addresses the low accuracy, poor timeliness, and insufficient processing capacity of multi-source heterogeneous data found in existing flood prediction methods, improving the accuracy and reliability of flood predictions and providing strong support for flood prevention and disaster reduction decision-making.

Claims

1. A flood prediction and deduction method based on a CNN model, characterized in that: include: Step S1: Determine the river basin for flood prediction and deduction, obtain the main stream and tributaries of the river basin, and collect the rainfall in the river basin based on the rainfall monitoring stations set up within the river basin on both sides of the tributaries, as well as the three-dimensional coordinates of the corresponding rainfall monitoring stations distributed within the river basin; Step S2: Analyze the diffusion of rainfall around the rainfall monitoring station to the tributaries on both sides based on the terrain data and distance data between the rainfall monitoring station and the tributaries, and calculate the distribution weight coefficients of the rainfall collected by the rainfall monitoring station to the tributaries on both sides; Step S3: Calculate the contribution rainfall to the tributaries on both sides according to the allocation weight coefficient and the rainfall, and obtain the contribution rainfall data contributed by the rainfall monitoring stations on both sides of the tributary to the tributary; Step S4: Collect historical rainfall data from rainfall monitoring stations on both sides of each tributary during the flood disaster process, construct a tributary rainfall data matrix, and collect flood water level rise data that changes with time after the historical tributary floods flow into the main stream to form a training data set; Step S5: construct a CNN convolutional neural network model, train the CNN convolutional neural network model using a training data set, determine whether the CNN convolutional neural network model has converged based on a loss function, correct the weight coefficients and bias of the CNN convolutional neural network model, and output the trained CNN convolutional neural network model; Step S6: Collect real-time rainfall data from rain monitoring stations on both sides of the tributary, calculate the actual contribution rainfall data diffused to the tributary, input it into the trained CNN convolutional neural network model, and calculate the water level rise height deduced from the main stream flood forecast point.

2. The flood prediction and deduction method based on the CNN model according to claim 1 is characterized in that: The step S2 comprises: Step S21: Evenly select several reference points on the tributaries on both sides of the rainfall monitoring station so that the river lengths between the reference points are equal, and obtain the three-dimensional coordinates of each reference point , i is the number of the reference point; Draw a perpendicular line from the location of the rainfall monitoring station to the tributary, obtain the perpendicular point between the perpendicular line and the tributary, and traverse the reference points on both sides of the perpendicular point along the tributary based on the perpendicular point, and calculate the altitude difference between the traversed reference points and the rainfall monitoring station , ; Step S22: Filter out all altitude differences The reference point u is connected to the rainfall monitoring station, and several path points are taken on the connection line to obtain the altitude z of each path point. u , if there is a path point with an altitude , then delete the reference point corresponding to the line where the path point is located, and e is the number of the path point; At the same time, the distance threshold L0 between the rainfall monitoring station and the reference point is set; like , then it is determined that the rainfall around the rainfall monitoring station cannot spread to the tributary section where the reference point u is located, and the reference point u is also deleted; is the three-dimensional coordinate of the reference point u, is the three-dimensional coordinate of rainfall monitoring station n; Step S23: Obtain the number U of rainfall diffusion paths established between the remaining reference points and the rainfall monitoring stations, and the average length of the rainfall diffusion paths established between the remaining reference points and the rainfall monitoring stations. ; Step S24: Repeat steps S21-S23 to obtain the distribution weight coefficients of the number of diffusion paths and the length of the diffusion paths when the rainfall collected by rainfall monitoring station n diffuses to the tributaries on both sides; ； in, are the distribution weight coefficients of the number of diffusion paths when diffusing to the two tributaries, are the distribution weight coefficients of the diffusion path length when diffusing to the two tributaries, are the number of diffusion paths corresponding to the tributaries on both sides, are the average values ​​of the diffusion path lengths corresponding to the tributaries on both sides; Step S25: Based on the average value of the altitude differences between the remaining reference points and the rainfall monitoring stations , calculate the distribution weight coefficient of the terrain factor when the rainfall at the rainfall monitoring station n spreads to the tributaries on both sides ; ； in, are the altitudes of the remaining reference points on both sides of the tributaries.

3. The flood prediction and deduction method based on the CNN model according to claim 2 is characterized in that: The step S3 comprises: Step S31: Calculate the rainfall according to the distribution weight coefficient The contribution of rainfall to the tributaries on both sides under the conditions ; ； Step S32: Obtain the rainfall collected by the rainfall monitoring stations distributed on both sides of the tributary Contributing rainfall data for tributaries ; j is the number of the tributary, The contribution rainfall corresponding to the Nth rainfall monitoring station distributed on both sides of the tributary, N is the number of rainfall monitoring stations distributed on both sides of the tributary; ； in, For rainfall The distribution weight coefficient for the number of diffusion paths when diffusing to the jth tributary, For rainfall The distribution weight coefficient of the diffusion path length when diffusing to the jth tributary, For rainfall The distribution weight coefficient of terrain factors when diffusing to the jth tributary, The contribution rainfall corresponding to the nth rainfall monitoring station distributed on both sides of the tributary.

4. The flood prediction and deduction method based on the CNN model according to claim 3 is characterized in that: The step S4 comprises: Step S41: Collect historical rainfall data from rainfall monitoring stations on both sides of each tributary during the flood disaster, calculate the contribution of each rainfall data to the tributary, and construct the rainfall data matrix of the tributary. ; ； Among them, T is the time when historical rainfall data was collected, is the contribution rainfall corresponding to the Nth rainfall monitoring station on both sides of the tributary at time T; Step S42: Collect the flood level rise data of the flood flowing into the main stream at each moment during the flood disaster in the historical period 1-T. , , is the average time that flood water flows in the tributary, is the height of the flood water level at time t; Step S43: Rainfall data matrix As input data, flood water level rise height data As output data, a training dataset is formed.

5. The flood prediction and deduction method based on the CNN model according to claim 4 is characterized in that: The step S5 comprises: Step S51: Construct a CNN convolutional neural network model. The convolution kernel size of the CNN convolutional neural network model is , the activation function of the CNN convolutional neural network model is: ； in, is the output function, is the input function, is the size of the convolution kernel, is the contribution rainfall corresponding to the nth rainfall monitoring station on both sides of the tributary at time t, is the weight coefficient, is bias; Step S52: Initialize the weight coefficients of the CNN convolutional neural network model, and input the training data set into the CNN convolutional neural network model for training. The loss function L of the training process is: ； in, is the number of training samples, is the predicted value; Step S53: Setting the loss function threshold ,but , the CNN convolutional neural network model has converged, and the trained CNN convolutional neural network model is output; otherwise, the weight coefficients and biases are updated, and the process returns to step S52 to train the CNN convolutional neural network model with the updated weight coefficients and biases until the CNN convolutional neural network model converges; The update method of weight coefficient and bias is: ； Among them, g is the number of weight coefficient and bias update iterations, are the weight coefficients and biases after updating iterations, are the weight coefficient and bias before updating iteration respectively, Represents the weight coefficient Including weight coefficient and weight coefficient , Indicates bias Including bias and bias , is the learning rate.

6. The flood prediction and deduction method based on the CNN model according to claim 5 is characterized in that: The step S6 comprises: Step S61: Collect the real-time rainfall data from the rain monitoring stations on both sides of the tributary, and execute steps S31 and S32 to calculate the actual contribution rainfall data to the tributary, and input it into the trained CNN convolutional neural network model to output the time The predicted rise in the main stream water level ; Step S62: Calculate the flood forecast point at time M based on the number of tributaries M upstream of the flood forecast point on the main stream. The subsequent water level rise .

7. A flood prediction and deduction system for executing the flood prediction and deduction method based on the CNN model according to any one of claims 1 to 6, characterized in that: include: A rainfall monitoring system installed in a watershed, the rainfall monitoring system comprising a plurality of rainfall monitoring stations distributed in the watershed, and a wireless communication module provided in each monitoring station, wherein rainfall data collected by the monitoring stations is uploaded to the rainfall monitoring system via the wireless communication module; The data analysis module is equipped with a CNN convolutional neural network model. The rainfall monitoring system transmits rainfall data to the data analysis module, which uses the CNN convolutional neural network model to predict the flood level and output the height of the flood level rise.

8. A computer-readable storage medium, characterized in that A computer program is stored, and when the computer program is run by a processor, the flood prediction and deduction method based on the CNN model as described in any one of claims 1 to 6 is executed.

9. A terminal device, characterized in that: The terminal device is used to call and run a computer program from a memory, so that the terminal device executes the flood prediction and deduction method based on the CNN model as described in any one of claims 1 to 6.

Images